HIGH SPEED DESIGN TECHNIQUES

HIGH SPEED DESIGN TECHNIQUES
HIGH SPEED
DESIGN TECHNIQUES
PREFACE
HIGH SPEED OP AMPS
HIGH SPEED OP AMP APPLICATIONS
RF/IF SUBSYSTEMS
HIGH SPEED SAMPLING AND HIGH SPEED ADCs
HIGH SPEED ADC APPLICATIONS
HIGH SPEED DACs AND DDS SYSTEMS
HIGH SPEED HARDWARE DESIGN TECHNIQUES
INDEX
ANALOG DEVICES TECHNICAL REFERENCE BOOKS
PUBLISHED BY PRENTICE HALL
Analog-Digital Conversion Handbook
Digital Signal Processing Applications Using the ADSP-2100 Family
(Volume 1:1992, Volume 2:1994)
Digital Signal Processing in VLSI
DSP Laboratory Experiments Using the ADSP-2101
ADSP-2100 Family User's Manual
PUBLISHED BY ANALOG DEVICES
High Speed Design Techniques
Practical Analog Design Techniques
Linear Design Seminar
ADSP-21000 Family Applications Handbook
System Applications Guide
Applications Reference Manual
Amplifier Applications Guide
Mixed Signal Design Seminar Notes
High-Speed Design Seminar Notes
Nonlinear Circuits Handbook
Transducer Interfacing Handbook
Synchro & Resolver Conversion
THE BEST OF Analog Dialogue, 1967-1991
HOW TO GET INFORMATION FROM ANALOG DEVICES
Analog Devices publishes data sheets and a host of other technical literature
supporting our products and technologies. Follow the instructions below for
worldwide access to this information.
FOR DATA SHEETS
U.S.A. and Canada
n
Fax Retrieval. Telephone number 800-446-6212. Call this number
and use a faxcode corresponding to the data sheet of your choice
for a fax-on-demand through our automated AnalogFax™ system.
Data sheets are available 7 days a week, 24 hours a day.
Product/faxcode cross reference listings are available by calling
the above number and following the prompts. There is a short
index with just part numbers, faxcodes, page count and revision
for each data sheet. There is also a longer index sorted by
product type with short descriptions.
n
World Wide Web and Internet. Our address is http://www.analog.com.
Use the browser of your choice and follow the prompts. We also
provide extensive DSP literature support on an Internet FTP site. Type
ftp ftp.analog.com or ftp 137.71.23.11. Log in as anonymous using your
e-mail address for your password.
n
Analog Devices Literature Distribution Center. Call 800-262-5643 and
select option two from the voice prompts, or call 617-461-3392 for
direct access, or fax your request to 617-821-4273.
n
DSP Bulletin Board Service. For the latest updates, call 617-461-4258,
8 data bits, 1 stop bit, no parity, 300bps to 14.4kbps.
Europe and Israel
n
Fax Retrieval. Telephone number 49-8765-9300-xxxx, where xxxx is
the faxcode. For a list of faxcodes dial 49-8765-9300-1000. (1000 is
the code for the faxcode cross-reference listing.)
n
World Wide Web. Our address is http://www.analog.com. use the
browser of your choice and follow the prompts.
n
Analog Devices Sales Offices. Call your local sales office and request
a data sheet. A Worldwide Sales Directory including telephone listings
is on pages 218 and 219 of the 1996 Short Form Designers' Guide.
n
DSP Support Center. Fax requests to **49-89-57005-200 or e-mail
[email protected] The Bulletin Board Service is at
**43-1-8887656.
Australia and New Zealand
n
Fax Retrieval. Telephone number 61-59-864377. Follow the voice
prompts.
n
Call 91-80-526-3606 or fax 91-80-526-3713 and request the data
sheet of interest.
India
Other Locations
n
World Wide Web. Our address is http://www.analog.com. Use the
browser of your choice and follow the prompts.
n
Analog Devices Sales Offices. Call your local sales office and request
a data sheet. A Worldwide Sales Directory including telephone numbers
is listed on pages 218 and 219 of the 1996 Short Form Designers' Guide.
TECHNICAL SUPPORT AND CUSTOMER SERVICE
n
In the U.S.A. and Canada, call 800-ANALOGD, (800-262-5643).
For technical support on all products, select option one, then select
the product area of interest. For price and delivery, select option three.
For literature and samples, select option two.
Non-800 Number: 617-937-1428
HIGH SPEED
DESIGN TECHNIQUES
a
ACKNOWLEDGMENTS
Thanks are due the many technical staff members of Analog Devices in Engineering and
Marketing who provided invaluable inputs during this project. Particular credit is due the
individual authors whose names appear at the beginning of their material.
Special thanks go to Adolfo Garcia, Walter G. Jung, and Ed Grokulsky for thoroughly
reviewing the material for content and accuracy.
Linda Grimes Brandon of Brandon's WordService prepared the new illustrations and typeset
the text. Ernie Lehtonen of the Analog Devices' art department supplied many camera-ready
drawings. Judith Douville compiled the index, and printing was done by R. R. Donnelley and
Sons, Inc.
Walt Kester
1996
Copyright  1996 by Analog Devices, Inc.
Printed in the United States of America
All rights reserved. This book, or parts thereof, must not be reproduced in any form without
permission of the copyright owner.
Information furnished by Analog Devices, Inc., is believed to be accurate and reliable.
However, no responsibility is assumed by Analog Devices, Inc., for its use.
Analog Devices, Inc., makes no representation that the interconnections of its circuits as
described herein will not infringe on existing or future patent rights, nor do the descriptions
contained herein imply the granting of licenses to make, use, or sell equipment constructed in
accordance therewith.
Specifications are subject to change without notice.
ISBN-0-916550-17-6
HIGH SPEED
DESIGN TECHNIQUES
PREFACE
SECTION 1
HIGH SPEED OP AMPS
n
Voltage Feedback Op Amps
n
Current Feedback Op Amps
n
Effects of Feedback Capacitance
n
High-Speed Current-to-Voltage Converters, and
the Effects of Inverting Input Capacitance
n
Noise Comparisons between Voltage Feedback
Op Amps and Current Feedback Op Amps
n
DC Characteristics of High Speed Op Amps
SECTION 2
HIGH SPEED OP AMP APPLICATIONS
n
Optimizing the Feedback Network for Maximum
Bandwidth Flatness in Wideband CFB Op Amps
n
Driving Capacitive Loads
n
Cable Drivers and Receivers
n
A High Performance Video Line Driver
n
Differential Line Drivers/Receivers
n
High Speed Clamping Amplifiers
n
Single-Supply/Rail-to-Rail Considerations
n
High Speed Video Multiplexing with Op Amps Using
Disable Function
n
Video Programmable Gain Amplifier
n
Video Multiplexers and Crosspoint Switches
n
High Power Line Drivers and ADSL
n
High Speed Photodiode Preamps
SECTION 3
RF/IF SUBSYSTEMS
n
Dynamic Range Compression
n
Linear VCAs
n
Log/Limiting Amplifiers
n
Receiver Overview
n
Multipliers, Modulators, and Mixers
n
Modulation / Demodulation
n
Receiver Subsystems
SECTION 4
HIGH SPEED SAMPLING AND HIGH SPEED ADCs
n
Fundamentals of High Speed Sampling
n
Baseband Antialiasing Filters
n
Undersampling
n
Antialiasing Filters in Undersampling Applications
n
Distortion and Noise in an Ideal N-bit ADC
n
Distortion and Noise in Practical ADCs
n
High Speed ADC Architectures
SECTION 5
HIGH SPEED ADC APPLICATIONS
n
Driving ADC Inputs for Low Distortion and Wide
Dynamic Range
n
Applications of High Speed ADCs in CCD Imaging
n
High Speed ADC Applications in Digital Receivers
SECTION 6
HIGH SPEED DACs AND DDS SYSTEMS
n
Introduction to DDS
n
Aliasing in DDS Systems
n
125MSPS DDS System (AD9850)
n
DDS Systems as ADC Clock Drivers
n
Amplitude Modulation in a DDS System
n
The AD9831/AD9832 Complete DDS System
n
Spurious Free Dynamic Range Considerations in
DDS Systems
n
High Speed Low Distortion DAC Architectures
n
Improving SFDR Using Sample-and-Hold Deglitchers
n
High Speed Interpolating DACs
n
QPSK Signal Generation Using DDS (AD9853)
SECTION 7
HIGH SPEED HARDWARE DESIGN TECHNIQUES
INDEX
n
Analog Circuit Simulation
n
Prototyping Analog Circuits
n
Evaluation Boards
n
Grounding in High Speed Systems
n
Power Supply Noise Reduction and Filtering
n
Power Supply Regulation/Conditioning
n
Thermal Management
n
EMI/RFI Considerations
n
Shielding Concepts
HIGH SPEED
DESIGN TECHNIQUES
PREFACE
HIGH SPEED OP AMPS
HIGH SPEED OP AMP APPLICATIONS
RF/IF SUBSYSTEMS
HIGH SPEED SAMPLING AND HIGH SPEED ADCs
HIGH SPEED ADC APPLICATIONS
HIGH SPEED DACs AND DDS SYSTEMS
HIGH SPEED HARDWARE DESIGN TECHNIQUES
INDEX
PREFACE:
HIGH SPEED DESIGN TECHNIQUES
High speed integrated circuits, both analog, digital, and mixed-signal are used in all
types of electronic equipment today. This book examines high speed linear ICs both
from the theoretical and practical application point of view.
Figure P.1 shows some of the typical applications for high speed integrated circuits
by market segment. Many applications can be filled using standard linear IC
products, while others may be better served with specially designed chipsets (see
Figure P.2).
All of these high speed linear ICs depend upon a broad base of high speed core
competencies shown in Figure P.3. Analog Devices has been a leader in real-world
signal processing for over 30 years and has the required expertise in each critical
competency area. Regardless of how complex or highly integrated mixed-signal ICs
may become, there is no escaping the requirement for these basic building blocks.
An understanding of these building blocks is required for the customer to
successfully specify, select, and apply new high speed products at the system level.
While a detailed knowledge of the internal circuits is not required, an overall
understanding of the operation of the devices is critical to success.
This book is not intended to be a system design manual. Instead, it covers the theory
and application of many high speed analog signal processing building blocks such as
amplifiers, ADCs, DACs, etc. System applications are presented when they are of
broad general interest or illustrate emerging market trends.
The proper application of high speed devices also requires a thorough knowledge of
good hardware design techniques, such as simulation, prototyping, layout,
decoupling, and grounding. The last section in the book focuses on these issues as
well as EMI and RFI design considerations.
1
HIGH SPEED PRODUCTS: TYPICAL APPLICATIONS
VIDEO
IMAGING
COMMUNICATIONS
INSTRUMENTATION
u Cameras
u Medical
u Cellular:
u Oscilloscopes
Broadband
Narrowband
u Mixing
u Distribution
u
Scanners
u Direct Broadcast
u Copiers
u Hybrid Fiber Coax
Satellite
(HFC)
u Video
u Lasers
u CATV
Conferencing
u Spectrum
Analyzers
u Frequency
Synthesizers
u Automatic Test
Equipment
u Displays
u CCD
u ADSL/HDSL
u MPEG Systems
uRadar/So
nar
u Data Recovery and
uData Acquisition
Retiming
a
P.1
ADI HIGH SPEED INTEGRATED / CHIPSET SOLUTIONS
n
Cellular Communications: GSM, DECT,
AMPS, PCS, etc. (Handsets and Basestations)
n
ADSL/HDSL
n
CCD Imaging
n
Video Signal Processing (MPEG, etc.)
n
Fiber Optic and Disk Drive Data Recovery
n
Direct Broadcast Satellite Receivers
n
High Speed Modems
n
Multimedia Sound and Video Processing
a
P.2
2
CORE COMPETENCIES: "DC TO LIGHT"
n
Amplifiers:
Op Amps, VCAs, PGAs, Log Amps,
Sample-and-Hold Amplifiers
n
Switches and Multiplexers
n
Analog-to-Digital Converters (ADCs)
n
Digital-to-Analog Converters (DACs)
n
Analog Signal Processing
Multipliers, RMS-DC Converters, etc.
n
RF/IF Signal Processing
n
DSP
a
P.3
3
SECTION 1
HIGH SPEED OPERATIONAL AMPLIFIERS
Walt Kester
INTRODUCTION
High speed analog signal processing applications, such as video and
communications, require op amps which have wide bandwidth, fast settling time,
low distortion and noise, high output current, good DC performance, and operate at
low supply voltages. These devices are widely used as gain blocks, cable drivers,
ADC pre-amps, current-to-voltage converters, etc. Achieving higher bandwidths for
less power is extremely critical in today's portable and battery-operated
communications equipment. The rapid progress made over the last few years in
high-speed linear circuits has hinged not only on the development of IC processes
but also on innovative circuit topologies.
The evolution of high speed processes by using amplifier bandwidth as a function of
supply current as a figure of merit is shown in Figure 1.1. (In the case of duals,
triples, and quads, the current per amplifier is used). Analog Devices BiFET process,
which produced the AD712 and OP249 (3MHz bandwidth, 3mA current), yields
about 1MHz per mA. The CB (Complementary Bipolar) process (AD817, AD847,
AD811, etc.) yields about 10MHz/mA of supply current. Ft's of the CB process PNP
transistors are about 700MHz, and the NPN's about 900MHz.
The latest generation complementary bipolar process from Analog Devices is a
high speed dielectrically isolated process called XFCB (eXtra Fast Complementary
Bipolar). This process (2-4 GHz Ft matching PNP and NPN transistors), coupled
with innovative circuit topologies allow op amps to achieve new levels of costeffective performance at astonishing low quiescent currents. The approximate figure
of merit for this process is typically 100MHz/mA, although the AD8011 op amp is
capable of 300MHz bandwidth on 1mA of supply current due to its unique two-stage
current-feedback architecture.
1
AMPLIFIER BANDWIDTH VERSUS SUPPLY CURRENT
FOR ANALOG DEVICES' PROCESSES
1000
AD8001
AD8011
BANDWIDTH (MHz)
300
A
Rm
E
zP
MH
100
AD811
100
B:
FC
X
30
AD847
AD817
A
Rm
H
10M
B:
10
E
zP
C
ER
3
OP482
:
ET
BiF
1M
P
Hz
1
0.3
1
mA
AD712
OP249
741
3
10
30
SUPPLY CURRENT (PER AMPLIFIER), mA
a
1.1
In order to select intelligently the correct op amp for a given application, an
understanding of the various op amp topologies as well as the tradeoffs between
them is required. The two most widely used topologies are voltage feedback (VFB)
and current feedback (CFB). The following discussion treats each in detail and
discusses the similarities and differences.
VOLTAGE FEEDBACK (VFB) OP AMPS
A voltage feedback (VFB) op amp is distinguished from a current feedback (CFB) op
amp by circuit topology. The VFB op amp is certainly the most popular in low
frequency applications, but the CFB op amp has some advantages at high
frequencies. We will discuss CFB in detail later, but first the more traditional VFB
architecture.
Early IC voltage feedback op amps were made on "all NPN" processes. These
processes were optimized for NPN transistors, and the "lateral" PNP transistors had
relatively poor performance. Lateral PNPs were generally only used as current
sources, level shifters, or for other non-critical functions. A simplified diagram of a
typical VFB op amp manufactured on such a process is shown in Figure 1.2.
2
VOLTAGE FEEDBACK (VFB) OP AMP
DESIGNED ON AN "ALL NPN" IC PROCESS
+VS
I
T
"LATERAL" PNP
I
I
T
+
i = v•gm
T
VBIAS
2
2
Q3
Q2
Q1
A
"gm"
STAGE
v
Q4
-
C
P
I
I
T
R
vOUT
T
T
2
-V S
gm = Ic •
q
kT
vOUT =
a
i
jω
ωCP
=v•
gm
jω
ωCP
@ HF
1.2
The input stage is a differential pair consisting of either a bipolar pair (Q1, Q2) or a
FET pair. This "gm" (transconductance) stage converts the small-signal differential
input voltage, v, into a current, i, i.e., it's transfer function is measured in units of
conductance, 1/Ω, (or mhos). The small signal emitter resistance, re, is
approximately equal to the reciprocal of the small-signal gm. The formula for the
small-signal gm of a single bipolar transistor is given by the following equation:
gm =
q
q  IT 
1
IC ) =
=
(

 , or
re kT
kT  2 
 1   IT 
gm ≈ 

.
 26mV   2 
IT is the differential pair tail current, IC is the collector bias current (IC = IT/2), q is
the electron charge, k is Boltzmann's constant, and T is absolute temperature. At
+25°C, VT = kT/q= 26mV (often called the Thermal Voltage, VT).
As we will see shortly, the amplifier unity gain-bandwidth product, fu, is equal to
gm/2πCp, where the capacitance Cp is used to set the dominant pole frequency. For
this reason, the tail current, IT,is made proportional to absolute temperature
(PTAT). This current tracks the variation in re with temperature thereby making
gm independent of temperature. It is relatively easy to make Cp reasonably
constant over temperature.
The output of one side of the gm stage drives the emitter of a lateral PNP transistor
(Q3). It is important to note that Q3 is not used to amplify the signal, only to level
3
shift, i.e., the signal current variation in the collector of Q2 appears at the collector
of Q3. The output collector current of Q3 develops a voltage across high impedance
node A. Cp sets the dominant pole of the frequency response. Emitter follower Q4
provides a low impedance output.
The effective load at the high impedance node A can be represented by a resistance,
RT, in parallel with the dominant pole capacitance, Cp. The small-signal output
voltage, vout, is equal to the small-signal current, i, multiplied by the impedance of
the parallel combination of RT and Cp.
Figure 1.3 shows a simple model for the single-stage amplifier and the
corresponding Bode plot. The Bode plot is constructed on a log-log scale for
convenience.
MODEL AND BODE PLOT FOR A VFB OP AMP
i = v • gm
+
+
v
gm
-
vOUT
X1
-
RT
CP
vin
NOISE GAIN = G
=1+
R1
R2
AO
fO
fO =
fu =
6dB/OCTAVE
fCL =
1+
R2
R1
R2
R1
1
2π
πR TC P
gm
2π
πC P
fu
1 + R2
R1
=
fu
G
f u = UNITY GAIN FREQUENCY
f
1
f CL = CLOSED LOOP BANDWIDTH
a
1.3
The low frequency breakpoint, fo,is given by:
fo =
1
.
2πR TCp
Note that the high frequency response is determined solely by gm and Cp:
v out = v ⋅
gm
.
jωCp
The unity gain-bandwidth frequency, fu, occurs where |vout|=|v|. Solving the
above equation for fu,assuming |vout|=|v|:
4
fu =
gm
.
2πCp
We can use feedback theory to derive the closed-loop relationship between the
circuit's signal input voltage, vin,and it's output voltage, vout:
v out
=
v in
R2
R1
.
jωCp 
R2 
1+
1 +

gm 
R1 
1+
At the op amp 3dB closed-loop bandwidth frequency, fcl, the following is true:
2πf cl Cp 
R 2
1 +
 = 1 , and hence
gm 
R1 


gm  1 

 , or
f cl =
R2 
2πCp 
1 +


R1 
f cl =
fu
.
R2
1+
R1
This demonstrates a fundamental property of VFB op amps: The closed-loop
bandwidth multiplied by the closed-loop gain is a constant, i.e., the VFB op amp
exhibits a constant gain-bandwidth product over most of the usable frequency range.
Some VFB op amps (called de-compensated) are unstable at unity gain and are
designed to be operated at some minimum amount of closed-loop gain. For these op
amps, the gain-bandwidth product is still relatively constant over the region of
allowable gain.
Now, consider the following typical example: IT = 100µA, Cp = 2pF. We find that:
I / 2 50µA
1
=
=
gm = T
VT
26mV 520Ω
fu =
gm
1
=
= 153MHz .
2πCp 2π(520)( 2 ⋅ 10 −12 )
Now, we must consider the large-signal response of the circuit. The slew-rate, SR, is
simply the total available charging current, IT/2, divided by the dominant pole
capacitance, Cp. For the example under consideration,
5
I=C
dv dv
I
,
= SR , SR =
dt dt
C
I / 2 50µA
SR = T
=
= 25V / µs .
Cp
2pF
The full-power bandwidth (FPBW) of the op amp can now be calculated from the
formula:
FPBW =
25V / µs
SR
=
= 4 MHz, ,
2πA
2π ⋅ 1V
where A is the peak amplitude of the output signal. If we assume a 2V peak-to-peak
output sinewave (certainly a reasonable assumption for high speed applications),
then we obtain a FPBW of only 4MHz, even though the small-signal unity gainbandwidth product is 153MHz! For a 2V p-p output sinewave, distortion will begin
to occur much lower than the actual FPBW frequency. We must increase the SR by
a factor of about 40 in order for the FPBW to equal 153MHz. The only way to do this
is to increase the tail current, IT,of the input differential pair by the same factor.
This implies a bias current of 4mA in order to achieve a FPBW of 160MHz. We are
assuming that Cp is a fixed value of 2pF and cannot be lowered by design.
6
VFB OP AMP BANDWIDTH AND SLEW RATE CALCULATION
n
n
n
n
n
Assume that IT = 100µA, Cp = 2pF
I
50µ A
1
gm = c =
=
VT
26mV 520Ω
gm
fu =
= 153MHz
2π Cp
I /2
Slew Rate = SR = T
= 25 V / µ s
Cp
BUT FOR 2V PEAK-PEAK OUTPUT (A = 1V)
n
SR
= 4MHz
2π A
Must increase IT to 4mA to get FPBW = 160MHz!!
n
Reduce gm by adding emitter degeneration resistors
FPBW =
a
1.4
In practice, the FPBW of the op amp should be approximately 5 to 10 times the
maximum output frequency in order to achieve acceptable distortion performance
(typically 55-80dBc @ 5 to 20MHz, but actual system requirements vary widely).
Notice, however, that increasing the tail current causes a proportional increase in
gm and hence fu. In order to prevent possible instability due to the large increase in
fu, gm can be reduced by inserting resistors in series with the emitters of Q1 and Q2
(this technique, called emitter degeneration, also serves to linearize the gm transfer
function and lower distortion).
A major inefficiency of conventional bipolar voltage feedback op amps is their
inability to achieve high slew rates without proportional increases in quiescent
current (assuming that Cp is fixed, and has a reasonable minimum value of 2 or
3pF). This of course is not meant to say that high speed op amps designed using this
architecture are deficient, it's just that there are circuit design techniques available
which allow equivalent performance at lower quiescent currents. This is extremely
important in portable battery operated equipment where every milliwatt of power
dissipation is critical.
VFB Op Amps Designed on Complementary Bipolar Processes
With the advent of complementary bipolar (CB) processes having high quality PNP
transistors as well as NPNs, VFB op amp configurations such as the one shown in
the simplified diagram (Figure 1.5) became popular.
7
VFB OP AMP USING TWO GAIN STAGES
+VS
D1
Q3
Q4
+
Q1
Q2
OUTPUT
BUFFER
X1
CP
-
I
T
-VS
a
1.5
Notice that the input differential pair (Q1, Q2) is loaded by a current mirror (Q3 and
D1). We show D1 as a diode for simplicity, but it is actually a diode-connected PNP
transistor (matched to Q3) with the base and collector connected to each other. This
simplification will be used in many of the circuit diagrams to follow in this section.
The common emitter transistor, Q4, provides a second voltage gain stage. Since the
PNP transistors are fabricated on a complementary bipolar process, they are high
quality and matched to the NPNs and suitable for voltage gain. The dominant pole
of the amplifier is set by Cp, and the combination of the gain stage,Q4, and Cp is
often referred to as a Miller Integrator. The unity-gain output buffer is usually a
complementary emitter follower.
The model for this two-stage VFB op amp is shown in Figure 1.6. Notice that the
unity gain-bandwidth frequency, fu, is still determined by the gm of the input stage
and the dominant pole capacitance, Cp. The second gain stage increases the DC
open-loop gain, but the maximum slew rate is still limited by the input stage tail
current: SR = IT/Cp.
8
MODEL FOR TWO STAGE VFB OP AMP
CP
i = v•gm
-
+
v
-
gm
-
+
IT
v in
v out
a
X1
+
VREF
R1
R2
fu =
fu
gm
fCL =
2π
πCP
1+
R2
R1
IT
SR =
CP
a
1.6
The two-stage topology is widely used throughout the IC industry in VFB op amps,
both precision and high speed.
Another popular VFB op amp architecture is the folded cascode as shown in Figure
1.7. An industry-standard video amplifier family (the AD847) is based on this
architecture. This circuit takes advantage of the fast PNPs available on a CB
process. The differential signal currents in the collectors of Q1 and Q2 are fed to the
emitters of a PNP cascode transistor pair (hence the term folded cascode). The
collectors of Q3 and Q4 are loaded with the current mirror, D1 and Q5, and Q4
provides voltage gain. This single-stage architecture uses the junction capacitance at
the high-impedance node for compensation (and some variations of the design bring
this node to an external pin so that additional external capacitance can be added).
9
AD847-FAMILY FOLDED CASCODE SIMPLIFIED CIRCUIT
+VS
2IT
2IT
CCOMP
IT
IT
Q4
Q3
+
Q2
Q1
VBIAS
X1
CSTRAY
Q5
2IT
AC GROUND
D1
-VS
a
1.7
With no emitter degeneration resistors in Q1 and Q2, and no additional external
compensating capacitance, this circuit is only stable for high closed-loop gains.
However, unity-gain compensated versions of this family are available which have
the appropriate amount of emitter degeneration.
The availability of JFETs on a CB process allows not only low input bias current but
also improvements in the tradeoff which must be made between gm and IT found in
bipolar input stages. Figure 1.8 shows a simplified diagram of the AD845 16MHz op
amp. JFETs have a much lower gm per mA of tail current than a bipolar transistor.
This allows the input tail current (hence the slew rate) to be increased without
having to increase Cp to maintain stability. The unusual thing about this seemingly
poor performance of the JFET is that it is exactly what is needed on the input stage.
For a typical JFET, the value of gm is approximately Is/1V (Is is the source current),
rather than Ic/26mV for a bipolar transistor, i.e., about 40 times lower. This allows
much higher tail currents (and higher slew rates) for a given gm when JFETs are
used as the input stage.
10
AD845 BiFET 16MHz OP AMP SIMPLIFIED CIRCUIT
+VS
D1
Q5
Q6
Q3
Q4
Q1
+
C
P
VBIAS
X1
Q2
-
-V S
a
1.8
A New VFB Op Amp Architecture for "Current-on-Demand" Performance,
Lower Power, and Improved Slew Rate
Until now, op amp designers had to make the above tradeoffs between the input gm
stage quiescent current and the slew-rate and distortion performance. Analog
Devices' has patented a new circuit core which supplies current-on-demand to charge
and discharge the dominant pole capacitor, Cp, while allowing the quiescent current
to be small. The additional current is proportional to the fast slewing input signal
and adds to the quiescent current. A simplified diagram of the basic core cell is
shown in Figure 1.9.
11
"QUAD-CORE" VFB gm STAGE FOR CURRENT-ON-DEMAND
+VS
Q5
Q7
Q1
Q2
CP1
-
+
X1
CP2
Q3
Q4
Q8
Q6
-VS
a
1.9
The quad-core (gm stage) consists of transistors Q1, Q2, Q3, and Q4 with their
emitters connected together as shown. Consider a positive step voltage on the
inverting input. This voltage produces a proportional current in Q1 which is
mirrored into Cp1 by Q5. The current through Q1 also flows through Q4 and Cp2.
At the dynamic range limit, Q2 and Q3 are correspondingly turned off. Notice that
the charging and discharging current for Cp1 and Cp2 is not limited by the quad
core bias current. In practice, however, small current-limiting resistors are required
forming an "H" resistor network as shown. Q7 and Q8 form the second gain stage
(driven differentially from the collectors of Q5 and Q6), and the output is buffered by
a unity-gain complementary emitter follower.
The quad core configuration is patented (Roy Gosser, U.S. Patent 5,150,074 and
others pending), as well as the circuits which establish the quiescent bias currents
(not shown in the diagram). A number of new VFB op amps using this proprietary
configuration have been released and have unsurpassed high frequency low
distortion performance, bandwidth, and slew rate at the indicated quiescent current
levels (see Figure 1.10). The AD9631, AD8036, and AD8047 are optimized for a gain
of +1, while the AD9632, AD8037, and AD8048 for a gain of +2. The same quad-core
architecture is used as the second stage of the AD8041 rail-to-rail output, zero-volt
input single-supply op amp. The input stage is a differential PNP pair which allows
the input common-mode signal to go about 200mV below the negative supply rail.
The AD8042 and AD8044 are dual and quad versions of the AD8041.
"QUAD-CORE" TWO STAGE XFCB VFB OP AMPS
AC CHARACTERISTICS VERSUS SUPPLY CURRENT
12
PART #
ISY / AMP
BANDWIDTH
SLEW RATE
DISTORTION
AD9631/32
17mA
320MHz
1300V/µs
–[email protected]
AD8036/37 Clamped
20mA
240MHz
1200V/µs
–[email protected]
AD8047/48
5.8mA
250MHz
750V/µs
–[email protected]
AD8041 (1)
5.2mA
160MHz
160V/µs
–[email protected]
AD8042 (2)
5.2mA
160MHz
200V/µs
–[email protected]
AD8044 (4)
2.75mA
150MHz
170V/µs
–[email protected]
AD8031 (1)
0.75mA
80MHz
30V/µs
–[email protected]
AD8032 (2)
0.75mA
80MHz
30V/µs
–[email protected]
Number in ( ) indicates single, dual, or quad
a
1.10
CURRENT FEEDBACK (CFB) OP AMPS
We will now examine the current feedback (CFB) op amp topology which has
recently become popular in high speed op amps. The circuit concepts were
introduced many years ago, however modern high speed complementary bipolar
processes are required to take full advantage of the architecture.
It has long been known that in bipolar transistor circuits, currents can be switched
faster than voltages, other things being equal. This forms the basis of nonsaturating emitter-coupled logic (ECL) and devices such as current-output DACs.
Maintaining low impedances at the current switching nodes helps to minimize the
effects of stray capacitance, one of the largest detriments to high speed operation.
The current mirror is a good example of how currents can be switched with a
minimum amount of delay.
The current feedback op amp topology is simply an application of these fundamental
principles of current steering. A simplified CFB op amp is shown in Figure 1.11. The
non-inverting input is high impedance and is buffered directly to the inverting input
through the complementary emitter follower buffers Q1 and Q2. Note that the
inverting input impedance is very low (typically 10 to 100Ω), because of the low
emitter resistance. In the ideal case, it would be zero. This is a fundamental
difference between a CFB and a VFB op amp, and also a feature which gives the
CFB op amp some unique advantages.
13
SIMPLIFIED CURRENT FEEDBACK (CFB) OP AMP
+VS
Q3
Q1 i
i
X1
-
+
Q2
CP
RT
i
Q4
-VS
R2
R1
a
1.11
The collectors of Q1 and Q2 drive current mirrors which mirror the inverting input
current to the high impedance node, modeled by RT and Cp. The high impedance
node is buffered by a complementary unity gain emitter follower. Feedback from the
output to the inverting input acts to force the inverting input current to zero, hence
the term Current Feedback. (In the ideal case, for zero inverting input impedance, no
small signal voltage can exist at this node, only small signal current).
Consider a positive step voltage applied to the non-inverting input of the CFB op
amp. Q1 immediately sources a proportional current into the external feedback
resistors creating an error current which is mirrored to the high impedance node by
Q3. The voltage developed at the high impedance node is equal to this current
multiplied by the equivalent impedance. This is where the term transimpedance op
amp originated, since the transfer function is an impedance, rather than a unitless
voltage ratio as in a traditional VFB op amp.
Note that the error current is not limited by the input stage bias current, i.e., there
is no slew-rate limitation in an ideal CFB op amp. The current mirrors supply
current-on-demand from the power supplies. The negative feedback loop then forces
the output voltage to a value which reduces the inverting input error current to zero.
The model for a CFB op amp is shown in Figure 1.12 along with the corresponding
Bode plot. The Bode plot is plotted on a log-log scale, and the open-loop gain is
expressed as a transimpedance, T(s), with units of ohms.
14
CFB OP AMP MODEL AND BODE PLOT
VIN
VOUT
X1
RT
CP
i
X1
RO
R1
R2
fO
1
RT
fCL =
2πR2CP
6dB/OCTAVE
|T(s)|
≈
(Ω)
fCL
1
2πR2CP
R
( 1 + R2O +
RO
R1
)
FOR
RO << R1
RO << R2
R2
12dB/OCTAVE
RO
1.12
a
The finite output impedance of the input buffer is modeled by Ro. The input error
current is i. By applying the principles of negative feedback, we can derive the
expression for the op amp transfer function:
v out
=
v in
1+
R2
R1
Ro Ro 

1 + jωCpR21 +
+


R2 R1 
.
At the op amp 3db closed-loop bandwidth frequency, fcl, the following is true:
Ro Ro 

2πf cl CpR21 +
+
 =1.

R2 R1 
Solving for fcl:
f cl =
1
Ro Ro 

2πCpR21 +
+


R2 R1 
.
For the condition Ro << R2 and R1, the equation simply reduces to:
f cl =
1
2πCpR2
15
Examination of this equation quickly reveals that the closed-loop bandwidth of a
CFB op amp is determined by the internal dominant pole capacitor, Cp, and the
external feedback resistor R2, and is independent of the gain-setting resistor, R1. This
ability to maintain constant bandwidth independent of gain makes CFB op amps
ideally suited for wideband programmable gain amplifiers.
Because the closed-loop bandwidth is inversely proportional to the external feedback
resistor, R2, a CFB op amp is usually optimized for a specific R2. Increasing R2 from
it's optimum value lowers the bandwidth, and decreasing it may lead to oscillation
and instability because of high frequency parasitic poles.
The frequency response of the AD8011 CFB op amp is shown in Figure 1.13 for
various closed-loop values of gain (+1, +2, and +10). Note that even at a gain of +10,
the closed loop bandwidth is still greater than 100MHz. The peaking which occurs at
a gain of +1 is typical of wideband CFB op amps when used in the non-inverting
mode and is due primarily to stray capacitance at the inverting input. The peaking
can be reduced by sacrificing bandwidth and using a slightly larger feedback
resistor. The AD8011 CFB op amp represents state-of-the-art performance, and key
specifications are shown in Figure 1.14.
AD8011 FREQUENCY RESPONSE
G = +1, +2, +10
+5
+4
NORMALIZED GAIN - dB
G = +1
RF = 1kΩ
VS = +5V OR ±5V
VOUT = 200mV p-p
+3
+2
G = +2
RF = 1kΩ
+1
0
-1
G = +10
RF = 500Ω
-2
-3
-4
-5
1
10
100
500
FREQUENCY - MHz
a
1.13
AD8011 CFB OP AMP KEY SPECIFICATIONS
n
n
n
n
n
1mA Power Supply Current (+5V or ±5V)
300MHz Bandwidth (G = +1)
2000 V/µs Slew Rate
29ns Settling Time to 0.1%
Video Specifications (G = +2)
Differential Gain Error 0.02%
16
n
n
Differential Phase Error 0.06°
25MHz 0.1dB Bandwidth
Distortion
–70dBc @ 5MHz
–62dBc @ 20MHz
Fully Specified for ±5V or +5V Operation
a
1.14
Traditional current feedback op amps have been limited to a single gain stage, using
current-mirrors as previously described. The AD8011 (and also others in this family:
AD8001, AD8002, AD8004, AD8005, AD8009, AD8013, AD8072, AD8073), unlike
traditional CFB op amps uses a two-stage gain configuration as shown in Figure
1.15. Until now, fully complementary two-gain stage CFB op amps have been
impractical because of their high power dissipation. The AD8011 employs a second
gain stage consisting of a pair of complementary amplifiers (Q3 and Q4). Note that
they are not connected as current mirrors but as grounded-emitters. The detailed
design of current sources (I1 and I2), and their respective bias circuits (Roy Gosser,
patent-applied-for) are the key to the success of the two-stage CFB circuit; they
keep the amplifier's quiescent power low, yet are capable of supplying current-ondemand for wide current excursions required during fast slewing.
SIMPLIFIED TWO-STAGE CFB OP AMP
+VS
I1
Q3
Q1
CC/2
+
X1
CC/2
CP
Q2
Q4
I2
-V S
a
1.15
A further advantage of the two-stage amplifier is the higher overall bandwidth (for
the same power), which means lower signal distortion and the ability to drive
heavier external loads.
17
Thus far, we have learned several key features of CFB op amps. The most important
is that for a given complementary bipolar IC process, CFB generally always yields
higher FPBW (hence lower distortion) than VFB for the same amount of quiescent
supply current. This is because there is practically no slew-rate limiting in CFB.
Because of this, the full power bandwidth and the small signal bandwidth are
approximately the same.
The second important feature is that the inverting input impedance of a CFB op amp
is very low. This can be advantageous when using the op amp in the inverting mode
as an I/V converter, because there is much less sensitivity to inverting input
capacitance than with VFB.
The third feature is that the closed-loop bandwidth of a CFB op amp is determined
by the value of the internal Cp capacitor and the external feedback resistor R2 and is
relatively independent of the gain-setting resistor R1. We will now examine some
typical applications issues and make further comparisons between CFBs and VFBs.
CURRENT FEEDBACK OP AMP FAMILY
PART
ISY/AMP
BANDWIDTH
SLEW RATE
DISTORTION
AD8001 (1)
5.5mA
880MHz
1200V/µs
–[email protected]
AD8002 (2)
5.0mA
600MHz
1200 V/µs
–[email protected]
AD8004 (4)
3.5mA
250MHz
3000 V/µs
–[email protected]
AD8005 (1)
0.4mA
180MHz
500 V/µs
–[email protected]
AD8009 (1)
11mA
1000MHz
7000 V/µs
–[email protected]
AD8011 (1)
1mA
300MHz
2000 V/µs
–[email protected]
AD8012 (2)
1mA
300MHz
1200 V/µs
–[email protected]
AD8013 (3)
4mA
140MHz
1000 V/µs
∆G=0.02%, ∆φ=0.06 °
AD8072 (2)
5mA
100MHz
500 V/µs
∆G=0.05%, ∆φ=0.1 °
AD8073 (3)
5mA
100MHz
500 V/µs
∆G=0.05%, ∆φ=0.1 °
Number in ( ) Indicates Single, Dual, Triple, or Quad
a
1.16
SUMMARY: CURRENT FEEDBACK OP AMPS
n
CFB yields higher FPBW and lower distortion than
VFB for the same process and power dissipation
18
n
Inverting input impedance of a CFB op amp is low,
non-inverting input impedance is high
n
Closed-loop bandwidth of a CFB op amp is determined
by the internal dominant-pole capacitance and the
external feedback resistor, independent of the gainsetting resistor
a
1.17
EFFECTS OF FEEDBACK CAPACITANCE IN OP AMPS
At this point, the term noise gain needs some clarification. Noise gain is the amount
by which a small amplitude noise voltage source in series with an input terminal of
an op amp is amplified when measured at the output. The input voltage noise of an
op amp is modeled in this way. It should be noted that the DC noise gain can also be
used to reflect the input offset voltage (and other op amp input error sources) to the
output.
Noise gain must be distinguished from signal gain. Figure 1.18 shows an op amp in
the inverting and non-inverting mode. In the non-inverting mode, notice that noise
gain is equal to signal gain. However, in the inverting mode, the noise gain doesn't
change, but the signal gain is now –R2/R1. Resistors are shown as feedback
elements, however, the networks may also be reactive.
NOISE GAIN AND SIGNAL GAIN COMPARISON
VIN
NON-INVERTING
INVERTING
VOUT
+
VOUT
+
VN
VN
-
VIN
R2
R1
R1
R2
SIGNAL GAIN = 1 +
R2
R1
SIGNAL GAIN =
NOISE GAIN = = 1 +
R2
R1
NOISE GAIN = 1 +
- R2
R1
R2
R1
FOR VFB OP AMP:
CLOSED-LOOP BW =
fu
fCL =
G
UNITY GAIN BANDWIDTH FREQUENCY
NOISE GAIN
a
1.18
19
Two other configurations are shown in Figure 1.19 where the noise gain has been
increased independent of signal gain by the addition of R3 across the input
terminals of the op amp. This technique can be used to stabilize de-compensated op
amps which are unstable for low values of noise gain. However, the sensitivity to
input noise and offset voltage is correspondingly increased.
INCREASING THE NOISE GAIN
WITHOUT AFFECTING SIGNAL GAIN
VIN
NON-INVERTING
VOUT
+
R3
INVERTING
VOUT
+
VN
R3
-
VN
-
VIN
R2
R1
R1
R2
SIGNAL GAIN = 1 +
R2
R1
SIGNAL GAIN =
NOISE GAIN = = 1 +
R2
R1||R3
NOISE GAIN = 1 +
- R2
R1
R2
R1||R3
1.19
a
Noise gain is often plotted as a function of frequency on a Bode plot to determine the
op amp stability. If the feedback is purely resistive, the noise gain is constant with
frequency. However, reactive elements in the feedback loop will cause it to change
with frequency. Using a log-log scale for the Bode plot allows the noise gain to be
easily drawn by simply calculating the breakpoints determined by the frequencies of
the various poles and zeros. The point of intersection of the noise gain with the openloop gain not only determines the op amp closed-loop bandwidth, but also can be
used to analyze stability.
An excellent explanation of how to make simplifying approximations using Bode
plots to analyze gain and phase performance of a feedback networks is given in
Reference 4.
Just as signal gain and noise gain can be different, so can the signal bandwidth and
the closed-loop bandwidth. The op amp closed-loop bandwidth, fcl, is always
determined by the intersection of the noise gain with the open-loop frequency
response. The signal bandwidth is equal to the closed-loop bandwidth only if the
feedback network is purely resistive.
20
It is quite common to use a capacitor in the feedback loop of a VFB op amp to shape
the frequency response as in a simple single-pole lowpass filter (see Figure 1.20a).
The resulting noise gain is plotted on a Bode plot to analyze stability and phase
margin. Stability of the system is determined by the net slope of the noise gain and
the open loop gain where they intersect. For unconditional stability, the noise gain
plot must intersect the open loop response with a net slope of less than 12dB/octave.
In this case, the net slope where they intersect is 6dB/octave, indicating a stable
condition. Notice for the case drawn in Figure 1.20a, the second pole in the
frequency response occurs at a considerably higher frequency than fu.
NOISE GAIN STABILITY ANALYSIS FOR VFB AND CFB
OP AMPS WITH FEEDBACK CAPACITOR
C2
R1
A
B
R2
-
VFB OP AMP
|A(s)|
+
1
1
fp =
R2
1 + R1
CFB OP AMP
|T(s)|
(Ω)
fp =
2π
πR2C2
2π
πR2C2
R2
fCL
fu
RO
f
1
f
UNSTABLE
a
1.20
In the case of the CFB op amp (Figure 1.20b), the same analysis is used, except that
the open-loop transimpedance gain, T(s), is used to construct the Bode plot. The
definition of noise gain (for the purposes of stability analysis) for a CFB op amp,
however, must be redefined in terms of a current noise source attached to the
inverting input (see Figure 1.21). This current is reflected to the output by an
impedance which we define to be the "current noise gain" of a CFB op amp:
" CURRENT NOISE GAIN"
Ro 

≡ Ro + Z21 +


Z1 
.
21
CURRENT "NOISE GAIN" DEFINITION
FOR CFB OP AMP FOR USE IN STABILITY ANALYSIS
T(s)
X1
VOUT
X1
i
RO
Z1
Z2
VOUT
Z2
CURRENT
"NOISE GAIN"
i
RO
=
VOUT
i
R
= RO +Z2 1+ O
Z1
(
Z1
a
)
1.21
Now, return to Figure 1.20b, and observe the CFB current noise gain plot. At low
frequencies, the CFB current noise gain is simply R2 (making the assumption that
Ro is much less than Z1 or Z2. The first pole is determined by R2 and C2. As the
frequency continues to increase, C2 becomes a short circuit, and all the invertng
input current flows through Ro (refer back to Figure 1.21).
The CFB op amp is normally optimized for best performance for a fixed feedback
resistor, R2. Additional poles in the transimpedance gain, T(s), occur at frequencies
above the closed loop bandwidth, fcl, (set by R2). Note that the intersection of the
CFB current noise gain with the open-loop T(s) occurs where the slope of the T(s)
function is 12dB/octave. This indicates instability and possible oscillation.
It is for this reason that CFB op amps are not suitable in configurations which
require capacitance in the feedback loop, such as simple active integrators or lowpass
filters. They can, however, be used in certain active filters such as the Sallen-Key
configuration shown in Figure 1.22 which do not require capacitance in the feedback
network.
22
EITHER CFB OR VFB OP AMPS CAN BE USED IN
THE SALLEN-KEY FILTER CONFIGURATION
+
-
R2
R1
R2 FIXED FOR CFB OP AMP
1.22
a
VFB op amps, on the other hand, make very flexible active filters. A multiple
feedback 20MHz lowpass filter using the AD8048 is shown in Figure 1.23.
MULTIPLE FEEDBACK 20MHz LOWPASS FILTER
USING THE AD8048 VFB OP AMP
VIN
R1
154Ω
Ω
R4
154Ω
Ω
+5V
C1
50pF
R3
78.7Ω
Ω
100Ω
Ω
0.1µ
µF
1
7
2
C2
100pF
µF
10µ
AD8048
5
3
6
VOUT
µF
0.1µ
4
10µ
µF
-5V
a
1.23
23
In general, the amplifier should have a bandwidth which is at least ten times the
bandwidth of the filter if problems due to phase shift of the amplifier are to be
avoided. (The AD8048 has a bandwidth of over 200MHz in this configuration). The
filter is designed as follows:
Choose:
Fo = Cutoff Frequency = 20MHz
∝ = Damping Ratio = 1/Q = 2
H = Absolute Value of Circuit Gain
= |–R4/R1| = 1
k = 2πFoC1
C2 =
4C1( H + 1)
= 100pF , for C1 = 50pF
α2
α
= 159.2Ω , use 154Ω
2Hk
α
R3 =
= 79.6Ω , use 78.7Ω
2k( H + 1)
R4 = H·R1 = 159.2Ω, use 154Ω
R1 =
HIGH SPEED CURRENT-TO-VOLTAGE CONVERTERS, AND
THE EFFECTS OF INVERTING INPUT CAPACITANCE
Fast op amps are useful as current-to-voltage converters in such applications as high
speed photodiode preamplifiers and current-output DAC buffers. A typical
application using a VFB op amp as an I/V converter is shown in Figure 1.24.
24
COMPENSATING FOR INPUT CAPACITANCE IN A
CURRENT-TO-VOLTAGE CONVERTER USING VFB OP AMP
C2
R2
-
i
C1
VFB
+
fp =
1
2π
πR2C1
UNCOMPENSATED
fx =
2π
πR2C2
COMPENSATED
fx =
fp • fu
|A(s)|
NOISE
GAIN
fx
C1
f
1
fp
1
C2 =
2π
πR2 • f u
FOR 45º PHASE MARGIN
fu
1.24
a
The net input capacitance, C1, forms a pole at a frequency fp in the noise gain
transfer function as shown in the Bode plot, and is given by:
fp =
1
.
2πR2C1
If left uncompensated, the phase shift at the frequency of intersection, fx, will cause
instability and oscillation. Introducing a zero at fx by adding feedback capacitor C2
stabilizes the circuit and yields a phase margin of about 45 degrees. The location of
the zero is given by:
fx =
1
.
2 πR2C2
Although the addition of C2 actually decreases the pole frequency slightly, this effect
is negligible if C2 << C1. The frequency fx is the geometric mean of fp and the unitygain bandwidth frequency of the op amp, fu,
fx =
fp ⋅ fu .
These equations can be solved for C2:
C2 =
C1
.
2πR2 ⋅ f u
25
This value of C2 will yield a phase margin of about 45 degrees. Increasing the
capacitor by a factor of 2 increases the phase margin to about 65 degrees (see
References 4 and 5).
In practice, the optimum value of C2 may be optimized experimentally by varying it
slightly to optimize the output pulse response.
A similar analysis can be applied to a CFB op amp as shown in Figure 1.25. In this
case, however, the low inverting input impedance, Ro, greatly reduces the sensitivity
to input capacitance. In fact, an ideal CFB with zero input impedance would be
totally insensitive to any amount of input capacitance!
COMPENSATING FOR INPUT CAPACITANCE IN A
CURRENT-TO-VOLTAGE CONVERTER USING CFB OP AMP
C2
R2
i
C1
RO
+
fp =
1
2π
πRO||R2•C1
|T(s)|
fx =
UNCOMPENSATED
≈
1
2π
πROC1
1
2π
πR2C2
fx =
fp • f CL
C2 =
RO
R2
COMPENSATED
fx
f
R2
fp
•
C1
2π
π R2•fCL
FOR 45º PHASE MARGIN
fCL
1.25
a
The pole caused by C1 occurs at a frequency fp:
fp =
1
1
≈
.
2π( Ro||R2)C1 2πRoC1
This pole frequency will be generally be much higher than the case for a VFB op
amp, and the pole can be ignored completely if it occurs at a frequency greater than
the closed-loop bandwidth of the op amp.
We next introduce a compensating zero at the frequency fx by inserting the
capacitor C2:
fx =
1
.
2 πR2C2
26
As in the case for VFB, fx is the geometric mean of fp and fcl:
fx =
fp ⋅ fu .
Solving the equations for C2 and rearranging it yields:
C2 =
Ro
C1
⋅
.
R2
2πR2 ⋅ f cl
There is a significant advantage in using a CFB op amp in this configuration as can
be seen by comparing the similar equation for C2 required for a VFB op amp. If the
unity-gain bandwidth product of the VFB is equal to the closed-loop bandwidth of
the CFB (at the optimum R2), then the size of the CFB compensation capacitor, C2,
is reduced by a factor of R2 / Ro .
A comparison in an actual application is shown in Figure 1.26. The full scale output
current of the DAC is 4mA, the net capacitance at the inverting input of the op amp
is 20pF, and the feedback resistor is 500Ω. In the case of the VFB op amp, the pole
due to C1 occurs at 16MHz. A compensating capacitor of 5.6pF is required for 45
degrees of phase margin, and the signal bandwidth is 57MHz.
LOW INVERTING INPUT IMPEDANCE OF CFB
OP AMP MAKES IT RELATIVELY INSENSITIVE TO INPUT
CAPACITANCE WHEN USED AS A
CURRENT-TO-VOLTAGE CONVERTER
C1
20pF
C2
R2
R2
500Ω
Ω
-
4mA
C2
500Ω
Ω
CFB
C1
20pF
VFB
+
-
4mA
+
fu = 200MHz
f CL = 200MHz
RO = 50Ω
Ω
fp =
1
2π
π R2C1
fp =
= 16MHz
1
2π
πRO C1
C2 = 5.6pF
C2 = 1.8pF
fx = 57MHz
fx = 176MHz
a
= 160MHz
1.26
For the CFB op amp, however, because of the low inverting input impedance (Ro =
50Ω), the pole occurs at 160Mhz, the required compensation capacitor is about
1.8pF, and the corresponding signal bandwidth is 176MHz. In actual practice, the
27
pole frequency is so close to the closed-loop bandwidth of the op amp that it could
probably be left uncompensated.
It should be noted that a CFB op amp's relative insensitivity to inverting input
capacitance is when it is used in the inverting mode. In the non-inverting mode,
even a few picofarads of stray capacitance on the inverting input can cause
significant gain-peaking and potential instability.
Another advantage of the low inverting input impedance of the CFB op amp is when
it is used as an I/V converter to buffer the output of a high speed current output
DAC. When a step function current (or DAC switching glitch) is applied to the
inverting input of a VFB op amp, it can produce a large voltage transient until the
signal can propagate through the op amp to its output and negative feedback is
regained. Back-to-back Schottky diodes are often used to limit this voltage swing as
shown in Figure 1.27. These diodes must be low capacitance, small geometry devices
because their capacitance adds to the total input capacitance.
A CFB op amp, on the other hand, presents a low impedance (Ro) to fast switching
currents even before the feedback loop is closed, thereby limiting the voltage
excursion without the requirement of the external diodes. This greatly improves the
settling time of the I/V converter.
LOW INVERTING INPUT IMPEDANCE OF CFB OP AMP
HELPS REDUCE AMPLITUDE OF FAST DAC TRANSIENTS
CURRENT-OUTPUT
DAC
I
+
R2
VFB
* SCHOTTKY
CATCH
DIODES
*
NOT REQUIRED FOR CFB OP AMP
BECAUSE OF LOW INVERTING INPUT IMPEDANCE
a
1.27
NOISE COMPARISONS BETWEEN VFB AND CFB OP AMPS
28
Op amp noise has two components: low frequency noise whose spectral density is
inversely proportional to the square root of the frequency and white noise at
medium and high frequencies. The low-frequency noise is known as 1/f noise (the
noise power obeys a 1/f law - the noise voltage or noise current is proportional to
1/√f). The frequency at which the 1/f noise spectral density equals the white noise is
known as the "1/f Corner Frequency" and is a figure of merit for the op amp, with
the low values indicating better performance. Values of 1/f corner frequency vary
from a few Hz for the most modern low noise low frequency amplifiers to several
hundreds, or even thousands of Hz for high-speed op amps.
In most applications of high speed op amps, it is the total output rms noise that is
generally of interest. Because of the high bandwidths, the chief contributor to the
output rms noise is the white noise, and that of the 1/f noise is negligible.
In order to better understand the effects of noise in high speed op amps, we use the
classical noise model shown in Figure 1.28. This diagram identifies all possible white
noise sources, including the external noise in the source and the feedback resistors.
The equation allows you to calculate the total output rms noise over the closed-loop
bandwidth of the amplifier. This formula works quite well when the frequency
response of the op amp is relatively flat. If there is more than a few dB of high
frequency peaking, however, the actual noise will be greater than the predicted
because the contribution over the last octave before the 3db cutoff frequency will
dominate. In most applications, the op amp feedback network is designed so that the
bandwidth is relatively flat, and the formula provides a good estimate. Note that
BW in the equation is the equivalent noise bandwidth which, for a single-pole
system, is obtained by multiplying the closed-loop bandwidth by 1.57.
OP AMP NOISE MODEL FOR A
FIRST-ORDER CIRCUIT WITH RESISTIVE FEEDBACK
R2
V
R1J
V
R2J
V
n
R1
VON
V
RPJ
InRp
+
BW = 1.57fcl
fcl = CLOSED LOOP BANDWIDTH
In+
VON =
BW
R2
In- 2 R 22 + I n+2 RP2 1 +
R1
2
+ Vn2 1 +
a
R2
R1
2
+ 4kTR 2 + 4kTR 1
2
R2
+ 4kTR P 1 +
R1
R1
R2
2
1.28
29
Figure 1.29 shows a table which indicates how the individual noise contributors are
referred to the output. After calculating the individual noise spectral densities in
this table, they can be squared, added, and then the square root of the sum of the
squares yields the RSS value of the output noise spectral density since all the
sources are uncorrelated. This value is multiplied by the square root of the noise
bandwidth (noise bandwidth = closed-loop bandwidth multiplied by a correction
factor of 1.57) to obtain the final value for the output rms noise.
REFERRING ALL NOISE SOURCES TO THE OUTPUT
NOISE SOURCE EXPRESSED AS
A VOLTAGE
Johnson Noise in Rp:
MULTIPLY BY THIS FACTOR TO REFER
TO THE OP AMP OUTPUT
R2
R1
R2
Noise Gain = 1 +
R1
Noise Gain = 1 +
4kTRp
Non-Inverting Input Current Noise
Flowing in Rp:
In+Rp
Input Voltage Noise:
Vn
Johnson Noise in R1:
R2
R1
–R2/R1 (Gain from input of R1 to Output)
Noise Gain = 1 +
4kTR1
Johnson Noise in R2:
1
4kTR2
Inverting Input Current Noise
Flowing in R2:
In-R2
1
a
1.29
Typical high speed op amps with bandwidths greater than 150MHz or so, and
bipolar input stages have input voltage noises ranging from about 2 to 20nV/√Hz. To
put voltage noise in perspective, let's look at the Johnson noise spectral density of a
resistor:
v n = 4 kTR ⋅ BW ,
where k is Boltzmann's constant, T is the absolute temperature, R is the resistor
value, and BW is the equivalent noise bandwidth of interest. (The equivalent noise
bandwidth of a single-pole system is 1.57 times the 3dB frequency). Using the
formula, a 100Ω resistor has a noise density of 1.3nV/√Hz, and a 1000Ω resistor
about 4nV/√Hz (values are at room temperature: 27°C, or 300K).
The base-emitter in a bipolar transistor has an equivalent noise voltage source
which is due to the "shot noise" of the collector current flowing in the transistor's
30
(noiseless) incremental emitter resistance, re. The current noise is proportional to
the square root of the collector current, Ic. The emitter resistance, on the other
hand, is inversely proportional to the collector current, so the shot-noise voltage is
inversely proportional to the square root of the collector current. (Reference 5, Section
9).
Voltage noise in FET-input op amps tends to be larger than for bipolar ones, but
current noise is extremely low (generally only a few tens of fA/√Hz) because of the
low input bias currents. However, FET-inputs are not generally required for op amp
applications requiring bandwidths greater than 100MHz.
Op amps also have input current noise on each input. For high-speed FET-input op
amps, the gate currents are so low that input current noise is almost always
negligible (measured in fA/√Hz).
For a VFB op amp, the inverting and non-inverting input current noise are typically
equal, and almost always uncorrelated. Typical values for wideband VFB op amps
range from 0.5pA/√Hz to 5pA/√Hz. The input current noise of a bipolar input stage
is increased when input bias-current cancellation generators are added, because
their current noise is not correlated, and therefore adds (in an RSS manner) to the
intrinsic current noise of the bipolar stage.
The input voltage noise in CFB op amps tends to be lower than for VFB op amps
having the same approximate bandwidth. This is because the input stage in a CFB
op amp is usually operated at a higher current, thereby reducing the emitter
resistance and hence the voltage noise. Typical values for CFB op amps range from
about 1 to 5nV/√Hz.
The input current noise of CFB op amps tends to be larger than for VFB op amps
because of the generally higher bias current levels. The inverting and non-inverting
current noise of a CFB is usually different because of the unique input architecture,
and are specified separately. In most cases, the inverting input current noise is the
larger of the two. Typical input current noise for CFB op amps ranges from 5 to
40pA/√Hz.
The general principle of noise calculation is that uncorrelated noise sources add in a
root-sum-squares manner, which means that if a noise source has a contribution to
the output noise of a system which is less than 20% of the amplitude of the noise
from other noise source in the system, then its contribution to the total system noise
will be less than 2% of the total, and that noise source can almost invariably be
ignored - in many cases, noise sources smaller than 33% of the largest can be
ignored. This can simplify the calculations using the formula, assuming the correct
decisions are made regarding the sources to be included and those to be neglected.
The sources which dominate the output noise are highly dependent on the closedloop gain of the op amp. Notice that for high values of closed loop gain, the op amp
voltage noise will tend be the chief contributor to the output noise. At low gains, the
effects of the input current noise must also be considered, and may dominate,
especially in the case of a CFB op amp.
Feedforward/feedback resistors in high speed op amp circuits may range from less
than 100Ω to more than 1kΩ, so it is difficult to generalize about their contribution
31
to the total output noise without knowing the specific values and the closed loop
gain. The best way to make the calculations is to write a simple computer program
which performs the calculations automatically and include all noise sources. In most
high speed applications, the source impedance noise can be neglected for source
impedances of 100Ω or less.
Figure 1.30 shows an example calculation of total output noise for the AD8011
(300MHz, 1mA) CFB op amp. All six possible sources are included in the calculation.
The appropriate multiplying factors which reflect the sources to the output are also
shown on the diagram. For G=2, the close-loop bandwidth of the AD8011 is
180MHz. The correction factor of 1.57 in the final calculation converts this singlepole bandwidth into the circuits equivalent noise bandwidth.
AD8011 OUTPUT NOISE ANALYSIS
5pA/√
√ Hz
2nV/√
√ Hz
1.8nV/√
√ Hz
(G • R S)
0.5nV/√
√ Hz
(G)
+
4nV/√
√ Hz
fCL = 180MHz
AD8011
√ Hz
0.9nV/√
(G)
G=1+
R2
R1
RS
50 Ω
R2
1k Ω
√ Hz
4nV/√
√ Hz
5pA/√
4nV/√
√ Hz
R1
1k Ω
(1)
√ Hz
4nV/√
(R2)
√ Hz
5nV/√
(-R2/R1)
√ Hz
4nV/√
OUTPUT NOISE SPECTRAL DENSITY = 8.7nV/√
√ Hz
TOTAL NOISE = 8.7 1.57 X 180 X 106 = 146µ
µV rms
1.30
a
In communications applications, it is common to specify the noise figure (NF) of an
amplifier. Figure 1.31 shows the definition. NF is the ratio of the total integrated
output noise from all sources to the total output noise which would result if the op
amp were "noiseless" (this noise would be that of the source resistance multiplied by
the gain of the op amp using the closed-loop bandwidth of the op amp to make the
calculation). Noise figure is expressed in dB. The value of the source resistance must
be specified, and in most RF systems, it is 50Ω. Noise figure is useful in
communications receiver design, since it can be used to measure the decrease in
signal-to-noise ratio. For instance, an amplifier with a noise figure of 10dB following
a stage with a signal-to-noise ratio of 50dB reduces the signal-to-noise ratio to 40dB.
32
NOISE FIGURE OF AN OP AMP
NOISE FIGURE = 20log
TOTAL OUTPUT NOISE
OUTPUT NOISE DUE TO RS
G
√ Hz
1.8nV/√
+
√ Hz
0.9nV/√
AD8011
RS
50 Ω
TOTAL = 8.7nV/√
√Hz
R1
R2
1k Ω
1k Ω
NOISE FIGURE = 20log
8.7
1.8
= 13.7dB
a
1.31
The ratio is commonly expressed in dB and is useful in signal chain analysis. In the
previous example, the total output voltage noise was 8.7nV/√Hz. Integrated over the
closed loop bandwidth of the op amp (180MHz), this yielded an output noise of
146µV rms. The noise of the 50Ω source resistance is 0.9nV/√Hz. If the op amp were
noiseless (with noiseless feedback resistors), this noise would appear at the output
multiplied by the noise gain (G=2) of the op amp, or 1.8nV/√Hz. The total output
rms noise just due to the source resistor integrated over the same bandwidth is
30.3µV rms. The noise figure is calculated as:
 146 
NF = 20 log10 
 = 13.7dB .
 30.3 
The same result can be obtained by working with spectral densities, since the
bandwidths used for the integration are the same and cancel each other in the
equation.
 8.7 
NF = 20 log10   = 13.7dB .
 1.8 
HIGH SPEED OP AMP NOISE SUMMARY
n
Voltage Feedback Op Amps:
u
u
n
Voltage Noise: 2 to 20nV/√
√ Hz
Current Noise: 0.5 to 5pA/√
√ Hz
Current Feedback Op Amps:
33
u
u
Voltage Noise: 1 to 5nV/√
√ Hz
Current Noise: 5 to 40pA/√
√ Hz
n
Noise Contribution from Source Negligible if < 100Ω
Ω
n
Voltage Noise Usually Dominates at High Gains
n
Reflect Noise Sources to Output and Combine (RSS)
n
Errors Will Result if there is Significant
High Frequency Peaking
a
1.32
DC CHARACTERISTICS OF HIGH SPEED OP AMPS
High speed op amps are optimized for bandwidth and settling time, not for precision
DC characteristics as found in lower frequency op amps such as the industry
standard OP27. In spite of this, however, high speed op amps do have reasonably
good DC performance. The model shown in Figure 1.33 shows how to reflect the
input offset voltage and the offset currents to the output.
MODEL FOR CALCULATING TOTAL
OP AMP OUTPUT VOLTAGE OFFSET
VOS
R1
R2
-
Ib-
VO
R3
+
Ib+
VO = ±VOS 1 +
R2
R2
+ I b+R3 1 +
- I b-R2
R1
R1
IF Ib+ = Ib- AND R3 = R1||R2
V O = ±VOS 1 +
R2
R1
a
1.33
34
Input offset voltages of high speed bipolar input op amps are rarely trimmed, since
offset voltage matching of the input stage is excellent, typically ranging from 1 to
3mV, with offset temperature coefficients of 5 to 15µV/°C.
Input bias currents on VFB op amps (with no input bias current compensation
circuits) are approximately equal for (+) and (–) inputs, and can range from 1 to 5µA.
The output offset voltage due to the input bias currents can be nulled by making the
effective source resistance, R3, equal to the parallel combination of R1 and R2.
This scheme will not work, however, with bias-current compensated VFB op amps
which have additional current generators on their inputs. In this case, the net input
bias currents are not necessarily equal or of the same polarity. Op amps designed for
rail-to-rail input operation (parallel PNP and NPN differential stages as described
later in this section) have bias currents which are also a function of the commonmode input voltage. External bias current cancellation schemes are ineffective with
these op amps also. It should be noted, however, that it is often desirable to match
the source impedance seen by the (+) and (–) inputs of VFB op amps to minimize
distortion.
CFB op amps generally have unequal and uncorrelated input bias currents because
the (+) and (–) inputs have completely different architectures. For this reason,
external bias current cancellation schemes are also ineffective. CFB input bias
currents range from 5 to 15µA, being generally higher at the inverting input
OUTPUT OFFSET VOLTAGE SUMMARY
n
High Speed Bipolar Op Amp Input Offset Voltage:
u
Ranges from 1 to 3mV for VFB and CFB
u
Offset TC Ranges from 5 to 15µV/°C
n
High Speed Bipolar Op Amp Input Bias Current:
u
For VFB Ranges from 1 to 5µA
u
For CFB Ranges from 5 to 15µA
n
Bias Current Cancellation Doesn't Work for:
u
Bias Current Compensated Op Amps
u
Current Feedback Op Amps
a
1.34
PSRR CHARACTERISTICS OF HIGH SPEED OP AMPS
As with most op amps, the power supply rejection ratio (PSRR) of high speed op
amps falls off rapidly at higher frequencies. Figure 1.35 shows the PSRR for the
AD8011 CFB 300MHz CFB op amp. Notice that at DC, the PSRR is nearly 60dB,
however, at 10MHz, it falls to only 20dB, indicating the need for excellent external
LF and HF decoupling. These numbers are fairly typical of most high speed VFB or
CFB op amps, although the DC PSRR may range from 55 to 80dB depending on the
op amp.
35
AD8011 POWER SUPPLY REJECTION RATIO
+10
0
-10
PSRR - dB
-20
-PSRR
VS = +5V OR ±5V
G = +2
RF = 1kΩ
+PSRR
-30
-40
-50
-60
-70
-80
-90
100k
1M
10M
100M
500M
FREQUENCY - Hz
1.35
a
The power pins of op amps must be decoupled directly to a large-area ground plane
with capacitors which have minimal lead length. It is generally recommended that a
low-inductance ceramic surface mount capacitor (0.01µF to 0.1µF) be used for the
high frequency noise. The lower frequency noise can be decoupled with lowinductance tantalum electrolytic capacitors (1 to 10µF).
36
REFERENCES
1.
Thomas M. Frederiksen, Intuitive Operational Amplifiers,
McGraw-Hill, 1988.
2.
Sergio Franco, Current Feedback Amplifiers, EDN, Jan.5, 1989.
3.
Roy Gosser, U.S Patent 5,150,074.
4.
James L. Melsa and Donald G. Schultz, Linear Control Systems,
McGraw-Hill, 1969, pp. 196-220.
5.
Amplifier Applications Guide, Analog Devices, Inc., 1992,
Section 3.
6.
Walter G. Jung, IC Op amp Cookbook, Third Edition,
Howard Sams & Co., 1986, ISBN: 0-672-22453-4.
7.
Paul R. Gray and Robert G. Meyer, Analysis and Design of Analog
Integrated Circuits, Third Edition, John Wiley, 1993.
8.
J. K. Roberge, Operational Amplifiers-Theory and Practice,
John Wiley, 1975.
9.
Henry W. Ott, Noise Reduction Techniques in Electronic Systems,
Second Edition, John Wiley, Inc., 1988.
10.
Lewis Smith and Dan Sheingold, Noise and Operational Amplifier Circuits,
Analog Dialogue 25th Anniversary Issue, pp. 19-31, 1991.
11.
D. Stout, M. Kaufman, Handbook of Operational Amplifier Circuit
Design, New York, McGraw-Hill, 1976.
12.
Joe Buxton, Careful Design Tames High-Speed Op Amps, Electronic
Design, April 11, 1991.
13.
J. Dostal, Operational Amplifiers, Elsevier Scientific Publishing,
New York, 1981.
14.
Barrie Gilbert, Contemporary Feedback Amplifier Design,
15.
Sergio Franco, Design with Operational Amplifiers and Analog ICs,
McGraw-Hill Book Company, 1988.
16.
Jerald Graeme, Photodiode Amplifiers-Op Amp Solutions,
Gain Technology Corporation, 2700 W. Broadway Blvd., Tucson,
AZ 85745, 1996.
37
SECTION 2
HIGH SPEED OP AMP APPLICATIONS
Walt Kester, Walt Jung
OPTIMIZING THE FEEDBACK NETWORK FOR MAXIMUM
BANDWIDTH FLATNESS IN WIDEBAND CFB OP AMPS
Achieving the highest 0.1dB bandwidth flatness is important in many video
applications. Because of the critical relationship between the feedback resistor and
the bandwidth of a CFB op amp, optimum bandwidth flatness is highly dependent
on the feedback resistor value, the resistor parasitics, as well as the op amp package
and PCB parasitics. Figure 2.1 shows the fine scale (0.1dB/division) flatness plotted
versus the feedback resistance for the AD8001 in a non-inverting gain of 2. These
plots were made using the AD8001 evaluation board with surface mount resistors.
AD8001 CFB OP AMP BANDWIDTH FLATNESS OPTIMIZED BY
PROPER SELECTION OF FEEDBACK RESISTOR
0.1
RF =
649Ω
0
G = +2
R F = 698Ω
-0.1
AD8001
RG = RF
OUTPUT - dB
RF
-0.2
G = +2
R F = 750Ω
-0.3
-0.4
-0.5
-0.6
-0.7
-0.8
-0.9
1M
10M
FREQUENCY - Hz
a
100M
2.1
It is recommended that once the optimum resistor values have been determined, 1%
tolerance values should be used. In addition, resistors of different construction have
different associated parasitic capacitance and inductance. Surface mount resistors
are the optimum choice, and it is not recommended that leaded components be used
with high speed op amps at these frequencies because of their parasitics.
Slightly different resistor values may be required to achieve optimum performance
in the DIP versus the SOIC packages (see Figure 2.2). The SOIC package exhibits
slightly lower parasitic capacitance and inductance than the DIP. The data shows
the optimum feedback (RG) and feedforward (RF) resistors for highest 0.1dB
bandwidth for the AD8001 in the DIP and the SOIC packages. As you might
1
suspect, the SOIC package can be optimized for slightly higher 0.1dB bandwidth
because of its lower parasitics.
OPTIMUM VALUES OF RF AND RG FOR AD8001
DIP AND SOIC PACKAGES (MAXIMUM 0.1dB BANDWIDTH)
AD8001AN (DIP) GAIN
Component
–1
+1
+2
RF
649Ω
Ω
1050Ω
Ω
750Ω
Ω
RG
649Ω
Ω
-
750Ω
Ω
0.1dB Flatness
105MHz
70MHz
105MHz
AD8001AR (SOIC) GAIN
Component
–1
+1
+2
RF
604Ω
Ω
953Ω
Ω
681Ω
Ω
RG
604Ω
Ω
-
681Ω
Ω
0.1dB Flatness
130MHz
100MHz
120MHz
a
2.2
As has been discussed, the CFB op amp is relatively insensitive to capacitance on
the inverting input when it is used in the inverting mode (as in an I/V application).
This is because the low inverting input impedance is in parallel with the external
capacitance and tends to minimize its effect. In the non-inverting mode, however,
even a few picofarads of stray inverting input capacitance may cause peaking and
instability. Figure 2.3 shows the effects of adding summing junction capacitance to
the inverting input of the AD8004 (SOIC package) for G = +2. Note that only 1pF of
added inverting input capacitance (CJ) causes a significant increase in bandwidth
and an increase in peaking. For G = –2, however, 5pF of additional inverting input
capacitance causes only a small increase in bandwidth and no significant increase in
peaking.
High speed VFB op amps are sensitive to stray inverting input capacitance when
used in either the inverting or non-inverting mode.
2
AD8004 CFB OP AMP SENSITIVITY TO
INVERTING INPUT CAPACITANCE FOR G = +2, G = -2
+2
CJ = 1pF
CJ = 0
+2
NORMALIZED GAIN - dB, G = -2
0
-2
G = -2
-4
0
-6
-2
-4
-6
VIN = 50mV rms
RL = 100Ω
±5VS
-8
CJ = 5.1pF
-10
-12
-8
CJ = 0
-10
-14
NORMALIZED GAIN - dB, G = +2
G = +2
-12
-14
1
10
40
FREQUENCY - MHz
a
100
500
2.3
DRIVING CAPACITIVE LOADS
From system and signal fidelity points of view, transmission line coupling between
stages is best, and is described in some detail in the next section. However, complete
transmission line system design may not always be possible or practical. In addition,
various other parasitic issues need careful consideration in high performance
designs. One such problem parasitic is amplifier load capacitance, which potentially
comes into play for all wide bandwidth situations which do not use transmission line
signal coupling.
A general design rule for wideband linear drivers is that capacitive loading (cap
loading) effects should always be considered. This is because PC board capacitance
can build up quickly, especially for wide and long signal runs over ground planes
insulated by a thin, higher K dielectric. For example, a 0.025” PC trace using a G-10
dielectric of 0.03” over a ground plane will run about 22pF/foot (Reference 1). Even
relatively small load capacitance (i.e., <100 pF) can be troublesome, since while not
causing outright oscillation, it can still stretch amplifier settling time to greater
than desirable levels for a given accuracy.
The effects of cap loading on high speed amplifier outputs are not simply
detrimental, they are actually an anathema to high quality signals. However,
before-the-fact designer knowledge still allows high circuit performance by
employing various tricks of the trade to combat the capacitive loading. If it is not
driven via a transmission line, remote signal circuitry should be checked for
capacitive loading very carefully, and characterized as best possible. Drivers which
face poorly defined load capacitance should be bullet-proofed accordingly with an
appropriate design technique from the options list below.
3
Short of a true matched transmission line system, a number of ways exist to drive a
load which is capacitive in nature while maintaining amplifier stability.
Custom capacitive load (cap load) compensation includes two possible options,
namely a); overcompensation, and b); an intentionally forced-high loop noise gain
allowing crossover in a stable region. Both of these steps can be effective in special
situations, as they reduce the amplifier’s effective closed loop bandwidth, so as to
restore stability in the presence of cap loading.
Overcompensation of the amplifier, when possible, reduces amplifier bandwidth so
that the additional load capacitance no longer represents a danger to phase margin.
As a practical matter however, amplifier compensation nodes to allow this are
available on few high speed amplifiers. One such useful example is the AD829,
compensated by a single capacitor at pin 5. In general, almost any amplifier using
external compensation can always be over compensated to reduce bandwidth. This
will restore stability against cap loads, by lowering the amplifier’s unity gain
frequency.
Forcing a high noise gain, is shown in Figure 2.4, where the capacitively loaded
amplifier with a noise gain of unity at the left is seen to be unstable, due to a 1/β open loop rolloff intersection on the Bode diagram in an unstable –12dB/octave
region. For such a case, quite often stability can be restored by introducing a higher
noise gain to the stage, so that the intersection then occurs in a stable –6dB/octave
region, as depicted at the diagram right Bode plot.
CAPACITIVE LOADING ON OP AMP GENERALLY REDUCES
PHASE MARGIN AND MAY CAUSE INSTABILITY,
BUT INCREASING THE NOISE GAIN OF THE CIRCUIT
IMPROVES STABILITY
RO
...
RO
...
CL
CL
R2
R1
NOISE GAIN = 1
GAIN
(dB)
NG = 1 +
R2
R1
NOISE GAIN = 1 +
R2
R1
STABLE
UNSTABLE
0
NG = 1
LOG FREQUENCY
LOG FREQUENCY
a
2.4
To enable a higher noise gain (which does not necessarily need to be the same as the
stage’s signal gain), use is made of resistive or RC pads at the amplifier input, as in
Figure 2.5. This trick is more broad in scope than overcompensation, and has the
advantage of not requiring access to any internal amplifier nodes. This generally
4
allows use with any amplifier setup, even voltage followers. The technique adds an
extra resistor RD, which works against RF to force the noise gain of the stage to a
level appreciably higher than the signal gain (which is unity in both cases here).
Assuming that CL is a value which produces a parasitic pole near the amplifier’s
natural crossover, this loading combination would likely lead to oscillation due to the
excessive phase lag. However with RD connected, the higher amplifier noise gain
produces a new 1/β - open loop rolloff intersection, about a decade lower in
frequency. This is set low enough that the extra phase lag from CL is no longer a
problem, and amplifier stability is restored.
RAISING NOISE GAIN (DC OR AC) FOR
FOLLOWER OR INVERTER STABILITY
(B) INVERTER
(A) FOLLOWER
RF
RD =
RF/10
VIN
-
-
VOUT
VOUT
RD =
RF/10
CD
+
VIN
RF
RIN
+
RL
CL
CD
a
RL
CL
2.5
A drawback to this trick is that the DC offset and input noise of the amplifier are
raised by the value of the noise gain, when the optional CD is not present. But,
when CD is used in series with RD, the offset voltage of the amplifier is not raised,
and the gained-up AC noise components are confined to a frequency region above
1/(2π•RD•CD). A further caution is that the technique can be somewhat tricky
when separating these operating DC and AC regions, and should be applied
carefully with regard to settling time (Reference 2). Note that these simplified
examples are generic, and in practice the absolute component values should be
matched to a specific amplifier.
“Passive” cap load compensation, shown in Figure 2.6, is the most simple (and most
popular) isolation technique available. It uses a simple “out-of-the-loop” series resistor
RX to isolate the cap load, and can be used with any amplifier, current or voltage
feedback, FET or bipolar input.
5
OPEN-LOOP SERIES RESISTANCE ISOLATES CAPACITIVE
LOAD FOR AD811 CURRENT FEEDBACK OP AMP
(CIRCUIT BANDWIDTH = 13.5MHz)
RIN
30.9kΩ
RF
750Ω
RX
12Ω
2 -
AD811
VIN
7
3 +
RB
1kΩ
0.1µF
4
+12V
100µF/
25V
VOUT
6
0.1µF
CL
1nF
RL
500Ω
100µF/
25V
-12V
a
2.6
As noted, this technique can be applied to virtually any amplifier, which is a major
reason why it is so useful. It is shown here with a current feedback amplifier
suitable for high current line driving, the AD811, and it consists of just the simple
(passive) series isolation resistor, RX. This resistor’s minimum value for stability
will vary from device to device, so the amplifier data sheet should be consulted for
other ICs. Generally, information will be provided as to the amount of load
capacitance tolerated, and a suggested minimum resistor value for stability
purposes.
Drawbacks of this approach are the loss of bandwidth as RX works against CL, the
loss of voltage swing, a possible lower slew rate limit due to IMAX and CL, and a
gain error due to the RX-RL division. The gain error can be optionally compensated
with RIN, which is ratioed to RF as RL is to RX. In this example, a ±100mA output
from the op amp into CL can slew VOUT at a rate of 100V/µs, far below the intrinsic
AD811 slew rate of 2500V/µs. Although the drawbacks are serious, this form of cap
load compensation is nevertheless useful because of its simplicity. If the amplifier is
not otherwise protected, then an RX resistor of 50-100Ω should be used with
virtually any amplifier facing capacitive loading. Although a non-inverting amplifier
is shown, the technique is equally applicable to inverter stages.
With very high speed amplifiers, or in applications where lowest settling time is
critical, even small values of load capacitance can be disruptive to frequency
response, but are nevertheless sometimes inescapable. One case in point is an
amplifier used for driving ADC inputs. Since high speed ADC inputs quite often look
capacitive in nature, this presents an oil/water type problem. In such cases the
amplifier must be stable driving the capacitance, but it must also preserve its best
6
bandwidth and settling time characteristics. To address this type of cap load case,
Rs and CL performance data for a specified settling time is most appropriate.
Some applications, in particular those that require driving the relatively high
impedance of an ADC, do not have a convenient back termination resistor to
dampen the effects of capacitive loading. At high frequencies, an amplifier’s output
impedance is rising with frequency and acts like an inductance, which in
combination with CL causes peaking or even worse, oscillation. When the bandwidth
of an amplifier is an appreciable percentage of device Ft,the situation is complicated
by the fact that the loading effects are reflected back into its internal stages. In spite
of this, the basic behavior of most very wide bandwidth amplifiers such as the
AD8001 is very similar.
In general, a small damping resistor (Rs) placed in series with CL will help restore
the desired response (see Figure 2.7). The best choice for this resistor’s value will
depend upon the criterion used in determining the desired response. Traditionally,
simply stability or an acceptable amount of peaking has been used, but a more strict
measure such as 0.1% (or even 0.01%) settling will yield different values. For a given
amplifier, a family of Rs - CL curves exists, such as those of Figure 2.7. These data
will aid in selecting Rs for a given application.
AD8001 RS REQUIRED FOR VARIOUS CL VALUES
40
G = +2, 0.1% SETTLING
RS
30
RS
AD8001
(Ω)
20
CL
G = +2,
20% OVERSHOOT
G = +10,
20% OVERSHOOT
10
0
20
40
60
80
100
CL (pF)
a
2.7
The basic shape of this curve can be easily explained. When CL is very small, no
resistor is necessary. When CL increases to some threshold value an Rs becomes
necessary. Since the frequency at which the damping is required is related to the
Rs•CL time constant, the Rs needed will initially increase rapidly from zero, and
then will decrease as CL is increased further. A relatively strict requirement, such as
for 0.1%, settling will generally require a larger Rs for a given CL, giving a curve
falling higher (in terms of Rs) than that for a less stringent requirement, such as
7
20% overshoot. For the common gain configuration of +2, these two curves are
plotted in the figure for 0.1% settling (upper-most curve) and 20% overshoot (middle
curve). It is also worth mentioning that higher closed loop gains lessen the problem
dramatically, and will require less Rs for the same performance. The third (lowermost) curve illustrates this, demonstrating a closed loop gain of 10 Rs requirement
for 20% overshoot for the AD8001 amplifier. This can be related to the earlier
discussion associated with Figure 2.5.
The recommended values for Rs will optimize response, but it is important to note
that generally CL will degrade the maximum bandwidth and settling time
performance which is achievable. In the limit, a large Rs•CL time constant will
dominate the response. In any given application, the value for Rs should be taken as
a starting point in an optimization process which accounts for board parasitics and
other secondary effects.
Active or “in-the-loop” cap load compensation can also be used as shown in Figure
2.8, and this scheme modifies the passive configuration to provide feedback correction
for the DC & low frequency gain error associated with RX. In contrast to the passive
form, active compensation can only be used with voltage feedback amplifiers, because
current feedback amplifiers don’t allow the integrating connection of CF.
ACTIVE "IN-LOOP" CAPACITIVE LOAD COMPENSATION
CORRECTS FOR DC AND LF GAIN ERRORS
+VS
VIN
+
1kΩ
33Ω
-VS
RF
VOUT
RX
U1
AD845
CF
CL
1nF
RL
500Ω
50pF
2.5kΩ
RIN
2.5kΩ
a
2.8
This circuit returns the DC feedback from the output side of isolation resistor RX,
thus correcting for errors. AC feedback is returned via CF, which bypasses RX/RF at
high frequencies. With an appropriate value of CF (which varies with CL, for fixed
resistances) this stage can be adjusted for a well damped transient response
(Reference 2,3). There is still a bandwidth reduction, a headroom loss, and also
(usually) a slew rate reduction, but the DC errors can be very low. A drawback is the
need to tune CF to CL, as even if this is done well initially, any change to CL will
8
alter the response away from flat. The circuit as shown is useful for voltage feedback
amplifiers only, because capacitor CF provides integration around U1. It also can be
implemented in inverting fashion, by driving the bottom end of RIN.
Internal cap load compensation involves the use of an amplifier which has topological
provisions for the effects of external cap loading. To the user, this is the most
transparent of the various techniques, as it works for any feedback situation, for any
value of load capacitance. Drawbacks are that it produces higher distortion than does
an otherwise similar amplifier without the network, and the compensation against
cap loading is somewhat signal level dependent.
The internal cap load compensated amplifier sounds at first like the best of all
possible worlds, since the user need do nothing at all to set it up. Figure 2.9, a
simplified diagram of an AD817 amplifier with internal cap load compensation,
shows how it works. The cap load compensation is the CF -resistor network shown
around the unity gain output stage of the amplifier - note that the dotted connection
of this network underscores the fact that it only makes its presence felt for certain
load conditions.
AD817 SIMPLIFIED SCHEMATIC ILLUSTRATES INTERNAL
COMPENSATION FOR DRIVING CAPACITIVE LOADS
+VS
CF
OUTPUT
-IN
+IN
-VS
a
NULL 1
NULL 8
2.9
Under normal (non-capacitive or light resistive) loading, there is limited
input/output voltage error across the output stage, so the CF network then sees a
relatively small voltage drop, and has little or no effect on the AD817s high
impedance compensation node. However when a capacitor (or other heavy) load is
present, the high currents in the output stage produce a voltage difference across
the CF network, which effectively adds capacitance to the compensation node. With
this relatively heavy loading, a net larger compensation capacitance results, and
reduces the amplifier speed in a manner which is adaptive to the external
9
capacitance, CL. As a point of reference, note that it requires 6.3mA peak current to
support a 2Vp-p swing across a 100pF load at 10MHz.
Since this mechanism is resident in the amplifier output stage and it affects the
overall compensation characteristics dynamically, it acts independent of the specific
feedback hookup, as well as size of the external cap loading. In other words, it can
be transparent to the user in the sense that no specific design conditions need be set
to make it work (other than selecting an IC which employs it). Some amplifiers
using internal cap load compensation are the AD847 and the AD817, and their dual
equivalents, AD827 and AD826.
There are, however, some caveats also associated with this internal compensation
scheme. As with the passive compensation techniques, bandwidth decreases as the
device slows down to prevent oscillation with higher load currents. Also, this
adaptive compensation network has its greatest effect when enough output current
flows to produce significant voltage drop across the CF network. Conversely, at
small signal levels, the effect of the network on speed is less, so greater ringing may
actually be possible for some circuits for lower-level outputs.
RESPONSE OF INTERNAL CAP LOAD
COMPENSATED AMPLIFIER VARIES WITH SIGNAL LEVEL
(A) VOUT = 10V p-p
Vertical Scale: 5V/div
5V
(B) VOUT = 200mV p-p
Vertical Scale: 100mV/div
500ns
100mV
500ns
100mv
5V
Horizontal Scale: 500ns/div
AD817 INVERTER
RF = RIN = 1kΩ
a
RL = 1kΩ , CL = 1nF, VS = ±15V
2.10
The dynamic nature of this internal cap load compensation is illustrated in Figure
2.10, which shows an AD817 unity gain inverter being exercised at both high and
low output levels, with common conditions of Vs = ±15V, RL = 1kΩ, CL = 1nF, and
using 1kΩ input/feedback resistors. In both photos the input signal is on the top
trace and the output signal is on the bottom trace, and the time scale is fixed. In the
10Vp-p output (A) photo at the left, the output has slowed down appreciably to
accommodate the capacitive load, but settling is still relatively clean, with a small
percentage of overshoot. This indicates that for this high level case, the bandwidth
reduction due to CL is most effective. However, in the (B) photo at the right, the
10
200mVp-p output shows greater overshoot and ringing, for the lower level signal.
The point is that the performance of the cap load compensated amplifier is signal
dependent, but is always stable with any cap load.
Finally, because the circuit is based on a nonlinear principle, the internal network
affects distortion performance and load drive ability, and these factors influence
amplifier performance in video applications. Though the network’s presence does not
by any means make devices like the AD847 or AD817 unusable for video, it does not
permit the very lowest levels of distortion and differential gain and phase which are
achievable with otherwise comparable amplifiers (for example, the AD818 which is
an AD817 without the internal compensating network).
While the individual techniques for countering cap loading outlined above have
various specific tradeoffs as noted, all of the techniques have a common drawback of
reducing speed (both bandwidth and slew rate). If these parameters cannot be
sacrificed, then a matched transmission line system is the solution, and is discussed
in more detail later in the chapter. As for choosing among the cap load compensation
schemes, it would seem on the surface that amplifiers using the internal form offer
the best possible solution to the problem- just pick the right amplifier and forget
about it. And indeed, that would seem the “panacea” solution for all cap load
situations - if you use the “right” amplifier you never need to think about cap
loading again. Could there be more to it?
Yes! The “gotcha” of internal cap load compensation is subtle, and lies in the fact
that the dynamic adaptive nature of the compensation mechanism actually can
produce higher levels of distortion, vis-à-vis an otherwise similar amplifier, without
the CF -resistor network. Like the old saying about no free lunches, if you care about
attaining top-notch levels of high frequency AC performance, you should give the
issue of whether to use an internally compensated cap load amplifier more serious
thought than simply picking a trendy device.
On the other hand, if you have no requirements for the lowest levels of distortion,
then such an amplifier could be a good choice. Such amplifiers are certainly easier to
use, and relatively forgiving about loading issues. Some applications of this chapter
illustrate the distortion point specifically, quoting performance in a driver circuit
with/without the use of an internal cap load compensated amplifiers.
CABLE DRIVERS AND RECEIVERS
High quality video signals are best transmitted over terminated coaxial cable having
a controlled characteristic impedance. The characteristic impedance is given by the
equation Zo = √(L/C) where L is the distributed inductance per foot, and C is the
distributed capacitance per foot. Popular values are 50, 75, and 93 or 100Ω.
If a length of coaxial cable is terminated, it presents a resistive load to the driver. If
left unterminated, however, it may present a predominately capacitive load to the
driver depending on the output frequency. If the length of an unterminated cable is
much less than the wavelength of the output frequency of the driver, then the load
appears approximately as a lumped capacitance. For instance, at the audio
frequency of 20kHz (wavelength ≈ 50,000 feet, or 9.5miles), a 5 foot length of
unterminated 50Ω coaxial cable would appear as a lumped capacitance of
11
approximately 150pF (the distributed capacitance of coaxial cable is about 30pF/ft).
At 100MHz (wavelength ≈ 10 feet), however, the unterminated coax must be treated
as a transmission line in order to calculate the standing wave pattern and the
voltage at the unterminated cable output.
Because of skin effect and wire resistance, coaxial cable exhibits a loss which is a
function of frequency. This varies considerably between cable types. For instance the
attenuation in at 100MHz of RG188A/U is 8dB/100ft, RG58/U is 5.5dB/100ft, and
RG59/U 3.6dB/100ft (Reference 4).
Skin effect also affects the pulse response of long coaxial cables. The response to a
fast pulse will rise sharply for the first 50% of the output swing, then taper off
during the remaining portion of the edge. Calculations show that the 10 to 90%
waveform risetime is 30 times greater than the 0 to 50% risetime when the cable is
skin effect limited (see Reference 5).
DRIVING CABLES
n
All Interconnections are Really Transmission Lines Which
Have a Characteristic Impedance (Even if Not Controlled)
n
The Characteristic Impedance is Equal to L/C, where L
and C are the Distributed Inductance and Capacitance
n
Correctly Terminated Transmission Lines Have Impedances Equal
to Their Characteristic Impedance
n
Unterminated Transmission Lines Behave Approximately as
Lumped Capacitance if the Wavelength of the Output Frequency
is Much Greater than the Length of the Cable
u
Example: At 20kHz (Wavelength = 9.5 miles), 5 feet of
Unterminated 50Ω
Ω Cable (30pF/ft) Appears Like 150pF Load
u
Example: At 100MHz (Wavelength = 10 feet), 5 feet of
50Ω
Ω Cable Must be Properly Terminated to
PreventReflections and Standing Waves!!!!
a
2.11
It is useful to examine what happens for conditions of proper and improper cable
source/load terminations. To illustrate the behavior of a high speed op amp driving a
coaxial cable, consider the circuit of Figure 2.12. The AD8001 drives 5 feet of 50Ω
coaxial cable which is load-end terminated in the characteristic impedance of 50Ω.
No termination is used at the amplifier (driving) end. The pulse response is also
shown in the figure.
The output of the cable was measured by connecting it directly to the 50Ω input of a
500MHz Tektronix 644A digitizing oscilloscope. The 50Ω resistor termination is
12
actually the input of the scope. The 50Ω load is not a perfect termination (the scope
input capacitance is about 10pF), so some of the pulse is reflected out of phase back
to the source. When the reflection reaches the op amp output, it sees the closed-loop
output impedance of the op amp which, at 100MHz, is approximately 100Ω. Thus, it
is reflected back to the load with no phase reversal, accounting for the negativegoing "blip" which occurs approximately 16ns after the leading edge. This is equal to
the round-trip delay of the cable (2•5ft•1.6 ns/ft=16ns). In the frequency domain
(not shown), the cable mismatch will cause a loss of bandwidth flatness at the load.
PULSE RESPONSE OF AD8001 DRIVING
5 FEET OF LOAD-TERMINATED 50Ω COAXIAL CABLE
649Ω
649Ω
PULSE
INPUT
53.6Ω
SCOPE
+5V
5ft
AD8001
+
8ns
-5V
VERTICAL
SCALE: 200mV/div
50Ω
10pF
SCOPE
OUTPUT
HORIZONTAL
SCALE: 10ns/div
a
2.12
Figure 2.13 shows a second case, the results of driving the same coaxial cable, but
now used with both a 50Ω source-end as well as a 50Ω load-end termination. This
case is the preferred way to drive a transmission line, because a portion of the
reflection from the load impedance mismatch is absorbed by the amplifier’s source
termination resistor. The disadvantage is that there is a 2× gain reduction, because
of the voltage division between the equal value source/load terminations. However, a
major positive attribute of this configuration, with matched source and load
terminations in conjunction with a low-loss cable, is that the best bandwidth
flatness is ensured, especially at lower operating frequencies. In addition, the
amplifier is operated under near optimum loading conditions, i.e., a resistive load.
13
PULSE RESPONSE OF AD8001 DRIVING 5 FEET
OF SOURCE AND LOAD-TERMINATED 50Ω COAXIAL CABLE
649Ω
649Ω
PULSE
INPUT
53.6Ω
SCOPE
+5V
50Ω
5ft
AD8001
+
8ns
-5V
VERTICAL
SCALE: 100mV/div
50Ω
10pF
SCOPE
OUTPUT
HORIZONTAL
SCALE: 10ns/div
a
2.13
Source-end (only) terminations can also be used as shown in Figure 2.14, where the
op amp is source terminated by the 50Ω resistor which drives the cable. The scope is
set for 1MΩ input impedance, representing an approximate open circuit. The initial
leading edge of the pulse at the op amp output sees a 100Ω load (the 50Ω source
resistor in series with the 50Ω coax impedance. When the pulse reaches the load, a
large portion is reflected in phase because of the high load impedance, resulting in a
full-amplitude pulse at the load. When the reflection reaches the source-end of the
cable, it sees the 50Ω source resistance in series with the op amp closed loop output
impedance (approximately 100Ω at the frequency represented by the 2ns risetime
pulse edge). The reflected portion remains in phase, and appears at the scope input
as the positive-going "blip" approximately 16ns after the leading edge.
14
PULSE RESPONSE OF AD8001 DRIVING 5 FEET
OF SOURCE-TERMINATED 50Ω COAXIAL CABLE
649Ω
PULSE
INPUT
53.6Ω
649Ω
-
SCOPE
+5V
50Ω
5ft
AD8001
+
8ns
-5V
VERTICAL
SCALE: 200mV/div
1MΩ
10pF
SCOPE
OUTPUT
HORIZONTAL
SCALE: 10ns/div
a
2.14
From these experiments, one can easily see that the preferred method for minimum
reflections (and therefore maximum bandwidth flatness) is to use both source and
load terminations and try to minimize any reactance associated with the load. The
experiments represent a worst-case condition, where the frequencies contained in
the fast edges are greater than 100MHz. (Using the rule-of-thumb that bandwidth =
0.35/risetime). At video frequencies, either load-only, or source-only terminations
may give acceptable results, but the data sheet should always be consulted to
determine the op amp's closed-loop output impedance at the maximum frequency of
interest. A major disadvantage of the source-only termination is that it requires a
truly high impedance load (high resistance and minimal parasitic capacitance) for
minimum absorption of energy. It also places a burden on this amplifier to maintain
a low output impedance at high frequencies.
Now, for a truly worst case, let us replace the 5 feet of coaxial cable with an
uncontrolled-impedance cable (one that is largely capacitive with little inductance).
Let us use a capacitance of 150pF to simulate the cable (corresponding to the total
capacitance of 5 feet of coaxial cable whose distributed capacitance is about
30pF/foot). Figure 2.15 shows the output of the AD8001 driving a lumped 160pF
capacitance (including the scope input capacitance of 10pF). Notice the overshoot
and ringing on the pulse waveform due to the capacitive loading. This example
illustrates the need to use good quality controlled-impedance coaxial cable in the
transmission of high frequency signals.
15
PULSE RESPONSE OF AD8001 DRIVING 160pF || 50 Ω LOAD
649Ω
Ω
PULSE
INPUT
649Ω
Ω
-
53.6Ω
Ω
SCOPE
+5V
DIRECT
CONNECTION
AD8001
++
150pF
50Ω
Ω
-5V
VERTICAL
SCALE: 200mV/div
10pF
SCOPE
OUTPUT
HORIZONTAL
SCALE: 10ns/div
a
2.15
A HIGH PERFORMANCE VIDEO LINE DRIVER
The AD8047 and AD8048 VFB op amps have been optimized to offer outstanding
performance as video line drivers. They utilized the "quad core" gm stage as
previously described for high slew rate and low distortion. The AD8048 (optimized
for G = +2) has a differential gain of 0.01% and a differential phase of 0.02°, making
it suitable for HDTV applications. In the configuration shown in Figure 2.16, the
0.1dB bandwidth is 50MHz for ±5V supplies, slew rate is 1000V/µs, and 0.1%
settling time is 13ns. Total quiescent current is 6mA (±5V), and quiescent power
dissipation 60mW.
16
VIDEO LINE DRIVER USING AD8047/AD8048:
∆G = 0.01%, ∆ φ = 0.02°, 50MHz 0.1dB BANDWIDTH, 6mA (±5V)
200Ω
200Ω
+VS
10µ F
0.1µ F
2
75Ω
CABLE
AD8047/
AD8048
3
VIN
7
4
75Ω
75Ω
CABLE
6
VOUT
0.1µ F
75Ω
75Ω
10µF
-VS
a
2.16
DIFFERENTIAL LINE DRIVERS/RECEIVERS
Many applications require gain/phase matched complementary or differential
signals. Among these are analog-digital-converter (ADC) input buffers, where
differential operation can provide lower levels of 2nd harmonic distortion for certain
converters. Other uses include high frequency bridge excitation, and drivers for
balanced transmission twisted pair lines such as in ADSL and HDSL.
The transmission of high quality signals across noisy interfaces (either between
individual PC boards or between racks) has always been a challenge to design
engineers. Differential techniques using high common-mode-rejection-ratio (CMRR)
instrumentation amplifiers largely solves the problem at low frequencies.
At audio frequencies, transformers, or products such as the SSM-2142 balanced line
driver and SSM-2141/SSM-2143 line receiver offer outstanding CMRRs and the
ability to transmit low-level signals in the presence of large amounts of noise. At
high frequencies, small toroid transformers using bifilar windings are effective.
The problem of signal transmission at video frequencies is complex. Transformers
are not effective, because the baseband video signal has low-frequency components
down to a few tens of Hz. Video signals are generally single-ended, and therefore
don't adapt easily to balanced transmission line techniques. In addition, shielded
twin-conductor coaxial cable with good bandwidth is usually somewhat bulky and
expensive. Finally, designing high bandwidth, low distortion differential video
drivers and receivers with high CMRRs at high frequencies is an extremely difficult
task.
Even with the above problems, there are differential techniques available now which
offer distinct advantages over single-ended methods. Some of these techniques make
17
use of discrete components, while others utilize the latest in state-of-the-art video
differential amplifiers.
Two solutions to the problem of differential transmission and reception are shown in
Figure 2.17. The first represents the ideal case, where a balanced differential line
driver drives a balanced twin-conductor coaxial cable which in turn drives a
differential line receiver. This circuit, however, is difficult to implement fully at
video frequencies for the reasons previously discussed.
TWO APPROACHES FOR
DIFFERENTIAL LINE DRIVING/RECEIVING
RO/2
+
+
RO
RO/2
-
--
VNOISE
GND A
+
GND B
RO
+
RO
-
-
VNOISE
GND A
a
GND B
2.17
The second and most often used approach makes use of a single-ended driver which
drives a source-terminated coaxial cable. The shield of the coaxial cable is grounded
at the transmitting end. At the receiving end, the coaxial cable is terminated in its
characteristic impedance, but the shield is left floating in order to prevent a ground
loop between the two systems. The common mode ground noise is rejected by the
CMRR of the differential line receiver. The success of this approach depends upon
the characteristics of the line receiver.
Inverter-Follower Differential Driver
The circuit of Figure 2.18 is useful as a high speed differential driver for driving
high speed 10-12 bit ADCs, differential video lines, and other balanced loads at
levels of 1-4Vrms. As shown it operates from ±5V supplies, but it can also be
adapted to supplies in the range of ±5 to ±15V. When operated directly from ±5V as
here, it minimizes potential for destructive ADC overdrive when higher supply
voltage buffers drive a ±5V powered ADC, in addition to minimizing driver power.
18
DIFFERENTIAL DRIVER USING INVERTER/FOLLOWER
+5V
VIN
+
RTA
U1A
AD812
RIN
VOUTA
75Ω
-
83.5Ω(75Ω)
53.6Ω (50Ω)
RG1
RFB1
205Ω
549Ω
RG2
RFB2
RTB
715Ω
715Ω
75Ω
VOUTB
R4
NOTE:
U1B
301W
AD812
DECOUPLING
NOT SHOWN
+
-5V
a
2.18
In many of these differential drivers the performance criteria is high. In addition to
low output distortion, the two signals should maintain gain/phase flatness. In this
topology, two sections of an AD812 dual current feedback amplifier are used for the
channel A & B buffers, U1A & U1B. This can provide inherently better open-loop
bandwidth matching than the use of two singles (where bandwidth varies between
devices from different manufacturing lots).
The two buffers here operate with precise gains of ±1, as defined by their respective
feedback and input resistances. Channel B buffer U1B is conventional, and uses a
matched pair of 715Ω resistors- the value for using the AD812 on ±5V supplies.
In channel A, non-inverting buffer U1A has an inherent signal gain of 1, by virtue of
the bootstrapped feedback network RFB1 and RG1(Reference 5). It also has a higher
noise gain, for phase matching. Normally a current feedback amplifier operating as
a simple unity gain follower would use one (optimum) resistor RFB1, and no gain
resistor at all. Here, with input resistor RG1 added, a U1A noise gain like that of
U1B results. Due to the bootstrap connection of RFB1-RG1, the signal gain is
maintained at unity. Given the matched open loop bandwidths of U1A and U1B,
similar noise gains in the A-B channels provide closely matched output bandwidths
between the driver sides, a distinction which greatly impacts overall matching
performance.
In setting up a design for the driver, the effects of resistor gain errors should be
considered for RG2-RFB2. Here a worst case 2% mis-match will result in less than
0.2dB gain error between channels A and B. This error can be improved simply by
specifying tighter resistor ratio matching, avoiding trimming.
19
If desired, phase matching is trimmed via RG1, so that the phase of channel A
closely matches that of B. This can be done for new circuit conditions, by using a
pair of closely matched (0.1% or better) resistors to sum the A and B channels, as
RG1 is adjusted for the best null conditions at the sum node. The A-B gain/phase
matching is quite effective in this driver, with test results of the circuit as shown
0.04dB and 0.1° between the A and B output signals at 10MHz, when operated into
dual 150Ω loads. The 3dB bandwidth of the driver is about 60MHz.
Net input impedance of the circuit is set to a standard line termination value such
as 75Ω (or 50Ω), by choosing RIN so that the desired value results with RIN in
parallel with RG2. In this example, an RIN value of 83.5Ω provides a standard input
impedance of 75Ω when paralleled with 715Ω. For the circuit just as shown, dual
voltage feedback amplifier types with sufficiently high speed and low distortion can
also be used. This allows greater freedom with regard to resistor values using such
devices as the AD826 and AD828.
Gain of the circuit can be changed if desired, but this is not totally straightforward.
An easy step to satisfy diverse gain requirements is to simply use a triple amplifier
such as the AD813, with the third channel as a variable gain input buffer.
Cross-Coupled Differential Driver
Another differential driver approach uses cross-coupled feedback to get very high
CMR and complementary outputs at the same time. In Figure 2.19, by connecting
AD8002 dual current feedback amplifier sections as cross-coupled inverters, their
outputs are forced equal and opposite, assuring zero output common mode voltage.
CROSS-COUPLED DIFFERENTIAL DRIVER
PROVIDES BALANCED OUTPUTS AND 250MHz BANDWIDTH
C1
VIN
0.9pF (see text)
R1
R2
511Ω
511Ω
+5V
U1A
AD8002
VOUTA
+
-
A OUT
49.9Ω
RX
VOUT
511W
VOUTB
RX
511Ω
R3
(see text)
RTA
RTB
B OUT
49.9Ω
RX
511Ω
NOTE:
ALL RESISTORS 1%
DECOUPLING NOT SHOWN
RX
511Ω
-
U1B
R4
a
100Ω
AD8002
+
-5V
20
2.19
The gain cell which results, U1A and U1B plus cross-coupling resistances RX, is
fundamentally a differential input and output topology, but it behaves as a voltage
feedback amplifier with regard to the feedback port at the U1A (+) node. The gain of
the stage from VIN to VOUT is:
G=
VOUT 2R2
=
VIN
R1
where VOUT is the differential output, equal to VOUTA – VOUTB.
This relationship may not be obvious, so it can be derived as follows:
Using the conventional inverting op amp gain equation, the input voltage VIN
develops an output voltage VOUTB given by:
VOUTB = − VIN
R2
.
R1
Also, VOUTA = –VOUTB,
because VOUTA is inverted by U1B.
However, VOUT = VOUTA – VOUTB = –2VOUTB.
Therefore,
R2 
R2

VOUT = −2 − VIN
, and
 = 2VIN

R1 
R1
VOUT 2 R2
=
.
VIN
R1
This circuit has some unique benefits. First, differential gain is set by a single
resistor ratio, so there is no necessity for side-side resistor matching with gain
changes, as is the case for conventional differential amplifiers (see line receivers,
below). Second, because the (overall) circuit emulates a voltage feedback amplifier,
these gain resistances are not as restrictive as in the case of a conventional current
feedback amplifier. Thus, they are not highly critical as to value as long as the
equivalent resistance seen by U1A is reasonably low (≤1kΩ in this case). Third, the
cell bandwidth can be optimized to the desired gain by a single optional resistor, R3,
as follows. If for instance, a net gain of 20 is desired (R2/R1=10), the bandwidth
would otherwise be reduced by roughly this amount, since without R3, the cell
operates with a constant gain-bandwidth product (working in the voltage feedback
mode). With R3 present however, advantage can be taken of the AD8002 current
feedback amplifier characteristics. Additional internal gain is added by the
connection of R3, which, given an appropriate value, effectively raises gainbandwidth to a level so as to restore the bandwidth which would otherwise be lost
by the higher closed loop gain.
21
In the circuit as shown, no R3 is necessary at the low working gain of 2 times
differential, since the 511Ω RX resistors are already optimized for maximum
bandwidth. Note that these four matched RX resistances are somewhat critical, and
will change in absolute value with the use of another current feedback amplifier. At
higher gain closed loop gains as set by R2/R1, R3 can be chosen to optimize the
working transconductance in the input stages of U1A and U1B, as follows:
R3 ≅
Rx
(R2 / R1) − 1
As in any high speed inverting feedback amplifier, a small high-Q chip type feedback
capacitance, C1, may be needed to optimize flatness of frequency response. In this
example, a 0.9pF value was found optimum for minimizing peaking. In general,
provision should be made on the PC layout for an NPO chip capacitor in the range of
0.5-2pF. This capacitor is then value selected at board characterization for optimum
frequency response.
For the dual trace, 1-500MHz swept frequency response plot of Figure 2.20, output
levels were 0dBm into matched 50Ω loads, through back termination resistances
RTA and RTB, at VOUTA and VOUTB. In this plot the vertical scale is 2dB/div, and
it shows the 3dB bandwidth of the driver measuring about 250MHz, with peaking
about 0.1dB. The four RX resistors along with RTA and RTB control low frequency
amplitude matching, which was within 0.1dB in the lab tests, using 511Ω 1%
resistor types. For tightest amplitude matching, these resistor ratios can be more
closely controlled.
FREQUENCY RESPONSE OF AD8002 CROSS-COUPLED
DRIVER IS >250MHz (C1 = 0.9pF ± 0.1pF)
RELATIVE
RESPONSE
(dB)
0
-3dB BW = 260 MHz
-2
-4
A OUT
B OUT
-6
-8
-10
-12
-14
-16
-18
10
20
40
100
200
400
FREQUENCY (MHz)
a
2.20
Due to the high gain-bandwidths involved with the AD8002, the construction of this
circuit should follow RF rules, with the use of a ground plane, chip bypass capacitors
22
of zero lead length at the ±5V supply pins, and surface mount resistors for lowest
inductance.
4 Resistor Differential Line Receiver
Figure 2.21 shows a low cost, medium performance line receiver using a high speed
op amp rated for video use. It is actually a standard 4 resistor difference amplifier
optimized for high speed, with a differential to single-ended gain of R2/R1. Using low
value, DC accurate/AC trimmed resistances for R1-R4 and a high speed, high CMR
op amp provides the good performance.
SIMPLE VIDEO LINE RECEIVER USING THE AD818 OP AMP
R2
1kΩ
C2
5pF
+VS
0.1µF
R1
1kΩ
7
2
VIN
U1
AD818
R3
1kΩ
3
VOUT
6
4
G=
0.1µF
R4
1kΩ
R2
,
R1
R1
R
= 3
R2
R4
C1
5pF
-VS
AC CMR
ADJUST
2.21
a
Practically speaking however, at low frequencies resistor matching can be more
critical to overall CMR than the rated CMR of the op amp. For example, the worst
case CMR (in dB) of this circuit due to resistor mismatch is:
 1 + R2 

R1  .
CMR = 20 log10 

 4 Kr 


In this expression the term “Kr” is a single resistor tolerance in fractional form
(1%=0.01, etc.), and it is assumed the amplifier has significantly higher CMR
(≥100dB). Using discrete 1% metal films for R1/R2 and R3/R4 yields a worst case
CMR of 34dB, 0.1% types 54dB, etc. Of course 4 random 1% resistors will on the
average yield a CMR better than 34dB, but not dramatically so. A single substrate
dual matched pair thin film network is preferred, for reasons of best noise rejection
and simplicity. One type suitable is the Ohmtek 1005, (Reference 6) which has a
ratio match of 0.1%, which will provide a worst case low frequency CMR of 66 dB.
23
This circuit has an interesting and desirable side property. Because of the resistors
it divides down the input voltage, and the amplifier is protected against overvoltage.
This allows CM voltages to exceed ±5V supply rails in some cases without hazard.
For operation with ±15V supplies, inputs should not exceed the supply rails.
At frequencies above 1MHz, the bridge balance is dominated by AC effects, and a
C1-C2 capacitive balance trim should be used for best performance. The C1
adjustment is intended to allow this, providing for the cancellation of stray layout
capacitance(s) by electrically matching the net C1-C2 values. In a given PC layout
with low and stable parasitic capacitance, C1 is best adjusted once in 0.5pF
increments, for best high frequency CMR. Using designated PC pads, production
values then would use the trimmed value. Good AC matching is essential to
achieving good CMR at high frequencies. C1-C2 should be types similar physically,
such as NPO (or other stable) ceramic chip style capacitors.
While the circuit as shown has unity gain, it can be gain-scaled in discrete steps, as
long as the noted resistor ratios are maintained. In practice, this means using taps
on a multi-ratio network for gain change, so as to raise both R2 and R4, in identical
proportions. There is no other simple way to change gain in this receiver circuit.
Alternately, a scheme for continuous gain control without interaction with CMR is to
follow this receiver with a scaling amplifier/driver with adjustable gain. The similar
AD828 dual amplifier allows this with the addition of only two resistors.
Video gain/phase performance of this stage is dependent upon the device used for U1
and the operating supply voltages. Suitable voltage feedback amplifiers work best at
supplies of ±10 - ±15V, which maximizes op amp bandwidth. And, while many high
speed amplifiers function in this circuit, those expressly designed with low distortion
video operation perform best. The circuit as shown can be used with supplies of ±5 to
±15V, but lowest NTSC video distortion occurs for supplies of ±10V or more, where
differential gain/differential phase errors are less than 0.01%/0.05°. Operating at
±5V supplies, the distortion rises somewhat, but the lowest power drain of 70mW
occurs.
One drawback to this circuit is that it does load a 75Ω video line to some extent, and
so should be used with this loading taken into account. On the plus side, it has wide
dynamic range for both signal and CM voltages, plus the inherent overvoltage
protection.
Active Feedback Differential Line Receiver
Fully integrating the line receiver function eliminates the resistor-related
drawbacks of the 4 resistor line receiver, improving CMR performance, ease of use,
and overall circuit flexibility. An IC designed for this function is the AD830 active
feedback amplifier (Reference 7,8). Its use as a differential line receiver with gain is
illustrated in Figure 2.22.
24
VIDEO LOOP-THROUGH CONNECTION USING THE AD830
VP
V1
1
2
0.1µF
A= 1
-
RT
75Ω
7
RL
75Ω
3
4
NC
6
+
C
GM
V4
R3
249Ω
8
GM
V2
V3
AD830
+
R2
499Ω
-
5
0.1µF
VN
CA (see text)
R1
499Ω
CA = 5.1pF (±15V)
CA = 12pF (±5V)
ZCM
a
2.22
The AD830 operates as a feedback amplifier with two sets of fully differential
inputs, available at pins 1-2 and 3-4, respectively. Internally, the outputs of the two
stages are summed and drive a buffer output stage. Both input stages have high
CMR, and can handle differential signals up to ±2V, and CM voltages can range up
to –Vs+3V or +Vs–2.1V, with a ±1V differential input applied. While the AD830 does
not normally need protection against CM voltages, if sustained transient voltage
beyond the rails is encountered, an optional pair of equal value (≅200Ω) resistances
can be used in series with pins 1-2.
In this device the overall feedback loop operates so that the differential voltages V12 and V3-4 are forced to be equal. Feedback is taken from the output back to one
input differential pair, while the other pair is driven by a differential input signal.
An important point of this architecture is that high CM rejection is provided by the
two differential input pairs, so CMR isn’t dependent on resistor bridges and their
associated matching problems. The inherently wideband balanced circuit and the
quasi-floating operation of the driven input provide the high CMR, which is typically
100dB at DC.
The general expression for the U1 stage’s gain “G” is like a non-inverting op amp, or:
G=
VOUT
R2
= 1+
VIN
R1
For lowest DC offset, balancing resistor R3 is used (equal to R1|| R2).
In this example of a video “loop-through” connection, the input signal tapped from a
coax line and applied to one input stage at pins 1-2, with the scaled output signal
25
tied to the second input stage between pins 3-4. With the R1-R2 feedback
attenuation of 2/1, the net result is that the output of U1, is then equal to 2•VIN,
i.e., a gain of 2.
Functionally, the input and local grounds are isolated by the CMR of the AD830,
which is typically 75dB at frequencies below 1MHz, 60dB at 4.43MHz, and
relatively supply independent.
With the addition of an output source termination resistor RT, this circuit has an
overall loaded gain of unity at the load termination, RL. It is a ground isolating
video repeater, driving the terminated 75Ω output line, delivering a final output
equal to the original input, VIN.
NTSC video performance will be dependent upon supplies. Driving a terminated line
as shown, the circuit has optimum video distortion levels for Vs = ±15V, where
differential gain is typically 0.06%, and differential phase 0.08°. Bandwidth can be
optimized by the optional 5.1pF (or 12pF) capacitor, CA, which allows a 0.1dB
bandwidth of 10MHz with ±15V operation. The differential gain and phase errors
are about 2× at ±5V.
HIGH SPEED CLAMPING AMPLIFIERS
There are many situations where it is desirable to clamp the output of an op amp to
prevent overdriving the circuitry which follows. Specially designed high speed, fast
recovery clamping amplifiers offer an attractive alternative to designing external
clamping/protection circuits. The AD8036/AD8037 low distortion, wide bandwidth
clamp amplifiers represent a significant breakthrough in this technology. These
devices allow the designer to specify a high (VH) and low (VL) clamp voltage. The
output of the device clamps when the input exceeds either of these two levels. The
AD8036/AD8037 offer superior clamping performance compared to competing
devices that use output-clamping. Recovery time from overdrive is less than 5ns.
The key to the AD8036 and AD8037's fast, accurate clamp and amplifier
performance is their proprietary input clamp architecture. This new design reduces
clamp errors by more than 10x over previous output clamp based circuits, as well as
substantially increasing the bandwidth, precision, and versatility of the clamp
inputs.
Figure 2.23 is an idealized block diagram of the AD8036 connected as a unity gain
voltage follower. The primary signal path comprises A1 (a 1200V/µs, 240MHz high
voltage gain, differential to single-ended amplifier) and A2 (a G=+1 high current
gain output buffer). The AD8037 differs from the AD8036 only in that A1 is
optimized for closed-loop gains of two or greater.
26
AD8036/AD8037 CLAMP AMPLIFIER EQUIVALENT CIRCUIT
RF
140Ω
-VIN
+VIN
+
A1
-
A
+1
A2
+1
VOUT
S1
VH
+1
VL
+1
B
C
S1
VIN > VH
+
CH
-
A B C
0
1
0
VL ≤ VIN ≤ VH 1
0
0
0
1
VIN < VL
0
+
CL
-
a
2.23
The input clamp section is comprised of comparators CH and CL, which drive switch
S1 through a decoder. The unity-gain buffers in series with the +VIN, VH, and VL
inputs isolate the input pins from the comparators and S1 without reducing
bandwidth or precision.
The two comparators have about the same bandwidth as A1 (240MHz), so they can
keep up with signals within the useful bandwidth of the AD8036. To illustrate the
operation of the input clamp circuit, consider the case where VH is referenced to
+1V, VL is open, and the AD8036 is set for a gain of +1 by connecting its output
back to its inverting input through the recommended 140Ω feedback resistor. Note
that the main signal path always operates closed loop, since the clamping circuit
only affects A1's noninverting input.
If a 0V to +2V voltage ramp is applied to the AD8036's +VIN for the connection just
described, VOUT should track +VIN perfectly up to +1V, then should limit at exactly
+1V as +VIN continues to +2V.
In practice, the AD8036 comes close to this ideal behavior. As the +VIN input
voltage ramps from zero to 1V, the output of the high limit comparator CH starts in
the off state, as does the output of CL. When +VIN just exceeds VH (practically, by
about 18mV), CH changes state, switching S1 from "A" to "B" reference level. Since
the + input of A1 is now connected to VH, further increases in +VIN have no effect
on the AD8036's output voltage. The AD8036 is now operating as a unity-gain
buffer for the VH input, as any variation in VH, for VH > 1V, will be faithfully
produced at VOUT.
Operation of the AD8036 for negative input voltages and negative clamp levels on
VL is similar, with comparator CL controlling S1. Since the comparators see the
27
voltage on the +VIN pin as their common reference level, the voltage VH and VL are
defined as "High" or "Low" with respect to +VIN. For example, if VIN is set to zero
volts, VH is open, and VL is +1V, comparator CL will switch S1 to "C", so the
AD8036 will buffer the voltage on VL and ignore +VIN.
The performance of the AD8036/AD8037 closely matches the ideal just described.
The comparator's threshold extends from 60mV inside the clamp window defined by
the voltages on VL and VH to 60mV beyond the window's edge. Switch S1 is
implemented with current steering, so that A1's + input makes a continuous
transition from say, VIN to VH as the input voltage traverses the comparator's
input threshold from 0.9V to 1.0V for VH = 1.0V.
The practical effect of the non-ideal operation is to soften the transition from
amplification to clamping modes, without compromising the absolute clamp limit set
by the input clamping circuit. Figure 2.24 is a graph of VOUT versus VIN for the
AD8036 and a typical output clamp amplifier. Both amplifiers are set for G=+1 and
VH = +1V.
COMPARISON BETWEEN INPUT AND OUTPUT CLAMPING
OUTPUT VOLTAGE - VOUT
1.6
1.4
1.2
CLAMP ERROR - 25mV
AD8036
1.0
AD8036
OUTPUT CLAMP AMP
0.8
0.6
0.6
a
CLAMP ERROR - >200mV
OUTPUT CLAMP
0.8
1.0
1.2
1.4
INPUT VOLTAGE +VIN
1.6
1.8 02
.
2.24
The worst case error between VOUT (ideally clamped) and VOUT (actual) is
typically 18mV times the amplifier closed-loop gain. This occurs when VIN equals
VH (or VL). As VIN goes above and/or below this limit, VOUT will stay within 5mV
of the ideal value.
In contrast, the output clamp amplifier's transfer curve typically will show some
compression starting at an input of 0.8V, and can have an output voltage as far as
200mV over the clamp limit. In addition, since the output clamp causes the
amplifier to operate open-loop in the clamp mode, the amplifier's output impedance
will increase, potentially causing additional errors, and the recovery time is
significantly longer.
28
It is important that a clamped amplifier such as the AD8036/AD8037 maintain low
levels of distortion when the input signals approach the clamping voltages. Figure
2.25 shows the second and third harmonic distortion for the amplifiers as the output
approaches the clamp voltages. The input signal is 20MHz, the output signal is 2V
peak-to-peak, and the output load is 100Ω.
Recovery from step voltage which is two times over the clamping voltage is shown in
Figure 2.26. The input step voltage starts at +2V and goes to 0V (left-hand traces on
scope photo). The input clamp voltage (VH) is set at +1V. The right-hand trace
shows the output waveform. The key specifications for the AD8036/AD8037 clamped
amplifiers are summarized in Figure 2.27.
AD8036/AD8037 DISTORTION NEAR CLAMPING REGION,
OUTPUT = 2V p-p, LOAD = 100Ω , f = 20MHz
-80
-75
HARMONIC DISTORTION - dBc
-70
AD8037 3RD
HARMONIC
-65
-60
-55
-50
-45
AD8037 2ND
HARMONIC
AD8036 3RD
HARMONIC
AD8036 2ND
HARMONIC
AD8036
-40
-35
-30
0.6
0.65
0.7
0.75
AD8037
VH
+1V
+0.5V
VL
-1V
-0.5V
G
+1
+2
0.8
0.85
0.9
0.95
1.0
ABSOLUTE VALUE OF OUTPUT VOLTAGE - Volts
a
2.25
29
AD8036/AD8037 OVERDRIVE (2x) RECOVERY
INPUT +2V
OUTPUT
+1V
0V
1ns
REF
HORIZONTAL SCALE: 1ns/div
a
2.26
AD8036/AD8037 SUMMARY SPECIFICATIONS
n
Proprietary Input Clamping Circuit with Minimized Nonlinear
Clamping Region
n
Small Signal Bandwidth: 240MHz (AD8036), 270MHz (AD8037)
n
Slew Rate: 1500V/µs
n
1.5ns Overdrive Recovery
n
Low Distortion: -72dBc @ 20MHz (500Ω
Ω load)
n
Low Noise: 4.5nv/√
√ Hz, 2pA/√
√ Hz
n
20mA Supply Current on ±5V
a
2.27
Figure 2.28 shows the AD9002 8-bit, 125MSPS flash converter driven by the
AD8037 (240MHz bandwidth) clamping amplifier. The clamp voltages on the
AD8037 are set to +0.55 and –0.55V, referenced to the ±0.5V input signal, with the
external resistive dividers. The AD8037 also supplies a gain of two, and an offset of
–1V (using the AD780 voltage reference), to match the 0 to –2V input range of the
AD9002 flash converter. The output signal is clamped at +0.1V and –2.1V. This
multi-function clamping circuit therefore performs several important functions as
30
well as preventing damage to the flash converter which occurs if its input exceeds
+0.5V, thereby forward biasing the substrate diode. The 1N5712 Schottky diode
adds further protection during power-up.
AD9002 8-BIT, 125MSPS FLASH CONVERTER
DRIVEN BY AD8037 CLAMP AMPLIFIER
0.1µF
BIPOLAR
SIGNAL
±0.5V
+5V
806Ω
IN5712
+
RT
75Ω
VH = +0.55V
AD8037
CLAMPING
AMP
+5V
750Ω
10µF
AD780
+2.5V
REF
0.1µF
49.9Ω
AD9002
FLASH CONVERTER
(8-BITS, 125MSPS)
VIN = -1 ±1V
VL = -0.55V
-
+
100Ω
100Ω
806Ω
SUBSTRATE
DIODE
R3
-5.2V
0.1µF
-5.2V
R2
301Ω
R1
499Ω
a
0.1µF
AD8037 OUTPUT
CLAMPS AT +0.1V, -2.1V
R1 R3 = R2
2.5 R1
R1 + R3
= 1 VOLT
2.28
The feedback resistor, R2 = 301Ω, is selected for optimum bandwidth per the data
sheet recommendation. For a gain of two, the parallel combination of R1 and R3
must also equal R2:
R1 ⋅ R3
= R2 = 301Ω
R1 + R3
(nearest 1% standard resistor value).
In addition, the Thevenin equivalent output voltage of the AD780 +2.5V reference
and the R3/ R1 divider must be +1V to provide the –1V offset at the output of the
AD8037.
2.5 ⋅ R1
= 1volt
R1 + R3
Solving the equations yields R1 = 499Ω, R3 = 750Ω (using the nearest 1% standard
resistor values).
Other input and output voltages ranges can be accommodated by appropriate
changes in the external resistors.
Further examples of applications of these fast clamping op amps are given in
Reference 9.
31
SINGLE-SUPPLY/RAIL-TO-RAIL CONSIDERATIONS
The market is driving high speed amplifiers to operate at lower power on lower
supply voltages. High speed bipolar processes, such as Analog Devices' CB and
XFCB, are basically 12V processes, and circuits designed on these processes are
generally limited to ±5V power supplies (or less). This is ideal for high speed video,
IF, and RF signals, which rarely exceed 5V peak-to-peak.
The emphasis on low power, battery-operated portable communications and
instrumentation equipment has brought about the need for ICs which operate on
single +5V, and +3V, and lower supplies. The term single-supply has various
implications, some of which are often further confused by marketing hype.
There are many obvious reasons for lower power dissipation, such as the ability to
function without fans, reliability issues, etc. There are, therefore, many applications
for single-supply devices other than in systems which have only one supply voltage.
For example, the lower power dissipation of a single-supply ADC may be the reason
for its selection, rather than the fact that it requires just one supply.
There are also systems which truly operate on a single power supply. In such cases,
it can often be difficult to maintain DC coupling from a transducer all the way
through to the ADC. In fact, AC coupling is often used in single-supply systems,
with DC restoration preceding the ADC. This may be required to prevent the loss of
dynamic range which would otherwise occur because of the need to provide adequate
headroom to an AC coupled signal of arbitrary duty cycle. In the AC-coupled
portions of such systems, a "false-ground" is often created, usually centered between
the rails.
There are other disadvantages associated with lower power supply voltages. Signal
swings are limited, therefore high-speed single-supply circuits tend to be more
sensitive to corruption by wideband noise, etc. The single-supply op amp and ADC
usually utilize the same power bus that supplies the digital circuits, making proper
filtering and decoupling extremely critical.
In order to maximize the signal swing in single-supply circuits, it is desirable that a
high speed op amp utilize as much of the supply range as possible on both the input
and output. Ideally, a true rail-to-rail input op amp has an input common-mode
range that includes both supply rails, and an output range which does likewise. This
makes for some interesting tradeoffs and compromises in the circuit design of the op
amp.
In many cases, an op amp may be fully specified for both dual ±5V and single-supply
operation but neither its input nor its output can actually swing closer than about
1V to either supply rail. Such devices must be used in applications where the input
and output common-mode restrictions are not violated. This generally involves
offsetting the inputs using a false ground reference scheme.
To summarize, there are many tradeoffs involved in single-supply high-speed
designs. In many cases, using devices specified for operation on +5V, but without
true rail inclusive input/output operation can give good performance. Amplifiers are
32
also becoming available that are true single supply rail-to-rail devices.
Understanding single-supply rail-to-rail input and output limitations is easy if you
understand a few basics about the circuitry inside the op amp. We shall consider
input and output stages separately.
HIGH SPEED SINGLE SUPPLY AMPLIFIERS
n
Single Supply Offers:
u
Lower Power
u
Battery Operated Portable Equipment
u
Simplifies Power Supply Requirements (one voltage)
n
Design Tradeoffs:
u
Limited Signal Swings Increase Sensitivity to Noise
u
Usually Share Noisy Digital Power Supply
u
DC Coupling Throughout is Difficult
u
Rail-to-Rail Input and Output Increases Signal Swing,
but not Required in All Applications
u
Many Op Amps Specified for Single Supply, but do not
have Rail-to-Rail Inputs or Outputs
a
2.29
There is some demand for high-speed op amps whose input common-mode voltage
includes both supply rails. Such a feature is undoubtedly useful in some
applications, but engineers should recognize that there are relatively few
applications where it is absolutely essential. These should be carefully distinguished
from the many applications where common-mode range close to the supplies or one
that includes one of the supplies is necessary, but input rail-to-rail operation is not.
In many single-supply applications, it is required that the input go to only one of the
supply rails (usually ground). Amplifiers which will handle zero-volt inputs are
relatively easily designed using PNP differential pairs (or N-channel JFET pairs) as
shown in Figure 2.30 (circuit used in the AD8041, AD8042, AD8044). The input
common-mode range of such an op amp extends from about 200mV below the
negative supply to within about 1V of the positive supply. If the stage is designed
with N-channel JFETs (AD820/AD822/AD823/AD824), the input common-mode
range would also include the negative rail.
33
OP90 AND OPX93 INPUT STAGE ALLOWS
INPUT TO GO TO THE NEGATIVE RAIL
+VS
+VBIAS
-VS
a
2.30
The input stage could also be designed with NPN transistors (or P-channel JFETs),
in which case the input common-mode range would include the positive rail and to
within about 1V of the negative rail; however, this requirement typically occurs in
applications such as high-side current sensing, a low-frequency measurement
application. The OP282/OP482 input stage uses the P-channel JFET input pair
whose input common-mode range includes the positive rail.
True rail-to-rail input stages require two long-tailed pairs (see Figure 2.31), one of
NPN bipolar transistors (or N-channel JFETs), the other of PNP transistors (or
N-channel JFETs). These two pairs exhibit different offsets and bias currents, so
when the applied input common-mode voltage changes, the amplifier input offset
voltage and input bias current does also. In fact, when both current sources (I1 and
I2) remain active throughout the entire input common-mode range, amplifier input
offset voltage is the average offset voltage of the NPN pair and the PNP pair. In
those designs where the current sources are alternatively switched off at some point
along the input common-mode voltage, amplifier input offset voltage is dominated by
the PNP pair offset voltage for signals near the negative supply, and by the NPN
pair offset voltage for signals near the positive supply.
34
RAIL-TO-RAIL INPUT STAGE TOPOLOGY
+VS
R2
R1
I1
Ib
Q2
Q1
+IN
Q3
Q4
VOS
-IN
R3
I2
R4
-VS
a
2.31
Amplifier input bias current, a function of transistor current gain, is also a function
of the applied input common-mode voltage. The result is relatively poor commonmode rejection (CMR), and a changing common-mode input impedance over the
common-mode input voltage range, compared to familiar dual-supply devices. These
specifications should be considered carefully when choosing a rail-rail input op amp,
especially for a non-inverting configuration. Input offset voltage, input bias current,
and even CMR may be quite good over part of the common-mode range, but much
worse in the region where operation shifts between the NPN and PNP devices and
vice versa.
True rail-to-rail amplifier input stage designs must transition from one differential
pair to the other differential pair somewhere along the input common-mode voltage
range. Devices like the AD8031/AD8032 (specified for ±5V, +5V, +3V, and +2.5V)
have a common-mode crossover threshold at approximately 1V below the positive
supply. The PNP differential input stage is active from about 200mV below the
negative supply to within about 1V of the positive supply. Over this common-mode
range, amplifier input offset voltage, input bias current, CMR, input noise
voltage/current are primarily determined by the characteristics of the PNP
differential pair. At the crossover threshold, however, amplifier input offset voltage
becomes the average offset voltage of the NPN/PNP pairs and can change rapidly.
Also, amplifier bias currents, dominated by the PNP differential pair over most of
the input common-mode range, change polarity and magnitude at the crossover
threshold when the NPN differential pair becomes active.
Applications which require true rail-rail inputs should therefore be carefully
evaluated, and the amplifier chosen to ensure that its input offset voltage, input bias
current, common-mode rejection, and noise (voltage and current) are suitable.
35
Figure 2.32 shows two typical high-speed op amp output stages. The emitterfollower stage is widely used, but its output voltage range is limited to within about
1V of either supply voltage. This is sufficient for many applications, but the
common-emitter stage (used in the AD8041/8042/8044/8031/8032 and others) allows
the output to swing to within the transistor saturation voltage, VCE(SAT), of the
rails. For small amounts of load current (less than 100µA), the saturation voltage
may be as low as 5 to 20mV, but for higher load currents, the saturation voltage can
increase to several hundred millivolts (for example, 500mV at 50mA). This is
illustrated in Figure 2.33 for the AD8042 (zero-volts in, rail-to-rail output). The solid
curves show the output saturation voltage of the PNP transistor (output sourcing
current), and the dotted curves the NPN transistor (sinking current). The saturation
voltage increases with increasing temperature as would be expected.
HIGH SPEED SINGLE SUPPLY OP AMP OUTPUT STAGES
COMMON EMITTER
EMITTER FOLLOWER
+V S
+VS
OUTPUT
OUTPUT
-V S
-VS
a
2.32
36
AD8042 OUTPUT SATURATION VOLTAGE
VERSUS LOAD CURRENT
0.80
VS = +5V
OUTPUT SATURATION VOLTAGE - V
0.70
+5V - VOH (+125°C)
+5V - VOH (+25°C)
0.60
+5V - VOH (-55°C)
0.50
0.40
0.30
0.20
+VOL
(+125°C)
+VOL
(+25°C)
+VOL (-55°C)
0.10
0
0
5
10
15
20
25
30
LOAD CURRENT - mA
35
40
45
a
50
2.33
An output stage constructed of CMOS FETs can provide true rail-to-rail
performance, but only under no-load conditions, and in much lower frequency
amplifiers. If the output must source or sink current, the output swing is reduced by
the voltage dropped across the FETs internal "on" resistance (typically 100Ω).
SINGLE SUPPLY OP AMP APPLICATIONS
The following section illustrates a few applications of op amps in single-supply
circuits. All of the op amps are fully specified for both ±5V and +5V (and +3V where
the design supports it). Both rail-to-rail and non-rail-to-rail applications are shown.
A Single-Supply 10-bit 20MSPS ADC Direct-Coupled Driver Using the
AD8011
The circuit in Figure 2.34 shows the AD8011 op amp driving the AD876 10-bit,
20MSPS ADC in a direct-coupled application. The input and output common-mode
voltage of the AD8011 must lie between approximately +1 and +4V when operating
on a single +5V supply. The input range of the AD876 is 2V peak-to-peak centered
around a common-mode value of +2.6V, well within the output voltage range of the
AD8011. The upper and lower range setting voltages are +1.6V and +3.6V and are
supplied externally to the AD876. They are easily derived from a resistor divider
driven by a reference such as the REF198 (+4.096V). The two taps on the resistor
divider should be buffered using precision single-supply op amps such as the AD822
(dual).
37
DC COUPLED SINGLE SUPPLY DRIVER FOR
AD876 10-BIT, 20MSPS ADC
R3
+5V +3.6V
2000Ω
Ω
REF*
R2
1000Ω
Ω
R1
499Ω
Ω
RT
88.7Ω
Ω
75 Ω
+FS REF
+5V
AD876*
+5V
-
100Ω
Ω
50 Ω
AD8011
5pF
+
0 TO +2V
+1.6V
+5V
+1.6V
REF*
*COMPLETE CIRCUIT
-FS REF
NOT SHOWN
a
2.34
The source is represented as a 2V video signal referenced to ground. (The equivalent
of a current generator of 0 to 27mA in parallel with the 75Ω source resistor. The
termination resistor, RT, is selected such that the parallel combination of RT and R1
is 75Ω. The peak-to-peak swing at the termination resistor is 1V, so the AD8011
must supply a gain of two.
The non-inverting input of the AD8011 is biased to a common-mode voltage of +1.6V
(well within it's allowable common-mode range). R3 is calculated as follows:
When the source voltage is zero-volts, there is a current of 3.0mA flowing through
R1 (499Ω) and into 40.6Ω to ground (the equivalent parallel combination of the 75Ω
source and the 88.7Ω termination resistor is 40.6Ω). The output of the AD8011
should be +3.6V under these conditions. This means that 2mA must flow through
R2. Therefore R3 (connected to the +3.6V source) must supply 1.0mA into the
summing junction (+1.6V), and therefore its value must be 2000Ω.
The input of the AD876 has a series MOSFET switch that turns on and off at the
sampling frequency. This MOSFET is connected to a hold capacitor internal to the
device. The on impedance of the MOSFET is about 50Ω, while the hold capacitor is
about 5pF.
In a worst case condition, the input voltage to the AD876 will change by a full-scale
value (2V) in one sampling cycle. When the input MOSFET turns on, the output of
the op amp will be connected to the charged hold capacitor through the series
resistance of the MOSFET. Without any other series resistance, the instantaneous
current that flows would be 40mA. This causes settling problems for the op amp.
38
The series 100Ω resistor limits the instantaneous current to about 13mA. This
resistor cannot be made too large, or the high frequency performance will be
affected. In practice, the optimum value is often determined experimentally.
The sampling MOSFET of the AD876 is closed for half of each cycle (25ns when
sampling at 20MSPS). Approximately 7 time constants are required for settling to
10 bits. The series 100Ω resistor along with the 50Ω on resistance and the 5pF hold
capacitor form a time constant of about 750ps. These values leave a comfortable
margin for settling. Overall, the AD8011 provides adequate buffering for the AD876
ADC without introducing distortion greater than that of the ADC itself.
A 10-Bit, 40MSPS ADC Low-Distortion Single-Supply ADC Driver Using the
AD8041 Op Amp
A DC coupled application which requires the rail-to-rail output capability of the
AD8041 is shown in Figure 2.35 as a driver for the AD9050 10-bit, 40MSPS singlesupply ADC. The input range of the AD9050 is 1V p-p centered around +3.3V. The
maximum input signal is therefore +3.8V. The non-inverting input of the AD8041 is
driven with a common-mode voltage of +1.65V which is derived from the unused
differential input of the AD9050. This allows the op amp to act as a level shifter for
the ground-referenced bipolar input 1V p-p signal, with unity gain as determined by
the 1kΩ resistors, R1 and R2.
DC COUPLED SINGLE-SUPPLY DRIVER FOR
AD9050 10-BIT, 40MSPS ADC
+5V
R2
1000Ω
Ω
R1
Ω
1000Ω
VIN
-0.5 TO +0.5V
AD8041
+
+3.8V
TO +2.8V
8k Ω
VIN(A)
16kΩ
Ω
+
+5V
0.1µ
µF
VIN(B)
VCM = +1.65V
+5V
1000Ω
Ω
AD9050
+5V
+5V
Ω
1000Ω
+
AD820
-
+3.3V
8k Ω
-
16kΩ
Ω
0.1µ
µF
a
2.35
Op amps with complementary emitter follower outputs such as the AD8011
(operating on +5V) generally will exhibit high frequency distortion for sinewaves
with full-scale amplitudes of 1V p-p centered at +3.3V. Because of its common
emitter output stage, however, the AD8041 is capable of driving the AD9050, while
maintaining a distortion floor of greater than 66dB with a 4.9MHz fullscale input
(see Figure 2.36).
39
FFT OUTPUT OF AD9050 CIRCUIT WITH 4.9MHz INPUT
AND 40MSPS SAMPLING FREQUENCY
0
-10
-20
-30
-40
F1 = 4.9MHz
FUNDAMENTAL = 0.6dB
2nd HARMONIC = 66.9dB
3rd HARMONIC = 74.7dB
SNR = 55.2dB
NOISE FLOOR = - 86.1dB
ENCODE FREQUENCY = 40MHz
-50
-60
-70
-80
-90
-100
a
2.36
Single-Supply RGB Buffer
Op amps such as the AD8041/AD8042/ and AD8044 can provide buffering of RGB
signals that include ground while operating from a single +3V or +5V supply. The
signals that drive an RGB monitor are usually supplied by current output DACs
that operate from a single +5V supply. Examples of such are triple video DACs like
the ADV7120/21/22 from Analog Devices.
During the horizontal blanking interval, the current output of the DACs goes to
zero, and the RGB signals are pulled to ground by the termination resistors. If more
than one RGB monitor is desired, it cannot simply be connected in parallel because
it will provide an additional termination. Therefore, buffering must be provided
before connecting a second monitor.
Since the RGB signals include ground as part of their dynamic output range, it has
previously been required to use a dual supply op amp to provide this buffering. In
some systems, this is the only component that requires a negative supply, so it can
be quite inconvenient to incorporate this multiple monitor feature.
Figure 2.37 shows a diagram of one channel of a single supply gain-of-two buffer for
driving a second RGB monitor. No current is required when the amplifier output is
at ground. The termination resistor at the monitor helps pull the output down at low
voltage levels.
40
SINGLE SUPPLY RGB BUFFER OPERATES ON +3V OR +5V
CURRENT OUTPUT
VIDEO DAC
+3V OR +5V
R, G, OR B
75 Ω
0mA
TO
27mA
AD8041
75 Ω
1k Ω
1k Ω
75 Ω
75 Ω
SECOND RGB
MONITOR
PRIMARY RGB
MONITOR
a
2.37
Figure 2.38 shows the output of such a buffer operating from a single +3V supply
and driven by the Blue signal of a color bar pattern. Note that the input and output
are at ground during the horizontal blanking interval. The RGB signals are specified
to output a maximum of 700mV peak. The output of the AD8041 is +1.4V with the
termination resistors providing a divide-by-two. The Red and Green signals can be
buffered in the same manner with a duplication of this circuit. Another possibility is
to use the quad AD8044 single-supply op amp.
INPUT/OUTPUT OF SINGLE SUPPLY RGB BUFFER
OPERATING ON +3V
500mV
5µs
100
VIN 90
GND
VOUT
10
GND
0%
500mV
a
2.38
41
Single-Supply Sync Stripper
Some RGB monitors use only three cables total and carry the synchronizing signals
and the Green (G) signal on the same cable. The sync signals are pulses that go in
the negative direction from the blanking level of the G signal.
In some applications, such as prior to digitizing component video signals with ADCs,
it is desirable to remove or strip the sync portion from the G signal. Figure 2.39 is a
circuit using the AD8041 running on a single +5V supply that performs this
function.
SINGLE SUPPLY VIDEO SYNC STRIPPER
+5V
75 Ω
+
VIN
75 Ω
GREEN WITH SYNC
VOUT
AD8041
-
75 Ω
R2
1k Ω
+0.4V
0V
VBLANK
R1
1k Ω
0V
0.8V
GREEN WITHOUT SYNC
(2 X VBLANK)
a
2.39
The upper waveform in Figure 2.40 shows the Green plus sync signal that is output
from an ADV7120, a single supply triple video DAC. Because the DAC is single
supply, the lowest level of the sync tip is at ground or slightly above. The AD8041 is
set from a gain of two to compensate for the divide-by-two of the output
terminations. The reference voltage for R1 should be twice the DC blanking level of
the G signal. If the blanking level is at ground and the sync tip is negative, as in
some dual supply systems, then R1 can be tied to ground. In either case, the output
will have the sync removed and have the blanking level at ground.
42
INPUT/OUTPUT OF SINGLE SUPPLY SYNC STRIPPER
500mV
10µs
100
VIN
90
VOUT
10
0%
500mV
a
2.40
A Single-Supply Video Line Driver with Zero-Volt Output,
Eamon Nash
When operated with a single supply, the AD8031 80MHz rail-to-rail voltage
feedback op amp has optimum distortion performance when the signal has a
common mode level of Vs/2, and when there is about 500mV of headroom to each
rail. If low distortion is required for signals which swing close to ground, an emitter
follower can be used at the op amp output.
Figure 2.41 shows the AD8031 configured as a single supply gain-of-two line driver.
With the output driving a back terminated 50Ω line, the overall gain is unity from
Vin to Vout. In addition to minimizing reflections, the 50Ω back termination resistor
protects the transistor from damage if the cable is short circuited. The emitter
follower, which is inside the feedback loop, ensures that the output voltage from the
AD8031 stays about 700mV above ground. Using this circuit excellent distortion is
obtained even when the output signal swings to within 50mV of ground. The circuit
was tested at 500kHz and 2MHz using a single +5V supply. For the 500kHz signal,
THD was 68dBc with a peak-to-peak swing at Vout of 1.85V (50mV to +1.9V). This
corresponds to a signal at the emitter follower output of 3.7V p-p (100mV to 3.8V).
Data was taken with an output signal of 2MHz, and a THD of 55dBc was measured
with a Vout of 1.55V p-p (50mV to 1.6V).
43
LOW DISTORTION ZERO-VOLT OUTPUT
SINGLE SUPPLY LINE DRIVER USING AD8031
+5V
+
10 µF
THD = 68dBc @ 500kHz FOR
VOUT = 1.85Vp-p (50mV TO 1.9V)
VIN
0.1µ
µF
+
AD8031
49.9Ω
Ω
THD = 55dBc @ 2MHz FOR
VOUT = 1.55Vp-p (50mV TO 1.6V)
2N3904
2.49kΩ
Ω
49.9Ω
Ω
VOUT
2.49kΩ
Ω
200Ω
Ω
49.9Ω
Ω
a
2.41
This circuit can also be used to drive the analog input of a single supply high speed
ADC whose input voltage range is ground-referenced. In this case, the emitter of the
external transistor is connected directly to the ADC input. A peak positive voltage
swing of approximately 3.8V is possible before significant distortion begins to occur.
Headroom Considerations in AC-Coupled Single-Supply Circuits
The AC coupling of arbitrary waveforms can actually introduce problems which
don’t exist at all in DC coupled or DC restored systems. These problems have to do
with the waveform duty cycle, and are particularly acute with signals which
approach the rails, as they can in low supply voltage systems which are AC coupled.
In Figure 2.42 (A), an example of a 50% duty cycle square wave of about 2Vp-p level
is shown, with the signal swing biased symmetrically between the upper and lower
clip points of a 5V supply amplifier. Assume that the amplifier has a complementary
emitter follower output and can only swing to the limited DC levels as marked,
about 1V from either rail. In cases (B) and (C), the duty cycle of the input waveform
is adjusted to both low and high duty cycle extremes while maintaining the same
peak-to-peak input level. At the amplifier output, the waveform is seen to clip either
negative or positive, in (B) and (C), respectively.
44
WAVEFORM DUTY CYCLE TAXES
HEADROOM IN AC COUPLED AMPLIFIERS
(A)
50%
DUTY CYCLE
NO CLIPPING
4.0V (+) CLIPPING
2Vp-p
2.5V
1.0V (-) CLIPPING
(B)
LOW
DUTY CYCLE
CLIPPED
POSITIVE
4.0V (+) CLIPPING
2Vp-p
2.5V
1.0V (-) CLIPPING
4.0V (+) CLIPPING
(C)
HIGH
DUTY CYCLE
CLIPPED
2Vp-p
NEGATIVE
2.5V
1.0V (-) CLIPPING
a
2.42
Since standard video waveforms do vary in duty cycle as the scene changes, the
point is made that low distortion operation on AC coupled single supply stages must
take the duty cycle headroom degradation effect into account. If a stage has a 3Vp-p
output swing available before clipping, and it must cleanly reproduce an arbitrary
waveform, then the maximum allowable amplitude is less than 1/2 of this 3Vp-p
swing, that is <1.5Vp-p. An example of violating this criteria is the 2Vp-p waveform
of Figure 2.42, which is clipping for both the high and low duty cycles. Note that the
criteria set down above is based on avoiding hard clipping, while subtle distortion
increases may in fact take place at lower levels. This suggests an even more
conservative criteria for lowest distortion operation such as composite NTSC video
amplifiers.
Figure 2.43 shows a single supply gain-of-two composite video line driver using the
AD8041. Since the sync tips of a composite video signal extend below ground, the
input must be AC coupled and shifted positively to prevent clipping during negative
excursions. The input is terminated in 75Ω and AC coupled via the 47µF to a voltage
divider that provides the DC bias point to the input. Setting the optimal commonmode bias voltage requires some understanding of the nature of composite video
signals and the video performance of the AD8041.
45
SINGLE SUPPLY AC COUPLED COMPOSITE VIDEO
LINE DRIVER HAS ∆G = 0.06% AND ∆φ = 0.06°°
+5V
+
10µ
µF
COMPOSITE
VIDEO
75 Ω
+
4.99kΩ
Ω
10µ
µF
4.99kΩ
Ω
+
47µ
µF
7.87kΩ
Ω
RBIAS
VCM
0.1µ
µF
+
AD8041
1000µ
µF 75Ω
Ω
+
VOUT
75 Ω
1k Ω
RBIAS OPTIMIZES
VCM = +2.2V
0.1µ
µF
1k Ω
+
220µ
µF
a
2.43
As discussed above, signals of bounded peak-to-peak amplitude that vary in duty
cycle require larger dynamic swing capability than their peak-to-peak amplitude
after AC coupling. As a worst case, the dynamic signal swing required will approach
twice the peak-to-peak value. The two bounding cases are for a duty cycle that is
mostly low, but occasionally goes high at a fraction of a percent duty cycle, and vice
versa.
Composite video is not quite this demanding. One bounding extreme is for a signal
that is mostly black for an entire frame, but occasionally has a white (full intensity),
minimum width spike at least once per frame.
The other extreme is for a video signal that is full white everywhere. The blanking
intervals and sync tips of such a signal will have negative going excursions in
compliance with composite video specifications. The combination of horizontal and
vertical blanking intervals limit such a signal to being at its highest level (white) for
only about 75% of the time.
As a result of the duty cycle variations between the two extremes presented above, a
1V p-p composite video signal that is multiplied by a gain-of-two requires about 3.2V
p-p of dynamic voltage swing at the output for the op amp to pass a composite video
signal of arbitrary duty cycle without distortion.
The AD8041 not only has ample signal swing capability to handle the dynamic
range required, but also has excellent differential gain and phase when buffering
these signals in an AC coupled configuration.
To test this, the differential gain and phase were measured for the AD8041 while
the supplies were varied. As the lower supply is raised to approach the video signal,
the first effect is that the sync tips become compressed before the differential gain
46
and phase are adversely affected. Thus, there must be adequate swing in the
negative direction to pass the sync tips without compression.
As the upper supply is lowered to approach the video, the differential gain and phase
were not significantly adversely affected until the difference between the peak video
output and the supply reached 0.6V. Thus, the highest video level should be kept at
least 0.6V below the positive supply rail.
Taking the above into account, it was found that the optimal point to bias the noninverting input was at +2.2V DC. Operating at this point, the worst case differential
gain was 0.06% and the differential phase 0.06°.
The AC coupling capacitors used in the circuit at first glance appear quite large. A
composite video signal has a lower frequency band edge of 30Hz. The resistances at
the various AC coupling points - especially at the output - are quite small. In order
to minimize phase shifts and baseline tilt, the large value capacitors are required.
For video system performance that is not to be of the highest quality, the value of
these capacitors can be reduced by a factor of up to five with only a slight observable
change in the picture quality.
Single-Supply AC Coupled Single-Ended-to-Differential Driver
The circuit shown in Figure 2.44 provides a flexible solution to differential line
driving in a single-supply application and utilizes the dual AD8042. The basic
operation of the cross-coupled configuration has been described earlier in this
section. The input, VIN, is a single-ended signal that is capacitively coupled into the
feedforward resistor, R1. The non-inverting inputs of each half of the AD8042 are
biased at +2.5V. The gain from single-ended input to differential output is equal to
2R2/R1. The gain can be varied by changing one resistor (either R1 or R2).
SINGLE SUPPLY AC COUPLED DIFFERENTIAL DRIVER
R2
R1
VIN
+5V 1k Ω
+
0.1µ
µF
1k Ω
1/2 U
-
1k Ω
1k Ω
1k Ω
2VIN •
R2
R1
1k Ω
+5V
2.49kΩ
Ω
+2.5V
1/2 U
+
2.49kΩ
Ω
U = AD8042
0.1µ
µF
a
2.44
47
HIGH SPEED VIDEO MULTIPLEXING WITH OP AMPS
UTILIZING DISABLE FUNCTION
A common video circuit function is the multiplexer, a stage which selects one of "N"
video inputs and transmits a buffered version of the selected signal to the output. A
number of video op amps (AD810, AD813, AD8013) have a disable mode which,
when activated by applying the appropriate level to a pin on the package, disables
the op amp output stage and drops the power to a lower value.
In the case of the AD8013 (triple current-feedback op amp), asserting any one of the
disable pins about 1.6V from the negative supply will put the corresponding
amplifier into a disabled, powered-down state. In this condition, the amplifier's
quiescent current drops to about 0.3mA, its output becomes a high impedance, and
there is a high level of isolation from the input to the output. In the case of the gainof-two line driver, for example, the impedance at the output node will be about equal
to the sum of the feedback and feedforward resistors (1.6kΩ) in parallel with about
12pF capacitance. Input-to-output isolation is about 66dB at 5MHz.
Leaving the disable pin disconnected (floating) will leave the corresponding amplifier
operational, in the enabled state. The input impedance of the disable pin is about
40kΩ in parallel with 5pF. When driven to 0V, with the negative supply at –5V,
about 100µA flows into the disable pin.
When the disable pins are driven by CMOS logic, on a single +5V supply, the disable
and enable times are about 50ns. When operated on dual supplies, level shifting will
be required from standard logic outputs to the disable pins.
The AD8013's input stages include protection from the large differential voltages
that may be applied when disabled. Internal clamps limit this voltage to about ±3V.
The high input-to-output isolation will be maintained for voltages below this limit.
Wiring the amplifier outputs together as shown in Figure 2.45 will form a 3:1
multiplexer with about 50ns switching time between channels. The 0.1dB
bandwidth of the circuit is 35MHz, and the OFF channel isolation is 60dB at
10MHz. The simple logic level-shifting circuit shown on the diagram does not
significantly affect switching time.
The resistors were chosen as follows. The feedback resistor R2 of 845Ω was chosen
first for optimum bandwidth of the AD8013 current feedback op amp. When any
given channel is ON, it must drive both the termination resistor RL, and the net
dummy resistance, RX/2, where RX is an equivalent series resistance equal to R1 +
R2 + R3. To provide a net overall gain of unity plus an effective source resistance of
75Ω, the other resistor values must be as shown.
48
3:1 VIDEO MULTIPLEXER SWITCHES IN 50ns
R1, 665Ω
Ω
DISABLE
DRIVERS
(ONE SHOWN)
R2, 845Ω
Ω
+5V
R3
84 Ω
1/3 U1
VIN1
+
1 = ENABLE
0 = DISABLE
75 Ω
U1 = AD8013
DISABLE 1
845Ω
Ω
665Ω
Ω
FROM CMOS
+5V
VIN2
1/3 U1
2N3904
+
4k Ω
10kΩ
Ω
VOUT
84 Ω
-
8k Ω
75 Ω
TO
DISABLE
DISABLE 2
665Ω
Ω
-5V
84 Ω
VIN3
75 Ω
845Ω
Ω
1/3 U1
+
-5V
75 Ω
DISABLE 3
a
2.45
Configuring two amplifiers as unity gain followers and using the third to set the
gain results in a high performance 2:1 multiplexer as shown in Figure 2.46. The
circuit takes advantage of the very low crosstalk between the amplifiers and
achieves the OFF channel isolation shown in Figure 2.47. The differential gain and
phase performance of the circuit is 0.03% and 0.07°, respectively.
2:1 VIDEO MULTIPLEXER
2k Ω
+5V
10 Ω
1/3U1
+
VIN1
75 Ω
U1 = AD8013
DISABLE 1
VOUT
+
1/3U1
-
VIN2
-5V
10 Ω
+
1/3U1
75 Ω
-
845Ω
Ω
845Ω
Ω
DISABLE 2
2k Ω
a
2.46
49
2:1 MULTIPLEXER ON-CHANNEL GAIN AND MUX
OFF-CHANNEL FEEDTHROUGH VS. FREQUENCY
2
1
GAIN
-1
-2
-3
-30
-4
-40
-50
-5
FEEDTHROUGH
-6
-60
-7
-70
-8
1M
100M
10M
FEEDTHROUGH - dB
CLOSED-LOOP GAIN - dB
0
-80
1G
FREQUENCY - Hz
a
2.47
VIDEO PROGRAMMABLE GAIN AMPLIFIER USING THE
AD813 TRIPLE CURRENT FEEDBACK OP AMP
Closely related to the multiplexers described above is a programmable gain video
amplifier, or PGA, as shown in Figure 2.48. In the case of the AD813, the individual
op amps are disabled by pulling the disable pin about 2.5V below the positive
supply. This puts the corresponding amplifier in its powered down state. In this
condition, the amplifier's quiescent supply current drops to about 0.5mA, its output
becomes a high impedance, and there is a high level of isolation between the input
and the output. Leaving the disable pin disconnected (floating) will leave the
amplifier operational, in the enabled state. The input impedance of the disable pins
is about 35kΩ in parallel with 5pF. When grounded, about 50µA flows out of a
disable pin when operating on ±5V supplies. The switching threshold is such that
the disable pins can be driven directly from +5V CMOS logic with no level shifting
(as was required in the previous example).
50
PROGRAMMABLE GAIN AMPLIFIER USING
AD813 TRIPLE CFB OP AMP
R1
750Ω
+5V
VIN
VOUT
1
+
75Ω
U2
74HC238
A0
A1
A2
U1
AD813
SELECT 1
Y0
2
+
Y1
E1
E2
E3
OUTPUT TABLE
R2
649Ω
R3, 649Ω
A0
A1
VOUT / VIN
L
L
1
H
L
2, (1 + R2/R3)
L
H
4, (1 + R4/R5)
H
H
0, (OFF)
SELECT 2
Y2
R4
301Ω
R5, 100Ω
+5V
NOTE:
3
DECOUPLING
NOT SHOWN
+
-5V
SELECT 3
a
2.48
With a two-line digital control input, this circuit can be set up to provide 3 different
gain settings. This makes it a useful circuit in various systems which can employ
signal normalization or gain ranging prior to A/D conversion, such as CCD systems,
ultrasound, etc. The gains can be binary related as here, or they can be arbitrary.
An extremely useful feature of the AD813 CFB current feedback amplifier is the fact
that the bandwidth does not reduce as gain is increased. Instead, it stays relatively
constant as gain is raised. Thus more useful bandwidth is available at the higher
programmed gains than would be true for a fixed gain-bandwidth product VFB
amplifier type.
In the circuit, channel 1 of the AD813 is a unity gain channel, channel 2 has a gain
of 2, and channel 3 a gain of 4, while the fourth control state is OFF. As is indicated
by the table, these gains can varied by adjustment of the R2/R3 or R4/R5 ratios. For
the gain range and values shown, the PGA will be able to maintain a 3dB
bandwidth of about 50MHz or more for loading as shown (a high impedance load of
1kΩ or more is assumed). Fine tuning the bandwidth for a given gain setting can be
accomplished by lowering the resistor values at the higher gains, as shown in the
circuit, where for G=1, R1=750Ω, for G=2, R2=649Ω, and for G=4, R4=301Ω.
VIDEO MULTIPLEXERS AND CROSSPOINT SWITCHES
Traditional CMOS switches and multiplexers suffer from several disadvantages at
video frequencies. Their switching time (typically 100ns or so) is not fast enough for
today's applications, and they require external buffering in order to drive typical
video loads. In addition, the small variation of the CMOS switch "on" resistance with
signal level (called Ron modulation) introduces unwanted distortion and degradation
51
in differential gain and phase. Multiplexers based on complementary bipolar
technology offer a better solution at video frequencies.
Functional block diagrams of the AD8170/8174/8180/8182 bipolar video multiplexer
are shown in Figure 2.49. These devices offer a high degree of flexibility and are
ideally suited to video applications, with excellent differential gain and phase
specifications. Switching time for all devices in the family is 10ns to 0.1%. The
AD8170/8174 muxes include an on-chip current feedback op amp output buffer
whose gain can be set externally. Off channel isolation and crosstalk are typically
greater than 80dB for the entire family. Key specifications are shown in Figure 2.50.
AD8170/8174/8180/8182 BIPOLAR VIDEO MULTIPLEXERS
AD8170
8
1
SELECT
AD8180
VOUT
IN0
1
8
SELECT
7
ENABLE
6
OUT
5
-Vs
+1
LOGIC
7
+
2
-Vs
-VIN
GND
2
_
GND
DECODER
3
6
+Vs
IN1
3
+1
IN0
4
IN0
1
GND
2
+1
AD8174
3
GND
4
IN2
5
+Vs
IN1
14
+Vs
_
13
VOUT
+1
12
11
2
10
+1
4
AD8182
+
+1
IN1
5
+1
-VIN
S/D
ENABLE
IN0A
1
GND
2
IN1A
3
+Vs
4
IN1B 5
14 SELECT A
+1
DECODER
+1
13 ENABLE A
12 OUT A
11 -Vs
10 OUT B
+1
LOGIC
-Vs
6
IN3
7
+1
2
9
A0
8
A1
a
GND
6
IN0B
7
DECODER
+1
9
ENABLE B
8
SELECT B
2.49
AD817X AND AD818X MULTIPLEXER KEY SPECIFICATIONS
n
10ns Switching Time
n
Wide Bandwidth (-3dB BW):
u
200MHz (AD817X)
u
600MHz (AD818X)
n
Gain Flatness (0.1dB):
u
80MHz (AD817X)
u
150MHz (AD818X)
n
0.02% / 0.02° Differential Gain and Phase (AD817X, RL = 150Ω
Ω)
52
n
0.02% / 0.03° Differential Gain and Phase (AD818X, RL = 1kΩ
Ω)
n
Off-Channel Isolation and Crosstalk > –80dB @ 10MHz
n
Low Power (±5V Supplies):
u
u
u
u
AD8170 - 65mW
AD8180 - 35mW
AD8174 - 85mW
AD8182 - 70mW
a
2.50
Figure 2.51 shows an application circuit for three AD8170 2:1 muxes where the RGB
monitor can be switched between two computers. The AD8174 4:1 mux is used in
Figure 2.52 to allow a single high speed ADC to digitize the RGB outputs of a
scanner. Figure 2.53 shows two AD8174 4:1 muxes expanded into an 8:1 mux.
DUAL SOURCE RGB MULTIPLEXER USING THREE 2:1 MUXES
CHANNEL
SELECT
COMPUTER
R
G
B
IN0
R
IN1
IN0
G
MONITOR
IN1
IN0
B
IN1
R
G
B
THREE AD8170 2:1 MUXES
COMPUTER
a
2.51
53
DIGITIZING RGB SIGNALS WITH ONE ADC AND A 4:1 MUX
AD8174
R
SCANNER
IN0
G
4:1
MUX
IN1
ADC
A1
B
IN2
A0
IN3
CHANNEL SELECT
a
2.52
EXPANDING TWO 4:1 MUXES INTO AN 8:1 MUX
CHANNEL SELECT
TWO AD8174
4:1 MUXES
IN0
A0
IN1
IN2
A1
EN
VIDEO
OUTPUT
IN3
VIDEO
INPUTS
IN0
IN1
IN2
A0
A1
EN
IN3
a
2.53
The AD8116 extends the concepts above to yield a 16×16 buffered video crosspoint
switch matrix (Figure 2.54). The 3dB bandwidth is greater than 200MHz, and the
0.1dB gain flatness extends to greater than 40MHz. Channel switching time is less
than 30ns to 0.1%. Crosstalk is 70dB and isolation is 90dB (both measured at
10MHz). Differential gain and phase is 0.01% and 0.01° for a 150Ω load. Total
power dissipation is 900mW on ±5V supplies.
54
The AD8116 includes output buffers which can be put into a high impedance state
for paralleling crosspoint stages so that the off channels do not load the output bus.
The channel switching is performed via a serial digital control which can
accommodate "daisy chaining" of several devices. The AD8116 is packaged in a 128pin TQFP package. Key specifications for the device are summarized in Figure 2.55.
AD8116 16×16 BUFFERED VIDEO CROSSPOINT SWITCH
AD8116
SERIAL
CLOCK
SERIAL
CLOCK
SERIAL
DATA IN
LATCH
SERIAL
DATA OUT
80 Bit SHIFT REG.
LATCH
80
C ENABLE
PARALLEL LATCH
80
DECODE
16x5:16 Decoders
256
SE T
IN DI VDU AL
OR
RESET ALL OUTPUTS
TO "OFF"
RESET
C ENABLE
16
OUTPUT
BUFFER
RESET
+1
+1
+1
+1
+1
+1
SWITCH
MATRIX
+1
+1
+1
+1
+1
ENABLE/DISABLE
+1
16 ANALOG
INPUTS
16 ANALOG
OUTPUTS
+1
+1
+1
+1
a
2.54
55
AD8116 CROSSPOINT SWITCH KEY SPECIFICATIONS
n
16×16 Buffered Inputs and Outputs
n
Output Buffer Disable Feature Allows Expansion
n
3dB Bandwidth 200MHz, 0.1dB Bandwidth 40MHz
n
30ns Switching to 0.1%
n
Differential Gain 0.01%, Differential Phase 0.01°
n
Power Dissipation: 900mW (±5V Supplies)
n
128-pin TQFP, 0.36 Square Inches Area
a
2.55
HIGH POWER LINE DRIVERS AND ADSL
ADSL (Asymmetric Digital Subscriber Line) uses the current subscriber line
connection to the central office to transmit data as high as 8Mbps, almost 300 times
the speed of the fastest traditional modem. ADSL uses the entire bandwidth
(approximately 1MHz) of the connection in addition for the modulation scheme
called Discrete Multi Tone (DMT).
Although high-speed fiber links already exist, it is still too difficult and expensive to
bring them directly to every residence. ADSL uses the existing infrastructure for
"the last mile" connecting the home and the local central office (which already has a
high-speed fiber link to the national network).
Many applications are uneven (asymmetric) in their bandwidth needs - sending
more information in one direction than the other. Typically, a user will request a
video channel, ask for information from a central database, or view complex
graphical images on a web page. All of these applications require considerable
bandwidth. In contrast, the user may only send commands or files back up to the
server. Realizing this, ADSL was designed to deliver a bigger downstream capacity
to the home, while having a smaller two-way capacity.
Key to the ADSL system is the requirement for a low-distortion differential drive
amplifier which delivers approximately 40V p-p into a 60Ω differential load
impedance. The AD815 dual high current driver can deliver 40V p-p differential into
a 50Ω load (corresponding to 400mA peak current!) using the application circuit
shown in Figure 2.56. Low harmonic distortion is also required for ADSL
applications, since it affects system bit error rates. The typical distortion of the
device is shown in Figure 2.57 for 50Ω and 200Ω differential loads.
56
ADSL DIFFERENTIAL LINE DRIVER USING THE AD815
+15V
1/2
AD815
100Ω
R1 = 15Ω
AMP1
499Ω
VIN =
4Vp-p
125Ω
RL
120Ω
VD =
40Vp-p
G = +10
VOUT =
40Vp-p
499Ω
R2 = 15Ω
100Ω
AMP2
1/2
AD815
1:2
TRANSFORMER
-15V
2.56
a
THD VS. FREQUENCY FOR AD815 DIFFERENTIAL DRIVER
TOTAL HARMONIC DISTORTION - dBc
-40
-50
VS = ±15V
G = +10
VOUT = 40V p-p
-60
-70
-80
RL = 50Ω
(DIFFERENTIAL)
RL = 200Ω
(DIFFERENTIAL)
-90
-100
-110
100
a
1k
10k
100k
FREQUENCY - Hz
1M
10M
2.57
There are three AD815 models, two are available in a 15-pin power package, and
the third as a 24-pin thermally enhanced SOIC. The 15-pin power package
(AD815AY-through hole and AD815AVR-surface mount) has a low thermal
resistance (θJA = 41°C/W) which can be reduced considerably ( to θJA = 16 °C/W) by
connecting the package to an area of copper which acts as a heat sink. The AD815
incorporates a thermal shutdown circuit to protect the die from thermal overload.
57
The AD815 also has applications as a general purpose high current coil,
transformer, or twisted pair cable driver, a CRT convergence adjustment control, or
a video signal distribution amplifier. Each amplifier in the AD815 is capable of
driving 6 back-terminated 75Ω video loads with a differential gain and phase of
0.05% and 0.45° respectively.
HIGH SPEED PHOTODIODE PREAMPS
Photodiodes generate a small current which is proportional to the level of
illumination. They have many applications ranging from precision light meters to
high-speed fiber optic receivers.
The equivalent circuit for a photodiode is shown in Figure 2.58. One of the standard
methods for specifying the sensitivity of a photodiode is to state its short circuit
photocurrent (Isc) at a given light level from a well defined light source. The most
commonly used source is an incandescent tungsten lamp running at a color
temperature of 2850K. At 100fc (foot-candles) illumination (approximately the light
level on an overcast day), the short circuit current is usually in the picoamps to
hundreds of microamps range for small area (less than 1mm2) diodes.
PHOTODIODE EQUIVALENT CIRCUIT
RS << RSH
INCIDENT
LIGHT
RSH(T)
IDEAL
DIODE
PHOTO
CURRENT
CJ
≈100kΩ TO
100GΩ
2.58
a
The short circuit current is very linear over 6 to 9 decades of light intensity, and is
therefore often used as a measure of absolute light levels. The open circuit forward
voltage drop across the photodiode varies logarithmically with light level, but,
because of its large temperature coefficient, the diode voltage is seldom used as an
accurate measure of light intensity.
The shunt resistance is usually in the order of several hundred kΩ to more than
1GΩ at room temperature, and decreases by a factor of two for every 10°C rise in
58
temperature. Diode capacitance is a function of junction area and the diode bias
voltage. A value of 10 to 50pF at zero bias is typical for small area diodes.
Photodiodes may either be operated with zero bias (photovoltaic mode) or reverse
bias (photoconductive mode) as shown in Figure 2.59. The most precise linear
operation is obtained in the photovoltaic mode, while higher switching speeds are
realizable when the diode is operated in the photoconductive mode. Under reverse
bias conditions, a small amount of current called dark current will flow even when
there is no illumination. There is no dark current in the photovoltaic mode. In the
photovoltaic mode, the diode noise is basically the thermal noise generated by the
shunt resistance. In the photoconductive mode, shot noise due to conduction is an
additional source of noise. Photodiodes are usually optimized during the design
process for use in either the photovoltaic mode or the photoconductive mode, but not
both.
PHOTODIODE MODES OF OPERATION
-VBIAS
PHOTOCONDUCTIVE
PHOTOVOLTAIC
· Zero Bias
· No Dark Current
· Precision Applications
· Low Noise (Johnson)
· Reverse Bias
· Dark Current Exists
· High Speed Applications
· Higher Noise (Johnson + Shot))
a
2.59
Optimizing photodiode preamplifiers is probably one of the most challenging of
design problems, especially if high bandwidth and direct coupling is required. Figure
2.60 shows a basic photodiode preamp designed with an op amp connected as a
current-to-voltage converter.
59
HIGH BANDWIDTH PHOTODIODE PREAMP
EQUIVALENT CIRCUIT
ID = IS + IDARK
C2
C1 = CD + CIN
RSH >>R2
R2
VN
IB
VOUT = R2ID
fu
RSH
ID
CD
CIN
fu = OP AMP GBW PRODUCT
-VBIAS
FOR R2 = 100kΩ
Ω
SIGNAL BW =
ID = 100µ
µA
1
2πR2C2
VOUT = 10V
a
2.60
The sensitivity of the circuit is determined by the amount of photodiode current
multiplied by the feedback resistor R2. The key parameters of the diode (see Figure
2.61) are its sensitivity (output current as a function of illumination level), dark
current (the amount of current which flows due to the reverse bias voltage when the
diode is not illuminated), risetime, shunt capacitance, and shunt resistance.
The key parameters of the op amp are its input voltage and current noise, bias
current, unity gain-bandwidth product, fu, and input capacitance, Cin.
The HP 5082-4204 PIN Photodiode will be used as an example for our discussion. Its
characteristics are given in Figure 2.61. It is typical of many commercially available
PIN photodiodes. As in most high-speed photodiode applications, the diode is
operated in the reverse-biased or photoconductive mode. This greatly lowers the
diode junction capacitance, but causes a small amount of dark current to flow even
when the diode is not illuminated (we will show a circuit which compensates for the
dark current error later in the section).
HP 5082-4204 PHOTODIODE
n
Sensitivity: 350µA @ 1mW, 900nm
n
Maximum Linear Output Current: 100µA
n
Area: 0.002cm2 (0.2mm2)
n
Capacitance: 4pF @ 10V reverse bias
60
n
Shunt Resistance: 1011Ω
n
Risetime: 10ns
n
Dark Current: 600pA @ 10V reverse bias
a
2.61
This photodiode is linear with illumination up to approximately 50 to 100µA of
output current. The dynamic range is limited by the total circuit noise and the diode
dark current (assuming no dark current compensation).
Using the simple circuit shown in Figure 2.60, assume that we wish to have a full
scale output of 10V for a diode current of 100µA. This determines the value of the
feedback resistor R2 to be 10V/100µA = 100kΩ.
Analysis of Frequency Response and Stability
The photodiode preamp model is the classical second-order system shown in Figure
2.62, where the I/V converter has a total input capacitance C1 (the sum of the diode
capacitance and the op amp input capacitance). The shunt resistance of the
photodiode is neglected since it is much greater than R2, the feedback resistor.
COMPENSATING FOR INPUT CAPACITANCE IN A
CURRENT-TO-VOLTAGE CONVERTER USING VFB OP AMP
C2
R2
-
i
C1
VFB
+
fp =
1
2π
πR2C1
UNCOMPENSATED
fx =
1
2π
π R2C2
COMPENSATED
fx =
fp • fu
|A(s)|
fx
C1
f
1
fp
C2 =
2π
π R2 • f u
FOR 45º PHASE MARGIN
fu
a
2.62
The net input capacitance, C1, forms a pole at a frequency fp in the noise gain
transfer function as shown in the Bode plot.
61
fp =
1
.
2πR2C1
Note that we are neglecting the effects of the compensation capacitor C2 and are
assuming that it is small relative to C1 and will not significantly affect the pole
frequency fp when it is added to the circuit. In most cases, this approximation yields
results which are close enough, considering the other variables in the circuit.
If left uncompensated, the phase shift at the frequency of intersection, fx, will cause
instability and oscillation. Introducing a zero at fx by adding the feedback capacitor
C2 stabilizes the circuit and yields a phase margin of about 45 degrees.
fx =
1
2 πR2C2
Since fx is the geometric mean of fp and the unity-gain bandwidth frequency of the
op amp, fu,
fx =
fp ⋅ fu .
These equations can be solved for C2:
C2 =
C1
.
2πR2 ⋅ f u
This value of C2 will yield a phase margin of about 45 degrees. Increasing the
capacitor by a factor of 2 increases the phase margin to about 65 degrees (see
References 4 and 5).
In practice, the optimum value of C2 should be optimized experimentally by varying
it slightly to optimize the output pulse response.
Selection of the Op Amp
The photodiode preamp should be a wideband FET-input one in order to minimize
the effects of input bias current and allow low values of photocurrents to be detected.
In addition, if the equation for the 3dB bandwidth, fx, is rearranged in terms of fu,
R2, and C1, then
fx =
fu
,
2πR2C1
where C1 = CD + Cin
By inspection of this equation, it is clear that in order to maximize fx, the FET-input
op amp should have both a high unity gain-bandwidth product, fu, and a low input
capacitance, Cin. In fact, the ratio of fu to Cin is a good figure-of-merit when
evaluating different op amps for this application. Figure 2.63 compares a number of
FET-input op amps suitable for photodiode preamps.
62
FET-INPUT OP AMP COMPARISON TABLE
FOR WIDE BANDWIDTH PHOTODIODE PREAMPS
Unity
GBW
Product,
fu (MHz)
Input
Capacitanc
eCin (pF)
fu/Cin
(MHz/pF)
Input Bias
Current
Ib (pA)
Voltage Noise
@10kHz
(nV/√
√ Hz)
AD823
16
1.8
8.9
3
16
AD843
34
6
5.7
600
19
AD744
13
5.5
2.4
100
16
AD845
16
8
2
500
18
AD745*
20
20
1
250
2.9
AD645
1
1
1
1.5
8
AD820
1.9
2.8
0.7
2
13
AD743
4.5
20
0.2
250
2.9
* Stable for Noise Gains > 5, Usually the Case, Since
High Frequency Noise Gain = 1 + C1/C2, and C1Usually > 4C2.
a
2.63
By inspection, the AD823 op amp has the highest ratio of unity gain-bandwidth
product to input capacitance, in addition to relatively low input bias current. For
these reasons, it was chosen for the wideband photodiode preamp design.
Using the diode capacitance, CD=4pF, and the AD823 input capacitance, Cin=1.8pF,
the value of C1 = CD+Cin = 5.8pF. Solving the above equations using C1=5.8pF,
R2=100kΩ, and fu=16MHz, we find that:
fp
C2
fx
=
=
=
274kHz
0.76pF
2.1MHz.
In the final design (Figure 2.64), note that the 100kΩ resistor is replaced with three
33.2kΩ film resistors to minimize stray capacitance. The feedback capacitor, C2, is a
variable 1.5pF ceramic and is adjusted in the final circuit for best bandwidth/pulse
response. The overall circuit bandwidth is approximately 2MHz.
The full scale output voltage of the preamp for 100µA diode current is 10V, and the
error (RTO) due to the photodiode dark current of 600pA is 60mV. The dark current
63
error can be canceled using a second photodiode of the same type in the noninverting input of the op amp as shown in Figure 2.64.
2MHz BANDWIDTH PHOTODIODE PREAMP
WITH DARK CURRENT COMPENSATION
≈ 0.8pF
C2
C1 = CD + CIN
CD = 4pF, CIN = 1.8pF
Ω
Ω 33.2kΩ
33.2kΩ
Ω 33.2kΩ
R2 = 100kΩ
Ω
+15V
-10V
D1
-
C1
5.8pF
AD823
D2
+
D1, D2: HP-5082-4204
-15V
0.1µ
µF
LOW LEAKAGE
POLYSTYRENE
Ω
100kΩ
a
2.64
Photodiode Preamp Noise Analysis
As in most noise analyses, only the key contributors need be identified. Because the
noise sources combine in an RSS manner, any single noise source that is at least
three or four times as large as any of the others will dominate.
In the case of the wideband photodiode preamp, the dominant sources of output
noise are the input voltage noise of the op amp, Vni, and the resistor noise due to
R2, VnR2. The input current noise of the FET-input op amp is negligible. The shot
noise of the photodiode (caused by the reverse bias) is negligible because of the
filtering effect of the shunt capacitance C1. The resistor noise is easily calculated by
knowing that a 1kΩ resistor generates about 4nV/√Hz, therefore, a 100kΩ resistor
generates 40nV/√Hz. The bandwidth for integration is the signal bandwidth,
2.1MHz, yielding a total output rms noise of:
V nR2(OUT) = 40 157
. ⋅ 21
. ⋅ 10 6 = 73µVrms .
The factor of 1.57 converts the approximate single-pole bandwidth of 2.1MHz into
the equivalent noise bandwidth.
The output noise due to the input voltage noise is obtained by multiplying the noise
gain by the voltage noise and integrating the entire function over frequency. This
would be tedious if done rigorously, but a few reasonable approximations can be
made which greatly simplify the math. Obviously, the low frequency 1/f noise can be
64
neglected in the case of the wideband circuit. The primary source of output noise is
due to the high-frequency noise-gain peaking which occurs between fp and fu. If we
simply assume that the output noise is constant over the entire range of frequencies
and use the maximum value for AC noise gain [1+(C1/C2)], then
C1 

V ni(OUT) ≈ V ni 1 +
. f x = 250µVrms .
 157

C2 
The total rms noise referred to the output is then the RSS value of the two
components:
V n( TOTAL) =
(73)2 + (250) 2
= 260µVrms .
The total output dynamic range can be calculated by dividing the full scale output
signal (10V) by the total output rms noise, 260µVrms, and converting to dB, yielding
approximately 92dB.
EQUIVALENT CIRCUIT FOR OUTPUT NOISE ANALYSIS
C2
R2
Vni
Vni = 16nV/√
√Hz
C1 = 5.8pF
C2 = 0.76pF
R2 = 100kΩ
Ω
VnR2
-
C1
AD823
Vn (TOTAL)
+
1+
C1
C2
1
C1
Vni(out) ≈ Vni ( 1+
C2
NOISE GAIN
)
1.57fx
fp
fx
fu
274kHz
2.1MHz
16MHz
= 250µ
µV rms
VnR2(out) ≈ 4kTR2 • 1.57f x
= 73µ
µV rms
Vn(TOTAL) = 2502 + 732
= 260µ
µV rms
DYNAMIC RANGE = 20 log
10V
260µ
µV
= 92dB
2.65
a
65
REFERENCES
1.
Walt Kester, Maintaining Transmission Line Impedances on the
PC Board, within Chapter 11 of System Applications Guide,
Analog Devices, 1993.
2.
Joe Buxton, Careful Design Tames High-Speed Op Amps,
Electronic Design, April 11, 1991.
3.
Walt Jung, Op Amps in Line-Driver and Receiver Circuits, Part 1,
Analog Dialogue, Vol. 26-2, 1992.
4.
William R. Blood, Jr., MECL System Design Handbook
(HB205, Rev.1), Motorola Semiconductor Products, Inc., 1988.
5.
Dave Whitney, Walt Jung, Applying a High-Performance Video
Operational Amplifier, Analog Dialogue, 26-1, 1992.
6.
Ohmtek, Niagara Falls, NY, (716) 283-4025.
7.
Walt Kester, Video Line Receiver Applications Using the AD830
Active Feedback Amplifier Topology, within Chapter 11 of System
Applications Guide, Analog Devices, 1993.
8.
Walt Jung, Analog-Signal-Processing Concepts Get More Efficient,
Electronic Design Analog Applications Issue, June 24, 1993.
9.
Peter Checkovich, Understanding and Using High-Speed Clamping
Amplifiers, Analog Dialogue, Vol. 29-1, 1995.
10.
Walt Jung, Scott Wurcer, Design Video Circuits Using High-Speed
Op-Amp Systems, Electronic Design Analog Applications Issue,
November 7, 1994.
11.
W. A. Kester, PCM Signal Codecs for Video Applications,
SMPTE Journal, No. 88, November 1979, pp. 770-778.
12.
IEEE Standard for Performance Measurements of A/D and D/A
Converters for PCM Television Circuits, IEEE Standard 746-1984.
13.
Practical Analog Design Techniques, Chapters 1, 2, and 4,
1995, Analog Devices.
14.
Amplifier Applications Guide, 1992, Analog Devices.
15.
Jerald G. Graeme, Photodiode Amplifiers: Op Amp
Solutions,, McGraw Hill, 1995.
66
SECTION 3
RF/IF SUBSYSTEMS
Walt Kester, James Bryant, Bob Clarke,
Barrie Gilbert
DYNAMIC RANGE COMPRESSION
In many cases, a wide dynamic range is an essential aspect of a signal, something to
be preserved at all costs. This is true, for example, in the high-quality reproduction
of music and in communications systems. However, it is often necessary to compress
the signal to a smaller range without any significant loss of information.
Compression is often used in magnetic recording, where the upper end of the
dynamic range is limited by tape saturation, and the lower end by the granularity of
the medium. In professional noise-reduction systems, compression is "undone" by
precisely-matched nonlinear expansion during reproduction. Similar techniques are
often used in conveying speech over noisy channels, where the performance is more
likely to be measured in terms of word-intelligibility than audio fidelity. The
reciprocal processes of compressing and expanding are implemented using
"compandors", and many schemes have been devised to achieve this function.
There is a class of linear dynamic range compression systems where the gain of the
amplifiers in the signal processing chain is independent of the instantaneous
amplitude of the signal, but is controlled by a closed loop system in such a way as to
render the output (that is the peak, or rms value) essentially constant. The
harmonic distortion is relatively low. These systems use what are often called
variable-gain amplifiers. While correct, this lacks precision, because nonlinear
amplifiers (such as log amps) also exhibit variable gain, but in direct response to the
signal magnitude. The term voltage controlled amplifier (VCA) is preferred in this
context; it clearly describes the way in which the gain control is implemented, while
allowing latitude in regard to the actual circuit means used to achieve the function.
The gain may be controlled by a current within the circuit, but usually a voltage.
Analog multipliers may be used as VCAs, but there are other topologies which will
be discussed later in this section.
Logarithmic amps find applications where signals having wide dynamic ranges
(perhaps greater than 100dB) must be processed by elements, such as ADCs, which
may have limited dynamic ranges. Log amps have maximum incremental gain for
small signals; the gain decreases in inverse proportion to the magnitude of the
input. This permits the amplifier to accept signals with a wide input dynamic range
and compress them substantially.
Log amps provide nonlinear dynamic range compression and are used in
applications where low harmonic distortion is not a requirement. All types of log
amps produce a low dynamic range output without the need to first acquire some
measure of the signal amplitude for use in controlling gain.
1
We will first examine linear compression techniques using voltage-controlled
amplifiers within automatic-gain-control (AGC) loops. Nonlinear signal compression
using log amps is then discussed.
Both AGC loops using VCAs and log amps make excellent building blocks for highly
integrated RF/IF subsystems for signal processing in communications systems as
will be demonstrated.
RF / IF SUBSYSTEM BUILDING BLOCKS
n
n
Signal Dynamic Range Compression Techniques
u
Linear: Automatic Gain Control Loop (AGC) using
Voltage Controlled Amplifier (VCA) and Detector
u
Non-Linear: Demodulating / Limiting Logarithmic
Amplifiers
Modulation / Demodulation: In-Phase and Quadrature (I/Q)
and Polar (Amplitude and Phase)
u
Dynamic Range Compression Required
u
IF Subsystems: AGC, Log / Limiting, RSSI, Mixers
a
3.1
AUTOMATIC GAIN CONTROL (AGC) AND VOLTAGECONTROLLED AMPLIFIERS (VCAS)
In radio systems, the received energy exhibits a large dynamic range due to the
variability of the propagation path, requiring dynamic-range compression in the
receiver. In this case, the wanted information is in the modulation envelope
(whatever the modulation mode), not in the absolute magnitude of the carrier. For
example, a 1MHz carrier modulated at 1kHz to a 30% modulation depth would
convey the same information, whether the received carrier level is at 0dBm or –
120dBm. Some type of automatic gain control (AGC) in the receiver is generally
utilized to restore the carrier amplitude to some normalized reference level, in the
presence of large input fluctuations. AGC circuits are dynamic-range compressors
which respond to some metric of the signal – often its mean amplitude – acquired
over an interval corresponding to many periods of the carrier. Consequently, they
require time to adjust to variations in received signal level. The time required to
respond to a sudden increase in signal level can be reduced by using peak detection
methods, but with some loss of robustness, since transient noise peaks can now
activate the AGC detection circuits. Nonlinear filtering and the concept of "delayed
AGC" can be useful in optimizing an AGC system. Many tradeoffs are found in
practice; Figure 3.2 shows a basic system.
2
A TYPICAL AUTOMATIC GAIN CONTROL (AGC) SYSTEM
VOLTAGE CONTROLLED AMP
VXsinω
ωt
VRsinω
ωt
VCA
INPUT: UNKNOWN
AMPLITUDE
OUTPUT: FIXED
AMPLITUDE
CONTROL
VOLTAGE
MEASURES
SIGNAL
LEVEL
RECTIFIER
DETECTOR
RMS/DC CONVERTER
PEAK DETECTOR
DIFFERENCE
AMP
LPF
+
VREF
a
3.2
It is interesting to note that an AGC loop actually has two outputs. The obvious
output is the amplitude-stabilized signal. The less obvious output is the control
voltage to the VCA, which is in reality, a measure of the average amplitude of the
input signal. If the system is precisely scaled, the control voltage may be used as a
measure of the input signal, sometimes referred to as a received signal strength
indicator (RSSI).
VOLTAGE CONTROLLED AMPLIFIERS (VCAS)
An analog multiplier can be used as a variable-gain amplifier as shown in Figure
3.3. The control voltage is applied to one input, and the signal to the other. In this
configuration, the gain is directly proportional to the control voltage.
3
USING A MULTIPLIER AS A
VOLTAGE-CONTROLLED AMPLIFIER (VCA)
VIN
+
VC
VIN
K
•
R2
R1
VXsinω
ωt
VO =
VO
-
CONTROL
INPUT
R
(1 + R21 )VC
a
3.3
Most VCAs made with analog multipliers have gain which is linear in volts with
respect to the control voltage, and they tend to be noisy. There is a demand,
however, for a VCA which combines a wide gain range with constant bandwidth and
phase, low noise with large signal-handling capabilities, and low distortion with low
power consumption, while providing accurate, stable, linear-in-dB gain. The AD600,
AD602, and AD603 achieve these demanding and conflicting objectives with a
unique and elegant solution - the X-AMP™ (for exponential amplifier). The concept
is simple: a fixed-gain amplifier follows a passive, broadband attenuator equipped
with special means to alter its attenuation under the control of a voltage (see Figure
3.4). The amplifier is optimized for low input noise, and negative feedback is used to
accurately define its moderately high gain (about 30 to 40dB) and minimize
distortion. Since this amplifier's gain is fixed, so also are its ac and transient
response characteristics, including distortion and group delay; since its gain is high,
its input is never driven beyond a few millivolts. Therefore, it is always operating
within its small signal response range.
4
SINGLE CHANNEL OF THE DUAL 30MHz AD600/AD602 X-AMP
GAT1
PRECISION
PASSIVE
INPUT
ATTENUATOR
SCALING
REFERENCE
C1HI
C1LO
GATING
INTERFACE
+
-
VG +
GAIN CONTROL
INTERFACE
0dB
A1OP
RF2
2.24kΩ (AD600)
694Ω (AD602)
-6.02dB -12.04dB
-18.06dB
-22.08dB
-33.1dB -36.12dB -42.14dB
A1HI
500Ω
62.5Ω
A1LO
R - 2R LADDER NETWORK
(RO = 100Ω ± 2%)
A1CM
FIXED GAIN
AMPLIFIER
41.07dB (AD600)
31.07dB (AD602)
a
3.4
The attenuator is a 7-section (8-tap) R-2R ladder network. The voltage ratio between
all adjacent taps is exactly 2, or 6.02dB. This provides the basis for the precise
linear-in-dB behavior. The overall attenuation is 42.14dB. As will be shown, the
amplifier's input can be connected to any one of these taps, or even interpolated
between them, with only a small deviation error of about ±0.2dB. The overall gain
can be varied all the way from the fixed (maximum) gain to a value 42.14dB less.
For example, in the AD600, the fixed gain is 41.07dB (a voltage gain of 113); using
this choice, the full gain range is –1.07dB to +41.07dB. The gain is related to the
control voltage by the relationship GdB = 32VG + 20 where VG is in volts. For the
AD602, the fixed gain is 31.07dB (a voltage gain of 35.8), and the gain is given by
GdB = 32VG + 10.
The gain at VG = 0 is laser trimmed to an absolute accuracy of ±0.2dB. The gain
scaling is determined by an on-chip bandgap reference (shared by both channels),
laser trimmed for high accuracy and low temperature coefficient. Figure 3.5 shows
the gain versus the differential control voltage for both the AD600 and the AD602.
5
GAIN OF THE AD600/AD602
AS A FUNCTION OF CONTROL VOLTAGE
45
40
35
30
AD600
25
GAIN - dB
20
15
AD602
10
5
0
-5
-10
-15
-800
-400
0
400
800
VG - Millivolts
a
3.5
In order to understand the operation of the X-AMP, consider the simplified diagram
shown in Figure 3.6. Notice that each of the eight taps is connected to an input of
one of eight bipolar differential pairs, used as current-controlled transconductance
(gm) stages; the other input of all these gm stages is connected to the amplifier's
gain-determining feedback network, RF1/RF2. When the emitter bias current, IE, is
directed to one of the 8 transistor pairs (by means not shown here), it becomes the
input stage for the complete amplifier.
CONTINUOUS INTERPOLATION BETWEEN
TAPS IN THE X-AMP IS PERFORMCE WITH CURRENTCONTROLLED gm STAGES
+
+
OUTPUT
(A1OP)
AOL → ∞
IE1
R
R
(A1HI)
RO=100Ω
IE2
2R
R
2R
IE5
IE4
IE3
R
2R
IE7
R
R
2R
IE6
2R
IE8
RF2
R
2R
R
IE
RF1
(A1LO)
(R = 50Ω )
a
3.6
6
When IE is connected to the pair on the left-hand side, the signal input is connected
directly to the amplifier, giving the maximum gain. The distortion is very low, even
at high frequencies, due to the careful open-loop design, aided by the negative
feedback. If IE were now to be abruptly switched to the second pair, the overall gain
would drop by exactly 6.02dB, and the distortion would remain low, because only
one gm stage remains active.
In reality, the bias current is gradually transferred from the first pair to the second.
When IE is equally divided between two gm stages, both are active, and the
situation arises where we have an op amp with two input stages fighting for control
of the loop, one getting the full signal, and the other getting a signal exactly half as
large.
Analysis shows that the effective gain is reduced, not by 3dB, as one might first
expect, but rather by 20log1.5, or 3.52dB. This error, when divided equally over the
whole range, would amount to a gain ripple of ±0.25dB; however, the interpolation
circuit actually generates a Gaussian distribution of bias currents, and a significant
fraction of IE always flows in adjacent stages. This smoothes the gain function and
actually lowers the ripple (see Reference 12). As IE moves further to the right, the
overall gain progressively drops.
The total input-referred noise of the X-AMP™ is 1.4nV/√Hz; only slightly more than
the thermal noise of a 100Ω resistor which is 1.29nV/√Hz at 25°C. The inputreferred noise is constant regardless of the attenuator setting, therefore the output
noise is always constant and independent of gain. For the AD600, the amplifier gain
is 113 and the output noise spectral density is therefore 1.4nV/√Hz×113, or 158nV/√
Hz. Referred to its maximum output of 2V rms, the signal-to-noise ratio would be
82dB in a 1MHz bandwidth. The corresponding signal-to-noise ratio of the AD602 is
10dB greater, or 92dB. Key features of the AD600/AD602 are summarized in Figure
3.7
KEY FEATURES OF THE AD600/AD602 X-AMPS
n
Precise Decibel-Scaled Gain Control
n
Accurate Absolute Gain Calibration
n
Low Input-Referred Noise (1.4nV/√
√ Hz)
n
Constant Bandwidth (dc to 35MHz)
n
Low Distortion: –60dBc THD at ±1V Output
n
Stable Group Delay (±2ns Over Gain Range)
n
Response Time: Less than 1µs for 40dB Gain Change
n
Low Power (125mW per channel maximum)
n
Differential Control Inputs
7
a
3.7
The AD603 X-AMP is a single version of the AD600/AD602 which provides 90MHz
bandwidth. There are two pin-programmable gain ranges: –11dB to +31dB with
90MHz bandwidth, and +9dB to +51dB with 9MHz bandwidth. Key specifications
for the AD603 are summarized in Figure 3.8.
KEY FEATURES OF THE AD603 X-AMP
n
Precise "Linear in dB" Gain Control
n
Pin Programmable Gain Ranges:
–11dB to +31dB with 90MHz Bandwidth
+9dB to + 51dB with 9MHz Bandwidth
n
Bandwidth Independent of Variable Gain
n
Low Input-Referred Noise (1.3nV/√
√ Hz)
n
±0.5dB Typical Gain Accuracy
n
Low Distortion:
n
Low Power (125mW)
n
8-pin Plastic SOIC or Ceramic DIP
–60dBc, 1V rms Output @ 10MHz
a
3.8
AN 80 dB RMS-LINEAR-dB MEASUREMENT SYSTEM
RMS/DC converters provide a means to measure the rms value of an arbitrary
waveform. They also may provide a low-accuracy logarithmic ("decibel-scaled")
output. However, they have a fairly small dynamic range – typically only 50dB.
More troublesome is that the bandwidth is roughly proportional to the signal level;
for example, the AD636 provides a 3dB bandwidth of 900kHz for an input of 100mV
rms, but only a 100kHz bandwidth for an input of 10mV rms. Its "raw" logarithmic
output is unbuffered, uncalibrated, and not stable over temperature, requiring
considerable support circuitry, including at least two adjustments and a special
high-TC resistor.
All of these problems can be eliminated using an RMS/DC converter (i.e.,AD636)
merely as the detector element in an AGC loop, in which the difference between the
rms output of the AD636 and a fixed DC reference is nulled in a loop integrator. The
8
dynamic range and the accuracy with which the signal can be determined are now
entirely dependent on the amplifier used in the AGC system. Since the input to the
RMS/DC converter is forced to a constant amplitude, close to its maximum input
capability, the bandwidth is no longer signal-dependent. If the amplifier has a
precise exponential ("linear-dB") gain-control law, its control voltage is forced by the
AGC loop to have the general form
VLOG = VS log10
VIN( RMS)
VZ
where VS is the logarithmic slope and VZ is the logarithmic intercept, that is, the
value of VIN for which VLOG is zero.
Figure 3.9 shows a practical wide-dynamic-range rms measurement system using
the AD600. It can handle inputs from 100µV to 1V rms (4 decades) with a constant
measurement bandwidth of 20Hz to 2MHz, limited primarily by the AD636 RMS/DC
converter. Its logarithmic output is a buffered voltage, accurately-calibrated to
100mV/dB, or 2V per decade, which simplifies the interpretation of the reading
when using a DVM, and is arranged to be –4V for an input of 100µV rms input, zero
for 10mV, and +4V for a 1V rms input. In terms of the above equation, VS is 2V and
VZ is 10mV.
A COMPLETE 80dB RMS-LINEAR-dB MEASUREMENT SYSTEM
VRMS
CAL 0dB
INPUT
1V RMS
MAX
(SINE WAVE) R1
AF/RF
OUTPUT
C1
0.1µF
C4
4.7µF
C1LO
16
A1HI
2
R2 200Ω
A1LO
3
GAT1
+
-
A1
+
U3A
-
5
6
7
C2LO
12
+
11
A2
10
8
9
-6V DEC
VPOS
VNEG
A2OP
+6V DEC
C2
-6V DEC 2µF
NC
NC
C2HI
0.1µF
14
+6V DEC
13
NC
VNEG
12
NC
4
CAVG
11
NC
5
VLOG
6
BFOP
LDLO
9
7
BFIN
VRMS
8
COMM
-6V DEC
R6
3.16kΩ
10
+U3B -
1/2
AD712
R5
16.2kΩ
VG
15.625mV/dB
VPOS
U2
AD636
3
A2CM
U1 AD600
1/2
AD712
2
VINP
A1OP
REF
A2HI
NC
14
13
GAT2
A2LO
1
A1CM
15
4
R3
133kΩ
FB
C1HI
1
115Ω
+6V
+6V DEC
+
R7
56.2kΩ
0.1µF
FB
-6V
POWER SUPPLY
DECOUPLINGNETWORK
+316.2mV
C3
1µF
VOUT
R4
3.01kΩ
+100mV/dB
0V = 0dB (AT 10mV RMS)
NC = NOCONNECT
a
3.9
Note that the peak "log-output" of ±4V requires the use of ±6V supplies for the dual
op-amp U3 (AD712), although lower supplies would suffice for the AD600 and
AD636. If only ±5V supplies are available, it will either be necessary to use a
reduced value for VS (say, 1V, in which case the peak output would be only ±2V), or
to restrict the dynamic range of the signal to about 60dB.
9
The two amplifiers of the AD600 are used in cascade. The modest bandwidth of the
unity-gain buffer U3A acts as a low pass filter, thus eliminating the risk of
instability at the highest gains. The buffer also allows the use of a high-impedance
coupling network (C1/R3) which introduces a high-pass corner at about 12Hz. An
input attenuator of 10dB (× 0.316) is now provided by R1 + R2 operating in
conjunction with the AD600's input resistance of 100Ω. The adjustment provides
exact calibration of VZ in critical applications, but R1 and R2 may be replaced by a
fixed resistor of 215Ω if very close calibration is not needed, since the input
resistance of the AD600 (and all the other key parameters of it and the AD636) are
already laser-trimmed for accurate operation. This attenuator allows inputs as large
as ±4V to be accepted, that is, signals with an rms value of 1V combined with a
crest-factor of up to 4.
The output of A2 is AC-coupled via another 12Hz high-pass filter formed by C2 and
the 6.7kΩ input resistance of the AD636. The averaging time-constant for the
RMS/DC converter is determined by C4. The unbuffered output of the AD636 (at pin
8) is compared with a fixed voltage of +316mV set by the positive supply voltage of
+6V and resistors R6 and R7. (VZ is proportional to this voltage, and systems
requiring greater calibration accuracy should replace the supply-dependent
reference with a more stable source. However, VS is independent of the supply
voltages, being determined by the band-gap reference in the X-AMP.) Any
difference in these voltages is integrated by the op-amp U3B, with a time-constant
of 3ms formed by the parallel sum of R6/R7 and C3.
If the gain of the AD600 is too high, VOUT will be greater than the "set-point" of
316mV, causing the output of U3B – that is, VLOG – to ramp up (note that the
integrator is non-inverting). A fraction of VLOG is connected to the inverting gaincontrol inputs of the AD600, causing the gain to be reduced, as required, until VOUT
is equal to 316mV (DC), at which time the AC voltage at the output of A2 is forced to
exactly 316mV (rms). This fraction is set by R4 and R5 such that a 15.625mV
change in the control voltages of A1 and A2 – which would change the gain of the
two cascaded amplifiers by 1 dB – requires a change of 100mV at VLOG. Since A2 is
forced to operate well below its limiting level, waveforms of high crest-factor can be
tolerated throughout the amplifier.
To verify the operation, assume an input of 10mV rms is applied to the input,
resulting in a voltage of 3.16mV rms at the input to A1 (due to the 10dB
attenuator). If the system performs as claimed, VLOG (and hence VG) should be
zero. This being the case, the gain of both A1 and A2 will be 20dB and the output of
the AD600 will be 100 times (40dB) greater than its input, 316mV rms. This is the
input required at the AD636 to balance the loop, confirming the basic operation.
Note that unlike most AGC circuits, (which often have a high gain/temperature
coefficient due to the internal "kT/q" scaling), the voltages and thus the output of
this measurement system are very stable over temperature. This behavior arises
directly from the exact exponential calibration of the ladder attenuator.
Typical results are shown for a sinewave input at 100kHz. Figure 3.10 shows that
the output is held very close to the set-point of 316mV rms over an input range in
excess of 80dB.
10
SIGNAL OUTPUT VOUT VERSUS INPUT LEVEL
450
425
400
375
VOUT - mV RMS
350
325
300
275
250
225
200
175
150
10µV
100µV
1mV
10mV
100mV
1V
10V
INPUT SIGNAL - V RMS
a
3.10
Figure 3.11 shows the "decibel" output voltage, VLOG, and Figure 3.12 shows that
the deviation from the ideal output logarithmic output is within ±1 dB for the 80dB
range from 80µV to 800mV.
THE LOGARITHMIC OUTPUT V LOG
VERSUS INPUT SIGNAL LEVEL
5
4
3
2
VOUT - V
1
0
-1
-2
-3
-4
-5
10µV
100µV
1mV
10mV
100mV
1V
10V
INPUT SIGNAL - V RMS
a
3.11
11
DEVIATION FROM THE IDEAL
LOGARITHMIC OUTPUT
2.5
2.0
OUTPUT ERROR - dB
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-2.0
-2.5
10µV
100µV
1mV
10mV
100mV
1V
10V
INPUT SIGNAL - V RMS
a
3.12
By suitable choice of the input attenuator, R1+R2, this could be centered to cover
any range from 25µV to 250mV to, say, 1mV to 10V, with appropriate correction to
the value of VZ. (Note that VS is not affected by the changes in the range). The gain
ripple of ±0.2dB seen in this curve is the result of the finite interpolation error of the
X-AMP. It occurs with a periodicity of 12dB – twice the separation between the tap
points in each amplifier section.
This ripple can be canceled whenever the X-AMP stages are cascaded by introducing
a 3dB offset between the two pairs of control voltages. A simple means to achieve
this is shown in Figure 3.13: the voltages at C1HI and C2HI are "split" by
±46.875mV, or ±1.5dB. Alternatively, either one of these pins can be individually
offset by 3dB, and a 1.5dB gain adjustment made at the input attenuator (R1+R2).
The error curve shown in Figure 3.14 demonstrates that over the central portion of
the range, the output voltage can be maintained very close to the ideal value. The
penalty for this modification is higher errors at both ends of the range.
12
METHOD FOR CANCELING THE
GAIN-CONTROL RIPPLE
16
15
14
13
U1
AD600
12
11
10
9
C1HI
1
VINP
A1CM
NC
2
-6V
DEC
A1OP
VPOS
3
VNEG
4
CAVG
NC
5
VLOG
NC
6
BFOP
7
BFIN
+6V DEC
VNEG
C2
2µF
-6V DEC
A2OP
U2
AD636
A2CM
C2HI
-46.875mV
+46.875mV
-6V
DEC
+6V
10kΩ
78.7Ω
78.7Ω
10kΩ
DEC
3dB OFFSET
NC = NO CONNECT
MODIFICATION
a
3.13
LOGARITHMIC ERROR USING THE
PREVIOUS CIRCUIT MODIFICATION
2.5
2.0
OUTPUT ERROR - dB
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-2.0
-2.5
10µV
100µV
1mV
10mV
100mV
1V
10V
INPUT SIGNAL - V RMS
a
3.14
Figure 3.15 shows the ease with which the AD603 (90MHz X-AMP) can be used as a
high speed AGC amplifier. The circuit uses few parts, has a linear-in-dB gain,
operates from a single supply, uses two cascaded amplifiers in sequential gain mode
for maximum S/N ratio (see the data sheet for the AD600/AD602, or AD603 for a
complete description of the methods for cascading X-AMPS), and external resistor
programs each amplifier's gain. It also uses a simple temperature-compensated
detector.
13
A 40MHz, 80dB, LOW-NOISE
AGC AMPLIFIER USING THE AD603
THIS CAPACITOR SETS
AGC TIM E CONSTANT
V AGC
C7
0.1µF
8
3
RT
100Ω
+10V
R1
2.49kΩ
C3
100µF
C4
0.1µF
C8
0.1µF
R13
2.49kΩ
A1
AD603
5
3
2
4
+10V
1
R2
2.49k Ω
+
C5
100 µF
C6
0.1µF
Q1
2N3904
R8
806Ω
6
A2
AD603
5
R12
4.99kΩ
7
C9
0.1µF
J2
2
4
R3
2.49kΩ
C AV
0.1µF
R14
2.49kΩ
8
7
C11
0.1µF
R11
3.83k Ω
+10V
C2
0.1µF
6
R10
1.24kΩ
Q2
2N3906
+10V
C1
0.1µF
J1
R9
1.54kΩ
C10
0.1µF
1
R4
2.49kΩ
AGC LINE
1V OFFSET FOR
SEQUENTIAL GAIN
R5
5.49kΩ
R7
3.48kΩ
+10V
5.5V
NOTES:
1 RT PR OVIDES A 50Ω INPUT IM PEDANCE
2 C3 AND C5 ARE TANTALU M
6.5V
R6
1.05kΩ
a
3.15
The circuit operates from a single +10V supply. Resistors R1, R2 and R3, R4 bias the
common pins of A1 and A2 at 5V. This pin is a low impedance point and must have a
low impedance path to ground, provided by the 100µF tantalum capacitor and the
0.1µF ceramic capacitors.
The cascaded amplifiers operate in sequential gain. The offset voltage between the
pins 2 (GNEG) of A1 and A2 is 1.05V (42.14dB x 25mV/dB), provided by a voltage
divider consisting of resistors R5, R6, and R7. Using standard values, the offset is
not exact but is not critical for this application.
The gain of both A1 and A2 is programmed by resistors R13 and R14, respectively,
to be about 42dB; thus the maximum gain of the circuit is twice that, or 84dB. The
gain-control range can be shifted up by as much as 20dB by appropriate choices of
R13 and R14.
The circuit operates as follows. A1 and A2 are cascaded. Capacitor C1 and the 100Ω
of resistance at the input of A1 form a time-constant of 10µs. C2 blocks the small DC
offset voltage at the output of A1 (which might otherwise saturate A2 at its
maximum gain) and introduces a high-pass corner at about 16kHz, eliminating low
frequency noise.
A half-wave detector is used based on Q1 and R8. The current into capacitor CAV is
the difference between the collector current of Q2 (biased to be 300µA at 27°C,
300K) and the collector current of Q1, which increases with the amplitude of the
output signal. The automatic gain control voltage, VAGC, is the time-integral of this
error current. In order for VAGC (and thus the gain) to remain insensitive to shortterm amplitude fluctuations in the output signal, the rectified current in Q1 must,
14
on average, exactly balance the current in Q2. If the output of A2 is too small to do
this, VAGC will increase, causing the gain to increase, until Q1 conducts sufficiently.
Consider the case where R8 is zero and the output voltage VOUT is a square wave
at, say 455kHz, that is, well above the corner frequency of the control loop.
During the time VOUT is negative with respect to the base voltage of Q1, Q1
conducts; when VOUT is positive, it is cut off. Since the average collector current of
Q1 is forced to be 300µA, and the square wave has a duty cycle of 1:1, Q1's collector
current when conducting must be 600µA. With R8 omitted, the peak amplitude of
VOUT is forced to be just the VBE of Q1 at 600µA, typically about 700mV, or 2VBE
peak-to-peak. This voltage, hence the amplitude at which the output stabilizes, has
a strong negative temperature coefficient (TC), typically –1.7mV/°C. Although this
may not be troublesome in some applications, the correct value of R8 will render the
output stable with temperature.
To understand this, first note that the current in Q2 is made to be proportional to
absolute temperature (PTAT). For the moment, continue to assume that the signal
is a square wave.
When Q1 is conducting, VOUT is now the sum of VBE and a voltage which is PTAT
and which can be chosen to have an equal but opposite TC to that of VBE. This is
actually nothing more than an application of the "bandgap voltage reference"
principle. When R8 is chosen such that the sum of the voltage across it and the VBE
of Q1 is close to the bandgap voltage of about 1.2V, VOUT will be stable over a wide
range of temperatures, provided, of course, that Q1 and Q2 share the same thermal
environment.
Since the average emitter current is 600µA during each half-cycle of the square
wave, a resistor of 833Ω would add a PTAT voltage of 500mV at 300K, increasing by
1.66mV/°C. In practice, the optimum value will depend on the type of transistor
used, and, to a lesser extent, on the waveform for which the temperature stability is
to be optimized; for the inexpensive 2N3904/2N3906 pair and sine wave signals, the
recommended value is 806Ω.
This resistor also serves to lower the peak current in Q1 when more typical signals
(usually sinusoidal) are involved, and the 1.8kHz lowpass filter it forms with CAV
helps to minimize distortion due to ripple in VAGC. Note that the output amplitude
under sine wave conditions will be higher than for a square wave, since the average
value of the current for an ideal rectifier would be 0.637 times as large, causing the
output amplitude to be 1.2V/0.637=1.88V, or 1.33V rms. In practice, the somewhat
nonideal rectifier results in the sine wave output being regulated to about 1.4Vrms,
or 3.6V p-p.
The bandwidth of the circuit exceeds 40MHz. At 10.7MHz, the AGC threshold is
100µV (–67dBm) and its maximum gain is 83dB, 20log(1.4V/100µV). The circuit
holds its output at 1.4V rms for inputs as low as –67dBm to +15dBm (82dB), where
the input signal exceeds the AD603's maximum input rating. For a +10dBm input at
10.7MHz, the second harmonic is 34dB down from the fundamental, and the third
harmonic is 35dB down.
15
LOGARITHMIC AMPLIFIERS
The term "Logarithmic Amplifier" (generally abbreviated to "log amp") is something
of a misnomer, and "Logarithmic Converter" would be a better description. The
conversion of a signal to its equivalent logarithmic value involves a nonlinear
operation, the consequences of which can be confusing if not fully understood. It is
important to realize that many of the familiar concepts of linear circuits are
irrelevant to log amps. For example, the incremental gain of an ideal log amp
approaches infinity as the input approaches zero, and a change of offset at the
output of a log amp is equivalent to a change of amplitude at its input - not a change
of input offset.
For the purposes of simplicity in our initial discussions, we shall assume that both
the input and the output of a log amp are voltages, although there is no particular
reason why logarithmic current, transimpedance, or transconductance amplifiers
should not also be designed.
If we consider the equation y = log(x) we find that every time x is multiplied by a
constant A, y increases by another constant A1. Thus if log(K) = K1, then log(AK) =
K1 + A1, log(A2K) = K1 + 2A1, and log(K/A) = K1 – A1. This gives a graph as shown
in Figure 3.16, where y is zero when x is unity, y approaches minus infinity as x
approaches zero, and which has no values for x for which y is negative.
GRAPH OF Y = LOG(X)
Y
X
X=1
a
3.16
On the whole, log amps do not behave in this way. Apart from the difficulties of
arranging infinite negative output voltages, such a device would not, in fact, be very
useful. A log amp must satisfy a transfer function of the form
Vout = Vy log(Vin/Vx)
over some range of input values which may vary from 100:1 (40dB) to over
1,000,000:1 (120dB).
16
With inputs very close to zero, log amps cease to behave logarithmically, and most
then have a linear Vin/Vout law. This behavior is often lost in device noise. Noise
often limits the dynamic range of a log amp. The constant, Vy,has the dimensions of
voltage, because the output is a voltage. The input, Vin, is divided by a voltage, Vx,
because the argument of a logarithm must be a simple dimensionless ratio.
A graph of the transfer characteristic of a log amp is shown in Figure 3.17. The scale
of the horizontal axis (the input) is logarithmic, and the ideal transfer characteristic
is a straight line. When Vin = Vx, the logarithm is zero (log 1 = 0). Vx is therefore
known as the intercept voltage of the log amp because the graph crosses the
horizontal axis at this value of Vin.
LOG AMP TRANSFER FUNCTION
VYLOG(VIN/VX)
IDEAL
ACTUAL
2VY
SLOPE = VY
VY
+
0
ACTUAL
-
VIN = VX
VIN = 10VX
VIN = 100VX
INPUT ON
LOG SCALE
IDEAL
a
3.17
The slope of the line is proportional to Vy. When setting scales, logarithms to the
base 10 are most often used because this simplifies the relationship to decibel
values: when Vin = 10Vx,the logarithm has the value of 1, so the output voltage is
Vy. When Vin = 100Vx, the output is 2Vy ,and so forth. Vy can therefore be viewed
either as the "slope voltage" or as the "volts per decade factor."
The logarithm function is indeterminate for negative values of x. Log amps can
respond to negative inputs in three different ways: (1) They can give a fullscale
negative output as shown in Figure 3.18. (2) They can give an output which is
proportional to the log of the absolute value of the input and disregards its sign as
shown in Figure 3.19. This type of log amp can be considered to be a full-wave
detector with a logarithmic characteristic, and is often referred to as a detecting log
amp. (3) They can give an output which is proportional to the log of the absolute
value of the input and has the same sign as the input as shown in Figure 3.20. This
type of log amp can be considered to be a video amp with a logarithmic
characteristic, and may be known as a logarithmic video (log video) amplifier or,
sometimes, a true log amp (although this type of log amp is rarely used in videodisplay-related applications).
17
BASIC LOG AMP
(SATURATES WITH NEGATIVE INPUT)
+
OUTPUT
-
+
INPUT
-
a
3.18
DETECTING LOG AMP
(OUTPUT POLARITY INDEPENDENT
OF INPUT POLARITY)
+
OUTPUT
-
+
INPUT
-
a
3.19
18
LOG VIDEO OR "TRUE LOG AMP"
(SYMMETRICAL RESPONSE
TO POSITIVE OR NEGATIVE SIGNALS)
+
OUTPUT
-
+
INPUT
-
a
3.20
There are three basic architectures which may be used to produce log amps: the
basic diode log amp, the successive detection log amp, and the "true log amp" which
is based on cascaded semi-limiting amplifiers.
The voltage across a silicon diode is proportional to the logarithm of the current
through it. If a diode is placed in the feedback path of an inverting op-amp, the
output voltage will be proportional to the log of the input current as shown in Figure
3.21. In practice, the dynamic range of this configuration is limited to 40-60dB
because of non-ideal diode characteristic, but if the diode is replaced with a diodeconnected transistor as shown in Figure 3.22, the dynamic range can be extended to
120dB or more. This type of log amp has three disadvantages: (1) both the slope and
intercept are temperature dependent; (2) it will only handle unipolar signals; and (3)
its bandwidth is both limited and dependent on signal amplitude.
Where several such log amps are used on a single chip to produce an analog
computer which performs both log and antilog operations, the temperature variation
in the log operations is unimportant, since it is compensated by a similar variation
in the antilogging. This makes possible the AD538, a monolithic analog computer
which can multiply, divide, and raise to powers. Where actual logging is required,
however, the AD538 and similar circuits require temperature compensation
(Reference 7). The major disadvantage of this type of log amp for high frequency
applications, though, is its limited frequency response - which cannot be overcome.
However carefully the amplifier is designed, there will always be a residual feedback
capacitance Cc (often known as Miller capacitance), from output to input which
limits the high frequency response.
What makes this Miller capacitance particularly troublesome is that the impedance
of the emitter-base junction is inversely proportional to the current flowing in it - so
that if the log amp has a dynamic range of 1,000,000:1, then its bandwidth will also
vary by 1,000,000:1. In practice, the variation is less because other considerations
19
limit the large signal bandwidth, but it is very difficult to make a log amp of this
type with a small-signal bandwidth greater than a few hundred kHz.
THE DIODE / OP-AMP LOG AMP
-VIN
+ V V=
I
RIN
I = IIN
IIN
I
kT
In
IO
q
EO
if I >> IO
+
V
IIN
IO
kT
EO =
In
q
0.06log
VIN
RINIO
if IIN >> IO
a
3.21
TRANSISTOR / OP-AMP LOG AMP
IC
IE
EO
IIN
+
IIN
kT
EO = q In
IES
a
3.22
For high frequency applications, therefore, detecting and true log architectures are
used. Although these differ in detail, the general principle behind their design is
common to both: instead of one amplifier having a logarithmic characteristic, these
designs use a number of similar cascaded linear stages having well-defined large
signal behavior.
Consider N cascaded limiting amplifiers, the output of each driving a summing
circuit as well as the next stage (Figure 3.23). If each amplifier has a gain of A dB,
20
the small signal gain of the strip is NA dB. If the input signal is small enough for
the last stage not to limit, the output of the summing amplifier will be dominated by
the output of the last stage.
BASIC MULTI-STAGE LOG AMP ARCHITECTURE
INPUT
Σ
a
OUTPUT
3.23
As the input signal increases, the last stage will limit. It will now make a fixed
contribution to the output of the summing amplifier, but the incremental gain to the
summing amplifier will drop to (N-1)A dB. As the input continues to increase, this
stage in turn will limit and make a fixed contribution to the output, and the
incremental gain will drop to (N-2)A dB, and so forth - until the first stage limits,
and the output ceases to change with increasing signal input.
The response curve is thus a set of straight lines as shown in Figure 3.24. The total
of these lines, though, is a very good approximation to a logarithmic curve, and in
practical cases, is an even better one, because few limiting amplifiers, especially
high frequency ones, limit quite as abruptly as this model assumes.
21
BASIC MULTI-STAGE LOG AMP RESPONSE
(UNIPOLAR CASE)
G=0
}
} G = (N-3)A dB
OUTPUT
G = (N-4)A dB
} G = (N-2)A dB
}
G = (N-1)A dB
G = NA dB
INPUT
a
3.24
The choice of gain, A, will also affect the log linearity. If the gain is too high, the log
approximation will be poor. If it is too low, too many stages will be required to
achieve the desired dynamic range. Generally, gains of 10 to 12dB (3x to 4x) are
chosen.
This is, of course, an ideal and very general model - it demonstrates the principle,
but its practical implementation at very high frequencies is difficult. Assume that
there is a delay in each limiting amplifier of t nanoseconds (this delay may also
change when the amplifier limits but let's consider first order effects!). The signal
which passes through all N stages will undergo delay of Nt nanoseconds, while the
signal which only passes one stage will be delayed only t nanoseconds. This means
that a small signal is delayed by Nt nanoseconds, while a large one is "smeared",
and arrives spread over Nt nanoseconds. A nanosecond equals a foot at the speed of
light, so such an effect represents a spread in position of Nt feet in the resolution of
a radar system-which may be unacceptable in some systems (for most log amp
applications this is not a problem).
A solution is to insert delays in the signal paths to the summing amplifier, but this
can become complex. Another solution is to alter the architecture slightly so that
instead of limiting gain stages, we have stages with small signal gain of A and large
signal (incremental) gain of unity (0dB). We can model such stages as two parallel
amplifiers, a limiting one with gain, and a unity gain buffer, which together feed a
summing amplifier as shown in Figure 3.25.
22
STRUCTURE AND PERFORMANCE OF
"TRUE" LOG AMP ELEMENT AND OF A
LOG AMP FORMED BY SEVERAL SUCH ELEMENTS
LIMITING
AMPLIFIER
GAIN = 3
INPUT
UNITY GAIN
AMPLIFIER
GAIN = 1
OUTPUT
Σ
OUTPUT
OUTPUT
UNITY GAIN
(LARGE SIGNAL)
INPUT
GAIN = 4
(SMALL SIGNAL)
INPUT
a
3.25
Figure 3.25 shows that such stages, cascaded, form a log amp without the necessity
of summing from individual stages. Both the multi-stage architectures described
above are video log amplifiers, or true log amplifiers, but the most common type of
high frequency log amplifier is the successive detection log amp architecture shown
in Figure 3.26.
SUCCESSIVE DETECTION LOGARITHMIC AMPLIFIER
WITH LOG AND LIMITER OUTPUTS
LIMITING AMPLIFIERS
LIMITER
OUTPUT
INPUT
DETECTORS
DETECTORS MAY BE FULL OR HALF WAVE
SHOULD BE CURRENT OUTPUT DEVICES (NOT
SIMPLE DIODES) SO THAT OUTPUTS MAY BE
SUMMED WITHOUT ADDITIONAL SUMMING
COMPONENTS BEING NECESSARY
a
3.26
The successive detection log amp consists of cascaded limiting stages as described
above, but instead of summing their outputs directly, these outputs are applied to
detectors, and the detector outputs are summed as shown in Figure 3.26. If the
23
detectors have current outputs, the summing process may involve no more than
connecting all the detector outputs together.
Log amps using this architecture have two outputs: the log output and a limiting
output. In many applications, the limiting output is not used, but in some (FM
receivers with "S"-meters, for example), both are necessary. The limited output is
especially useful in extracting the phase information from the input signal in polar
demodulation techniques.
The log output of a successive detection log amplifier generally contains amplitude
information, and the phase and frequency information is lost. This is not necessarily
the case, however, if a half-wave detector is used, and attention is paid to equalizing
the delays from the successive detectors - but the design of such log amps is
demanding.
The specifications of log amps will include noise, dynamic range, frequency response
(some of the amplifiers used as successive detection log amp stages have low
frequency as well as high frequency cutoff), the slope of the transfer characteristic
(which is expressed as V/dB or mA/dB depending on whether we are considering a
voltage- or current-output device), the intercept point (the input level at which the
output voltage or current is zero), and the log linearity. (See Figures 3.27 and 3.28)
KEY PARAMETERS OF LOG AMPS
n
NOISE: The Noise Referred to the Input (RTI) of the Log Amp.
It May Be Expressed as a Noise Figure or as a Noise Spectral
Density (Voltage, Current, or Both) or as a Noise Voltage, a Noise
Current, or Both
n
DYNAMIC RANGE: Range of Signal Over Which the Amplifier
Behaves in a Logarithmic Manner (Expressed in dB)
n
FREQUENCY RESPONSE: Range of Frequencies Over Which
the Log Amp Functions Correctly
n
SLOPE: Gradient of Transfer Characteristic in V/dB or mA/dB
n
INTERCEPT POINT: Value of Input Signal at Which Output is Zero
n
LOG LINEARITY: Deviation of Transfer Characteristic (Plotted on
log/lin Axes) from a Straight Line (Expressed in dB)
a
3.27
24
LOG LINEARITY
EO
(LINEAR)
LOG ERROR
Ei (dBm)
a
3.28
In the past, it has been necessary to construct high performance, high frequency
successive detection log amps (called log strips) using a number of individual
monolithic limiting amplifiers such as the Plessey SL-1521-series (see Reference 16).
Recent advances in IC processes, however, have allowed the complete log strip
function to be integrated into a single chip, thereby eliminating the need for costly
hybrid log strips.
The AD641 log amp contains five limiting stages (10dB per stage) and five full-wave
detectors in a single IC package, and its logarithmic performance extends from dc to
250MHz. Furthermore, its amplifier and full-wave detector stages are balanced so
that, with proper layout, instability from feedback via supply rails is unlikely. A
block diagram of the AD641 is shown in Figure 3.29. Unlike many previous
integrated circuit log amps, the AD641 is laser trimmed to high absolute accuracy of
both slope and intercept, and is fully temperature compensated. Key features of the
AD641 are summarized in Figure 3.30. The transfer function for the AD641 as well
as the log linearity is shown in Figure 3.31.
25
BLOCK DIAGRAM OF THE AD641 MONOLITHIC LOG AMP
RG1
COM
18
ATN OUT
19
1kΩ
17
RG0
16
1kΩ
RG2
LOG OUT
LOG COM
15
14
13
INTERCEPT POSITIONING BIAS
12 +VS
FULL-WAVE
DETECTOR
FULL-WAVE
DETECTOR
FULL-WAVE
DETECTOR
FULL-WAVE
DETECTOR
FULL-WAVE
DETECTOR
10dB
10dB
10dB
10dB
10dB
SIG +IN 20
SIG -IN
1
ATN LO 2
AMPLIFIER/LIMITER
AMPLIFIER/LIMITER
AMPLIFIER/LIMITER
AMPLIFIER/LIMITER
11
SIG +OUT
10
SIG -OUT
AMPLIFIER/LIMITER
27Ω
ATN COM 3
30Ω
9 BL2
270Ω
ATN COM 4
5
6
ATN IN
BL1
GAIN BIAS REGULATOR
7
SLOPE BIAS REGULATOR
8
ITC
-VS
a
3.29
AD641 KEY FEATURES
n
44dB Dynamic Range
n
Bandwidth dc to 250MHz
n
Laser-Trimmed Slope of 1mA/decade - Temperature Stable
n
Laser-Trimmed Intercept of 1mV - Temperature Stable
n
Less than 2dB Log Non-Linearity
n
Limiter Output: ±1.6dB Gain Flatness, ±2° Phase Variation
for -44dBm to 0dBm inputs @ 10.7MHz
n
Balanced Circuitry for Stability
n
Minimal External Component Requirement
a
3.30
26
DC LOGARITHMIC TRANSFER FUNCTION
AND ERROR CURVE FOR SINGLE AD641
2
2.4
1
2.2
0
1.8
1.6
1.4
ERROR - dB
OUTPUT CURRENT - mA
2.0
1.2
1.0
0.8
0.6
0.4
0.2
0
-0.2
-0.4
0.1
1.0
10.0
100.0
1000.0
INPUT VOLTAGE - mV
(EITHER SIGN)
a
3.31
Because of its high accuracy, the actual waveform driving the AD641 must be
considered when calculating responses. When a waveform passes through a log
function generator, the mean value of the resultant waveform changes. This does
not affect the slope of the response, but the apparent intercept is modified according
to Figure 3.32.
THE EFFECT OF WAVEFORM ON INTERCEPT POINT
INPUT
PEAK
INTERCEPT
ERROR (RELATIVE
WAVEFORM
OR RMS
FACTOR
TO A DC INPUT)
Square Wave
Either
1
0.00dB
Sine Wave
Peak
2
–6.02dB
Sine Wave
RMS
1.414 (√
√ 2)
–3.01dB
Triwave
Peak
2.718 (e)
–8.68dB
Triwave
RMS
1.569 (e/√
√ 3)
–3.91dB
Gaussian Noise
RMS
1.887
–5.52dB
a
3.32
The AD641 is calibrated and laser trimmed to give its defined response to a DC level
or a symmetrical 2kHz square wave. It is also specified to have an intercept of 2mV
for a sinewave input (that is to say a 2kHz sinewave of amplitude 2mV peak [not
27
peak-to-peak] gives the same mean output signal as a DC or square wave signal of
1mV).
The waveform also affects the ripple or nonlinearity of the log response. This ripple
is greatest for DC or square wave inputs because every value of the input voltage
maps to a single location on the transfer function, and thus traces out the full
nonlinearities of the log response. By contrast, a general time-varying signal has a
continuum of values within each cycle of its waveform. The averaged output is
thereby "smoothed" because the periodic deviations away from the ideal response, as
the waveform "sweeps over" the transfer function, tend to cancel. As is clear in
Figure 3.33, this smoothing effect is greatest for a triwave.
THE EFFECT OF WAVEFORM ON AD641 LOG LINEARITY
DEVIATION FROM EXACT LOGARITHMIC
TRANSFER FUNCTION - dB
2
0
SQUARE
WAVE INPUT
-2
-4
SINE WAVE
INPUT
-6
TRIWAVE
INPUT
-8
-10
-80
-70
-60
-50
-40
-30
-20
-10
INPUT AMPLITUDE IN dB ABOVE 1V, AT 10kHz
a
3.33
Each of the five stages in the AD641 has a gain of 10dB and a full-wave detected
output. The transfer function for the device was shown in Figure 3.21 along with the
error curve. Note the excellent log linearity over an input range of 1 to 100mV
(40dB). Although well suited to RF applications, the AD641 is dc-coupled
throughout. This allows it to be used in LF and VLF systems, including audio
measurements, sonar, and other instrumentation applications requiring operation to
low frequencies or even dc.
The limiter output of the AD641 has better than 1.6dB gain flatness (-44dBm to
0dBm @ 10.7MHz) and less than 2° phase variation, allowing it to be used as a polar
demodulator
The AD606 is a complete monolithic 50MHz bandwidth log amp using 9 stages of
successive detection, and is shown in Figure 3.34. Key specifications are
summarized in Figure 3.35. Seven of the amplifier/detector stages handle inputs
from –80dBm (32µV rms) up to about –14dBm (45mV rms). The noise floor is about
–83dBm (18µV rms). Another two parallel stages receive the input attenuated by
22.3dB, and respond to inputs up to +10dBm (707mV rms). The gain of each stage is
11.15dB and is accurately stabilized over temperature by a precise biasing system.
28
The AD606 provides both logarithmic and limited outputs. The logarithmic output is
from a three-pole post-demodulation lowpass filter and provides an output voltage of
+0.1V DC to +4V DC. The logarithmic scaling is such that the output is +0.5V for a
sinusoidal input of –75dBm, and +3.5V at an input of +5dBm. Over this range, the
log linearity is typically within ±0.4dB.
AD606 50MHz, 80dB LOG AMP BLOCK DIAGRAM
INHI
COMM
PRUP
16
15
14
VPOS
FIL1
13
FIL2
12
LADJ
11
LMHI
10
9
REFERENCE
AND POWER UP
30pF
360kΩ
X1
360kΩ
30pF
30kΩ
30kΩ
OFFSET-NULL
LOW-PASS FILTER
1.5kΩ
FINAL
LIMITER
MAIN SIGNAL PATH
11.15 dB/STAGE
250Ω
12µA/dB
1.5kΩ
HIGH-END
DETECTORS
ONE-POLE
FILTER
AD606
2
INLO
3
COMM
ISUM
TWO-POLE
SALLEN-KEY
FILTER
9.375kΩ
X2
2µA/dB
1
2pF
9.375kΩ
2pF
4
5
ILOG
BFIN
6
VLOG
7
OPCM
a
8
LMLO
3.34
AD606 LOG AMP KEY FEATURES
n
Dynamic Range: –75dBm to +5dBm (80dB)
n
Input Noise: < 1.5nV/√
√ Hz
n
Usable from 200Hz to Greater than 50MHz
n
Slope: 37.5mV/dB Voltage Output
n
On-Chip Lowpass Output Filter
n
Limiter Output: ±1.6dB Gain Flatness, ±2° Phase Variation
for -44dBm to 0dBm inputs @ 10.7MHz
n
+5V Single-Supply, 65mW Power Consumption
a
3.35
29
The AD606 can operate above and below these limits, with reduced linearity, to
provide as much as 90dB of conversion range. A second lowpass filter automatically
nulls the input offset of the first stage down to the submicrovolt level.
The AD606's limiter output provides a hard-limited signal output as a differential
current of ±1.2mA from open-collector outputs. In a typical application, both of these
outputs are loaded by 200Ω resistors to provide a voltage gain of more than 90dB
from the input. This limiting amplifier has exceptionally low amplitude-to-phase
conversion. The limiter output has ±1dB output flatness and ±3° phase stability over
an 80dB range at 10.7MHz.
30
RECEIVER OVERVIEW
Walt Kester, Bob Clarke
We will now consider how the previously discussed building blocks can be used in
designing a receiver. First, consider the analog superheterodyne receiver invented in
1917 by Major Edwin H. Armstrong (see Figure 3.36). This architecture represented
a significant improvement over single-stage direct conversion (homodyne) receivers
which had previously been constructed using tuned RF amplifiers, a single detector,
and an audio gain stage. A significant advantage of the superheterodyne receiver is
that it is much easier and more economical to have the gain and selectivity of a
receiver at fixed intermediate frequencies (IF) than to have the gain and frequencyselective circuits "tune" over a band of frequencies.
DUAL CONVERSION SUPERHET RECEIVER
(EXAMPLE FREQUENCIES)
900MHz
RF
IF STRIP
LO1
TUNED
1ST IF
240MHz
LO2
FIXED
2ND IF
10.7MHz
DEMODULATOR
a
3.36
The receiver shown is a dual conversion receiver with two intermediate frequency
(IF) stages. The frequencies chosen are typical in digital mobile radio (DMR), but the
principles apply to other systems as well. The 900MHz RF signal is mixed down to
the first IF frequency of 240MHz. Tuning is accomplished by the first local oscillator
(LO1). The LO1 frequency is chosen such that the output of the first mixer is at the
first IF frequency, 240MHz. Choosing a relatively high first IF frequency eases the
requirement on the image frequency rejection filter as will be discussed in the next
section on mixers. The first IF is then mixed down to the second IF frequency of
10.7MHz, where it is demodulated (either using analog or digital techniques).
Because of the wide dynamic range of the RF signal, such a receiver requires the use
of automatic gain control, voltage controlled amplifiers, and in some cases
(depending on the type of demodulation), logarithmic amplifiers.
31
Receiver design is a complicated art, and there are many tradeoffs that can be made
between IF frequencies, single-conversion vs. double-conversion or triple conversion,
filter cost and complexity at each stage in the receiver, demodulation schemes, etc.
There are many excellent references on the subject, and the purpose of this section is
only to acquaint the design engineer with some of the building block ICs which can
make receiver design much easier.
Before we look at further details of a receiver, the subject of mixing requires further
discussion.
32
MULTIPLIERS, MODULATORS, AND MIXERS
Barrie Gilbert, Bob Clarke
An idealized mixer is shown in Figure 3.37. An RF (or IF) mixer (not to be confused
with video and audio mixers) is an active or passive device that converts a signal
from one frequency to another. It can either modulate or demodulate a signal. It has
three signal connections, which are called ports in the language of radio engineers.
These three ports are the radio frequency (RF) input, the local oscillator (LO) input,
and the intermediate frequency (IF) output.
THE MIXING PROCESS
IDEAL MIXER
IF OUTPUT
RF INPUT
fRF
fRF + fLO
fRF - fLO
LO INPUT
fLO
a
3.37
A mixer takes an RF input signal at a frequency fRF, mixes it with a LO signal at a
frequency fLO, and produces an IF output signal that consists of the sum and
difference frequencies, fRF ± fLO. The user provides a bandpass filter that follows
the mixer and selects the sum (fRF + fLO) or difference (fRF – fLO) frequency.
Some points to note about mixers and their terminology:
• When the sum frequency is used as the IF, the mixer called an upconverter; when
the difference is used, the mixer is called a downconverter. The former is often used
in a transmit channel, the latter in a receive channel.
• In a receiver, when the LO frequency is below the RF, it is called low-side injection
and the mixer a low-side downconverter; when the LO is above the RF, it is called
high-side injection, and the mixer a high-side downconverter.
• Each of the outputs is only half the amplitude (one-quarter the power) of the
individual inputs; thus, there is a loss of 6dB in this ideal linear mixer. (In a
33
practical multiplier, the conversion loss may be greater than 6dB, depending on the
scaling parameters of the device. Here, we assume a mathematical multiplier,
having no dimensional attributes.
A mixer can be implemented in several ways, using active or passive techniques.
A brief review of the various classes of nonlinear elements that can be used for
frequency translation may be helpful in setting the context. We can identify three
subclasses of circuits, sharing certain similarities. All are in the class of signal
multipliers, producing at their output a signal which is, in one way or another, the
product of its two inputs. They are multipliers, modulators, and mixers.
An analog multiplier generally has two signal input ports, which can be called X and
Y, and generates an output W that is the linear product of the voltages applied to
these two ports. To retain dimensional consistency, the analog linear multiplication
function must invoke the use of a reference voltage, which we can call U, thus
W=XY/U. In some cases, U is actually a third input that can be used to implement
analog division.
There are three functional categories of multipliers: In single-quadrant multipliers,
X and Y must be unipolar; in two-quadrant multipliers, one of the inputs may be
bipolar; in four-quadrant multipliers, both X and Y may be bipolar. Analog Devices
produces a wide range of "linear" multipliers, including the AD534, AD538, AD539,
AD633, AD734, AD834 and AD835, providing the highest available accuracy
(±0.02% for the AD734) to the highest speed (more than 500MHz for the AD834).
Modulators (sometimes called balanced-modulators, doubly-balanced modulators or
even on occasions high level mixers) can be viewed as sign-changers. The two inputs,
X and Y, generate an output W, which is simply one of these inputs (say, Y)
multiplied by just the sign of the other (say, X), that is W = Ysign(X). Therefore, no
reference voltage is required. A good modulator exhibits very high linearity in its
signal path, with precisely equal gain for positive and negative values of Y, and
precisely equal gain for positive and negative values of X. Ideally, the amplitude of
the X input needed to fully switch the output sign is very small, that is, the X-input
exhibits a comparator-like behavior. In some cases, where this input may be a logic
signal, a simpler X-channel can be used. A highly-linear mixer such as the AD831 is
well-suited as a modulator.
A mixer is a modulator optimized for frequency-translation. Its place in the signal
path is usually close to the antenna, where both the wanted and (often large)
unwanted signals coexist at its signal input, usually called the RF port. Thus, the
mixer must exhibit excellent linearity in the sense that its output (at the IF port) is
expected to increase by the same number of dB as a test signal applied to the RF
port, up to as high as level as possible. This attribute is defined both by the 1dB
gain-compression and the 3rd-order intercept (later explained). The conversion
process is driven by an input applied to the LO port.
Noise and matching characteristics are crucial to achieving acceptable levels of
performance in a receiver’s mixer. It is desirable to keep the LO power to a
minimum to minimize cross-talk between the three ports, but this often conflicts
with other requirements. The gain from the RF port to its IF port at specified RF
and LO frequencies is called the conversion gain and in classical diode-bridge mixers
is less than –4dB. Active mixers provide higher conversion gain, and better port-port
34
isolation, but often at the expense of noise and linearity. It is not usually possibly
(nor even desirable) to describe mixer behavior using equations relating the
instantaneous values of inputs and outputs; instead, we generally seek to
characterize mixers in terms of their non-ideal cross-product terms at the output. In
this class, Analog Devices has the AD831, and mixers are found embedded in the
AD607, AD608 and other signal-processing ICs.
Thus far, we have seen that multipliers are linear in their response to the
instantaneous value of both of their input voltages; modulators are linear in their
response to one input, the other merely flipping the sign of this signal at regular
intervals, with virtually zero transition time, and beyond that having ideally no
other effect on the signal; mixers are a sort of RF half-breed, ideally being very
linear on the RF input, and ‘binary’ in their switching function in response to the LO
input, but in reality being nonideal in both respects; they are optimized for very low
noise and minimal intermodulation distortion.
Mixing Using an Ideal Analog Multiplier
Figure 3.38 shows a greatly simplified RF mixer by assuming the use of an analog
multiplier.
Ideally, the multiplier has no noise, no limit to the maximum signal amplitude, and
no intermodulation between the various RF signals (that is, no spurious
nonlinearities). Figure 3.39 shows the result of mixing (= multiplying) an RF input
of sinωRFt with (= by) a LO input of sinωLOt, where ωRF = 2π×11MHz and ωLO =
2π×10MHz.
Clearly, to better understand mixer behavior, we will need to consider not only the
time-domain waveforms, as shown here, but also the spectrum of the IF output.
Figure 3.40 shows the output spectrum corresponding to the above IF waveform.
35
"MIXING" USING AN ANALOG MULTIPLIER
RF INPUT
VX
IF OUTPUT
VX • VY
LO INPUT
VY
ANALOG MULTIPLIER, e.g., AD834
a
3.38
INPUTS AND OUTPUTS FOR MULTIPLYING MIXER
FOR fRF = 11MHz, fLO = 10MHz
RF
Horizontal:
200ns/div.
LO
IF
a
3.39
36
OUTPUT SPECTRUM FOR MULTIPLYING MIXER
FOR fRF = 11MHz, fLO = 10MHz
fLO = 10MHz
0.8
fRF = 11MHz
LINEAR
AMPLITUDE
LO AND RF FULLY SUPPRESSED
0.6
0.5
0.5
0.4
SUM AT
21MHz
DIFFERENCE
AT 1MHz
NO HARMONICS
0.2
0
0
10
20
30
40
50
60
FREQUENCY (MHz)
a
3.40
There is no mystery so far. The mathematics are simple. Neglecting scaling issues
(real signals are voltages; thus a practical multiplier needs an embedded voltage
reference, ignored here) the relationship is:
sinωRFt sinωLOt = 1/2 { cos(ωRF + ωLO)t + cos(ωRF – ωLO)t }
Eq. 1
The multiplier has thus transformed the RF input into two, equal-amplitude
cosinusoidal components at its output (the IF port), one at the sum frequency, ωRF +
ωLO, and the other at the difference frequency, ωRF – ωLO.
In practice, an analog multiplier would be a poor choice for a mixer because the two
linear inputs bring with them a serious noise penalty.
Image Response
A receiver using even this mathematically perfect mixer suffers a basic problem,
that of image response. Consider the use of a low-side downconverter. The wanted
output is found at the frequency ωIF = ωRF – ωLO. So we might suppose that the only
component of the RF spectrum that finds its way through the mixer “sieve” to the
narrow IF passband is the wanted component at ωRF. But we could have just as
easily written (1) as
sinωRFt sinωLOt = 1/2 { cos(ωRF + ωLO)t + cos(ωLO – ωRF)t }
Eq. 1a
because the cosine function is symmetric about t = 0. So there is another spectral
component at the RF input that falls in the IF passband, namely the one for which
ωIF = ωLO – ωRF, in this case, the image frequency.
37
Consider the above example, where fLO = 10MHz and fIF = 1MHz; the wanted
response is at the IF frequency, fIF = 1MHz for fRF = 11MHz. However, the mixer
produces the same IF in response to the image frequency, fIMAGE = 9MHz (see
Figure 3.41).
IMAGE RESPONSE
fIF
SIGNAL AT THE
IMAGE FREQUENCY
ALSO PRODUCES A
RESPONSE AT THE
IF FREQUENCY
fIMAGE
fRF
fLO
fIF
0
1
9
fIF
10
11
FREQUENCY (MHz)
3.41
The most practical solution to this dilemma is to carefully choose the IF frequency to
minimize the likelihood of image sensitivity and also include an image-reject filter at
the RF input, just ahead of the mixer. Another approach is to use a special type of
mixer circuit that does not respond to the image frequency. This approach requires
circuitry which is considerably more complex, and for this reason has generally been
unpopular, but it is becoming more practical in a modern IC implementation. It has
the further disadvantage of higher power consumption, since two mixer cells
operating in quadrature are required.
The Ideal Mixer
Ideally, to meet the low-noise, high-linearity objectives of a mixer we need some
circuit that implements a polarity-switching function in response to the LO input.
Thus, the mixer can be reduced to Figure 3.42, which shows the RF signal being
split into in-phase (0°) and anti-phase (180°) components; a changeover switch,
driven by the local oscillator (LO) signal, alternately selects the in-phase and
antiphase signals. Thus reduced to essentials, the ideal mixer can be modeled as a
sign-switcher.
38
AN IDEAL SWITCHING MIXER
+1
RF INPUT
IF OUTPUT
-1
SWITCH, fLO
a
3.42
In a perfect embodiment, this mixer would have no noise (the switch would have
zero resistance), no limit to the maximum signal amplitude, and would develop no
intermodulation between the various RF signals. Although simple in concept, the
waveform at the intermediate frequency (IF) output can be very complex for even a
small number of signals in the input spectrum. Figure 3.43 shows the result of
mixing just a single input at 11MHz with an LO of 10MHz.
The wanted IF at the difference frequency of 1MHz is still visible in this waveform,
and the 21MHz sum is also apparent. But the spectrum of this waveform is clearly
more complex than that obtained using the analog multiplier. How are we to
analyze this?
39
INPUTS AND OUTPUTS FOR IDEAL SWITCHING MIXER
FOR fRF = 11MHz, fLO = 10MHz
RF
Horizontal:
200ns/div.
LO
IF
a
3.43
We still have a product, but now it is that of a sinusoid (the RF input) at ωRF and a
variable that can only have the values +1 or –1, that is, a unit square wave at ωLO.
The latter can be expressed as a Fourier series
SLO = 4/π { sinωLOt - 1/3 sin3ωLOt + 1/5 sin5ωLOt – . . . . }
Eq. 2
Thus, the output of the switching mixer is its RF input, which we can simplify as
sinωRFt, multiplied by the above expansion for the square wave, producing
SIF = 4/π { sinωRFt sinωLOt – 1/3 sinωRFt sin3ωLOt
+ 1/5 sin5ωRFt sin5ωLOt – . . . . }
Eq. 3
Now expanding each of the products, we obtain
SIF = 2/π { sin(ωRF + ωLO)t + sin(ωRF – ωLO)t
– 1/3 sin(ωRF + 3ωLO)t – 1/3 sin(ωRF – 3ωLO)t
+ 1/5 sin(ωRF + 5ωLO)t + 1/5 sin(ωRF – 5 ωLO)t – . . . }
Eq. 4
or simply
SIF = 2/π { sin(ωRF + ωLO)t + sin(ωRF – ωLO)t + harmonics }
Eq. 5
The most important of these harmonic components are sketched in Figure 3.44 for
the particular case used to generate the waveform shown in Figure 3.43, that is, fRF
= 11MHz and fLO = 10MHz. Because of the 2/π term, a mixer has a minimum 3.92
dB insertion loss (and noise figure) in the absence of any gain.
40
OUTPUT SPECTRUM FOR SWITCHING MIXER
FOR fRF = 11MHz AND fLO = 10MHz
0.8
LINEAR
AMPLITUDE
0.6
0.637 = -3.9dB
0.212 = -13.5dB
0.127 = -17.9dB
0.090 = -20.9dB
0.637
0.637
WANTED IF
AT 1MHz
SUM AT
21MHz
0.4
0.212
0.2
0.212
0.127
0.127
0.09
0
0
10
20
30
40
50
60
FREQUENCY (MHz)
a
3.44
Note that the ideal (switching) mixer has exactly the same problem of image
response to ωLO – ωRF as the linear multiplying mixer. The image response is
somewhat subtle, as it does not immediately show up in the output spectrum: it is a
latent response, awaiting the occurrence of the "wrong" frequency in the input
spectrum.
Diode-Ring Mixer
For many years, the most common mixer topology for high-performance applications
has been the diode-ring mixer, one form of which is shown in Figure 3.45. The
diodes, which may be silicon junction, silicon Schottky-barrier or gallium-arsenide
types, provide the essential switching action. We do not need to analyze this circuit
in great detail, but note in passing that the LO drive needs to be quite high—often
a substantial fraction of one watt—in order to ensure that the diode conduction is
strong enough to achieve low noise and to allow large signals to be converted
without excessive spurious nonlinearity.
Because of the highly nonlinear nature of the diodes, the impedances at the three
ports are poorly controlled, making matching difficult. Furthermore, there is
considerable coupling between the three ports; this, and the high power needed at
the LO port, make it very likely that there will be some component of the (highlydistorted) LO signal coupled back toward the antenna. Finally, it will be apparent
that a passive mixer such as this cannot provide conversion gain; in the idealized
scenario, there will be a conversion loss of 2/π [as Eq. 4 shows], or 3.92dB. A
practical mixer will have higher losses, due to the resistances of the diodes and the
losses in the transformers.
41
DIODE-RING MIXER
RF
IN
LO
IN
IF
OUT
a
3.45
Users of this type of mixer are accustomed to judging the signal handling
capabilities by a “Level” rating. Thus, a Level-17 mixer needs +17dBm (50mW) of
LO drive and can handle an RF input as high as +10dBm (±1V). A typical mixer in
this class would be the Mini-Circuits LRMS-1H, covering 2-500MHz, having a
nominal insertion loss of 6.25dB (8.5dB max), a worst-case LO-RF isolation of 20dB
and a worst-case LO-IF isolation of 22dB (these figures for an LO frequency of 250500MHz). The price of this component is approximately $10.00 in small quantities.
Even the most expensive diode-ring mixers have similar drive power requirements,
high losses and high coupling from the LO port.
FET Mixers
A modern alternative to the diode-ring mixer is one in which the diodes are replaced
by FETs. The idea here is to reduce the distortion caused by the inherent
nonlinearities of junction diodes, whose incremental resistance varies with the
instantaneous signal current. To reduce this effect, the diodes are often driven to
very high current levels. Indeed, some users of diode-ring mixers push them to
extremes, operating at current levels close to those which will cause the diodes to
fail by over-dissipation. Thus, in commenting about a certain minor variation to the
diode-ring-mixer, we read:
“This helps the mixer to accept higher LO power without burning out the diodes!”
(From Wes Hayward, Solid State Design for the Radio Amateur, ARRL, 1986,
Chapter 6, p.120)
To avoid “burning out the diodes”, some mixers use two or four J-FETs in an
analogous way to that shown in Figure 3.45. The idea is that the channel resistance
42
of a large FET driven into its triode region of conduction can be as low as the
dynamic resistance of a diode, thus achieving similar conversion gain and noise
levels. But this low resistance arises without any current flow in the channel and it
is also more linear than that of the diodes when signal current does flow, thus
resulting in lower intermodulation, and hence a larger overall dynamic range. MOSFETs can also be used in a similar way.
This style of FET-based mixers is very attractive for many high-performance
applications. However, since the active devices are still used only as switches, they
do not provide power gain, and have typical insertion losses of 6 to 8dB.
Furthermore, the balance of these mixers is still critically dependent on such things
as transistor matching and transformer winding accuracy, large LO drives (volts)
are needed, and the overall matching requirements continue to be difficult to
achieve over the full frequency range. Finally, of course, they are not directly
amenable to monolithic integration.
Another popular circuit, widely used in many inexpensive receivers, is the dual-gate
MOS-FET mixer. In this type of mixer, the RF signal is applied to one gate of the
FET and the LO signal to the second gate. The multiplication process is not very
well-defined, but in general terms relies on the fact that both the first and second
gates influence the current in the channel. The structure can be modeled as two
FETs, where the drain of the lower FET (having the RF input applied to it) is
intimately connected to the source of the upper FET (having the LO input on its
gate). The lower FET operates in its triode region, and thus exhibits a gm that is a
function of its drain voltage, controlled by the LO. Though not readily modeled to
great accuracy, this mixer, like many others, can be pragmatically optimized to
achieve useful performance, though not without the support of many associated
passive components for biasing and matching.
Classic Active Mixer
The diode-ring mixer not only has certain performance limitations, but it is also not
amenable to fabrication using integrated circuit technologies, at least in the form
shown in Figure 3.45. In the mid ’sixties it was realized that the four diodes could be
replaced by four transistors to perform essentially the same switching function. This
formed the basis of the now-classical bipolar circuit shown in Figure 3.46, which is a
minimal configuration for the fully-balanced version. Millions of such mixers have
been made, including variants in CMOS and GaAs. We will limit our discussion to
the BJT form, an example of which is the Motorola MC1496, which, although quite
rudimentary in structure, has been a mainstay in semi-discrete receiver designs for
about 25 years.
43
CLASSIC ACTIVE MIXER
IF OUTPUT
Q3
Q4
Q6
Q5
LO
INPUT
Q1
Q2
RF
INPUT
IEE
a
3.46
The active mixer is attractive for the following reasons:
• It can be monolithically integrated with other signal processing circuitry.
• It can provide conversion gain, whereas a diode-ring mixer always has an insertion
loss. (Note: Active mixers may have gain. The analog Devices' AD831 active mixer,
for example, amplifies the result in Eq. 5 by π/2 to provide unity gain from RF to IF.)
• It requires much less power to drive the LO port.
• It provides excellent isolation between the signal ports.
• Is far less sensitive to load-matching, requiring neither diplexer nor broadband
termination.
Using appropriate design techniques it can provide trade-offs between third-order
intercept (3OI or IP3) and the 1dB gain-compression point (P1dB), on the one hand,
and total power consumption (PD) on the other. (That is, including the LO power,
which in a passive mixer is "hidden" in the drive circuitry.)
Basic Operation of the Active Mixer
Unlike the diode-ring mixer, which performs the polarity-reversing switching
function in the voltage domain, the active mixer performs the switching function in
the current domain. Thus the active mixer core (transistors Q3 through Q6 in Figure
3.46) must be driven by current-mode signals. The voltage-to-current converter
formed by Q1 and Q2 receives the voltage-mode RF signal at their base terminals
and transforms it into a differential pair of currents at the their collectors.
44
A second point of difference between the active mixer and diode ring mixer,
therefore, is that the active mixer responds only to magnitude of the input voltage,
not to the input power; that is, the active mixer is not matched to the source. (The
concept of matching is that both the current and the voltage at some port are used
by the circuitry which forms that port). By altering the bias current, IEE, the
transconductance of the input pair Q1-Q2 can be set over a wide range. Using this
capability, an active mixer can provide variable gain.
A third point of difference is that the output (at the collectors of Q3–Q6) is in the
form of a current, and can be converted back to a voltage at some other impedance
level to that used at the input, hence, can provide further gain. By combining both
output currents (typically, using a transformer) this voltage gain can be doubled.
Finally, it will be apparent that the isolation between the various ports, in
particular, from the LO port to the RF port, is inherently much lower than can be
achieved in the diode ring mixer, due to the reversed-biased junctions that exist
between the ports.
Briefly stated, though, the operation is as follows. In the absence of any voltage
difference between the bases of Q1 and Q2, the collector currents of these two
transistors are essentially equal. Thus, a voltage applied to the LO input results in
no change of output current. Should a small DC offset voltage be present at the RF
input (due typically to mismatch in the emitter areas of Q1 and Q2), this will only
result in a small feedthrough of the LO signal to the IF output, which will be
blocked by the first IF filter.
Conversely, if an RF signal is applied to the RF port, but no voltage difference is
applied to the LO input, the output currents will again be balanced. A small offset
voltage (due now to emitter mismatches in Q3–Q6) may cause some RF signal
feedthrough to the IF output; as before, this will be rejected by the IF filters. It is
only when a signal is applied to both the RF and LO ports that a signal appears at
the output; hence, the term doubly-balanced mixer.
Active mixers can realize their gain in one other way: the matching networks used
to transform a 50Ω source to the (usually) high input impedance of the mixer
provides an impedance transformation and thus voltage gain due to the impedance
step up. Thus, an active mixer that has loss when the input is terminated in a
broadband 50Ω termination can have “gain” when an input matching network is
used.
The AD831, 500MHz, Low Distortion Active Mixer
The AD831 is a low distortion, wide dynamic range, monolithic mixer for use in such
applications as RF to IF down conversion in HF and VHF receivers, the second
mixer in digital mobile radio base stations, direct-to-baseband conversion,
quadrature modulation and demodulation, and doppler-frequency shift detection in
ultrasound imaging applications. The mixer includes a local oscillator driver and a
low-noise output amplifier. The AD831 provides a +24dBm third-order intercept
point for –10dBm local oscillator power, thus improving system performance and
reducing system cost, compared to passive mixers, by eliminating the need for a high
power local oscillator driver and its associated shielding and isolation problems. A
45
simplified block diagram of the AD831 is shown in Figure 3.47, and key
specifications in Figure 3.48.
AD831 500MHz LOW DISTORTION ACTIVE MIXER
+5V
0.1µF
CF
82pF
CF
82pF
2
3
20
1
IFN
AN
19
IFP
VP
50 Ω
4
AP
50 Ω
COM
GND
18
51.1Ω
0.1µF
VFB
5 VN
C1
C2
-5V
RF
INPUT
RFP
17
110Ω
OUT
6
RT
IF
OUTPUT
BPF
16
RT
L1
7
VN
AD831
Top View
RFN
15
-5V
0.1µF
0.1µF
8
VN
-5V
BIAS
VP
9
LON
10
LOP
11
GND
13
0.1µF
51.1Ω
+5V
VP
12
14 NC
0.1µF
+5V
LO INPUT
-10 dBm
a
3.47
AD831 ACTIVE MIXER KEY SPECIFICATIONS
n
Doubly-Balanced Mixer, 10dB Noise Figure
n
Low Distortion (IF = 10.7MHz, RF to 200MHz):
u
+24dBm Third Order Intercept
u
+10dBm 1dB Compression Point
n
Low LO Drive Required: –10dBm
n
Bandwidth:
u
500MHz RF and LO Input Bandwidths
u
250MHz Differential Current IF Output
u
DC to > 200MHz Single-Ended Voltage IF Output
a
3.48
46
Noise Figure
Noise Figure (NF) is a figure of merit used to determine how a device degrades the
signal-to-noise ratio of its input. Note: in RF systems, the impedance is 50Ω unless
otherwise stated. Mathematically, noise figure is defined as:
NF = 20 log10
SI / N I
,
SO / N O
where SI/NI is the input signal-to-noise ratio, and SO/NO is the output signal-tonoise ratio.
Typical noise figures for passive mixers with post amplifiers are 12 to 15dB. The NF
of the AD831 is 10dB with a matched input, which is adequate for applications in
which there is gain in front of the mixer.
Noise Figure is used in a "cascaded noise figure calculation", which gives the overall
noise figure of a receiver. Basically, the noise figure of each stage is converted into a
noise factor (F = antilog NF/10) and plugged into a spreadsheet containing the Friis
Equation:
N
F − 1 F3 − 1
FK − 1
FRECEIVER = F1 + 2
+
+ ∑
,
K −1
G1
G1G 2
K =4
∏ GJ
J =1
where FN and GN are the noise factor and gain, respectively, of the Nth stage in the
receiver.
For a passive diode-ring mixer, the noise figure is the same as the insertion loss. For
an active mixer, however, noise is added to the signal by the active devices in the
signal path. The difference between the noise figure of a matched active mixer and
an unmatched active mixer can be several dB due to the “voltage gain” of the
impedance-matching network, which acts as a “noiseless” preamplifier (Figure 3.49).
In the case of the AD831, the noise figure for the matched circuit is 10 dB (at
70MHz) and the unmatched circuit with its input terminated with a 50Ω resistor is
16dB.
The noise figure is 11.7dB at 220 MHz using the external matching network shown
in Figure 3.49. The values shown are for 220 MHz and provide 10 dB of voltage
gain.
47
AD831 ACTIVE MIXER WITH 220MHz
EXTERNAL MATCHING NETWORK
AD831
C1
6
RF
INPUT
RFP
L1
C2
L2
7
RFN
L1: 100nH, COILCRAFT 1008CS-101
L2: 56nH, COILCRAFT 1008CS-560
C1, C2: 2-10pF CERAMIC VARIABLE
a
3.49
Intermodulation Distortion
Even before the ‘mixing’ process in the core, the entire signal spectrum co-exists
within the RF input stage. This part of the mixer is inevitably nonlinear, to a
greater or lesser extent, and, with or without the LO input operative, generates a
very large number of intermodulation products.
Thus, the key objectives in the design of a high-performance active mixer are to
achieve a very linear RF input section, followed by a near-ideal polarity-switching
stage, followed by a very linear IF output amplifier (if used) prior to the first filter.
1dB Compression Point and Third-Order Intercept Point
For a single-sinusoid input to a system, a point will be reached as the input
amplitude is increased at which the apparent gain becomes 1dB lower than that
observed at lower input amplitudes. This is called the 1dB gain compression level,
which we’ll abbreviate P1dB, and is usually quoted in dBm, or ‘decibels above 1mW,
that is, it is expressed as a power measurement. When using an active mixer with
an input matching network, the gain of the input matching network must be taken
into account when defining the system in terms of an active mixer’s 1dB
compression point, since the impedance transformation of the network increases the
input voltage to mixer.
Another metric used in characterizing mixers is the third-order intercept, known as
P3OI or IP3. If two tones of frequency f1 and f2 (representing two adjacent channels
in a communications system, for example) are applied to a non-linear system, there
will be a large number of intermodulation products generated. The third-order
48
distortion products which fall at 2f2-f1 and 2f1-f2 are particularly troublesome,
because they are close to the original frequencies (see Figure 3.50). If the two tones
represent true signals, then the third-order IMD products can interfere with signals
in the adjacent channels.
THIRD-ORDER INTERMODULATION DISTORTION
f1
f2
2f1 - f2
2f2 - f1
FREQUENCY
a
3.50
Rather than measuring the third-order distortion products for a variety of signal
amplitudes, the concept of third-order intercept can be used to extract the IMD
information and is often used as a figure of merit for mixers and amplifiers in RF
applications.
A plot (Figure 3.51) of the power levels at the output of the system for the
fundamental of the output frequency and for its third harmonic, plotted versus the
input power, will generally yield a pair of straight lines which eventually intersect
(at the 3rd order intercept point, IP3).
49
THIRD-ORDER INTERCEPT USING DATA FOR AD831
+24
+10
3RD ORDER INTERCEPT POINT
IP3
1 dB COMPRESSION POINT
1 dB
-10
IF = 10.7 MHz
RF = 100 MHz
LINEAR
OUTPUT
OUTPUT
POWER
(dBm)
3RD ORDER IMD
AT 2f1 - f2 AND 2f2 - f1
-78
INPUT POWER (dBm)
a
3.51
The problem with this metric is that it has meaning only for certain simple cases. In
particular, the 3rd harmonic is assumed to increase at three times the rate of the
fundamental. The appeal of P3OI lies in the fact that it is easily measured, or at
least, it is easy to obtain measurements. (The measurements are not hard to make,
but it will be found that the apparent P3OI is signal-dependent). Apply a low level
signal, at some known level PO (in dBm, see Figure 3.51), measure the output power
at the fundamental, P1 (in relative terms, dBc) and at the third harmonic, P3 (also in
dBc) and from simple geometry calculate
P3OI = PO+ 1/2 (P1–P3)
Eq. 6
The non-linearity in some classical circuits, such as the diode-ring mixer,
approximates a cubic function, and the above relationship holds, but in practice, the
P3OI can be quite misleading, for several reasons. First, other circuits may not, in
general, exhibit this type of non-linearity. This type of behavior could easily lead to
apparent third-order intercept values which were impressively high (theoretically
infinite, if ‘measured’ using signals of less than the critical amplitude).
A spur chart is a compilation of the nf1 ± mf2 products that result from the mixing
process. The spur chart is useful because it allows an engineer developing a
frequency plan for a radio to identify possible problems due to spurious signals
created in the mixer. However, the spur chart is also tedious to create; for n = m = 7,
a chart requires 112 measurements.
The compilation of results is the spur chart (also called a “mixer table”). Details of
making the spur chart measurements and results are given in the AD831 data sheet
(see Reference 17).
50
Mixer Summary
Mixers are a special kind of analog multiplier optimized for use in frequency
translation, having one linear input (that associated with the RF signal) and a
second (that associated with the LO input) which alternates the phase of the first
input by 0/180°. In integrating complete receivers in monolithic form, certain basic
circuit forms have proven useful. So far, we have considered a classic form, a sixtransistor circuit exemplified by the AD831. Compared to a diode-ring mixer, this
circuit has several advantages, including much better isolation between ports, the
ability to provide conversion gain (which may also be variable), the need for much
lower LO drive levels, and the elimination of special matching networks.
Often cited as a disadvantage of the active mixer is it’s poorer dynamic range: we
have just begun to examine what defines this, beginning with a consideration of the
linearity of the RF port, traditionally characterized by the 1dB gain-compression
input power, P1dB, and the third-order intercept, P3OI. The second of these measures
was shown to be meaningful only if the nonlinearity is essentially cubic in form,
which may not always be true. In passing, we pointed out that while inputs and
outputs are invariably characterized in terms of a power level of so-many-dBm,
active mixers respond to instantaneous signal voltages at their inputs, which are
usually not matched to their source, which can be confusing at times.
Now that we have examined each of the fundamental receiver building blocks, we
are ready to look at receiver subsystems.
51
RECEIVER SUBSYSTEMS
Bob Clarke, Walt Kester
In order to design a communications receiver, a clear understanding of the
modulation technique is essential. There are many types of modulation, ranging
from simple amplitude modulation (AM), phase modulation (PM), and frequency
modulation (FM) to multi-level quadrature-amplitude-modulation (QAM) where
both amplitude and phase are modulated. Most modern modulation schemes make
use of both signal amplitude and phase information. A complex signal can thus be
represented in two ways as shown in the diagrams in Figure 3.52. The left-hand
diagram represents the signal in rectangular coordinates as an inphase (I) and
quadrature (Q) signal of the form:
S(t) = I(t) + jQ(t).
The right hand diagram represents the same signal expressed in polar coordinates:
S(t) = A(t)ejØ(t).
The conversions between the two coordinate systems are:
S(t) = A(t)ejØ(t) = I(t) + jQ(t), where
A(t) =
I( t ) 2 + Q( t ) 2 ,
 Q( t) 
Ø(t) = arctan 
.
 I( t) 
52
RECTANGULAR AND POLAR REPRESENTATIONS OF
AMPLITUDE AND PHASE MODULATED SIGNAL
Q(t)
Q(t)
jφ
φ(t)
S(t) = A(t)e
S(t) = I(t) + jQ(t)
A(t)
φ(t)
I(t)
I(t)
RECTANGULAR
POLAR
a
3.52
Note that the signals are identical, only their representation is different.
In the case of the I/Q (rectangular) representation, a linear IF strip is required.
Variable gain is required because of the wide dynamic range, and amplitude and
phase information must be preserved. This type of IF strip often incorporates an I/Q
demodulator whose outputs drive baseband ADCs followed by a DSP. Linear IF
amplifiers are used in these systems.
For the case of the polar representation, the signal amplitude is derived from the
RSSI (log) output of a log/limiting amplifier and the phase information from the
limited output. This type of IF strip operates at a high fixed gain, retains the phase
information in the limited output, and often incorporates a phase demodulator.
In order to handle these two fundamental representations of modulation, ADI has
developed two IF subsystems, the AD607 and the AD608. These are used in such
applications as PHS, PCN, DECT, CT2, and GSM where the modulation mode is
some form of phase-shift keying (PSK).
The chose of demodulation technique depends on the receiver architecture. The
standard architecture in GSM and PHS uses a rectangular representation of the
signal, that is S(t) = I(t) + jQ(t) and requires a linear IF amplifier stage such as that
in the AD607. In this architecture, a baseband converter consisting of two signal
inputs; each with individual low-pass filters, digitizes the I(t) and Q(t) outputs of the
IF IC's quadrature demodulator. Further demodulation is performed digitally using
a DSP. An equalizer in the DSP then determines the correct manual gain control
(MGC) voltage (or digital signal) to change the IF gain to center the signal in the
dynamic range of the baseband ADCs. The equalizer calculates the RSSI value as
part of this process (see Figure 3.53).
53
RECEIVER BASED ON AD607 SUBSYSTEM
USING INPHASE/QUADRATURE MODULATION
900MHz
10.7MHz
LO2
BPF OR
LPF
I
LPF
LO1
0º
90dB RANGE
BASEBAND
ADCs AND
DSP
QVCO
90º
240MHz
AD607
LPF
Q
MGC
a
3.53
A detailed block diagram of the AD607 Mixer/AGC/RSSI 3V receiver IF subsystem is
shown in Figure 3.54. The RF input frequency can be as high as 500MHz, and the
IF frequency from 400kHz to 12MHz. It consists of a mixer, linear IF amplifiers, I
and Q demodulators, a phase-locked quadrature oscillator, AGC detector, and a
biasing system with external power-down. Total power on +3V is 25mW.
AD607 FUNCTIONAL BLOCK DIAGRAM
LOIP
RFHI
VMID
IFHI
MXOP
IOUT
BPF
IFOP
RFLO
BPF OR
LPF
FDIN
DMIP
VQFO
FLTR
VMID
IFLO
QOUT
MID-POINT
BIAS
GENERATOR
AGC
DETECTOR
GAIN /RSSI
VPS1
VPS2
PTAT
VOLTAGE
BIAS
GENERATOR
AD607
GREF
PRUP
COM1
COM2
a
3.54
54
The AD607's low noise, high intercept mixer is a doubly-balanced Gilbert cell type.
It has a nominal –15dBm input-referred 1dB compression point and a –8dBm inputreferred third-order intercept. The mixer section also includes a local oscillator
preamplifier, which lowers the required external LO drive to –16dBm.
The variable-gain mixer and the linear four-stage IF amplifier strip together provide
a voltage controlled gain range of more than 90dB. The I and Q demodulators, each
consisting of a multiplier followed by a 2-pole, 2MHz low-pass filter, are driven by a
phase-locked loop providing inphase and quadrature clocks. An internal AGC
detector is included, and the temperature stable gain control system provides an
accurate RSSI capability.
The I and Q demodulators provide inphase and quadrature baseband outputs to
interface with Analog Devices' AD7013 (IS54/IS136, TETRA, MSAT) and AD7015
(GSM) baseband converters.
Key specifications for the AD607 are summarized in Figure 3.55.
AD607 MIXER / AGC / RSSI 3V RECEIVER KEY FEATURES
n
Mixer:
u
u
u
u
u
–15dBm Input 1dB Compression Point
–8dBm Input Third Order Intercept Point
RF/LO Inputs to 500MHz
12dB Noise Figure, Matched Input
–16dBm LO Drive
n
Linear IF Amplifier:
u
45MHz Bandwidth
u
Linear-in-dB Gain Control Over 90dB Gain Range
u
–15dBm Input 1dB Compression Point
u
+18dBm Output Third Order Intercept Point
n
In-Phase and Quadrature Demodulators:
u
1.5MHz Output Bandwidth
u
Compatible with Baseband Converters (AD7013, AD7015)
n
25mW Total Power @ Single +3V Supply
a
3.55
For cases where the signal is represented in polar form, the AD608 is the proper
choice. The AD608 Mixer/Limiter/RSSI 3V Receiver IF Subsystem consists of a
mixer followed by a logarithmic amplifier; the logarithmic amplifier has both limited
output (phase information) and an RSSI output (amplitude information). This
architecture is useful in polar demodulation applications as shown in Figure 3.56.
A block diagram of the AD608 is shown in Figure 3.57, and key specifications in
Figure 3.58.
55
RECEIVER BASED ON AD608 SUBSYSTEM
USING POLAR DEMODULATION
10.7MHz
LO2
900MHz
RSSI
LPF
DETECTORS
LO1
DEMOD
AND
DSP
80dB RANGE
LIMITING
AMPLIFIERS
240MHz
PHASE
AD608
3.56
AD608 FUNCTIONAL BLOCK DIAGRAM
24dB MIXER GAIN
RFHI
5
MIXER
MXOP
≈
7
BPF
DRIVER
6
VMID
LO
PREAMP
MID-SUPPLY
IF BIAS
10.7MHz
BANDPASS
FILTER
330Ω
COM2
VPS1 COM1
1
+2.7V TO
5.5V
2
3
LO INPUT
-16dBm
4
7 FULL-WAVE
RECTIFIER CELLS
11
9
330Ω
10nF
IFLO
100Ω
COM3 12
VPS2
14 +2.7V TO 5.5V
15
FINAL
LIMITER
10
RSSI OUTPUT
20mV/dB
0.2V TO 1.8V
LMOP
5-STAGE IF AMPLIFIER
(16dB PER STAGE)
8
100nF
RSSI
≈
2MHz
LPF
IFHI
LIMITER
OUTPUT
400mVp-p
13
LOHI
BIAS
100dB LIMITER GAIN
90dB RSSI
IF INPUT
-75dBm TO
+15dBm 2
±6mA M AX OUTPUT
(±890mV INTO 165Ω)
RF INPUT
-95 TO
-15dBm 1
RFLO
3dB NOM INAL
INSERTION LOSS
18nF
FDBK
AD608
±50µA
PRUP
16
NOTES:
CM OS LOGIC
INPUT
1 -15dBm = ±56mV MAX FOR LINEAR OPERATION
2 39.76µV RMS TO 397.6mV RMS FOR ±1dB RSSI ACCURACY
a
3.57
AD608 MIXER / LIMITER / RSSI 3V RECEIVER KEY FEATURES
n
Mixer:
u
–15dBm Input 1dB Compression Point
56
u
u
u
u
–5dBm Input Third Order Intercept Point
RF/LO Inputs to 500MHz
12dB Noise Figure, Matched Input
–16dBm LO Drive
n
Logarithmic Amplifier / Limiter:
u
100dB Limiter Gain, 90dB RSSI
u
±1dB Log Linearity
u
±3° Phase Variation, –75dBm to +5dBm IF @ 10.7MHz
n
21mW Total Power @ Single +3V Supply
a
3.58
The log amp both measures the level of the signal (like the AD641 and AD606) and
limits the signal. The RSSI or Received Signal Strength Indicator output is
proportional to the log of the input signal. As a limiting amplifier, the AD608
removes any amplitude changes in the signal and keeps only the phase or frequency
changes. These phase or frequency changes are proportional to the modulating
signal and contain the intelligence in the signal. The AD608's limiting amplifier is a
5-stage log amp with more than 80dB of dynamic range.
In a typical mobile phone application, the RF signal (typically 900MHz or 1800MHz)
is mixed down to the first IF (typically 240MHz), is filtered, and enters the AD608,
where it is mixed down to a second IF at 10.7MHz, where it is amplified, limited,
and measured. The limited output is demodulated by an external frequency or phase
demodulator. The RSSI output is digitized by an ADC and used for active power
control in the phone system.
As a practical note, the cutoff frequency of the log amp's internal low pass filter
depends on what range of frequencies the log amp was designed for. In analog
cellular systems, where the modulation mode is narrow-band FM, the IF is typically
450kHz. The low pass filters in the IF ICs designed for these standards have a fairly
low cutoff frequency, and the filter's voltage output response provides a "slow" RSSI.
In GSM (Global System for Mobile Communications) and PHS (Personal Handy
System) applications, the IF is typically 10.7MHz or higher, and the filter's voltage
output response provides a "fast" RSSI. The cutoff frequency of the low pass filter in
the AD608 is 2MHz.
57
REFERENCES
1.
Barrie Gilbert, ISSCC Digest of Technical Papers 1968, pp. 114-115
February 16, 1968.
2.
Barrie Gilbert, Journal of Solid State Circuits, Vol. SC-3, December 1968,
pp. 353-372.
3.
C.L. Ruthroff, Some Broadband Transformers, Proc. I.R.E., Vol.47,
August, 1959, pp.1337-1342.
4.
James M. Bryant, Mixers for High Performance Radio, Wescon 1981:
Session 24 (Published by Electronic Conventions, Inc., Sepulveda Blvd.,
El Segundo, CA)
5.
P.E. Chadwick, High Performance IC Mixers, IERE Conference on Radio
Receivers and Associated Systems, Leeds, 1981, IERE Conference
Publication No. 50.
6.
P.E. Chadwick, Phase Noise, Intermodulation, and Dynamic Range,
RF Expo, Anaheim, CA, January, 1986.
7.
Daniel H. Sheingold, Editor, Nonlinear Circuits Handbook, Analog
Devices, Inc., l974.
8.
Richard Smith Hughes, Logarithmic Amplifiers, Artech House, Inc.,
Dedham, MA., 1986.
9.
William L. Barber and Edmund R. Brown, A True Logarithmic Amplifier for
Radar IF Applications, IEEE Journal of Solid State Circuits, Vol. SC-15,
No. 3, June, 1980, pp. 291-295.
10.
Broadband Amplifier Applications, Plessey Co. Publication P.S. 1938,
September, 1984.
11.
M. S. Gay, SL521 Application Note, Plessey Co., 1966.
12.
Amplifier Applications Guide, Analog Devices, Inc., 1992. Section 9.
13.
Charles Kitchen and Lew Counts, RMS-to-DC Conversion Application
Guide, Second Edition, Analog Devices, Inc., 1986.
14.
Barrie Gilbert, A Low Noise Wideband Variable-Gain Amplifier Using
an Interpolated Ladder Attenuator, IEEE ISSCC Technical Digest, 1991,
pp. 280, 281, 330.
15.
Barrie Gilbert, A Monolithic Microsystem for Analog Synthesis of
Trigonometric Functions and their Inverses, IEEE Journal of Solid
State Circuits, Vol. SC-17, No. 6, December, 1982, pp. 1179-1191.
16.
Linear Design Seminar, Analog Devices, 1995, Section 3.
58
17.
AD831 Data Sheet, Rev. B, Analog Devices.
59
SECTION 4
HIGH SPEED SAMPLING AND
HIGH SPEED ADCs, Walt Kester
INTRODUCTION
High speed ADCs are used in a wide variety of real-time DSP signal-processing
applications, replacing systems that used analog techniques alone. The major reason
for using digital signal processing are (1) the cost of DSP processors has gone down,
(2) their speed and computational power has increased, and (3) they are
reprogrammable, thereby allowing for system performance upgrades without
hardware changes. DSP offers solutions that cannot be achieved in the analog
domain, i.e. V.32 and V.34 modems.
However, in order for digital signal processing techniques to be effective in solving
an analog signal processing problem, appropriate cost effective high speed ADCs
must be available. The ADCs must be tested and specified in such a way that the
design engineer can relate the ADC performance to specific system requirements,
which can be more demanding than if they were used in purely analog signal
processing systems. In most high speed signal processing applications, AC
performance and wide dynamic range are much more important than traditional DC
performance. This requires that the ADC manufacturer not only design the right
ADCs but specify them as completely as possible to cover a wide variety of
applications.
Another important aspect of integrating ADCs into a high speed system is a
complete understanding of the sampling process and the distortion mechanisms
which ultimately limit system performance. High speed sampling ADCs first were
used in instrumentation and signal processing applications, where much emphasis
was placed on time-domain performance. While this is still important, applications
of ADCs in communications also require comprehensive frequency-domain
specifications.
Modern IC processes also allow the integration of more analog functionality into the
ADC, such as on-board references, sample-and-hold amplifiers, PGAs, etc. This
makes them easier to use in a system by minimizing the amount of support circuitry
required.
Another driving force in high speed ADC development is the trend toward lower
power and lower supply voltages. Most high speed sampling ADCs today operate on
either dual or single 5V supplies, and there is increasing interest in single-supply
converters which will operate on 3V or less for battery powered applications. Lower
supply voltages tend to increase a circuit's sensitivity to power supply noise and
ground noise, especially mixed-signal devices such as ADCs and DACs.
The trend toward lower cost and lower power has led to the development of a variety
of high speed ADCs fabricated on standard 0.6 micron CMOS processes. Making a
precision ADC on a digital process (no thin film resistors are available) is a real
challenge to the IC circuit designer. ADCs which require the maximum in
1
performance still require a high speed complementary bipolar process (such as
Analog Devices' XFCB) with thin film resistors.
The purpose of this section is to equip the engineer with the proper tools necessary
to understand and select ADCs for high speed systems applications. Making
intelligent tradeoffs in the system design requires a thorough understanding of the
fundamental capabilities and limitations of state-of-the-art high speed sampling
ADCs.
HIGH SPEED SAMPLING ADCs
n
Wide Acceptance in Signal Processing and Communications
n
Emphasis on Dynamic Performance
n
Trend to Low Power, Low Voltage, Single-Supply
n
More On-Chip Functionality: PGAs, SHA, Digital Filters, etc.
n
Process Technology:
u
Low Cost CMOS: Up to 12-bits @ 10MSPS
u
High Speed Complementary Bipolar: Up to 12-bits @ 70MSPS
u
Statistical Matching Techniques Rather than Thin Film
Laser Trimming
a
4.1
FUNDAMENTALS OF HIGH SPEED SAMPLING
The sampling process can be discussed from either the frequency or time domain or
both. Frequency-domain analysis is applicable to communications, so that's what we
will consider.
First consider the case of a single frequency sinewave of frequency fa sampled at a
frequency fs by an ideal impulse sampler (see top diagram in Figure 4.2). Also
assume that fs > 2fa as shown. The frequency-domain output of the sampler shows
aliases or images of the original signal around every multiple of fs, i.e. at frequencies
equal to
|± Kfs ± fa|, K = 1, 2, 3, 4, .....
2
ANALOG SIGNAL fa SAMPLED @ fs USING IDEAL SAMPLER
HAS IMAGES (ALIASES) AT |±Kfs ±fa|, K = 1, 2, 3, . . .
fa
I
I
fs
0.5fs
ZONE 3
fa
I
0.5fs
2fs
1.5fs
ZONE 2
ZONE 1
I
I
ZONE 4
I
1.5fs
fs
a
I
I
2fs
4.2
The Nyquist bandwidth is defined to be the frequency spectrum from DC to fs/2. The
frequency spectrum is divided into an infinite number of Nyquist zones, each having
a width equal to 0.5fs as shown. In practice, the ideal sampler is replaced by an
ADC followed by an FFT processor. The FFT processor only provides an output from
DC to fs/2, i.e., the signals or aliases which appear in the first Nyquist zone.
Now consider the case of a signal which is outside the first Nyquist zone (Figure 4.2,
bottom diagram) Notice that even though the signal is outside the first Nyquist
zone, its image (or alias), fs–fa, falls inside. Returning to Figure 4.2, top diagram, it
is clear that if an unwanted signal appears at any of the image frequencies of fa, it
will also occur at fa, thereby producing a spurious frequency component in the first
Nyquist zone. This is similar to the analog mixing process and implies that some
filtering ahead of the sampler (or ADC) is required to remove frequency components
which are outside the Nyquist bandwidth, but whose aliased components fall inside
it. The filter performance will depend on how close the out-of-band signal is to fs/2
and the amount of attenuation required.
BASEBAND ANTIALIASING FILTERS
Baseband sampling implies that the signal to be sampled lies in the first Nyquist
zone. It is important to note that with no input filtering at the input of the ideal
sampler, any frequency component (either signal or noise) that falls outside the
Nyquist bandwidth in any Nyquist zone will be aliased back into the first Nyquist
zone. For this reason, an antialiasing filter is used in almost all sampling ADC
applications to remove these unwanted signals.
3
Properly specifying the antialiasing filter is important. The first step is to know the
characteristics of the signal being sampled. Assume that the highest frequency of
interest is fa. The antialiasing filter passes signals from DC to fa while attenuating
signals above fa.
Assume that the corner frequency of the filter is chosen to be equal to fa. The effect
of the finite transition from minimum to maximum attenuation on system dynamic
range is illustrated in Figure 4.3.
EFFECTS OF ANTIALIASING FILTER
ON SYSTEM DYNAMIC RANGE
fs - fa
fa
DR
fa
fs - fa
fs
fs
f
2
STOPBAND ATTENUATION = DR
FILTER
SPECIFICATIONS
a
TRANSITION BAND: fa TO fs - fa
CORNER FREQUENCY: fa
4.3
Assume that the input signal has fullscale components well above the maximum
frequency of interest, fa. The diagram shows how fullscale frequency components
above fs – fa are aliased back into the bandwidth DC to fa. These aliased
components are indistinguishable from actual signals and therefore limit the
dynamic range to the value on the diagram which is shown as DR.
Some texts recommend specifying the antialiasing filter with respect to the Nyquist
frequency, fs/2, but this assumes that the signal bandwidth of interest extends from
DC to fs/2 which is rarely the case. In the example shown in Figure 4.3, the aliased
components between fa and fs/2 are not of interest and do not limit the dynamic
range.
The antialiasing filter transition band is therefore determined by the corner
frequency fa, the stopband frequency fs – fa, and the stopband attenuation, DR. The
required system dynamic range is chosen based on our requirement for signal
fidelity.
4
Filters have to become more complex as the transition band becomes sharper, all
other things being equal. For instance, a Butterworth filter gives 6dB attenuation
per octave for each filter pole. Achieving 60dB attenuation in a transition region
between 1MHz and 2MHz (1 octave) requires a minimum of 10 poles, not a trivial
filter, and definitely a design challenge.
Therefore, other filter types are generally more suited to high speed applications
where the requirement is for a sharp transition band and in-band flatness coupled
with linear phase response. Elliptic filters meet these criteria and are a popular
choice.
There are a number of companies which specialize in supplying custom analog
filters. TTE is an example of such a company (Reference 1). As an example, the
normalized response of the TTE, Inc., LE1182 11-pole elliptic antialiasing filter is
shown in Figure 4.4. Notice that this filter is specified to achieve at least 80dB
attenuation between fc and 1.2fc. The corresponding passband ripple, return loss,
delay, and phase response are also shown in Figure 4.4. This custom filter is
available in corner frequencies up to 100MHz and in a choice of PC board, BNC, or
SMA with compatible packages.
CHARACTERISTICS OF TTE, INC., LE1182-SERIES
11-POLE ELLIPTICAL FILTER
Normalized Response
1.0
1.5
2.0
80
90
100
12.6S
Ultimate guaranteed stopband
- refer to page 3-1
F/Fc Transition Ratio
.893 933 .972 1.01 1.052 1.092 1.131 1.171 1.211 1.250 1.29
6.6S
Normalized Delay & Variation from Linear O
o
O LE1182 0
175o
F/Fc Transition Ratio
35o/Div.
70
(Please refer to page 3-3)
60
1.2S/Div.
50
Relative Attenuation , (dB)
40
5
10
15
20
2.5
25
F/Fc Transition Ratio
3.0
.005 .105 .205 .304 .404 .504 .604 .704 .803 .903 1.00330
20
30
Return Loss, (dB)
LE1182
10
Passband Attn., (dB)
Normalized Passband: Amplitude & Return Loss
0
LE1182
0
0.5
0
o
.6S.005 .105 .205 .304 .404 .504 .604 .704 .803 .903 1.003350
Reprinted with Permission of
TTE, Inc., 2251 Barry Ave., Los Angeles, CA 90064
a
4.4
From this discussion, we can see how the sharpness of the antialiasing transition
band can be traded off against the ADC sampling frequency. Choosing a higher
sampling rate (oversampling) reduces the requirement on transition band sharpness
(hence, the filter complexity) at the expense of using a faster ADC and processing
data at a faster rate. This is illustrated in Figure 4.5 which shows the effects of
5
increasing the sampling frequency while maintaining the same analog corner
frequency, fa,and the same dynamic range, DR, requirement.
INCREASING SAMPLING FREQUENCY RELAXES
REQUIREMENT ON ANTIALIASING FILTER
fa
B
A
fa
fs - fa
fs - fa
DR
0.5f s
fs
0.5fs
fs
LOWPASS FILTER SPECIFICATIONS:
a
4.5
The above design process is started by choosing an initial sampling rate of 2 to 4
times fa. Determine the filter specifications based on the required dynamic range
and see if such a filter is realizable within the constraints of the system cost and
performance. If not, consider a higher sampling rate which may require using a
faster ADC.
The antialiasing filter requirements can be relaxed somewhat if it is certain that
there will never be a fullscale signal at the stopband frequency fs – fa. In many
applications, it is improbable that fullscale signals will occur at this frequency. If the
maximum signal at the frequency fs – fa will never exceed XdB below fullscale.
Then, the filter stopband attenuation requirement is reduced by that same amount.
The new requirement for stopband attenuation at fs – fa based on this knowledge of
the signal is now only DR – XdB. When making this type of assumption, be careful
to treat any noise signals which may occur above the maximum signal frequency fa
as unwanted signals which will also alias back into the signal bandwidth.
UNDERSAMPLING (HARMONIC SAMPLING, BANDPASS
SAMPLING, IF SAMPLING, DIRECT IF TO DIGITAL
CONVERSION)
Thus far we have considered the case of baseband sampling, i.e., all the signals of
interest lie within the first Nyquist zone. Figure 4.6A shows such a case, where the
band of sampled signals is limited to the first Nyquist zone, and images of the
original band of frequencies appear in each of the other Nyquist zones.
6
Consider the case shown in Figure 4.6B, where the sampled signal band lies entirely
within the second Nyquist zone. The process of sampling a signal outside the first
Nyquist zone is often referred to as undersampling, or harmonic sampling. Note that
the first Nyquist zone image contains all the information in the original signal, with
the exception of its original location (the order of the frequency components within
the spectrum is reversed, but this is easily corrected by re-ordering the output of the
FFT).
UNDERSAMPLING
A
ZONE 1
0.5fs
fs
1.5fs
2fs
2.5fs
3fs
3.5fs
fs
1.5fs
2fs
2.5fs
3fs
3.5fs
2fs
2.5fs
3fs
3.5fs
ZONE 2
B
I
0.5fs
ZONE 3
C
I
0.5fs
fs
1.5fs
a
4.6
Figure 4.6C shows the sampled signal restricted to the third Nyquist zone. Note that
the first Nyquist zone image has no frequency reversal. In fact, the sampled signal
frequencies may lie in any unique Nyquist zone, and the first Nyquist zone image is
still an accurate representation (with the exception of the frequency reversal which
occurs when the signals are located in even Nyquist zones). At this point we can
clearly state the Nyquist criteria:
A signal must be sampled at a rate equal to or greater than twice its bandwidth in
order to preserve all the signal information.
Notice that there is no mention of the absolute location of the band of sampled
signals within the frequency spectrum relative to the sampling frequency. The only
constraint is that the band of sampled signals be restricted to a single Nyquist zone,
i.e., the signals must not overlap any multiple of fs/2 (this, in fact, is the primary
function of the antialiasing filter).
Sampling signals above the first Nyquist zone has become popular in
communications because the process is equivalent to analog demodulation. It is
becoming common practice to sample IF signals directly and then use digital
techniques to process the signal, thereby eliminating the need for the IF
7
demodulator. Clearly, however, as the IF frequencies become higher, the dynamic
performance requirements on the ADC become more critical. The ADC input
bandwidth and distortion performance must be adequate at the IF frequency, rather
than only baseband. This presents a problem for most ADCs designed to process
signals in the first Nyquist zone, therefore an ADC suitable for undersampling
applications must maintain dynamic performance into the higher order Nyquist
zones.
ANTIALIASING FILTERS IN UNDERSAMPLING
APPLICATIONS
Figure 4.7 shows a signal in the second Nyquist zone centered around a carrier
frequency, fc, whose lower and upper frequencies are f1 and f2. The antialiasing
filter is a bandpass filter. The desired dynamic range is DR, which defines the filter
stopband attenuation. The upper transition band is f2 to 2fs–f2, and the lower is f1
to fs – f1. As in the case of baseband sampling, the antialiasing filter requirements
can be relaxed by proportionally increasing the sampling frequency, but fc must also
be increased so that it is always centered in the second Nyquist zone.
ANTIALIASING FILTER FOR UNDERSAMPLING
f s - f1
f1
f2
2fs - f2
fc
DR
SIGNALS
OF
INTEREST
IMAGE
0.5f S
0
IMAGE
IMAGE
fS
1.5f S
BANDPASS FILTER SPECIFICATIONS:
a
2f S
STOPBAND ATTENUATION = DR
TRANSITION BAND: f2 TO 2fs - f2
f1 TO fs - f1
CORNER FREQUENCIES: f1, f2
4.7
Two key equations can be used to select the sampling frequency, fs, given the carrier
frequency, fc, and the bandwidth of its signal, ∆f. The first is the Nyquist criteria:
fs > 2∆f
Eq. 1
The second equation ensures that fc is placed in the center of a Nyquist zone:
fs =
4f c
, Eq. 2
2NZ − 1
8
where NZ = 1, 2, 3, 4, .... and NZ corresponds to the Nyquist zone in which the
carrier and its signal fall (see Figure 4.8).
NZ is normally chosen to be as large as possible while still maintaining fs > 2∆f. This
results in the minimum required sampling rate. If NZ is chosen to be odd, then fc
and it's signal will fall in an odd Nyquist zone, and the image frequencies in the first
Nyquist zone will not be reversed. Tradeoffs can be made between the sampling
frequency and the complexity of the antialiasing filter by choosing smaller values of
NZ (hence a higher sampling frequency).
CENTERING AN UNDERSAMPLED SIGNAL
WITHIN A NYQUIST ZONE
ZONE NZ - 1
ZONE NZ
ZONE NZ + 1
I
∆f
I
fc
0.5f s
0.5fs
fs > 2∆
∆f
0.5f s
fs =
4fc
2NZ - 1
, NZ = 1, 2, 3, . . .
a
4.8
As an example, consider a 4MHz wide signal centered around a carrier frequency of
71MHz. The minimum required sampling frequency is therefore 8MSPS. Solving Eq.
2 for NZ using fc = 71MHz and fs = 8MSPS yields NZ = 18.25. However, NZ must be
an integer, so we round 18.25 to the next lowest integer, 18. Solving Eq. 2 again for
fs yields fs = 8.1143MSPS. The final values are therefore fs = 8.1143MSPS, fc =
71MHz, and NZ = 18.
Now assume that we desire more margin for the antialiasing filter, and we select fs
to be 10MSPS. Solving Eq. 2 for NZ, using fc = 71MHz and fs = 10MSPS yields NZ =
14.7. We round 14.7 to the next lowest integer, giving NZ = 14. Solving Eq. 2 again
for fs yields fs = 10.519MSPS. The final values are therefore fs = 10.519MSPS, fc =
71MHz, and NZ = 14.
The above iterative process can also be carried out starting with fs and adjusting the
carrier frequency to yield an integer number for NZ.
DISTORTION AND NOISE IN AN IDEAL N-BIT ADC
9
Thus far we have looked at the implications of the sampling process without
considering the effects of ADC quantization. We will now treat the ADC as an ideal
sampler, but include the effects of quantization.
The only errors (DC or AC) associated with an ideal N-bit ADC are those related to
the sampling and quantization processes. The maximum error an ideal ADC makes
digitizing a DC input signal is ±1/2LSB. Any AC signal applied to an ideal N-bit
ADC will produce quantization noise whose rms value (measured over the Nyquist
bandwidth, DC to fs/2) is approximately equal to the weight of the least significant
bit (LSB), q, divided by √12. (See Reference 2). This assumes that the signal is at
least a few LSBs in amplitude so that the ADC output always changes state. The
quantization error signal from a linear ramp input is approximated as a sawtooth
waveform with a peak-to-peak amplitude equal to q, and its rms value is therefore
q/√12 (see Figure 4.9).
IDEAL N-BIT ADC QUANTIZATION NOISE
DIGITAL
CODE
OUTPUT
ANALOG
INPUT
q = 1LSB
ERROR
RMS ERROR = q/√
√12
SNR = 6.02N + 1.76dB + 10log
a
fs
2•BW
FOR FS SINEWAVE
4.9
It can be shown that the ratio of the rms value of a full scale sinewave to the rms
value of the quantization noise (expressed in dB) is:
SNR = 6.02N + 1.76dB,
where N is the number of bits in the ideal ADC. This equation is only valid if the
noise is measured over the entire Nyquist bandwidth from DC to fs/2. If the signal
bandwidth, BW, is less than fs/2, then the SNR within the signal bandwidth BW is
increased because the amount of quantization noise within the signal bandwidth is
smaller. The correct expression for this condition is given by:
 fs 
SNR = 6.02N + 176
. dB + 10 log
.
 2 ⋅ BW 
10
The above equation reflects the condition called oversampling, where the sampling
frequency is higher than twice the signal bandwidth. The correction term is often
called processing gain. Notice that for a given signal bandwidth, doubling the
sampling frequency increases the SNR by 3dB.
Although the rms value of the noise is accurately approximated q/√12, its frequency
domain content may be highly correlated to the AC input signal. For instance, there
is greater correlation for low amplitude periodic signals than for large amplitude
random signals. Quite often, the assumption is made that the theoretical
quantization noise appears as white noise, spread uniformly over the Nyquist
bandwidth DC to fs/2. Unfortunately, this is not true. In the case of strong
correlation, the quantization noise appears concentrated at the various harmonics of
the input signal, just where you don't want them.
In most applications, the input to the ADC is a band of frequencies (usually summed
with some noise), so the quantization noise tends to be random. In spectral analysis
applications (or in performing FFTs on ADCs using spectrally pure sinewaves - see
Figure 4.10), however, the correlation between the quantization noise and the signal
depends upon the ratio of the sampling frequency to the input signal. This is
demonstrated in Figure 4.11, where an ideal 12-bit ADCs output is analyzed using a
4096-point FFT. In the left-hand FFT plot, the ratio of the sampling frequency to
the input frequency was chosen to be exactly 32, and the worst harmonic is about
76dB below the fundamental. The right hand diagram shows the effects of slightly
offsetting the ratio, showing a relatively random noise spectrum, where the SFDR is
now about 92dBc. In both cases, the rms value of all the noise components is q/√12,
but in the first case, the noise is concentrated at harmonics of the fundamental.
DYNAMIC PERFORMANCE ANALYSIS
OF AN IDEAL N-BIT ADC
fs
ANALOG
INPUT
fa
IDEAL
N-BIT
ADC
N
BUFFER
MEMORY
M-WORDS
a
M POINT
M-POINT
2
FFT
PROCESSOR SPECTRAL
OUTPUT
4.10
11
EFFECT OF RATIO OF SAMPLING CLOCK TO INPUT
FREQUENCY ON SFDR FOR IDEAL 12-BIT ADC
fs / fa = 32
0
500
1000
SFDR = 76dBc
M = 4096
1500
2000
0
a
fs / fa = 32.25196850394
500
1000
SFDR = 92dBc
1500
2000
4.11
Note that this variation in the apparent harmonic distortion of the ADC is an
artifact of the sampling process and the correlation of the quantization error with
the input frequency. In a practical ADC application, the quantization error generally
appears as random noise because of the random nature of the wideband input signal
and the additional fact that there is a usually a small amount of system noise which
acts as a dither signal to further randomize the quantization error spectrum. (For
further discussions on dither, see Section 5 of this book).
It is important to understand the above point, because single-tone sinewave FFT
testing of ADCs is a universally accepted method of performance evaluation. In
order to accurately measure the harmonic distortion of an ADC, steps must be taken
to ensure that the test setup truly measures the ADC distortion, not the artifacts
due to quantization noise correlation. This is done by properly choosing the
frequency ratio and sometimes by injecting a small amount of noise (dither) with the
input signal.
Now, return to Figure 4.11, and note that the average value of the noise floor of the
FFT is greater than 100dB below full scale, but the theoretical SNR of a 12-bit ADC
is 74dB. The FFT noise floor is not the SNR of the ADC, because the FFT acts like
an analog spectrum analyzer with a bandwidth of fs/M, where M is the number of
points in the FFT, rather than fs/2. The theoretical FFT noise floor is therefore
10log10(M/2)dB below the quantization noise floor due to the so-called processing
gain of the FFT (see Figure 4.12). In the case of an ideal 12-bit ADC with an SNR of
74dB, a 4096-point FFT would result in a processing gain of 10log10(4096/2) =
33dB, thereby resulting in an overall FFT noise floor of 74+33=107dBc. In fact, the
FFT noise floor can be reduced even further by going to larger and larger FFTs; just
as an analog spectrum analyzer's noise floor can be reduced by narrowing the
bandwidth.
12
NOISE FLOOR FOR AN IDEAL 12-BIT ADC
USING 4096-POINT FFT
0
ADC FULLSCALE
(dB)
N = 12-BITS
M = 4096
20
40
74dB = 6.02N + 1.76dB
60
RMS QUANTIZATION NOISE LEVEL
80
33dB = 10log
()
M
2
100
FFT NOISE FLOOR
120
BIN SPACING =
fs
fs
4096
2
a
4.12
DISTORTION AND NOISE IN PRACTICAL ADCS
A practical sampling ADC (one that has an integral sample-and-hold), regardless of
architecture, has a number of noise and distortion sources as shown in Figure 4.13.
The wideband analog front-end buffer has wideband noise, non-linearity, and also
finite bandwidth. The SHA introduces further non-linearity, bandlimiting, and
aperture jitter. The actual quantizer portion of the ADC introduces quantization
noise, and both integral and differential non-linearity. In this discussion, assume
that sequential outputs of the ADC are loaded into a buffer memory of length M and
that the FFT processor provides the spectral output. Also assume that the FFT
arithmetic operations themselves introduce no significant errors relative to the ADC.
However, when examining the output noise floor, the FFT processing gain
(dependent on M) must be considered.
13
ADC MODEL SHOWING NOISE AND DISTORTION SOURCES
fs
ANALOG
INPUT
ADC
••
•
N
SAMPLE
AND
HOLD
BUFFER
••
••
NOISE
DISTORTION
BAND LIMITING
NOISE
DISTORTION
BAND LIMITING
APERTURE JITTER
••
•
QUANTIZATION NOISE
DIFFERENTIAL NON-LINEARITY
INTEGRAL NON-LINEARITY
M POINT
2
N
TEST
SYSTEM
TO MEMORY
ENCODER
M-POINT
FFT
PROCESSOR
BUFFER
MEMORY
M
•
•
SPECTRAL
OUTPUT
PROCESSING G AIN = 10log
( M2 )
ROUNDOFF ERROR (NEGLIGIBLE)
a
4.13
Equivalent Input Referred Noise (Thermal Noise)
The wideband ADC internal circuits produce a certain amount of wideband rms
noise due to thermal effects. This noise is present even for DC input signals, and
accounts for the fact that the output of most wideband ADCs is a distribution of
codes, centered around the nominal value of a DC input (see Figure 4.14). To
measure its value, the input of the ADC is grounded, and a large number of output
samples are collected and plotted as a histogram (sometimes referred to as a
grounded-input histogram). Since the noise is approximately Gaussian, the standard
deviation of the histogram is easily calculated (see Reference 3), corresponding to the
effective input rms noise. It is common practice to express this rms noise in terms of
LSBs, although it can be expressed as an rms voltage.
14
CODE FREQUENCY
HISTOGRAM OF 5000 CONVERSIONS
FOR A DC INPUT SHOWS 5 LSB p-p OR
0.8LSB RMS EQUIVALENT INPUT NOISE
(X-2)
(X-1)
(X)
a
(X+1)
(X+2)
(X+3)
CODE
4.14
Integral and Differential Non-Linearity
The overall integral non-linearity of an ADC is due to the integral non-linearity of
the front-end and SHA as well as the overall integral non-linearity in the ADC
transfer function. However, differential non-linearity is due exclusively to the
encoding process and may vary considerably dependent on the ADC encoding
architecture. Overall integral non-linearity produces distortion products whose
amplitude varies as a function of the input signal amplitude. For instance, secondorder intermodulation products increase 2dB for every 1dB increase in signal level,
and third-order products increase 3dB for every 1dB increase in signal level.
QUANTIFYING ADC DYNAMIC PERFORMANCE
n
Harmonic Distortion
n
Worst Harmonic
n
Total Harmonic Distortion (THD)
n
Total Harmonic Distortion Plus Noise (THD + N)
n
Signal-to-Noise-and-Distortion Ratio (SINAD, or S/N +D)
n
Effective Number of Bits (ENOB)
n
Signal-to-Noise Ratio (SNR)
15
n
Analog Bandwidth (Full-Power, Small-Signal)
n
Spurious Free Dynamic Range (SFDR)
n
Two-Tone Intermodulation Distortion
n
Noise Power Ratio (NPR)
a
4.15
The differential non-linearity in the ADC transfer function produces distortion
products which not only depend on the amplitude of the signal but the positioning of
the differential non-linearity along the ADC transfer function. Figure 4.16 shows
two ADC transfer functions containing differential non-linearity. The left-hand
diagram shows an error which occurs at midscale. Therefore, for both large and
small signals, the signal crosses through this point producing a distortion product
which is relatively independent of the signal amplitude. The right-hand diagram
shows another ADC transfer function which has differential non-linearity errors at
1/4 and 3/4 full scale. Signals which are above 1/2 scale peak-to-peak will exercise
these codes, while those less and 1/2 scale peak-to-peak will not.
ADC DNL ERRORS
CODE
OUT
CODE
OUT
IN
IN
MIDSCALE DNL
1/4FS, 3/4FS DNL
a
4.16
The design of most high-speed ADCs is such that differential non-linearity is spread
across the entire ADC range. Therefore, for signals which are within a few dB of full
scale, the overall integral non-linearity of the transfer function determines the
distortion products. For lower level signals, however, the harmonic content becomes
dominated by the differential non-linearities and does not generally decrease
proportionally with decreases in signal amplitude.
16
Harmonic Distortion, Worst Harmonic, Total Harmonic Distortion (THD),
Total Harmonic Distortion Plus Noise (THD + N)
There are a number of ways to quantify the distortion of an ADC. An FFT analysis
can be used to measure the amplitude of the various harmonics of a signal as shown
in Figure 4.17. The harmonics of the input signal can be distinguished from other
distortion products by their location in the frequency spectrum. The figure shows a
7MHz input signal sampled at 20MSPS and the location of the first 9 harmonics.
Aliased harmonics of fa fall at frequencies equal to |±Kfs±nfa|, where n is the order
of the harmonic, and K = 0, 1, 2, 3,.... The second and third harmonics are generally
the only ones specified on a data sheet because they tend to be the largest, although
some data sheets may specify the value of the worst harmonic. Harmonic distortion
is normally specified in dBc (decibels below carrier), although at audio frequencies it
may be specified as a percentage. Harmonic distortion is specified with an input
signal near full scale (generally 0.5 to 1dB below full scale to prevent clipping). For
signals much lower than full scale, other distortion products (not direct harmonics)
may limit performance.
LOCATION OF HARMONIC DISTORTION PRODUCTS:
INPUT SIGNAL = 7MHz, SAMPLING RATE = 20MSPS
RELATIVE
AMPLITUDE
fa
HARMONICS AT: |±Kf s±nfa|
n = ORDER OF HARMONIC, K = 0, 1, 2, 3, . . .
3
2
6
1
2
9
8
5
3
4
5
4
7
6
7
8
9
10
FREQUENCY (MHz)
a
4.17
Total harmonic distortion (THD) is the ratio of the rms value of the fundamental
signal to the mean value of the root-sum-square of its harmonics (generally, only the
first 5 are significant). THD of an ADC is also generally specified with the input
signal close to full scale.
Total harmonic distortion plus noise (THD+ N) is the ratio of the rms value of the
fundamental signal to the mean value of the root-sum-square of its harmonics plus
all noise components (excluding DC). The bandwidth over which the noise is
measured must be specified. In the case of an FFT, the bandwidth is DC to fs/2. (If
17
the bandwidth of the measurement is DC to fs/2, THD+N is equal to SINAD - see
below).
Signal-to-Noise-and-Distortion Ratio (SINAD), Signal-to-Noise Ratio (SNR),
and Effective Number of Bits (ENOB)
SINAD and SNR deserve careful attention, because there is still some variation
between ADC manufacturers as to their precise meaning. Signal-to-noise-and
Distortion (SINAD, or S/N+D) is the ratio of the rms signal amplitude to the mean
value of the root-sum-square (RSS) of all other spectral components, including
harmonics, but excluding DC. SINAD is a good indication of the overall dynamic
performance of an ADC as a function of input frequency because it includes all
components which make up noise (including thermal noise) and distortion. It is often
plotted for various input amplitudes. SINAD is equal to THD+N if the bandwidth for
the noise measurement is the same. A typical plot for the AD9220 12-bit, 10MSPS
ADC is shown in Figure 4.19.
SINAD, ENOB, AND SNR
n
SINAD (Signal-to-Noise-and-Distortion Ratio):
The ratio of the rms signal amplitude to the mean value of
the root-sum-squares (RSS) of all other spectral components,
including harmonics, but excluding DC
n
ENOB (Effective Number of Bits):
ENOB =
n
SINAD − 1.76 dB
6.02
SNR (Signal-to-Noise Ratio, or Signal-to-Noise Ratio
Without Harmonics):
The ratio of the rms signal amplitude to the mean value of
the root-sum-squares (RSS) of all other spectral components,
excluding the first 5 harmonics and DC
a
4.18
18
AD9220 12-BIT, 10MSPS ADC SINAD AND ENOB
VS. INPUT FREQUENCY FOR SAMPLING RATE = 10MSPS:
SINGLE-ENDED DRIVE, V cm = +2.5V, INPUT SPAN = 2V p-p
80
13
75
12.2
-0.5dB
70
11.3
65
10.5
60
9.7
55
ENOBS
SINAD - dB
-6dB
8.8
-20dB
50
8
45
7.2
6.3
40
0.1
1.0
10.0
FREQUENCY - MHz
a
4.19
The SINAD plot shows where the AC performance of the ADC degrades due to highfrequency distortion and is usually plotted for frequencies well above the Nyquist
frequency so that performance in undersampling applications can be evaluated.
SINAD is often converted to effective-number-of-bits (ENOB) using the relationship
for the theoretical SNR of an ideal N-bit ADC: SNR = 6.02N + 1.76dB. The equation
is solved for N, and the value of SINAD is substituted for SNR:
ENOB =
SINAD − 176
. dB
.
6.02
Signal-to-noise ratio (SNR, or SNR-without-harmonics) is calculated the same as
SINAD except that the signal harmonics are excluded from the calculation, leaving
only the noise terms. In practice, it is only necessary to exclude the first 5 harmonics
since they dominate. The SNR plot will degrade at high frequencies also, but not as
rapidly as SINAD because of the exclusion of the harmonic terms.
Many current ADC data sheets somewhat loosely refer to SINAD as SNR, so the
engineer must be careful when interpreting these specifications.
Analog Bandwidth
The analog bandwidth of an ADC is that frequency at which the spectral output of
the fundamental swept frequency (as determined by the FFT analysis) is reduced by
3dB. It may be specified for either a small signal (SSBW- small signal bandwidth),
or a full scale signal (FPBW- full power bandwidth), so there can be a wide variation
in specifications between manufacturers.
19
Like an amplifier, the analog bandwidth specification of a converter does not imply
that the ADC maintains good distortion performance up to its bandwidth frequency.
In fact, the SINAD (or ENOB) of most ADCs will begin to degrade considerably
before the input frequency approaches the actual 3dB bandwidth frequency. Figure
4.20 shows ENOB and full scale frequency response of an ADC with a FPBW of
1MHz, however, the ENOB begins to drop rapidly above 100kHz.
ADC GAIN (BANDWIDTH) AND ENOB VERSUS FREQUENCY
SHOWS IMPORTANCE OF ENOB SPECIFICATION
FPBW = 1MHz
GAIN (FS INPUT)
ENOB (FS INPUT)
ENOB
ENOB (-20dB INPUT)
10
100
1k
10k
100k
1M
10M
ADC INPUT FREQUENCY (Hz)
a
4.20
Spurious Free Dynamic Range (SFDR)
Probably the most significant specification for an ADC used in a communications
application is its spurious free dynamic range (SFDR). The SFDR specification is to
ADCs what the third order intercept specification is to mixers and LNAs. SFDR of
an ADC is defined as the ratio of the rms signal amplitude to the rms value of the
peak spurious spectral content (measured over the entire first Nyquist zone, DC to
fs/2). SFDR is generally plotted as a function of signal amplitude and may be
expressed relative to the signal amplitude (dBc) or the ADC full scale (dBFS).
For a signal near full scale, the peak spectral spur is generally determined by one of
the first few harmonics of the fundamental. However, as the signal falls several dB
below full scale, other spurs generally occur which are not direct harmonics of the
input signal. This is because of the differential non-linearity of the ADC transfer
function as discussed earlier. Therefore, SFDR considers all sources of distortion,
regardless of their origin.
The AD9042 is a 12-bit, 41MSPS wideband ADC designed for communications
applications where high SFDR is important. The SFDR for a 19.5MHz input and a
sampling frequency of 41MSPS is shown in Figure 4.21. Note that a minimum of
20
80dBc SFDR is obtained over the entire first Nyquist zone (DC to 20MHz). The plot
also shows SFDR expressed as dBFS.
AD99042 12-BIT, 41MSPS ADC
SFDR VS. INPUT POWER LEVEL
WORST CASE SPURIOUS - dBc AND dBFS
100
90
dBFS
80
70
60
ENCODE = 41 MSPS
AIN = 19.5MHz
50
40
SFDR = 80dB
REFERENCE LINE
dBc
30
20
10
0
-80
-70
-60
-50
-40
-30
-20
-10
0
ANALOG INPUT POWER LEVEL - dBFS
a
4.21
SFDR is generally much greater than the ADCs theoretical N-bit SNR (6.02N +
1.76dB). For example, the AD9042 is a 12-bit ADC with an SFDR of 80dBc and a
typical SNR of 65dBc (theoretical SNR is 74dB). This is because there is a
fundamental distinction between noise and distortion measurements. The process
gain of the FFT (33dB for a 4096-point FFT) allows frequency spurs well below the
noise floor to be observed. Adding extra resolution to an ADC may serve to increase
its SNR but may or may not increase its SFDR.
Two Tone Intermodulation Distortion
Two tone IMD is measured by applying two spectrally pure sinewaves to the ADC at
frequencies f1 and f2, usually relatively close together. The amplitude of each tone is
set slightly more than 6dB below full scale so that the ADC does not clip when the
two tones add in-phase. The location of the second and third-order products are
shown in Figure 4.22. Notice that the second-order products fall at frequencies
which can be removed by digital filters. However, the third-order products 2f2–f1
and 2f1–f2 are close to the original signals and are more difficult to filter. Unless
otherwise specified, two-tone IMD refers to these third-order products. The value of
the IMD product is expressed in dBc relative to the value of either of the two original
tones, and not to their sum.
21
SECOND AND THIRD-ORDER INTERMODULATION
PRODUCTS FOR f1 = 5MHz, f2 = 6MHz
2 = SECOND ORDER IMD PRODUCTS
f1
3
f2
= THIRD ORDER IMD PRODUCTS
NOTE: f1 = 5MHz, f2 = 6MHz
2
2
f2 - f1
f2 + f1
2f2
2f1
3
3
2f1 - f2
1
4
6 7
3
2f2 + f1
2f1 + f2
2f2 - f1
5
3
3f1
10 11 12
3f2
15 16 17 18
FREQUENCY: MHz
a
4.22
Note, however, that if the two tones are close to fs/4, then the aliased third harmonic
of the fundamental can make the identification of the actual 2f2–f1 and 2f1–f2
products difficult. Similarly, if the two tones are close to fs/3, the aliased second
harmonic may interfere with the measurement.
The concept of second and third-order intercept points is not valid for an ADC,
because the distortion products do not vary in a predictable manner (as a function
of signal amplitude). The ADC does not gradually begin to compress signals
approaching full scale (there is no 1dB compression point), it acts as a hard limiter
as soon as the signal exceeds the ADC input range, thereby suddenly producing
extreme amounts of distortion because of clipping.
On the other hand, for signals much below full scale, the distortion floor remains
relatively constant and is independent of signal level. This is illustrated in Figure
4.23 for the AD9042, where two-tone SFDR is plotted as a function of signal level.
The plot indicates that the distortion floor ranges from 85 to 90dBFS regardless of
the input signal amplitude.
22
AD9042 12-BIT, 41MSPS ADC TWO-TONE SFDR
WORST CASE SPURIOUS - dBc AND dBFS
100
dBFS
90
80
70
60
50
ENCODE = 41 MSPS
F1 = 19.3MHz
F2 = 19.51MHz
dBc
SFDR = 80dB
REFERENCE LINE
40
30
20
10
0
-80
-70
-60
-50
-40
-30
-20
-10
0
INPUT POWER LEVEL (F1 = F2) - dBFS
a
4.23
Noise Power Ratio (NPR)
Noise power ratio testing has been used extensively to measure the transmission
characteristics of Frequency Division Multiplexed (FDM) communications links (see
Reference 4). In a typical FDM system, 4kHz wide voice channels are "stacked" in
frequency bins for transmission over coaxial, microwave, or satellite equipment. At
the receiving end, the FDM data is demultiplexed and returned to 4kHz individual
baseband channels. In an FDM system having more than approximately 100
channels, the FDM signal can be approximated by Gaussian noise with the
appropriate bandwidth. An individual 4kHz channel can be measured for
"quietness" using a narrow-band notch (bandstop) filter and a specially tuned
receiver which measures the noise power inside the 4kHz notch (see Figure 4.24).
23
NOISE POWER RATIO (NPR) MEASUREMENTS
GAUSSIAN
NOISE
SOURCE
GAUSSIAN
NOISE
SOURCE
LPF
NOTCH
FILTER
LPF
NOTCH
FILTER
TRANSMISSION
SYSTEM
NARROWBAND
RECEIVER
N
ADC
BUFFER
MEMORY
AND FFT
PROCESSOR
fs
RMS
NOISE
LEVEL
(dB)
NPR
FREQUENCY
a
0.5f s
4.24
Noise Power Ratio (NPR) measurements are straightforward. With the notch filter
out, the rms noise power of the signal inside the notch is measured by the
narrowband receiver. The notch filter is then switched in, and the residual noise
inside the slot is measured. The ratio of these two readings expressed in dB is the
NPR. Several slot frequencies across the noise bandwidth (low, midband, and high)
are tested to characterize the system adequately. NPR measurements on ADCs are
made in a similar manner except the analog receiver is replaced by a buffer memory
and an FFT processor.
NPR is usually plotted on an NPR curve. The NPR is plotted as a function of rms
noise level referred to the peak range of the system. For very low noise loading level,
the undesired noise (in non-digital systems) is primarily thermal noise and is
independent of the input noise level. Over this region of the curve, a 1dB increase in
noise loading level causes a 1dB increase in NPR. As the noise loading level is
increased, the amplifiers in the system begin to overload, creating intermodulation
products which cause the noise floor of the system to increase. As the input noise
increases further, the effects of "overload" noise predominate, and the NPR is
reduced dramatically. FDM systems are usually operated at a noise loading level a
few dB below the point of maximum NPR.
In a digital system containing an ADC, the noise within the slot is primarily
quantization noise when low levels of noise input are applied. The NPR curve is
linear in this region. As the noise level increases, there is a one-for-one
correspondence between the noise level and the NPR. At some level, however,
"clipping" noise caused by the hard-limiting action of the ADC begins to dominate. A
theoretical curve for 10, 11, and 12-bit ADCs is shown in Figure 4.25 (see Reference
5). Peak NPR and corresponding loading levels are shown in Figure 4.26.
24
THEORETICAL NPR FOR 10, 11, 12-BIT ADCs
ADC RANGE = ±VO
NPR
(dB)
k=
60
62.7dB
VO
σ
σ = RMS NOISE LEVEL
12
-B
IT
S
57.1dB
11
-B
IT
S
55
51.6dB
10
-B
IT
S
50
45
-30
-25
-20
-15
-10
RMS NOISE LOADING LEVEL = -20log(k) dB
a
4.25
THEORETICAL NPR SUMMARY
BITS
k OPTIMUM
k(dB)
MAX NPR (dB)
8
3.92
11.87
40.60
9
4.22
12.50
46.05
10
4.50
13.06
51.56
11
4.76
13.55
57.12
12
5.01
14.00
62.71
13
5.26
14.41
68.35
14
5.49
14.79
74.01
15
5.72
15.15
79.70
16
5.94
15.47
85.40
ADC Range = ±Vo
k = Vo / σ
σ = RMS Noise Level
a
4.26
In multi-channel high frequency communication systems, NPR can also be used to
simulate the distortion caused by a large number of individual channels, similar to
25
an FDM system. A notch filter is placed between the noise source and the ADC, and
an FFT output is used in place of the analog receiver. The width of the notch filter is
set for several MHz as shown in Figure 4.27 for the AD9042. NPR is the "depth" of
the notch. An ideal ADC will only generate quantization noise inside the notch,
however a practical one has additional noise components due to intermodulation
distortion caused by ADC non-linearity. Notice that the NPR is about 60dB
compared to 62.7dB theoretical.
AD9042 12-BIT, 41MSPS ADC NPR
MEASURES 60dB (62.7dB THEORETICAL)
POWER RELATIVE TO ADC FULL SCALE - dB
0
ENCODE = 41 MSPS
AIN = BROADBAND_NOISE
-20
-40
-60
-80
-100
-120
dc
20.5
4.1
8.2
FREQUENCY - MHz
a
12.3
16.4
4.27
Aperture Jitter and Aperture Delay
Another reason that the SNR of an ADC decreases with input frequency may be
deduced from Figure 4.28, which shows the effects of phase jitter (or aperture time
jitter) on the sampling clock of an ADC (or internal in the sample-and-hold). The
phase jitter causes a voltage error which is a function of slew rate and results in an
overall degradation in SNR as shown in Figure 4.29. This is quite serious, especially
at higher input/output frequencies. Therefore, extreme care must be taken to
minimize phase noise in the sampling/reconstruction clock of any sampled data
system. This care must extend to all aspects of the clock signal: the oscillator itself
(for example, a 555 timer is absolutely inadequate, but even a quartz crystal
oscillator can give problems if it uses an active device which shares a chip with noisy
logic); the transmission path (these clocks are very vulnerable to interference of all
sorts), and phase noise introduced in the ADC or DAC. A very common source of
phase noise in converter circuitry is aperture jitter in the integral sample-and-hold
(SHA) circuitry.
26
EFFECTS OF APERTURE AND SAMPLING CLOCK JITTER
∆v =
ANALOG
INPUT
dv
= SLOPE
dt
dv
dt • ∆t
∆vRMS = APERTURE JITTER ERROR
{
NOMINAL
HELD
OUTPUT
∆tRMS = APERTURE JITTER
HOLD
TRACK
a
4.28
SNR DUE TO APERTURE AND SAMPLING CLOCK JITTER
tj =
1ps
100
SNR = 20log10
1
2π
πftj
16
14
tj =
10p
s
80
12
ENOB
SNR
(dB)
tj =
100
60
10
ps
8
tj =
1ns
40
6
4
20
0
1
a
3
10
30
100
FULLSCALE SINEWAVE INPUT FREQUENCY (MHz)
4.29
A decade or so ago, sampling ADCs were built up from a separate SHA and ADC.
Interface design was difficult, and a key parameter was aperture jitter in the SHA.
Today, most sampled data systems use sampling ADCs which contain an integral
SHA. The aperture jitter of the SHA may not be specified as such, but this is not a
27
cause of concern if the SNR or ENOB is clearly specified, since a guarantee of a
specific SNR is an implicit guarantee of an adequate aperture jitter specification.
However, the use of an additional high-performance SHA will sometimes improve
the high-frequency ENOB of a even the best sampling ADC by presenting "DC" to
the ADC, and may be more cost-effective than replacing the ADC with a more
expensive one.
It should be noted that there is also a fixed component which makes up the ADC
aperture time. This component, usually called effective aperture delay time, does not
produce an error. It simply results in a time offset between the time the ADC is
asked to sample and when the actual sample takes place (see Figure 4.30), and may
be positive or negative. The variation or tolerance placed on this parameter from
part to part is important in simultaneous sampling applications or other
applications such as I and Q demodulation where two ADCs are required to track
each other.
EFFECTIVE APERTURE DELAY TIME
+FS
ZERO CROSSING
ANALOG INPUT
SINEWAVE
0V
-FS
-te
+te
SAMPLING
CLOCK
te
a
4.30
HIGH SPEED ADC ARCHITECTURES
Successive Approximation ADCs
The successive approximation (SAR) ADC architecture has been used for decades
and is still a popular and cost effective form of converter for sampling frequencies of
1MSPS or less. A simplified block diagram of a SAR ADC is shown in Figure 4.31.
On the START CONVERT command, all the bits of the successive approximation
register (SAR) are reset to "0" except the MSB which is set to "1". Bit 1 is then tested
in the following manner: If the DAC output is greater than the analog input, the
MSB is reset, otherwise it is left set. The next most significant bit is then tested by
setting it to "1". If the DAC output is greater than the analog input, this bit is reset,
28
otherwise it is left set. The process is repeated with each bit in turn. When all the
bits have been set, tested, and reset or not as appropriate, the contents of the SAR
correspond to the digital value of the analog input, and the conversion is complete.
SUCCESSIVE APPROXIMATION ADC
ANALOG
INPUT
EOC OR
DRDY
COMPARATOR
SHA
+
SAR*
START
CONVERT
DAC
*SUCCESSIVE
APPROXIMATION
REGISTER
DIGITAL
OUTPUT
a
4.31
An N-bit conversion takes N steps. It would seem on superficial examination that a
16-bit converter would have a conversion time that is twice as long as an 8-bit one,
but this is not the case. In an 8-bit converter, the DAC must settle to 8-bit accuracy
before the bit decision is made, whereas in a 16-bit converter, it must settle to 16-bit
accuracy, which takes a lot longer. In practice, 8-bit successive approximation ADCs
can convert in a few hundred nanoseconds, while 16-bit ones will generally take
several microseconds.
The classic SAR ADC is only a quantizer: no sampling takes place, and for an
accurate conversion, the input must remain constant for the entire conversion
period. Most modern SAR ADCs are sampling types and have an internal sampleand-hold so that they can process AC signals. They are specified for both AC and DC
applications. A SHA is required in a SAR ADC because the signal must remain
constant during the entire N-bit conversion cycle.
The accuracy of a SAR ADC depends primarily on the accuracy (differential and
integral linearity, gain, and offset) of the internal DAC. Until recently, this accuracy
was achieved using laser trimmed thin film resistors. Modern SAR ADCs utilize
CMOS switched capacitor charge redistribution DACs. This type of DAC depends on
the accurate ratio matching and stability of on-chip capacitors rather than thin film
resistors. For resolutions greater than 12-bits, on-chip autocalibration techniques
using an additional calibration DAC and the accompanying logic can accomplish the
same thing as thin film laser trimmed resistors, at much less cost. Therefore, the
entire ADC can be made on a standard sub-micron CMOS process.
29
The successive approximation ADC has a very simple structure, is low power, and
has reasonably fast conversion times (<1MSPS). It is probably the most widely used
ADC architecture, and will continue to be used for medium speed and medium
resolution applications.
Current 12-bit SAR ADCs achieve sampling rates up to about 1MSPS, and 16-bit
ones up to about 300kSPS. Examples of typical state-of-the-art SAR ADCs are the
AD7892 (12-bits at 600kSPS), the AD976/977 (16-bits at 100kSPS), and the AD7882
(16-bits at 300kSPS).
Flash Converters
Flash ADCs (sometimes called parallel ADCs) are the fastest type of ADC and use
large numbers of comparators. An N-bit flash ADC consists of 2N resistors and 2N–1
comparators arranged as in Figure 4.32. Each comparator has a reference voltage
which is 1 LSB higher than that of the one below it in the chain. For a given input
voltage, all the comparators below a certain point will have their input voltage
larger than their reference voltage and a "1" logic output, and all the comparators
above that point will have a reference voltage larger than the input voltage and a
"0" logic output. The 2N–1 comparator outputs therefore behave in a way analogous
to a mercury thermometer, and the output code at this point is sometimes called a
thermometer code. Since 2N–1 data outputs are not really practical, they are
processed by a decoder to an N-bit binary output.
FLASH OR PARALLEL ADC
STROBE
ANALOG
INPUT
+VREF
1.5R
R
R
PRIORITY
ENCODER
AND
LATCH
R
N
DIGITAL
OUTPUT
R
R
R
0.5R
a
4.32
30
The input signal is applied to all the comparators at once, so the thermometer
output is delayed by only one comparator delay from the input, and the encoder
N-bit output by only a few gate delays on top of that, so the process is very fast.
However, the architecture uses large numbers of resistors and comparators and it
limited to low resolutions, and if it is to be fast, each comparator must run at
relatively high power levels. Hence, the problems of flash ADCs include limited
resolution, high power dissipation because of the large number of high speed
comparators (especially at sampling rates greater than 50MSPS), and relatively
large (and therefore expensive) chip sizes. In addition, the resistance of the reference
resistor chain must be kept low to supply adequate bias current to the fast
comparators, so the voltage reference has to source quite large currents (>10 mA).
In practice, flash converters are available up to 10-bits, but more commonly they
have 8-bits of resolution. Their maximum sampling rate can be as high as
500 MSPS, and input full-power bandwidths in excess of 300 MHz.
But as mentioned earlier, full-power bandwidths are not necessarily full-resolution
bandwidths. Ideally, the comparators in a flash converter are well matched both for
DC and AC characteristics. Because the strobe is applied to all the comparators
simultaneously, the flash converter is inherently a sampling converter. In practice,
there are delay variations between the comparators and other AC mismatches which
cause a degradation in ENOB at high input frequencies. This is because the inputs
are slewing at a rate comparable to the comparator conversion time.
The input to a flash ADC is applied in parallel to a large number of comparators.
Each has a voltage-variable junction capacitance, and this signal-dependent
capacitance results in all flash ADCs having reduced ENOB and higher distortion at
high input frequencies. A model is shown in Figure 4.33, where the input
capacitance is modeled as a fixed 10pF capacitor in parallel with a variable capacitor
(modeled as a diode with a zero-bias junction capacitance of 6pF). As the input
changes from –FS to +FS, the total input capacitance changes from about 12.5 to
16pF. The wideband external drive amplifier is isolated from the flash converter by
a 50Ω series resistor. The distortion of this circuit degrades from about 70dBc at
1MHz to 35dBc at 100MHz.
31
SIGNAL-DEPENDENT INPUT CAPACITANCE CAUSES
DISTORTION AT HIGH FREQUENCIES
ANALOG
INPUT
50 Ω
A
-FS TO +FS
CJO = 6pF
10pF
+FS
THD
(dB)
FLASH INPUT MODEL
70
60
50
40
30
1
10
INPUT FREQUENCY (MHz)
a
100
4.33
High data rate digital communications applications such as set-top boxes for direct
broadcast satellites (DBS) require dual 6 or 8-bit high speed ADCs to perform
quadrature demodulation. A dual flash converter ensures good matching between
the two ADCs. The AD9066 (dual 6-bit, 60MSPS) flash converter is representative of
this type of converter. The AD9066 is fabricated on a BiCMOS process, operates on a
single +5V supply, and dissipates 400mW. The effective bit performance of the
device is shown in Figure 4.34. Note that the device maintains greater than 5
ENOBs up to 60MSPS analog input.
32
AD9066 DUAL 6-BIT, 60MSPS ADC ENOB
VS. ANALOG INPUT FREQUENCY
5.8
ENCODE = 60 MSPS
5.7
ENOB - Bits
5.6
5.5
5.4
5.3
5.2
1
10
100
MHz
a
4.34
Part of the reason for the excellent performance of the AD9066 is the use of an
interpolation scheme that reduces the number of differential amplifiers required by
a factor of two (see Reference 6). The architecture enables 64 possible quantization
levels to be determined with only 32 preamplifiers which drive 63 latches. This
keeps the input capacitance to a minimum (10pF) and reduces total power
dissipation of the device. The basic interpolation circuit is shown in Figure 4.35.
33
"INTERPOLATING" FLASH REDUCES THE NUMBER
OF PREAMPLIFIERS BY FACTOR OF TWO
ANALOG
INPUT
A2
V2
ANALOG
INPUT
B
+
B
-
V2
LATCH
2
V1A
DECODE
V1A =
V1 + V2
2
V1
LATCH
1A
B
A
+
A1
V1
-
A
A
LATCH
1
A
LATCH
STROBE
B
a
4.35
The preamplifiers are low-gain gm stages whose bandwidth is proportional to the
tail currents of the differential pairs. Consider the case for a positive-going ramp
input which is initially below the reference to AMP A1, V1. As the input signal
approaches V1, the differential output of A1 approaches zero (i.e., A = A ), and the
decision point is reached. The output of A1 drives the differential input of LATCH 1.
As the input signals continues to go positive, A continues to go positive, and B
begins to go negative. The interpolated decision point is determined when A = B . As
the input continues positive, the third decision point is reached when B = B . This
novel architecture reduces the ADC input capacitance and thereby minimizes its
change with signal level and the associated distortion. The input capacitance of the
AD9066 is only about 10pF. Key specifications for the device are summarized in
Figure 4.36.
AD9066 DUAL 6-BIT, 60MSPS FLASH ADC
KEY SPECIFICATIONS
n
Input Range: 500mV p-p
n
Input Impedance: 50kΩ
Ω || 10pF
n
ENOB: 5.7bits @ 15.5MHz Input
n
On-Chip Reference
n
Power Supply: Single +5V
n
Power Dissipation: 400mW
34
n
Package: 28-pin SOIC
n
Ideal for Quadrature Demodulation
a
4.36
Subranging (Pipelined) ADCs
Although it is not practical to make flash ADCs with high resolution, flash ADCs are
often used as subsystems in "subranging" ADCs (sometimes known as "half-flash
ADCs"), which are capable of much higher resolutions (up to 16-bits).
A block diagram of an 8-bit subranging ADC based upon two 4-bit flash converters is
shown in Figure 4.37. Although 8-bit flash converters are readily available at high
sampling rates, this example will be used to illustrate the theory. The conversion
process is done in two steps. The first four significant bits (MSBs) are digitized by
the first flash (to better than 8-bits accuracy), and the 4-bit binary output is applied
to a 4-bit DAC (again, better than 8-bit accurate). The DAC output is subtracted
from the held analog input, and the resulting residue signal is amplified and applied
to the second 4-bit flash. The outputs of the two 4-bit flash converters are then
combined into a single 8-bit binary output word. If the residue signal range does not
exactly fill the range of the second flash converter, non-linearities and perhaps
missing codes will result.
8-BIT SUBRANGING ADC
ANALOG
INPUT
GAIN
+
RESIDUE
SIGNAL
SHA
-
4-BIT
FLASH
4-BIT
FLASH
4-BIT
DAC
4
4
OUTPUT REGISTER
8
a
4.37
Modern subranging ADCs use a technique called digital correction to eliminate
problems associated with the architecture of Figure 4.37. A simplified block diagram
35
of a 12-bit digitally corrected subranging (DCS) ADC is shown in Figure 4.38. The
architecture is similar to that used in the AD9042 12-bit, 41MSPS ADC. Note that a
6-bit and an 7-bit ADC have been used to achieve an overall 12-bit output. These
are not flash ADCs, but utilize a magnitude-amplifier (MagAmp™) architecture
which will be described shortly.
AD9042 12-BIT, 41MSPS PIPELINED SUBRANGING ADC
WITH DIGITAL ERROR CORRECTION
ANALOG
INPUT
SHA
2
SHA
1
6-BIT
ADC
GAIN
+
-
SHA
3
7-BIT
ADC
6-BIT
DAC
6
BUFFER
REGISTER
7
6
ERROR CORRECTION LOGIC
12
OUTPUT REGISTERS
12
a
4.38
If there were no errors in the first-stage conversion, the 6-bit "residue" signal
applied to the 7-bit ADC by the summing amplifier would never exceed one-half of
the range of the 7-bit ADC. The extra range in the second ADC is used in
conjunction with the error correction logic (usually just a full adder) to correct the
output data for most of the errors inherent in the traditional uncorrected subranging
converter architecture. It is important to note that the 6-bit DAC must be better
than 12-bit accurate, because the digital error correction does not correct for DAC
errors. In practice, "thermometer" or "fully-decoded" DACs using one current switch
per level (63 switches in the case of a 6-bit DAC) are often used instead of a "binary"
DAC to ensure excellent differential and integral linearity and minimum switching
transients.
The second SHA delays the held output of the first SHA while the first-stage
conversion occurs, thereby maximizing throughput. The third SHA serves to deglitch
the residue output signal, thereby allowing a full conversion cycle for the 7-bit ADC
to make its decision (the 6 and 7-bit ADCs in the AD9042 are bit-serial MagAmp
ADCs which require more settling time than a flash converter).
This multi-stage conversion technique is sometimes referred to as "pipelining."
Additional shift registers in series with the digital outputs of the first-stage ADC
ensure that its output is ultimately time-aligned with the last 7 bits from the
second ADC when their outputs are combined in the error correction logic. A
36
pipelined ADC therefore has a specified number of clock cycles of latency, or pipeline
delay associated with the output data. The leading edge of the sampling clock (for
sample N) is used to clock the output register, but the data which appears as a
result of that clock edge corresponds to sample N – L, where L is the number of clock
cycles of latency. In the case of the AD9042, there are two clock cycles of latency.
Key specifications for the AD9042 are shown in Figure 4.39.
AD9042 12-BIT, 41MSPS ADC KEY SPECIFICATIONS
n
Input Range: 1V peak-to-peak, Vcm = +2.4V
n
Input Impedance: 250Ω
Ω to Vcm
n
Effective Input Noise: 0.33LSBs rms
n
SFDR at 20MHz Input: 80dB minimum
n
SINAD (S/N+D) at 20MHz Input = 67dB
n
Digital Outputs: TTL Compatible
n
Power Supply: Single +5V
n
Power Dissipation: 595mW
n
Fabricated on High Speed Dielectrically Isolated
Complementary Bipolar Process
a
4.39
The error correction scheme described above is designed to correct for errors made in
the first conversion. Internal ADC gain, offset, and linearity errors are corrected as
long as the residue signal fall within the range of the second-stage ADC. These
errors will not affect the linearity of the overall ADC transfer characteristic. Errors
made in the final conversion, however, do translate directly as errors in the overall
transfer function. Also, linearity errors or gain errors either in the DAC or the
residue amplifier will not be corrected and will show up as nonlinearities or nonmonotonic behavior in the overall ADC transfer function.
So far, we have considered only two-stage subranging ADCs, as these are easiest to
analyze. There is no reason to stop at two stages, however. Three-pass and four-pass
subranging pipelined ADCs are quite common, and can be made in many different
ways, usually with digital error correction.
A simplified block diagram of the AD9220 12-bit, 10MSPS single-supply, 250mW
CMOS ADC is shown in Figure 4.40. The AD9221 (1.25MSPS, 60mW) and the
AD9223 (3MSPS, 100mW) ADCs use the identical architecture but operate at lower
power and lower sampling rates. This is a four-stage pipelined architecture with an
additional bit in the second, third, and fourth stage for error correction. Because of
37
the pipelined architecture, these ADCs have a 3 clock-cycle latency (see Figure 4.41).
Key specifications for the AD9220/9221/9223 are given in Figure 4.42.
AD9220/9221/9223 12-BIT PIPELINED CMOS ADC
ANALOG
INPUT
SHA
1
SHA
2
+
+
SHA
3
-
-
4-BIT 4-BIT
ADC DAC
5-BIT 5-BIT
ADC DAC
SHA
4
3-BIT 3-BIT
ADC DAC
4
5
+
-
3-BIT
ADC
3
3
BUFFER REGISTERS AND
ERROR CORRECTION LOGIC
12
OUTPUT REGISTERS
12
a
4.40
LATENCY (PIPELINE DELAY)
OF AD9220/9221/9223 ADC
ANALOG
INPUT
N
N+1
N+2
N+3
SAMPLING
CLOCK
OUTPUT
DATA
DATA N - 3
DATA N - 2
DATA N - 1
DATA N
4.41
a
38
AD9220, AD9221, AD9223
CMOS 12-BIT ADCs KEY SPECIFICATIONS
n
Family Members:
AD9221 (1.25MSPS), AD9223 (3MSPS), AD9220 (10MSPS)
n
Power Dissipation: 60, 100, 250mW, Respectively
n
FPBW: 25, 40, 60MHz, Respectively
n
Effective Input Noise: 0.1LSB rms (Span = 5V)
n
SINAD: 71dB
n
SFDR: 88dBc
n
On-Chip Reference
n
Differential Non-Linearity: 0.3LSB
n
Single +5V Supply
n
28-Pin SOIC Package
a
4.42
Bit-Per-Stage (Serial, or Ripple) ADCs
Various architectures exist for performing A/D conversion using one stage per bit. In
fact, a multistage subranging ADC with one bit per stage and no error correction is
one form. Figure 4.43 shows the overall concept. The SHA holds the input signal
constant during the conversion cycle. There are N stages, each of which have a bit
output and a residue output. The residue output of one stage is the input to the
next. The last bit is detected with a single comparator as shown.
39
BIT-PER-STAGE, SERIAL, OR RIPPLE ADC
VREF
ANALOG
INPUT
SHA
STAGE
1
BIT 1
MSB
R1
STAGE
2
BIT 2
R2
STAGE
N-1
BIT N-1
+
BIT N
LSB
DECODE LOGIC AND OUTPUT REGISTERS
N
a
4.43
The basic stage for performing a single binary bit conversion is shown in Figure
4.44. It consists of a gain-of-two amplifier, a comparator, and a 1-bit DAC. The
comparator detects the zero-crossing of the input and is the binary bit output for
that stage. The comparator also switches a 1-bit DAC whose output is summed with
the output of the gain-of-two amplifier. The resulting residue output is then applied
to the next stage.
40
SINGLE-STAGE OF BINARY ADC
+VR
RESIDUE
INPUT
Σ
G=2
INPUT
0
+
-VR
-
+VR
+VR
-VR
RESIDUE
SWITCH POSITION
SHOWN FOR
NEGATIVE INPUT
0
-VR
BIT OUTPUT
(BINARY CODE)
a
4.44
A simplified 3-bit serial-binary ADC is shown in Figure 4.45, and the residue
outputs are shown in Figure 4.46. Each residue output signal has discontinuities
which correspond to the point where the comparator changes state and causes the
DAC to switch. The fundamental problem with this architecture is the discontinuity
in the residue output waveforms. Adequate settling time must be allowed for these
transients to propagate through all the stages and settle at the final comparator
input. The prospects of making this architecture operate at high speed are therefore
dismal.
41
3-BIT SERIAL ADC WITH BINARY OUTPUT
±VR
ANALOG
INPUT
R1
R2
STAGE
2
STAGE
1
SHA
BIT 1
+
-
BIT 2
BIT 3
OUTPUT REGISTER
3
a
4.45
INPUT AND RESIDUE WAVEFORMS OF
3-BIT BINARY RIPPLE ADC
+V R
0
INPUT
-V R
+V R
R1
0
-V R
+V R
R2
0
-V R
BINARY
CODE
000
001
010
011
100
101
110
111
4.46
a
A much better bit-per-stage architecture was developed by F.D. Waldhauer
(Reference 7) based on absolute value amplifiers (magnitude amplifiers, or simply
MagAmps™). This scheme has often been referred to as serial-Gray (since the output
coding is in Gray code), or folding converter (References 8, 9, 10). The basic stage is
42
shown functionally in Figure 4.47. The comparator detects the polarity of the input
signal and provides the Gray bit output for the stage. It also determines whether the
overall stage gain is +2 or –2. The reference voltage VR is summed with the switch
output to generate the residue signal which is applied to the next stage. The
transfer function for the folding stage is also shown in Figure 4.47.
MagAmp STAGE FUNCTIONAL EQUIVALENT CIRCUIT
VR
INPUT
+VR
G = +2
INPUT
Σ
RESIDUE
0
G = -2
SWITCH POSITION
SHOWN FOR
NEGATIVE INPUT
-VR
+VR
+
RESIDUE
0
BIT OUTPUT
(GRAY CODE)
-VR
4.47
a
A 3-bit MagAmp folding ADC is shown in Figure 4.48, and the corresponding
residue waveforms in Figure 4.49. Notice that there is no abrupt transition in any of
the folding stage output waveforms.
43
3-BIT MagAmp™ (FOLDING) ADC BLOCK DIAGRAM
±VR
ANALOG
INPUT
MAGAMP
2
MAGAMP
1
SHA
BIT 1
+
-
BIT 2
BIT 3
GRAY CODE REGISTER
3
GRAY-TO-BINARY CONVERTER
3
OUTPUT REGISTER
3
a
4.48
INPUT AND RESIDUE WAVEFORMS
FOR 3-BIT MagAmp ADC
INPUT
+V R
0
-V R
+V R
R1
0
-VR
+V R
R2
0
-V R
GRAY
CODE
000
001
011
010
110
111
101
100
4.49
a
The key to operating this architecture at high speeds is the folding stage. Early
designs (see References 7, 8, 9) used discrete op amps with diodes inside the
feedback loop to generate the folding transfer function. Modern IC circuit designs
implement the transfer function using current-steering open-loop gain techniques
44
which can be made to operate much faster. Fully differential stages (including the
SHA) also provide speed, lower distortion, and yield 8-bit accurate folding stages
with no requirement for thin film resistor laser trimming.
An example of a fully differential gain-of-two MagAmp folding stage is shown in
Figure 4.50 (see References 11, 12, 13). The differential input signal is applied to the
degenerated-emitter differential pair Q1,Q2 and the comparator. The differential
input voltage is converted into a differential current which flows in the collectors of
Q1, Q2. If +IN is greater than –IN, cascode-connected transistors Q3, Q6 are on, and
Q4, Q6 are off. The differential signal currents therefore flow through the collectors
of Q3, Q6 into level-shifting transistors Q7, Q8 and into the output load resistors,
developing the differential output voltage between +OUT and –OUT. The overall
differential voltage gain of the circuit is two.
If +IN is less than –IN (negative differential input voltage), the comparator changes
stage and turns Q4, Q5 on and Q3, Q6 off. The differential signal currents flow from
Q5 to Q7 and from Q4 to Q8, thereby maintaining the same relative polarity at the
differential output as for a positive differential input voltage. The required offset
voltage is developed by adding a current IOFF to the emitter current of Q7 and
subtracting it from the emitter current of Q8.
The differential residue output voltage of the stage drives the next stage input, and
the comparator output represents the Gray code output for the stage.
CIRCUIT DETAILS OF MagAmp STAGE
+5V
2I - IOFF
2I + IOFF
Q7
VBIAS
Q3
Q8
VBIAS
Q4
Q5
Q6
+OUT
-OUT
i+
GRAY
GRAY
-
+
+IN
i-
-IN
Q2
Q1
R
I
R
a
R
I
4.50
The MagAmp architecture can be extended to sampling rates previously dominated
by flash converters. The AD9059 8-bit, 60MSPS dual ADC is shown in Figure 4.51.
The first five bits (Gray code) are derived from five differential MagAmp stages. The
differential residue output of the fifth MagAmp stage drives a 3-bit flash converter,
rather than a single comparator. The Gray-code output of the five MagAmps and the
45
binary-code output of the 3-bit flash are latched, all converted into binary, and
latched again in the output data register. Key specifications for the AD9059 are
shown in Figure 4.52.
AD9059 DUAL 8-BIT, 60MSPS ADC FUNCTIONAL DIAGRAM
ANALOG
INPUT
SHA
MAGAMP
1
MAGAMP
2
MAGAMP
3
MAGAMP
4
MAGAMP
5
BIT
1
GRAY
BIT
2
GRAY
BIT
3
GRAY
BIT
4
GRAY
BIT
5
GRAY
DIFFERENTIAL
OUTPUTS ON
BITS 1 - 5
3-BIT
FLASH
ADC
3
BINARY
REGISTER
8
GRAY-TO-BINARY CONVERTER
8
OUTPUT REGISTER
8
a
4.51
AD9059 DUAL 8-BIT, 60MSPS ADC
KEY SPECIFICATIONS
n
Input Range: 1V p-p, Vcm = +2.5V
n
Input Impedance: 200kΩ
Ω || 5pF
n
ENOB: 7.3 @ 10.3MHz Input
n
On-Chip Reference
n
Power Supply: Single +5V Supply (+5 or +3V Digital)
n
Power Dissipation: 375mW (Power Down: 10mW)
n
Package: 28-lead SSOP
n
Ideal for Quadrature Demodulation in DBS
Set-Top Boxes
a
4.52
46
47
REFERENCES
1.
Active and Passive Electrical Wave Filter Catalog, Vol. 34, TTE,
Incorporated, 2251 Barry Avenue, Los Angeles, CA 90064.
2.
W. R. Bennett, “Spectra of Quantized Signals”, Bell System Technical
Journal, No. 27, July 1948, pp. 446-472.
3.
Steve Ruscak and Larry Singer, Using Histogram Techniques to
Measure A/D Converter Noise, Analog Dialogue, Vol. 29-2, 1995.
4.
M.J. Tant, The White Noise Book, Marconi Instruments, July 1974.
5.
G.A. Gray and G.W. Zeoli, Quantization and Saturation Noise due
to A/D Conversion, IEEE Trans. Aerospace and Electronic
Systems, Jan. 1971, pp. 222-223.
6.
Chuck Lane, A 10-bit 60MSPS Flash ADC, Proceedings of the 1989
Bipolar Circuits and Technology Meeting, IEEE Catalog No.
89CH2771-4, September 1989, pp. 44-47.
7.
F.D. Waldhauer, Analog to Digital Converter, U.S. Patent
3-187-325, 1965.
8.
J.O. Edson and H.H. Henning, Broadband Codecs for an Experimental
224Mb/s PCM Terminal, Bell System Technical Journal, 44,
November 1965, pp. 1887-1940.
9.
J.S. Mayo, Experimental 224Mb/s PCM Terminals, Bell System
Technical Journal, 44, November 1965, pp. 1813-1941.
10.
Hermann Schmid, Electronic Analog/Digital Conversions,
Van Nostrand Reinhold Company, New York, 1970.
11.
Carl Moreland, An 8-bit 150MSPS Serial ADC, 1995 ISSCC Digest
of Technical Papers, Vol. 38, p. 272.
12.
Roy Gosser and Frank Murden, A 12-bit 50MSPS Two-Stage A/D
Converter, 1995 ISSCC Digest of Technical Papers, p. 278.
13.
Carl Moreland, An Analog-to-Digital Converter Using SerialRipple Architecture, Masters' Thesis, Florida State University
College of Engineering, Department of Electrical Engineering, 1995.
14.
Practical Analog Design Techniques, Analog Devices, 1995, Chapter
4, 5, and 8.
15.
Linear Design Seminar, Analog Devices, 1995, Chapter 4, 5.
16.
System Applications Guide, Analog Devices, 1993, Chapter 12, 13,
15,16.
48
17.
Amplifier Applications Guide, Analog Devices, 1992, Chapter 7.
18.
Walt Kester, Drive Circuitry is Critical to High-Speed Sampling ADCs,
Electronic Design Special Analog Issue, Nov. 7, 1994, pp. 43-50.
19.
Walt Kester, Basic Characteristics Distinguish Sampling A/D Converters,
EDN, Sept. 3, 1992, pp. 135-144.
20.
Walt Kester, Peripheral Circuits Can Make or Break Sampling ADC
Systems, EDN, Oct. 1, 1992, pp. 97-105.
21.
Walt Kester, Layout, Grounding, and Filtering Complete Sampling
ADC System, EDN, Oct. 15, 1992, pp. 127-134.
22.
Robert A. Witte, Distortion Measurements Using a Spectrum Analyzer,
RF Design, September, 1992, pp. 75-84.
23.
Walt Kester, Confused About Amplifier Distortion Specs?, Analog
Dialogue, 27-1, 1993, pp. 27-29.
24.
System Applications Guide, Analog Devices, 1993, Chapter 16.
25.
Frederick J. Harris, On the Use of Windows for Harmonic Analysis
with the Discrete Fourier Transform, IEEE Proceedings, Vol. 66, No. 1,
Jan. 1978, pp. 51-83.
26.
Joey Doernberg, Hae-Seung Lee, David A. Hodges, Full Speed Testing
of A/D Converters, IEEE Journal of Solid State Circuits, Vol. SC-19,
No. 6, Dec. 1984, pp. 820-827.
27.
Brendan Coleman, Pat Meehan, John Reidy and Pat Weeks, Coherent
Sampling Helps When Specifying DSP A/D Converters, EDN, October 15,
1987, pp. 145-152.
28.
Robert W. Ramierez, The FFT: Fundamentals and Concepts,
Prentice-Hall, 1985.
29.
R. B. Blackman and J. W. Tukey, The Measurement of Power
Spectra, Dover Publications, New York, 1958.
30.
James J. Colotti, Digital Dynamic Analysis of A/D Conversion
Systems Through Evaluation Software Based on FFT/DFT Analysis,
RF Expo East 1987 Proceedings, Cardiff Publishing Co., pp. 245-272.
31.
HP Journal, Nov. 1982, Vol. 33, No. 11.
32.
HP Product Note 5180A-2.
33.
HP Journal, April 1988, Vol. 39, No. 2.
34.
HP Journal, June 1988, Vol. 39, No. 3.
49
35.
Dan Sheingold, Editor, Analog-to-Digital Conversion Handbook,
Third Edition, Prentice-Hall, 1986.
36.
Lawrence Rabiner and Bernard Gold, Theory and Application of
Digital Signal Processing, Prentice-Hall, 1975.
37.
Matthew Mahoney, DSP-Based Testing of Analog and Mixed-Signal
Circuits, IEEE Computer Society Press, Washington, D.C., 1987.
38.
IEEE Trial-Use Standard for Digitizing Waveform Recorders,
No. 1057-1988.
39.
Richard J. Higgins, Digital Signal Processing in VSLI, Prentice-Hall,
1990.
40.
M. S. Ghausi and K. R. Laker, Modern Filter Design: Active RC and
Switched Capacitors, Prentice Hall, 1981.
41.
Mathcad™ 4.0 software package available from MathSoft, Inc.,
201 Broadway, Cambridge MA, 02139.
42.
Howard E. Hilton, A 10MHz Analog-to-Digital Converter with 110dB
Linearity, H.P. Journal, October 1993, pp. 105-112.
50
SECTION 5
HIGH SPEED ADC APPLICATIONS
Walt Kester, Brad Brannon, Paul Hendricks
DRIVING ADC INPUTS FOR LOW DISTORTION AND WIDE
DYNAMIC RANGE
In order to achieve wide dynamic range in high speed ADC applications, careful
attention must be given to the analog interface. Many ADCs are designed so that
analog signals can be interfaced directly to their inputs without the necessity of a
drive amplifier. This is especially true in ADCs such as the AD9220/21/23 family
and the AD9042, where even a low distortion drive amplifier may result in some
degradation in AC performance. If a buffer amplifier is required, it must be carefully
selected so that its distortion and noise performance is better than that of the ADC.
Single-supply ADCs generally yield optimum AC performance when the commonmode input voltage is centered between the supply rails (although the optimum
common-mode voltage may be skewed slightly in either direction about this point
depending upon the particular design). This also eases the drive requirement on the
input buffer amplifier (if required) since even "rail-to-rail" output op amps give best
distortion performance if their output is centered about mid-supply, and the peak
signals are kept at least 1V from either rail.
Typical high speed single-supply ADC peak-to-peak input voltage ranges may vary
from about 0.5V to 5V, but in most cases, 1V to 2V peak-to-peak represents the
optimum tradeoff between noise and distortion performance.
In single-supply applications requiring DC coupling, careful attention must be given
to the input and output common-mode range of the driving amplifier. Level shifting
is often required in order to center a ground-referenced signal within the allowable
common-mode input range of the ADC.
Small RF transformers are quite useful in AC coupled applications, especially if the
ADC has differential inputs. Significant improvement in even-order distortion
products and common-mode noise rejection may be realized, depending upon the
characteristics of the ADC.
An understanding of the input structure of the ADC is therefore necessary in order
to properly design the analog interface circuitry. ADCs designed on CMOS processes
typically connect the sample-and-hold switches directly to the analog input, thereby
generating transient current pulses. These transients may significantly degrade
performance if the settling time of the op amp is not sufficiently fast. On the other
hand, ADCs designed on bipolar processes may present a relatively benign load to
the drive amplifier with minimal transient currents.
The data sheet for the ADC is the prime source an engineer should use in designing
the interface circuits. It should contain recommended interface circuits and spell out
relevant tradeoffs. However, no data sheet can substitute for a fundamental
understanding of what's inside the ADC.
1
HIGH SPEED ADC INPUT CONSIDERATIONS
n
Selection of Drive Amplifier (Only if Needed!)
n
Single Supply Implications
n
Input Range (Span): Typically 1V to 2V peak-to-peak
for best distortion / noise tradeoff
n
Input Common-Mode Range:
Vs / 2 (Nominally) for Single Supply ADCs
n
Differential vs. Single-Ended
n
AC Coupling Using Transformers
n
Input Transient Currents
a
5.1
Switched-Capacitor Input ADCs
The AD9220/21/23-series of ADCs are excellent examples of the progress that has
been made in utilizing low-cost CMOS processes to achieve a high level of
performance. A functional block diagram is shown in Figure 5.2. This family of
ADCs offers sampling rates of 1.25MSPS (AD9221), 3MSPS (AD9223), and 10MSPS
(AD9220) at power dissipations of 60, 100, and 250mW respectively. Key
specifications for the family of ADCs are given in Figure 5.3. The devices contain an
on-chip reference voltage which allows the full scale span to be set at 2V or 5V peakto-peak (full scale spans between 2V and 5V can be set by adding two external gain
setting resistors).
2
AD922X-SERIES ADC FUNCTIONAL DIAGRAM
CLK
AVDD
DVDD
SHA
VINA
MDAC1
GAIN = 16
VINB
A/D
CAPT
MDAC2
GAIN = 8
5
4
A/D
5
MDAC3
GAIN = 4
4
CAPB
3
A/D
A/D
3
3
DIGITAL CORRECTION LOGIC
12
VREF
OTR
OUTPUT BUFFERS
SENSE
BIT 1
(MSB)
1V
MODE
SELECT
AD9221/23/20
REFCOM
AVSS
DVSS
BIT 12
(LSB)
CML
a
5.2
AD9220, AD9221, AD9223
CMOS 12-BIT ADCs KEY SPECIFICATIONS
n
Family Members:
AD9221 (1.25MSPS), AD9223 (3MSPS), AD9220 (10MSPS)
n
Power Dissipation: 60, 100, 250mW, Respectively
n
FPBW: 25, 40, 60MHz, Respectively
n
Effective Input Noise: 0.1LSB rms (Span = 5V)
n
SINAD: 71dB
n
SFDR: 88dBc
n
On-Chip Reference
n
Differential Non-Linearity: 0.3LSB
n
Single +5V Supply
n
28-Pin SOIC Package
3
a
5.3
The input circuit of the AD9220/21/23-series of CMOS ADCs contains the
differential sample-and-hold as shown in Figure 5.4. The switches are shown in the
track mode. They open and close at the sampling frequency. The 16pF capacitors
represent the effective capacitance of switches S1 and S2 plus the stray input
capacitance. The Cs capacitors (4pF) are the sampling capacitors, and the CH
capacitors are the hold capacitors. Although the input circuit is completely
differential, the ADC can be driven either single-ended or differential. Optimum
SFDR, however, is obtained using a differential transformer drive.
SIMPLIFIED INPUT CIRCUIT OF AD922X ADC FAMILY
CH
S6
4pF
16pF
CP
S4
CS
S1
VINA
4pF
+
S3
A
CS
S2
VINB
S5
4pF
CP
16pF
CH
S7
4pF
SWITCHES SHOWN IN TRACK MODE
a
5.4
In the track mode, the differential input voltage is applied to the Cs capacitors.
When the circuit enters the hold mode, the voltage across the sampling capacitors is
transferred to the CH hold capacitors and buffered by the amplifier A. (The switches
are controlled by the appropriate phases of the sampling clock). When the SHA
returns to the track mode, the input source must charge or discharge the voltage
stored on Cs to the new input voltage. This action of charging and discharging Cs,
averaged over a period of time and for a given sampling frequency fs, makes the
input impedance appear to have a benign resistive component. However, if this
action is analyzed within a sampling period (1/fs), the input impedance is dynamic,
and hence certain precautions on the input drive source should be observed.
The resistive component to the input impedance can be computed by calculating the
average charge that is drawn by CH from the input drive source. It can be shown
that if Cs is allowed to fully charge to the input voltage before switches S1 and S2
are opened that the average current into the input is the same as if there were a
4
resistor equal to 1/(Csfs) connected between the inputs. Since Cs is only a few
picofarads, this resistive component is typically greater than several kΩ for an fs =
10MSPS.
If one considers the SHA's input impedance over a sampling period, it appears as a
dynamic load to the input drive source. When the SHA returns to the track mode,
the input source should ideally provide the charging current through the Ron of
switches S1 and S2 in an exponential manner. The requirement of exponential
charging means that the source impedance should be both low and resistive up to
and beyond the sampling frequency.
The output impedance of an op amp can be modeled as a series inductor and
resistor. When a capacitive load is switched onto the output of the op amp, the
output will momentarily change due to its effective high frequency output
impedance. As the output recovers, ringing may occur. To remedy this situation, a
series resistor can be inserted between the op amp and the SHA input. The optimum
value of this resistor is dependent on several factors including the sampling
frequency and the op amp selected, but in most applications, a 30 to 50Ω resistor is
optimum.
The input voltage span of the AD922X-family is set by pin-strap options using the
internal voltage reference (see Figure 5.5). The common-mode voltage can be set by
either pin strap or applying the common-mode voltage to the VINB pin. Tradeoffs
can be made between noise and distortion performance. Maximum input range
allowable is 5V peak-to-peak, in which case, the common-mode input voltage must
be one-half the supply voltage, or +2.5V. The minimum input range is 2V peak-topeak, in which case the common-mode input voltage can be set from +1V to +4V. For
best DC linearity and maximum signal-to-noise ratio, the ADC should be operated
with an input signal of 5V peak-to-peak. However, for best high frequency noise and
distortion performance, 2V peak-to-peak with a common-mode voltage of +2.5V is
preferred. This is because the CMOS FET on-resistance is a minimum at this
voltage, and the non-linearity caused by the signal-dependence of Ron (Ron
modulation effect) is also minimal.
AD922X ADC INPUT VOLTAGE RANGE OPTIONS
SINGLE-ENDED INPUT
Input Signal Range
Peak-to-Peak Signal
Common-Mode Voltage
(Volts)
(Volts)
(Volts)
0 to +2
2
+1
0 to +5
5
+2.5
+1.5 to +3.5
2
+2.5
DIFFERENTIAL INPUT
Input Signal Range
Peak-to-Peak Signal
5
Common-Mode Voltage
(Volts)
Differential (Volts)
(Volts)
+2 to +3
2
+2.5
+1.25 to +3.75
5
+2.5
a
5.5
Figure 5.6 shows the THD performance of the AD9220 for a 2V peak-to-peak input
signal span and common-mode input voltage of 2.5V and 1V. The data was taken
with a single-ended drive. Note that the performance is significantly better for Vcm
= +2.5V.
AD9220 THD VS. INPUT FREQUENCY: SINGLE-ENDED DRIVE
2V p-p INPUT, Vcm = +1V AND Vcm = +2.5V, fs = 10MSPS
50
THD
(dBc)
60
Vcm = +1V
70
Vcm = +2.5V
80
90
0.1
0.2
0.5
1
2
5
10
INPUT FREQUENCY (MHz)
a
5.6
A simple single-ended circuit for AC coupling into the inputs of the AD9220-family is
shown in Figure 5.7. Note that the common-mode input voltage is set for +2.5V by
the 4.99kΩ resistors. The input impedance is also balanced for optimum distortion
performance.
6
SINGLE-ENDED AC-COUPLED
DRIVE CIRCUIT FOR AD922X ADC
+5V
0.1µ
µF
+5V
AD922X
INPUT
10µ
µF
4.99kΩ
Ω
33 Ω
+
51.1Ω
Ω
VINA
4.99kΩ
Ω
+5V
4.99kΩ
Ω
+2.5V
VINB
+
10µ
µF
0.1µ
µF
33 Ω
4.99kΩ
Ω
a
5.7
If the input to the ADC is coming from a long coaxial cable run, it may be desirable
to buffer the transient currents at the ADC inputs from the cable to prevent
problems resulting from reflections, especially if the cable is not source-terminated.
The circuit shown in Figure 5.8 uses the low distortion AD8011 op amp as a buffer
which can optionally provide signal gain. In all cases, the feedback resistor should be
fixed at 1kΩ for best op amp performance, since the AD8011 is a current-feedback
type. In this type of arrangement, care must be taken to observe the allowable input
and output range of the op amp. The AD8011 input common-mode range (operating
on a single +5V supply) is from +1.5 to +3.5V, and its output +1V to +4V. The ADC
should be operated with a 2V peak-to-peak input range. The 33Ω series resistor is
required to isolate the output of the AD8011 from the effective input capacitance of
the ADC. The value was empirically determined to yield the best high-frequency
SINAD.
7
BUFFERED AC-COUPLED INPUT DRIVE
CIRCUIT FOR AD922X ADC
0.1µ
µF
10µ
µF
INPUT
+5V
+5V
4.99kΩ
Ω
+5V
+
1Vp-p
51.1Ω
Ω
2Vp-p
33 Ω
AD8011
4.99kΩ
Ω
AD922X
VINA
-
1k Ω
1k Ω
+
10µ
µF
0.1µ
µF
+5V
4.99kΩ
Ω
+2.5V
33 Ω
VINB
+
10µ
µF
0.1µ
µF
a
4.99kΩ
Ω
5.8
Direct coupling of ground-referenced signals using a single supply requires the use of
an op amp with an acceptable common-mode input voltage, such as the AD8041
(input can go to 200mV below ground). The circuit shown in Figure 5.9 level shifts
the ground-referenced bipolar input signal to a common-mode voltage of +2.5V at
the ADC input. The common-mode bias voltage of +2.5V is developed directly from
an AD780 reference, and the AD8041 common-mode voltage of +1.25V is derived
with a simple divider.
8
DIRECT-COUPLED LEVEL SHIFTER
FOR DRIVING AD922X ADC INPUT
1k Ω
+5V
+5V
-
±1V
AD922X
33 Ω
AD8041
V INA
52.3Ω
Ω
+2.5V
+
±
1k Ω
INPUT
1V
+1.25V
+5V
AD780
2.5V
REF.
1kΩ
Ω
10µ
µF
+
1kΩ
Ω
µF
0.1µ
+2.5V
33 Ω
VINB
+
10µ
µF
a
0.1µ
µF
5.9
Transformer coupling provides the best CMR and the lowest distortion. Figure 5.10
shows the suggested circuit. The transformer is a Mini-Circuits RF transformer,
model #T4-6T which has an impedance ratio of four (turns ratio of 2). The schematic
assumes that the signal source has a 50Ω source impedance. The 1:4 impedance
ratio requires the 200Ω secondary termination for optimum power transfer and
VSWR. The Mini-Circuits T4-6T has a 1dB bandwidth from 100kHz to 100MHz. The
center tap of the transformer provides a convenient means of level shifting the input
signal to the optimum common-mode voltage. The AD922X CML pin is used to
provide the +2.5 common-mode voltage.
9
TRANSFORMER COUPLING INTO AD922X ADC
+5V
RF TRANSFORMER:
MINI-CIRCUITS T4-6T
33 Ω
1:2
AD922X
VINA
49.9Ω
Ω
200Ω
Ω
2Vp-p
33 Ω
VINB
+2.5V
CML
0.1µ
µF
a
5.10
Transformers with other turns ratios may also be selected to optimize the
performance for a given application. For example, a given input signal source or
amplifier may realize an improvement in distortion performance at reduced output
power levels and signal swings. Hence, selecting a transformer with a higher
impedance ratio (i.e. Mini-Circuits #T16-6T with a 1:16 impedance ratio, turns ratio
1:4) effectively "steps up" the signal level thus reducing the driving requirements of
the signal source.
Note the 33Ω series resistors inserted between the transformer secondary and the
ADC input. These values were specifically selected to optimize both the SFDR and
the SNR performance of the ADC. They also provide isolation from transients at the
ADC inputs. Transients currents are approximately equal on the VINA and VINB
inputs, so they are isolated from the primary winding of the transformer by the
transformer's common-mode rejection.
Transformer coupling using a common-mode voltage of +2.5V provides the
maximum SFDR when driving the AD922X-series. By driving the ADC
differentially, even-order harmonics are reduced compared with the single-ended
circuit. Figure 5.11 shows a plot of SFDR and SNR for the transformer-coupled
differential drive circuit using 2V p-p and 5V p-p inputs and a common-mode voltage
of +2.5V. Note that the SFDR is greater than 80dBc for input signals up to full scale
with a 5MHz input signal.
10
AD9220 SFDR AND SNR FOR 5Vp-p AND 2Vp-p INPUT:
Vcm = +2.5V, 5MHz INPUT, f s = 10MSPS
TRANSFORMER-COUPLED DIFFERENTIAL DRIVE
90
SNR - dB AND SFDR - dB
80
SFDR - 5.0
Vp-p
70
SFDR - 2.0
Vp-p
60
50
40
SNR - 5.0
Vp-p
SNR - 2.0
Vp-p
30
20
-50
-40
-30
INPUT AMPLITUDE - dBFS
-20
a
-10
0
5.11
Figure 5.11 also shows differences between the SFDR and SNR performance for 2V
p-p and 5V p-p inputs. Note that the SNR with a 5V p-p input is approximately 2dB
to 3dB better than that for a 2V p-p input because of the additional dynamic range
provided by the larger input range. Also, the SFDR performance using a 5V p-p
input is 3 to 5dB better for signals between about –6dBFS and –36dBFS. This
improvement in SNR and SFDR for the 5V p-p input range may be advantageous in
systems which require more than 6dB headroom to minimize clipping of the ADC.
Driving Bipolar Input ADCs
Bipolar technology is typically used for extremely high performance ADCs with wide
dynamic range and high sampling rates such as the AD9042. The AD9042 is a stateof-the-art 12-bit, 41MSPS two stage subranging ADC consisting of a 6-bit coarse
ADC and a 7-bit residue ADC with one bit of overlap to correct for any DNL, INL,
gain or offset errors of the coarse ADC, and offset errors in the residue path. A block
diagram is shown in Figure 5.12 and key specifications in Figure 5.13. A proprietary
gray-code architecture is used to implement the two internal ADCs. The gain
alignments of the coarse and residue, likewise the subtraction DAC, rely on the
statistical matching of the devices on the process. As a result, 12-bit integral and
differential linearity is obtained without laser trim. The internal DAC consists of
126 interdigitated current sources. Also on the DAC reference, there are an
additional 20 interdigitated current sources to set the coarse gain, residue gain, and
full scale gain. The interdigitization removes the requirement for laser trim. The
AD9042 is fabricated on a high speed dielectrically isolated complementary bipolar
process. The total power dissipation is only 575mW when operating on a single +5V
supply.
11
AD9042 12-BIT, 41MSPS ADC BLOCK DIAGRAM
ANALOG
INPUT
SHA
2
SHA
1
GAIN
+
-
7-BIT
ADC
6-BIT
DAC
6-BIT
ADC
SHA
3
6
BUFFER
REGISTER
7
6
ERROR CORRECTION LOGIC
12
OUTPUT REGISTERS
12
a
5.12
AD9042 12-BIT, 41MSPS ADC KEY SPECIFICATIONS
n
Input Range: 1V peak-to-peak, Vcm = +2.4V
n
Input Impedance: 250Ω
Ω to Vcm
n
Effective Input Noise: 0.33LSBs rms
n
SFDR at 20MHz Input: 80dB
n
SINAD at 20MHz Input = 66dB
n
Digital Outputs: TTL Compatible
n
Power Supply: Single +5V
n
Power Dissipation: 575mW
n
Fabricated on High Speed Dielectrically Isolated
Complementary Bipolar Process
a
5.13
The outstanding performance of the AD9042 is partly due to the use of differential
techniques throughout the device. The low distortion input amplifier converts the
single-ended input signal into a differential one. If maximum SFDR performance is
12
desired, the signal source should be coupled directly into the input of the AD9042
without using a buffer amplifier. Figure 5.14 shows a method using capacitive
coupling. Transformer coupling can also be used if desired.
INPUT STRUCTURE OF AD9042 ADC IS
DESIGNED TO BE DRIVEN DIRECTLY FROM 50Ω
SOURCE FOR BEST SFDR
AD9042
250Ω
INPUT =
1V p-p
FROM 50Ω
SOURCE
250Ω
-
RT
61.9Ω
+
5.14
a
The AD9050 is a 10-bit, 40MSPS single supply ADC designed for wide dynamic
range applications such as ultrasound, instrumentation, digital communications,
and professional video. Like the AD9042, it is fabricated on a high speed
complementary bipolar process. A block diagram of the AD9050 (Figure 5.15)
illustrates the two-step subranging architecture, and key specifications are
summarized in Figure 5.16.
13
AD9050 10-BIT, 40MSPS SINGLE SUPPLY ADC
ENCODE
AMP
ARRAY
AIN
AIN
VREFOUT
BANDGAP
REFERENC
E
5-BIT
ADC
6-BIT
ADC
VREFIN
ERROR
DECODE CORRECTION DECODE
LOGIC
LOGIC
REFBP
10
AD9050
a
5.15
AD9050 10-BIT, 40MSPS ADC KEY SPECIFICATIONS
n
10-Bits, 40MSPS, Single +5V Supply
n
Selectable Digital Supply: +5V, or +3V
n
Low Power: 300mW on BiCMOS Process
n
On-Chip SHA and +2.5V reference
n
56dB S/(N+D), 9 Effective Bits, with 10.3MHz Input Signal
n
No input transients, Input Impedance 5kΩ
Ω, 5pF
n
Input Range +3.3V ±0.5V Single-Ended or Differential
n
28-pin SOIC / SSOP Packages
n
Ideal for Digital Beamforming Ultrasound Systems
a
5.16
The analog input circuit of the AD9050 (see Figure 5.17) is differential, but can be
driven either single-endedly or differentially with equal performance. The input
signal range of the AD9050 is ±0.5V centered around a common-mode voltage of
14
+3.3V, which makes single supply op amp selection more difficult since the amplifier
has to drive +3.8V peak signals with low distortion.
AD9050 SIMPLIFIED INPUT CIRCUIT
+5V
8kΩ
8kΩ
INPUT BUFFER
AIN(A)
AIN(B)
16kΩ
16kΩ
INPUT RANGE:
+3.3V ± 0.5V
GND
a
5.17
The input circuit of the AD9050 is a relatively benign and constant 5kΩ in parallel
with approximately 5pF. Because of its well-behaved input, the AD9050 can be
driven directly from 50, 75, or 100Ω sources without the need for a low-distortion
buffer amplifier. In ultrasound applications, it is normal to AC couple the signal
(generally between 1MHz and 15MHz) into the AD9050 differential inputs using a
wideband transformer as shown in Figure 5.18. The Mini-Circuits T1-1T
transformer has a 1dB bandwidth from 200kHz to 80MHz. Signal-to-noise plus
distortion (SINAD) values of 57dB (9.2 ENOB) are typical for a 10MHz input signal.
If the input signal comes directly from a 50, 75, or 100Ω single-ended source,
capacitive coupling as shown in Figure 5.18 can be used.
15
AC COUPLING INTO THE INPUT OF THE AD9050 ADC
+5V
8kΩ
+5V
8kΩ
8kΩ
8kΩ
T1
50Ω
50Ω
16kΩ
16kΩ
16kΩ
16kΩ
T1:
MINI-CIRCUITS
T1 - 1T
CAPACITIVE
COUPLING
TRANSFORMER
COUPLING
a
5.18
If DC coupling is required, the AD8041 (zero-volt in, rail-to-rail output) op amp can
be used as a low distortion driver. The circuit shown in Figure 5.19 level shifts a
ground-referenced video signal to fit the +3.3V ±0.5V input range of the AD9050.
The source is a ground-referenced 0 to +2V signal which is series-terminated in 75Ω.
The termination resistor, RT, is chosen such that the parallel combination of RT and
R1 is 75Ω. The AD8041 op amp is configured for a signal gain of –1. Assuming that
the video source is at zero volts, the corresponding ADC input voltage should be
+3.8V. The common-mode voltage, Vcm, is determined from the following equation:
 R s||R T + R1 
38.8 + 1000


V cm = 3.8
 = 3.8
 = 1.94 V
 38.8 + 1000 + 1000 
 R s||R T + R1 + R2 
The common-mode voltage, Vcm, is derived from the common-mode voltage at the
inverting input of the AD9050. The +3.3V is buffered by the AD820 single-supply
FET-input op amp. A divider network generates the required +1.94V for the
AD8011, and a potentiometer provides offset adjustment capability.
The AD8041 voltage feedback op amp was chosen because of its low power (26mW),
wide bandwidth (160MHz), and low distortion (–69dBc at 10MHz). It is fully
specified for both ±5V,+5V, and +3V operation. When operating on a single +5V
supply, the input common-mode range is –0.2V to +4V, and the output swing is
+0.1V to +4.9V. Distortion performance of the entire circuit including the ADC is
better than –60dBc for an input frequency of 10MHz and a sampling rate of
40MSPS.
16
DC-COUPLED SINGLE-SUPPLY DRIVE CIRCUIT FOR
AD9050 10-BIT, 40MSPS ADC USING AD8041 OP AMP
+5V
R2
1000Ω
Ω
72mV TO 1.072V
RS
75 Ω
0 TO
+2V
0.1µ
µF
+5V
R1
8k Ω
0.1µ
µF
AIN(A)
3.8V TO 2.8V
1000Ω
Ω
AD8041
RT
80.6Ω
Ω
AD9050
+
16kΩ
Ω
+5V
AIN(B)
0.1µ
µF
8k Ω
+3.3V
+
10µ
µF
0.1µ
µF
+5V
VCM = +1.94V
1000Ω
Ω
16kΩ
Ω
+
AD820
50 Ω
249Ω
Ω
-
365Ω
Ω
OFFSET
ADJUST
a
5.19
17
APPLICATIONS OF HIGH SPEED ADCS IN CCD IMAGING
Charge coupled devices (CCDs) contains a large number of small photocells called
photosites or pixels which are arranged either in a single row (linear arrays) or in a
matrix (area arrays). CCD area arrays are commonly used in video applications,
while linear arrays are used in facsimile machines, graphics scanners, and pattern
recognition equipment.
The linear CCD array consists of a row of image sensor elements (photosites, or
pixels) which are illuminated by light from the object or document. During one
exposure period each photosite acquires an amount of charge which is proportional
to its illumination. These photosite charge packets are subsequently switched
simultaneously via transfer gates to an analog shift register. The charge packets on
this shift register are clocked serially to a charge detector (storage capacitor) and
buffer amplifier (source follower) which convert them into a string of photodependent output voltage levels (see Figure 5.20). While the charge packets from
one exposure are being clocked out to the charge detector, another exposure is
underway. The analog shift register typically operates at frequencies between 1 and
10MHz.
LINEAR CCD ARRAY
EXPOSURE
CLOCKS
PHOTO-SITES (PIXELS)
TRANSFER
CLOCKS
TRANSFER GATE
RESET
LEVEL
+V
SHIFT
CLOCKS
CCD
OUTPUT
ANALOG TRANSPORT
SHIFT REGISTER
CH
SAMPLE VIDEO/
SAMPLE RESET
FET SWITCH
a
-V
5.20
The charge detector readout cycle begins with a reset pulse which causes a FET
switch to set the output storage capacitor to a known voltage. Switching the FET
causes capacitive feedthrough which results in a reset glitch at the output as shown
in Figure 5.21. The switch is then opened, isolating the capacitor, and the charge
from the last pixel is dumped onto the capacitor causing a voltage change. The
difference between the reset voltage and the final voltage (video level) shown in
Figure 5.21 represents the amount of charge in the pixel. CCD charges may be as
18
low as 10 electrons, and a typical CCD output sensitivity is 0.6µV/electron. Most
CCDs have a saturation output voltage of about 1V (see Reference 1).
CCD OUTPUT WAVEFORM
RESET
GLITCH
CCD
OUTPUT
VIDEO
LEVEL
RESET
LEVEL
RESET
LEVEL
∆V
RESET
LEVEL
VIDEO
LEVEL
VIDEO
LEVEL
t
PIXEL PERIOD
a
5.21
Since CCDs are generally fabricated on MOS processes, they have limited capability
to perform on-chip signal conditioning. Therefore, the CCD output is generally
processed by external conditioning circuits.
CCD output voltages are small and quite often buried in noise. The largest source of
noise is the thermal noise in the resistance of the FET reset switch. This noise may
have a typical value of 100 to 300 electrons rms (approximately 60 to 180mV rms).
This noise occurs as a sample-to-sample variation in the CCD output level and is
common to both the reset level and the video level for a given pixel period. A
technique called correlated double sampling (CDS) is often used to reduce the effect
of this noise. Figure 5.22 shows two circuit implementations of the CDS scheme. In
the top circuit, the CCD output drives both SHAs. At the end of the reset interval,
SHA1 holds the reset voltage level. At the end of the video interval, SHA2 holds the
video level. The SHA outputs are applied to a difference amplifier which subtracts
one from the other. In this scheme, there is only a short interval during which both
SHA outputs are stable, and their difference represents ∆V, so the difference
amplifier must settle quickly to the desired resolution.
Another arrangement is shown in the bottom half of Figure 5.22, which uses three
SHAs and allows either for faster operation or more time for the difference amplifier
to settle. In this circuit, SHA1 holds the reset level so that it occurs simultaneously
with the video level at the input to SHA2 and SHA3. When the video clock is applied
simultaneously to SHA2 and SHA3, the input to SHA2 is the reset level, and the
input to SHA3 the video level. This arrangement allows the entire pixel period (less
the acquisition time of SHA2 and SHA3) for the difference amplifier to settle.
19
CORRELATED DOUBLE SAMPLING (CDS)
MINIMIZES SWITCHING NOISE AT OUTPUT
METHOD #1
SHA 1
CCD
OUTPUT
RESET CLOCK
-
VIDEO CLOCK
+
OUTPUT
SHA 2
SHA 1
CCD
OUTPUT
RESET
CLOCK
SHA 2
METHOD #2
-
VIDEO
CLOCK
OUTPUT
+
SHA 3
a
5.22
The AD9807 is a complete CCD imaging decoder and signal processor on a single
chip (see Figure 5.23). The input of the AD9807 allows direct AC coupling of the
CCD outputs and includes all the circuitry to perform three-channel correlated
double sampling (CDS) and programmable gain adjustment (1X to 4X in 16
increments) of the CCD output. A 12-bit ADC quantizes the analog signal
(maximum sampling frequency 6MSPS). After digitization, the on-board DSP allows
pixel rate offset and gain correction. The DSP also corrects odd/even CCD register
imbalance errors. A parallel control bus provides a simple interface to 8-bit
microcontrollers. The device operates on a single +5V supply and dissipates 500mW.
The AD9807 comes in a space saving 64-pin plastic quad flat pack (PQFP). By
disabling the CDS, the AD9807 is also suitable for non-CCD applications that do not
require CDS. The AD9807 is also offered in a pin-compatible 10-bit version, the
AD9805, to allow upgradeability and simplify design issues across different scanner
models.
20
AVDD
AVSS
CAPB
CAPT
DVSS DVDD DRVDD DRVSS
CML
VREF
VINR
PGA
CDS
OFFSET<M:0> GAIN<N:0>
BANDGAP
REFERENCE
8-10
12-10
DIG ITAL
DIGIT AL
OEB
12
VING
CDS
PGA
12
12-BIT A/D
3:1 MUX
SUBTRACTOR
x
12
12
DOUT<11:0>
MULTI PLIER
8
CSB
INPUT
OF FS ET
REGISTER
VINB
CDS
PGA
RDB
ROD D
R E VEN
G
G O DD
G EV E N
B
BOD D
R
R
G
B
CONFI GURA TION
REGISTER
CONFI GURA TION
REGISTER
2
MPU
PORT
WRB
A0
B E VE N
A1
A2
CDSCLK1
CDSCLK2
STRTLN
ADCCLK
a
5.23
21
HIGH SPEED ADC APPLICATIONS IN DIGITAL
RECEIVERS
Introduction
Consider the analog superheterodyne receiver invented in 1917 by Major Edwin H.
Armstrong (see Figure 5.24). This architecture represented a significant
improvement over single-stage direct conversion (homodyne) receivers which had
previously been constructed using tuned RF amplifiers, a single detector, and an
audio gain stage. A significant advantage of the superhetrodyne receiver is that it is
much easier and more economical to have the gain and selectivity of a receiver at
fixed intermediate frequencies (IF) than to have the gain and frequency-selective
circuits "tune" over a band of frequencies.
U.S. ADVANCED MOBILE PHONE SERVICE (AMPS)
SUPERHETERODYNE ANALOG RECEIVER
AMPS: 416 CHANNELS ("A" OR "B" CARRIER)
30kHz WIDE, FM
12.5MHz TOTAL BANDWIDTH
1 CALLER/CHANNEL
RF
BPF
LNA
LO1
TUNED
70MHz
LO2
FIXED
10.7MHz
LO3
FIXED
455kHz
ANALOG
DEMOD, CHANNEL 1
FILTER
30kHz
1ST IF
2ND IF
SAME AS ABOVE
a
3RD IF
CHANNEL n
30kHz
5.24
The frequencies shown in Figure 5.24 correspond to the AMPS (Advanced Mobile
Phone Service) analog cellular phone system currently used in the U.S. The receiver
is designed for AMPS signals at 900MHz RF. The signal bandwidth for the "A" or
"B" carriers serving a particular geographical area is 12.5MHz (416 channels, each
30kHz wide). The receiver shown uses triple conversion, with a first IF frequency of
70MHz and a second IF of 10.7MHz, and a third of 455kHz. The image frequency at
the receiver input is separated from the RF carrier frequency by an amount equal to
twice the first IF frequency (illustrating the point that using relatively high first IF
frequencies makes the design of the image rejection filter easier).
The output of the third IF stage is demodulated using analog techniques
(discriminators, envelope detectors, synchronous detectors, etc.). In the case of
AMPS the modulation is FM. An important point to notice about the above scheme
22
is that there is one receiver required per channel, and only the antenna, prefilter, and
LNA can be shared.
It should be noted that in to make the receiver diagrams more manageable, the
interstage amplifiers are not shown. They are, however, an important part of the
receiver, and the reader should be aware that they must be present.
Receiver design is a complicated art, and there are many tradeoffs that can be made
between IF frequencies, single-conversion vs. double-conversion or triple conversion,
filter cost and complexity at each stage in the receiver, demodulation schemes, etc.
There are many excellent references on the subject, and the purpose of this section is
only to acquaint the design engineer with some of the emerging architectures,
especially in the application of digital techniques in the design of advanced
communications receivers.
A Receiver Using Digital Processing at Baseband
With the availability of high performance high speed ADCs and DSPs (such as
ADSP-2181 and the ADSP-21062), it is now becoming common practice to use
digital techniques in at least part of the receive and transmit path, and various
chipsets are available from Analog Devices to perform these functions for GSM and
other cellular standards. This is illustrated in Figure 5.25 where the output of the
last IF stage is converted into a baseband in-phase (I) and quadrature (Q) signal
using a quadrature demodulator. The I and Q signals are then digitized by two
ADCs. The DSPs then perform the additional signal processing. The signal can then
be converted into analog format using a DAC, or it can be processed, mixed with
other signals, upconverted, and retransmitted.
DIGITAL RECEIVER USING
BASEBAND SAMPLING AND DIGITAL PROCESSING
I
I
LPF
455kHz
ADC
SIN
CHANNEL
1
CHANNEL 1
QVCO
3RD IF
DSP
COS
Q
LPF
ADC
Q
CHANNEL n
SAME AS ABOVE
a
CHANNEL
n
5.25
23
At this point, we should make it clear that a digital receiver is not the same thing as
digital modulation. In fact, a digital receiver will do an excellent job of receiving an
analog signal such as AM or FM. Digital receivers can be used to receive any type of
modulation standard including analog (AM, FM) or digital (QPSK, QAM, FSK,
GMSK, etc.). Furthermore, since the core of a digital radio is its digital signal
processor (DSP), the same receiver can be used for both analog and digitally
modulated signals (simultaneously if necessary), assuming that the RF and IF
hardware in front of the DSP is properly designed. Since it is software that
determines the characteristics of the radio, changing the software changes the radio.
For this reason, digital receivers are often referred to as software radios.
The fact that a radio is software programmable offers many benefits. A radio
manufacturer can design a generic radio in hardware. As interface standards change
(as from FM to CDMA or TDMA), the manufacturer is able to make timely design
changes to the radio by reprogramming the DSP. From a user or service-providers
point of view, the software radio can be upgraded by loading the new software at a
small cost, while retaining all of the initial hardware investment. Additionally, the
receiver can be tailored for custom applications at very low cost, since only software
costs are involved.
A digital receiver performs the same function as an analog one with one difference;
some of the analog functions have been replaced with their digital equivalent. The
main difference between Figure 5.24 and Figure 5.25 is that the FM discriminator
in the analog radio has been replaced with two ADCs and a DSP. While this is a
very simple example, it shows the fundamental beginnings of a digital, or software
radio.
An added benefit of using digital techniques is that some of the filtering in the radio
is now performed digitally. This eliminates the requirement of tight tolerances and
matching for frequency-sensitive components such as inductors and capacitors. In
addition, since filtering is performed within the DSP, the filter characteristics can be
implemented in software instead of costly and sensitive SAW, ceramic, or crystal
filters. In fact, many filters can be synthesized digitally that could never be
implemented in a strictly analog receiver.
This simple example is only the beginning. With current technology, much more of
the receiver can be implemented in digital form. There are numerous advantages to
moving the digital portion of the radio closer to the antenna. In fact, placing the
ADC at the output of the RF section and performing direct RF sampling might seem
attractive, but does have some serious drawbacks, particularly in terms of selectivity
and out-of-band (image) rejection. However, the concept makes clear one key
advantage of software radios: they are programmable and require little or no
component selection or adjustments to attain the required receiver performance.
Narrowband IF-Sampling Digital Receivers
A reasonable compromise in many digital receivers is to convert the signal to digital
form at the output of the first or the second IF stage. This allows for out-of-band
signals to be filtered before reaching the ADC. It also allows for some automatic gain
control (AGC) in the analog stage ahead of the ADC to reduce the possibility of inband signals overdriving the ADC and allows for maximum signal gain prior to the
A/D conversion. This relieves some of the dynamic range requirements on the ADC.
24
Additionally, IF sampling and digital receiver technology reduce costs by elimination
of further IF stages (mixers, filters, and amplifiers) and adds flexibility by the
replacement of fixed analog filter components with programmable digital ones.
In analyzing an analog receiver design, much of the signal gain is after the first IF
stage. This prevents front-end overdrive due to out-of-band signals or strong in-band
signals. However, in an IF sampling digital receiver, all of the gain is in the front
end, and great care must be taken to prevent in-band and out-of-band signals from
saturating the ADC, which results in excessive distortion. Therefore, a method of
attenuation must be provided when large in-band signals occur. While additional
signal gain can be obtained digitally after the ADC, there are certain restrictions.
Gain provided in the analog domain improves the SNR of the signal and only
reduces the performance to the degree that the noise figure (NF) degrades noise
performance.
Figure 5.26 shows a detailed IF sampling digital receiver for the GSM system . The
receiver has RF gain, automatic gain control (AGC), a high performance ADC,
digital demodulator/filter, and a DSP.
NARROWBAND IF SAMPLING GSM DIGITAL RECEIVER
900MHz
200kHz WIDE CHANNELS
16 CALLERS/CHANNEL
LO1
TUNED
70MHz
ADC
1ST IF
DIGITAL
DEMODULATION
AND
AGC
DSP
CHANNEL
1
DECIMATION
RSSI
SAME AS ABOVE
a
FILTER
CHANNEL
n
5.26
The heart of the system is the AD6600 dual channel, gain ranging 11-bit, 20MSPS
ADC with RSSI (Received Signal Strength Indicator) and the AD6620 dual channel
decimating receiver. A detailed block diagram of the AD6600 is shown in Figure
5.27 and key specifications in Figure 5.28.
25
AD6600 DUAL CHANNEL GAIN RANGING ADC WITH
RECEIVED SIGNAL STRENGTH INDICATOR (RSSI)
EXTERNAL
FILTER
+
ATTEN
IF
A
MUX
A/B
PEAK
DETECTOR
T/H
12dB
11-BIT
ADC
11
+
IF
B
ATTEN
-
RSSI
CONTROL
3
a
5.27
AD6600 KEY SPECIFICATIONS
n
Dual Input, 11-bit, 20MSPS ADC Plus 3-bits RSSI
n
Dynamic Range > 100dB
u
11-bit ADC
→
62dB
u
3-bits RSSI
→
30dB (5 levels, 6dB / level)
u
Process Gain →
12dB (6.5MSPS Sampling, 200kHz Channel)
n
On-Chip Reference and Timing
n
Single +5V Supply, 400mW
n
44-pin TQFP Package
n
Optimum Design for Narrowband Digital Receivers
with IF Frequencies to 250MHz
a
5.28
The AD6600 is a mixed signal chip that directly samples narrow band signals at IF
frequencies up to 250MHz. The device includes an 11-bit, 20MSPS ADC, input
attenuators, automatic gain ranging circuitry, a 450MHz bandwidth track-and-hold,
26
digital RSSI outputs, references, and control circuitry. The device accepts two inputs
(for use with diversity antennas) which are multiplexed to the single ADC.
The AD6600 provides greater than 92dB dynamic range from the ADC and the auto
gain-ranging/RSSI circuits. The gain range is 36dB in 6dB increments (controlled by
a 3-bit word from the RSSI circuit). This sets the smallest input range at 31mV
peak-to-peak, and the largest at 2V peak-to-peak. SFDR is 70dBc @ 100MHz and
53dBc @ 250MHz. Channel isolation is 70dB @ 100MHz and 60dB @ 250MHz. The
SNR performance of the AD6600 is shown in Figure 5.29. The dynamic range of the
AD6600 is greater than the minimum GSM specification of 91dB.
AD6600 INPUT VS. SNR
+10
Input voltage range & RSSI
for AD6600 input ranges
RSSI=101, Vin>=.5Vpp
-2
RSSI=100, .25Vpp<=Vin<.5Vpp
-8
RSSI=011, .125Vpp<=Vin<.25Vpp
-14
RSSI=010, .0625Vpp<=Vin<.125Vpp
-20
RSSI=001, .03125Vpp<=Vin<.0625Vpp
-26
RSSI=000, Vin<.03125Vpp
Ain
-83
-88
0
4
8
12
16
20
24
28
32
36
40
44
48
52
56
60
SNR
a
5.29
The analog input to the AD6600 consists of two parallel attenuator stages followed
by an output selection multiplexer. The attenuation levels can be set either by the
on-chip automatic RSSI circuit (synchronous peak detector) or can be set digitally
with external logic. The ADC T/H input can also be accessed directly by by-passing
the front-end attenuators.
An external analog filter is required between the attenuator output and the trackand-hold input of the ADC section. This filter may be either a lowpass or a bandpass
depending on the system architecture. Since the input bandwidth of the ADC is
450MHz, the filter minimizes the wideband noise entering the track-and-hold. The
bandwidth of the filter should be set to allow sufficient settling time (1/2 the
sampling period) during the RSSI peak detection period.
27
The ADC is based on the high dynamic range AD9042 architecture covered
previously. The ADC input is designed to take advantage of the excellent smallsignal linearity of the track-and-hold. Therefore, the full scale input to the ADC
section is only 50mV peak-to-peak. The track-and-hold is followed by a gain block
with a 6dB gain-select to increase the signal level for digitization by the 11-bit ADC.
This amplifier only requires enough bandwidth to accurately settle to the next value
during the sampling period (77ns for fs = 13MSPS). Because of its reduced
bandwidth, any high frequency track-and-hold feed through is also minimized.
The RSSI peak detector function consists of a bank of 5 high speed comparators with
separate reference inputs. Each reference input is 6dB lower than the previous one.
Each comparator has 6dB of built-in hysteresis to eliminate level uncertainty at the
threshold points. Once one of the comparators is tripped, it stays in that state until
it is reset by the negative-going edge of the sampling clock. The 5 comparator
outputs are decoded into a 3-bit word that is used to select the proper input
attenuation.
The RSSI follows the IF signal one clock cycle before the conversion is made. During
this time period, the RSSI looks for the signal peaks. Prior to digitization, the RSSI
word selects the correct attenuator factor to prevent the ADC from over-ranging on
the following conversion cycle. The peak signal is set 6dB below the full scale range
of the 11-bit ADC. The RSSI word can be read via the RSSI pins. The 11-bit ADC
output functions as the mantissa, while the RSSI word is the exponent, and the
combination forms a floating point number.
The AD6600 is ideal for use in a GSM narrowband basestation. Figure 5.30 shows a
block diagram of the fundamental receiver. Two separate antennas and RF sections
are used (this is often called diversity) to reduce the signal strength variations due to
multipath effects. The IF output (approximately 70MHz) of each channel is digitized
by the AD6600 at a sampling rate of 6.5MSPS (one-half the master GSM clock
frequency of 13MHz). The two antennas need only be separated by a few feet to
provide the required signal strength diversity (the wavelength of a 900MHz signal is
about 1 foot). The DSP portion of the receiver selects the channel which has the
largest signal amplitude.
28
NARROWBAND GSM BASESTATION WITH DIVERSITY
CHANNEL A
ANTENNA
900MHz
69.875MHz
DUAL-CHANNEL DUAL-CHANNEL
ADC, RSSI
DEMODULATION
AND GAIN
AND
RANGING
DECIMATION
IF
LNA
1ST IF
CHANNEL B
ANTENNA
LNA
LO
TUNED
A
B
IF
AD6600
fs = 6.5MSPS
PER CHANNEL
SAME AS ABOVE
a
AD6620
DSP
CHANNEL
ADSP-2181,
n
ADSP-21062
CHANNEL
1
CHANNEL
n
5.30
The bandwidth of a single GSM channel is 200kHz, and each channel can handle up
to 8 simultaneous callers for full-rate systems and 16 simultaneous callers for the
newer one-half-rate systems. A typical basestation may be required to handle 50 to
60 simultaneous callers, thereby requiring 4 separate signal processing channels
(assuming a one-half-rate system).
The IF frequency is chosen to be 69.875MHz, thus centering the 200kHz signal in
the 22nd Nyquist zone (see Figure 5.31). The dual channel digital decimating
receiver (AD6620) reverses the frequency sense of the signal and shifts it down to
baseband.
29
NARROWBAND GSM RECEIVER BANDPASS SAMPLING
OF A 200kHz CHANNEL AT 6.5MSPS
ZONE
1
ZONE
2
ZONE
3
ZONE
22
0
3.25
6.50
12fs
11fs
fs
9.75
68.25
71.50
74.75
78.00
FREQUENCY (MHz)
IF = 69.875MHz ± 100kHz
a
5.31
We now have a 200kHz baseband signal (generated by undersampling) which is
being oversampled by a factor of approximately 16 .
The signal is then passed through a digital filter (part of the AD6620) which
removes all frequency components above 200kHz, including the quantization noise
which falls in the region between 200kHz and 3.25MHz (the Nyquist frequency) as
shown in Figure 5.32. The resultant increase in SNR is 12dB (processing gain).
There is now no information contained in the signal above 200kHz, and the output
data rate can be reduced (decimated) from 6.5MSPS to 406.25kSPS, a data rate
which the DSP can handle. The data corresponding to the 200kHz channel is
transmitted to the DSP over a simple 3-wire serial interface. The DSP performs such
functions as channel equalization, decoding, and spectral shaping.
30
DIGITAL FILTERING AND DECIMATION
OF THE 200kHz CHANNEL
QUANTIZATION
NOISE
fs/2
200kHz
0
fs = 6.5MSPS
3.25MHz
AFTER FREQUENCY TRANSLATION
0
3.25MHz
AFTER DIGITAL FILTERING
PROCESSING GAIN = 12dB
0
3.25MHz
fs = 406.25kSPS
AFTER DECIMATION (÷16)
0
3.25MHz
a
5.32
The concept of processing gain is common to all communications systems, analog or
digital. In a sampling system, the quantization noise produced by the ADC is spread
over the entire Nyquist bandwidth which extends from DC to fs/2. If the signal
bandwidth, BW, is less than fs/2, digital filtering can remove the noise components
outside this bandwidth, thereby increasing the effective SNR. The processing gain in
a sampling system can be calculated from the formula:
 fs 
Processing Gain = 10 log
.
 2 ⋅ BW 
The SINAD (noise and distortion measured over fs/2 bandwidth) of the ADC at the
bandwidth of the signal should be used to compute the actual SINAD by adding the
processing gain determined by the above equation. If the ADC is an ideal N-bit
converter, then its SNR (measured over the Nyquist bandwidth) is 6.02N + 1.76dB.
PROCESSING GAIN
n
Measure ADC SINAD (6.02N + 1.76dB Theoretical)
n
Sampling Frequency = fs
n
Signal Bandwidth = BW
n
Processing Gain =
f
10 log  s 
 2 ⋅ BW 
31
 f 
10 log  s 
 2 ⋅ BW 
n
SINAD in Signal Bandwidth = SINAD +
n
SINAD (Theoretical) = 6.02N + 1.76dB +
n
Processing Gain Increases 3dB each time fs is doubled
a

 f
10 log  s 
 2 ⋅ BW 
5.33
Notice that as shown in the previous narrowband receiver example, there can be
processing gain even if the original signal is an undersampled one. The only
requirement is that the signal bandwidth be less than fs/2, and that the noise
outside the signal bandwidth be removed with a digital filter.
Wideband IF-Sampling Digital Receivers
Thus far, we have avoided a detailed discussion of narrowband versus wideband
digital receivers. A digital receiver can be either, but more detailed definitions are
important at this point. By narrowband, we mean that sufficient pre-filtering has
been done such that all undesired signals have been eliminated and that only the
signal of interest is presented to the ADC input. This is the case for the GSM
basestation example previously discussed.
Wideband simply means that a number of channels are presented to input of the
ADC and further filtering, tuning, and processing is performed digitally. Usually, a
wideband receiver is designed to receive an entire band; cellular or other similar
wireless services such as PCS (Personal Communications Systems). In fact, one
wideband digital receiver can be used to receive all channels within the band
simultaneously, allowing almost all of the analog hardware (including the ADC) to
be shared among all channels as shown in Figure 5.34 which compares the
narrowband and the wideband approaches.
32
NARROWBAND VERSUS WIDEBAND DIGITAL RECEIVER
NARROWBAND
TUNED
LO
IF
RF
FRONT
END
ADC
DIGITAL
DECIMATION
FILTER
DSP
DSP
CHANNEL
1
ADC
DIGITAL
DECIMATION
FILTER
DSP
CHANNEL
n
DIGITAL
CHANNELIZER
DSP
CHANNEL
1
DIGITAL
CHANNELIZER
DSP
TUNED
LO
IF
IF
BW: 30-200kHZ
WIDEBAND
FIXED
LO
RF
FRONT
END
IF
ADC
BW: 5-25MHz
a
CHANNEL
n
5.34
Note that in the narrowband digital radio, there is one front-end LO and mixer
required per channel to provide individual channel tuning. In the wideband digital
radio, however, the first LO frequency is fixed, and the "tuning" is done in the
digital channelizer circuits following the ADC.
A typical wideband digital receiver may process a 5 to 25MHz band of signals
simultaneously. This approach is frequently called block conversion. In the wideband
digital receiver, the variable local oscillator in the narrowband receiver has been
replaced with a fixed oscillator, so tuning must be accomplished digitally. Tuning is
performed using a digital down converter (DDC) and filter chip frequently called a
channelizer. The term channelizer is used because the purpose of these chips is to
select one channel out of the many within the broadband spectrum actually present
in the ADC output. A typical channelizer is shown in Figure 5.35.
33
DIGITAL CHANNELIZER IN WIDEBAND RECEIVER
I
DECIMATION
FILTER
LOWPASS
FILTER
I
SIN
SERIAL
DATA
TO
DSP
DATA FROM
TUNING
CONTROL
TUNING
NCO
WIDEBAND
ADC
COS
Q
DECIMATION
FILTER
a
LOWPASS
FILTER
Q
5.35
It consists of an NCO (Numerically Controlled Oscillator) with tuning capability,
dual mixer, and matched digital filters. These are the same functions that would be
required in an analog receiver, but implemented in digital form. The digital output
from the channelizer is the demodulated signal in I and Q format, and all other
signals have been filtered and removed. Since the channelizer output consists of one
selected RF channel, one channelizer is required for each channel. The channelizer
also serves to decimate the output data rate such that it can be processed by a DSP
such as the ADSP-2181 or the ADSP-21062. The DSP extracts the signal
information from the I and Q data and performs further processing. Another effect of
the filtering provided by the channelizer is to increase the SNR by adding processing
gain.
In the case of an AMPS signal, there are 416 channels, each 30kHz wide, for a total
bandwidth of 12.5MHz (each of the two carriers in a given region are allocated
12.5MHz of the total 25MHz cellular band). Each channel carries one call, so there is
a clear advantage in using the wideband approach versus the narrowband one in an
AMPS basestation which must handle between 50 and 60 simultaneous calls. On
the other hand, a 200kHz GSM channel can carry 16 calls simultaneously (for halfrate systems), so only three or four channels are required in the typical GSM
basestation, and the narrowband approach is more cost-effective. Using today's
technology (1996), the break-even cost point between narrowband and wideband
ranges from two and eight channels.
In an ADC used for narrowband applications, the key specifications are SINAD,
SFDR, and SNR. The narrowband ADC can take advantage of automatic gain
ranging (as in the AD6600) to account for signal amplitude variations between
individual channels and thereby achieve extra dynamic range.
34
On the other hand, an ADC used in a wideband receiver must digitize all channels
simultaneously, thereby eliminating the possibility of per-channel analog gain
ranging. For example, the GSM (European Digital Cellular) system specification
requires the receiver to process signals between –13dBm and –104dBm (with a noise
floor of –114dBm) in the presence of many other signals. This is a dynamic range of
91dB! This implies that the SFDR of the ADC and the analog front end must be
approximately 95 to 100dBFS, allowing for additional headroom. In addition, the
GSM system has 124 channels, each having a bandwidth of 200kHz for a total
signal bandwidth of 25MHz. The minimum required sampling rate for an ADC
suitable for wideband GSM is therefore greater than 50MSPS.
SFDR is a very important specification when a mobile phone is near the basestation
because it is an indication of how strong signals interfere with signals in other
channels. Strong signals usually produce the largest spurs due to front-end
distortion, and these spurs can mask weaker signals from mobile phones near the
cell fringes. The SFDR for weak signals provides an indication of the overall noise
floor, or SINAD which can ultimately be related to the receiver bit error rate (BER).
When digitizing a wideband signal, full scale single-tone evaluations are no longer
sufficient. Two-tone and multiple-tone intermodulation testing in conjunction with
SFDR amplitude sweeps are better indicators of performance.
GSM VERSUS AMPS COMPARISONS
GSM
AMPS
Digital Receiver
Narrowband
Wideband
# of Channels
124
416
Channel BW
200kHz
30kHz
Total BW
25MHz
12.5MHz
Callers/Channel
16 (one-half rate)
1
ADC
11-bits with RSSI
12-bits
Requirements
6.5 MSPS
30.72 MSPS
92dB Dynamic Range
80dB SFDR
12dB
27dB
Process Gain
a
5.36
The AMPS cellular system basestation is ideally suited to the wideband digital
receiver design, and a simplified diagram of one is shown in Figure 5.37. The
AD9042 sampling frequency of 30.72MSPS is chosen to be a power-of-two multiple
35
of the channel bandwidth (30kHz x 1024). Another popular AMPS wideband
receiver sampling frequency is 40.96MSPS. The choice of IF frequency is flexible,
and a second IF stage may be required if lower IF frequencies are chosen.
AMPS WIDEBAND DIGITAL RECEIVER
416 CHANNELS
30kHz CHANNEL BW
1 CALLER/CHANNEL
900MHz
LNA
LO1
FIXED
fs = 30.72 MSPS
OR 40.96 MSPS
CHANNELIZER
DSP
CHANNELIZER
DSP
CHANNEL
1
IF
AD9042
ADC
BW: 12.5MHz
NOTE: THERE MAY BE
2 IF STAGES
a
CHANNEL
n
5.37
With a sampling frequency of 30.72MSPS, the 12.5MHz bandwidth signal can be
positioned in the first Nyquist zone (DC to 15.36MHz) with an IF frequency of
7.68MHz, or in the second Nyquist zone (15.36MHz to 30.72MHz) with an IF
frequency of 23.04MHz.
With a sampling frequency of 40.96MSPS, the 12.5 MHz bandwidth signal can be
positioned in the first Nyquist zone (DC to 20.48MHz) with an IF frequency of
10.24MHz, or in the second Nyquist zone (20.48MHz to 40.96MHz) with an IF
frequency of 30.72MHz.
The digital channelizers provide the receiver tuning and demodulate the signal into
the I and Q components. The output data rate to the DSPs after decimation is
60kSPS. The processing gain incurred is calculated as follows:
 30.72 
Processing Gain = 10 log
. dB.
 = 271
 2 × 0.03 
AMPS WIDEBAND RECEIVER PROCESS GAIN
n
fs = 30.72MSPS
n
Channel BW = 30kHz
36
(1024 · 30kHz)
n
Process Gain =
 f  = 27.1dB
10 log  s 
 2 ⋅ BW 
a
5.38
In addition to SFDR, two-tone and multi-tone intermodulation distortion is
important in an ADC for wideband receiver applications. Figure 5.39 shows two
strong signals in two adjacent channels at frequencies f1 and f2. If the ADC has
third-order intermodulation distortion, these products will fall at 2f2–f1 and 2f1–f2
and are indistinguishable from signals which might be present in these channels.
This is one reason the GSM system is difficult to implement using the wideband
approach, since the dynamic range requirement is greater than 91dB.
TWO-TONE INTERMODULATION DISTORTION
IN MULTICHANNEL SYSTEM
(GSM REQUIREMENTS SHOWN)
f1
f2
STRONG SIGNALS
-13dBm
-104dBm
2f2 - f1
2f1 - f2
WEAK SIGNAL
OR 3RD ORDER IMD?
-114dBm
f
a
5.39
The two-tone SFDR of the AD9042 is greater than 80dB with input tones at
15.3MHz and 19.5MHz as shown in Figure 5.40. Note than the amplitude of each
tone must be 6dB below full scale in order to prevent the ADC from being
overdriven. The two-tone SFDR as a function of input signal amplitude is shown in
Figure 5.41 for tone frequencies of 19.3MHz and 19.51MHz. The upper curve is in
dBFS, and the lower in dBc. Note that the SFDR is greater than 80dBFS for all
input amplitudes. Figure 5.42 shows a multitone FFT output for the AD9042, and
the ADC still maintains 85dBFS of SFDR.
37
AD9042 TWO-TONE FFT OUTPUT
F1 = 15.3MHz, F2 = 19.5MHz, f s = 41MSPS
POWER RELATIVE TO ADC FULL-SCALE - dB
0
-20
ENCODE = 41 MSPS
AIN = 15.3, 19.5MHz
-40
-60
-80
-100
-120
dc
4.1
8.2
12.3
16.4
20.5
FREQUENCY - MHz
a
5.40
AD9042 TWO-TONE SFDR
F1 = 19.3MHz, F2 = 19.51MHz, f s = 41MSPS
100
dBFS
WORST CASE SPURIOUS - dBc AND dBFS
90
80
70
60
50
ENCODE = 41 MSPS
F1 = 19.3MHz
F2 = 19.51MHz
dBc
SFDR = 80dB
REFERENCE LINE
40
30
20
10
0
-80
-70
-60
-50
-40
-30
-20
-10
0
INPUT POWER LEVEL (F1 = F2) - dBFS
a
5.41
38
AD9042 MULTITONE PERFORMANCE (4 TONES)
fs = 41MSPS
POWER RELATIVE TO ADC FULL-SCALE - dB
0
ENCODE = 41 MSPS
-20
-40
-60
3
6
9
7
4
2
5
8
-80
-100
-120
dc
4.1
8.2
12.3
16.4
20.5
FREQUENCY - MHz
a
5.42
Direct IF-to-Digital Considerations
The dynamic performance of the AD9042 extends well beyond 20MHz analog input
signals (see Figure 5.43). Therefore it can be used to perform direct IF-to-digital
conversions using a wide range of IF frequencies. These IF signals can be
undersampled as previously described, and the minimum sampling frequency
required is determined by the bandwidth of the IF signal. Figure 5.44 shows a
21.4MHz signal sampled at 10MSPS using the AD9042. Note that under these
conditions, the SFDR performance is greater than 80dBFS.
39
AD9042 SFDR VERSUS INPUT FREQUENCY
90
WORST SPUR - dBFS
80
70
60
50
40
30
1
2
4
10
20
40
ANALOG INPUT FREQUENCY - MHz
100
a
5.43
AD9042 OUTPUT FOR IF SAMPLED INPUT:
f s = 10MSPS, ANALOG INPUT = 21.4MHz
0
ENCODE = 10.0 MSPS
AIN = 21.4MHz
WORST SPUR - dBFS
-20
-40
-60
8
7
8
6
2
5
3
4
-80
-100
-120
dc
4.1
8.2
12.3
16.4
20.5
FREQUENCY - MHz
a
5.44
The AD6640 represents the next generation in IF sampling ADCs. Key
specifications for the AD6640 are summarized in Figure 5.45. The architecture is
similar to that of the AD9042, but the device is fabricated on a faster XFCB process.
The input structure is fully differential and designed for transformer coupling for
40
minimum distortion. Maximum sampling frequency is 65MSPS, and the SINAD
performance is 67dB at 60MHz analog input. SFDR is greater than 80dBFS for
frequencies up to 25MHz. This device allows direct IF sampling in wideband
communications systems having bandwidths up to 25MHz (such as the AMPS
system, where each carrier is allocated 12.5MHz of spectrum). For systems with
smaller bandwidths, the higher sampling frequency provided by the AD6640 will
allow analog antialiasing filter requirements to be relaxed and provide processing
gain. In undersampling applications, the device can be used to digitize 70MHz IF
signals which lie in the second or third Nyquist zone. For instance, a 30MHz
wideband signal bandwidth centered around a carrier frequency of 48.75MHz can be
digitized at 65MSPS as shown in Figure 5.46. In narrowband applications, the high
sampling frequency can be used to achieve additional processing gain.
AD6640 12-BIT, 65MSPS ADC KEY SPECIFICATIONS
n
12-bit, 65MSPS IF-SAMPLING ADC
n
Based on AD9042 architecture, but 1.5X faster CB process
n
Fully differential inputs for optimum distortion performance
n
SFDR Greater than 80dB up to 25MHz Input
n
68dB SINAD for 60MHz IF input
n
Single +5V Supply, 695mW
n
44-Lead TQFP Package
a
5.45
41
SAMPLING A 25MHz BW SIGNAL USING AD6640:
IF FREQUENCY = 48.75MHz, fs = 65MSPS
ZONE 1
ZONE 2
IF
0
32.5
ZONE 3
fs = 65MSPS
65
97.5
f(MHz)
IF = 48.75MHz
SIGNAL BW = 25MHz
5.46
a
Achieving Wide Dynamic Range in High Speed ADCs Using Dither
There are two fundamental limitations to maximizing SFDR in a high speed ADC.
The first is the distortion produced by the front-end amplifier and the sample-andhold circuit. The second is that produced by non-linearity in the actual transfer
function of the encoder portion of the ADC. The key to wide SFDR is to minimize the
non-linearity of each.
There is nothing that can be done externally to the ADC to significantly reduce the
inherent distortion caused by the ADC front end. However, the non-linearity in the
ADC encoder transfer function can be reduced by the proper use of dither (external
noise which is summed with the analog input signal to the ADC).
Dithering improves ADC SFDR under certain conditions. For example, even in a
perfect ADC, there is some correlation between the quantization noise and the input
signal. This can reduce the SFDR of the ADC, especially if the input signal is an
exact sub-multiple of the sampling frequency. Summing broadband noise (about 1/2
LSB rms in amplitude) with the input signal tends to randomize the quantization
noise and minimize this effect (see Figure 5.47). In most systems, however, there is
enough noise riding on top of the signal so that adding additional dither noise is not
required. Increasing the wideband rms noise level beyond an LSB will proportionally
reduce the ADC SNR.
42
USING DITHER TO RANDOMIZE ADC TRANSFER FUNCTION
LARGE
AMPLITUDE
SMALL
AMPLITUDE
INPUT
INPUT
ADC
+
ADC
+
+
ADDER
≈1/2 LSB RMS
NOISE
GENERATOR
RANDOM
NUMBER
GENERATOR
DAC
a
5.47
Other schemes have been developed which use larger amounts of dither noise to
randomize the transfer function of the ADC. Figure 5.47 also shows a dither noise
source comprised of a pseudo-random number generator which drives a DAC. This
signal is subtracted from the ADC input signal and then digitally added to the ADC
output, thereby causing no significant degradation in SNR. An inherent
disadvantage of this technique is that the allowable input signal swing is reduced as
the amplitude of the dither signal is increased. This reduction in signal amplitude is
required to prevent overdriving the ADC. It should be noted that this scheme does
not significantly improve distortion created by the front-end of the ADC, only that
produced by the non-linearity of the ADC encoder transfer function.
Another method which is easier to implement, especially in wideband receivers, is to
inject a narrowband dither signal outside the signal band of interest as shown in
Figure 5.48. Usually, there are no signal components located in the frequency range
near DC, so this low-frequency region is often used for such a dither signal. Another
possible location for the dither signal is slightly below fs/2. Because the dither signal
occupies only a small bandwidth relative to the signal bandwidth, there is no
significant degradation in SNR, as would occur if the dither was broadband.
43
INJECTING OUT-OF-BAND DITHER TO IMPROVE ADC SFDR
fs
INPUT
BPF
ADC
+
+
NOISE
GENERATOR
OUT-OF-BAND NOISE
OUT-OFBAND
FILTER
NEAR DC OR fs/2
a
5.48
A subranging ADC such as the AD9042 (see Figure 5.49) has small differential nonlinearity errors that occur at specific regions across the ADC range. For instance,
the AD9042 uses a 6-bit ADC followed by a 7-bit one. There are 64 decision points
associated with the main-range 6-bit ADC, and they occur every 15.625mV for a 1V
full scale input range. Figure 5.50 shows a greatly exaggerated representation of
these non-linearities.
44
AD9042 12-BIT, 41MSPS PIPELINED SUBRANGING ADC
WITH DIGITAL ERROR CORRECTION
ANALOG
INPUT
SHA
2
SHA
1
GAIN
+
-
7-BIT
ADC
6-BIT
DAC
6-BIT
ADC
SHA
3
6
BUFFER
REGISTER
7
6
ERROR CORRECTION LOGIC
12
OUTPUT REGISTERS
12
a
5.49
AD9042 SUBRANGING POINT DNL ERRORS
(EXAGGERATED)
OUTPUT
CODE
15.625mV
64 LSBs
ANALOG INPUT
a
5.50
The distortion components produced by the front end of the AD9042 up to about
20MHz analog input are negligible compared to those produced by the encoder. That
45
is, the static non-linearity of the AD9042 transfer function is the chief limitation to
SFDR.
The goal is to select the proper amount of out-of-band dither so that the effect of
these small DNL errors are randomized across the ADC input range, thereby
reducing the average DNL error. The first plot shown in Figure 5.51 shows the
undithered DNL over a small portion of the input signal range. The horizontal axis
has been expanded to show two of the subranging points which are spaced
15.625mV (64 LSBs) apart. The second plot shows the DNL after adding 5.3mV rms
(22 LSBs rms) of dither. This amount of dither corresponds to –32.5dBm (1V p-p full
scale corresponds to +4dBm). It was determined that further increases in dither
amplitude provided no improvement in the AD9042 SFDR and would only serve to
cause a loss in headroom and a decrease in SNR.
AD9042 UNDITHERED AND DITHERED DNL
UNDITHERED
21.3 LSBs DITHER
1.5
1.5
1.0
1.0
0.5
0.5
0
0
DNL
(LSBs)
-0.5
-FS
-0.5
-FS
+FS
a
+FS
5.51
The dither signal was generated using a voltage feedback op amp (AD8048,
3.8nV/√Hz input voltage noise, 200MHz gain-bandwidth product) as the noise source
(see Figure 5.52). The op amp is configured for a gain of +26, and the output noise
spectral density is about 100nV/√Hz over an 8MHz bandwidth. The output of the
noise generator is then amplified by the AD600 dual wideband VCA which provides
a gain (in dB) which is proportional to the control voltage. The control voltage can be
fixed, or programmed using a DAC as shown. The gain of the AD600 can be set from
0dB to 80dB by varying the control voltage from 0 to +1V. The bandwidth of the
noise is limited to about 300kHz with a lowpass filter. The filter can be either
passive or active, but requires at least 4 poles in order to attenuate the out-of-band
noise. The output of the lowpass filter is buffered with the AD797 low-noise op amp
which also provides a gain of +2. The filtered noise is summed directly into the input
circuit of the AD9042 through a capacitor and a 1kΩ series resistor. The net input
impedance of the AD9042 is 50Ω (61.9Ω in parallel with the 250Ω AD9042 internal
impedance).
46
DITHER NOISE GENERATOR
100nV/√
√Hz
BW≈
≈8MHz
10 Ω
+
0 TO +1V, 0 TO 80dB GAIN
DAC
AD8048
1µ
µF
FILM
-
10 Ω
+
100 Ω
+
1µ
µF
FILM
A1
-
249 Ω
A2
-
A1, A2:
1/2 AD600
NOISE
300kHz
LPF
+
0.1µ
µF
1000Ω
Ω
0.1µ
µF
AD9042
Z = 250Ω
Ω
AD797
-
SIGNAL
301Ω
Ω
301Ω
Ω
a
61.9 Ω
5.52
The dramatic improvement in SFDR obtained with out-of-band dither is shown in
Figure 5.53 using a 4k FFT, where the AD9042 is sampling a 19.5MHz signal (–
29dBFS) at 41MSPS. Note that the SFDR without dither is approximately 80dBFS
compared to 94dBFS with dither, representing a 14dB improvement! This
improvement is also shown in the SFDR amplitude sweeps shown in Figure 5.54.
Note the similar improvement.
47
AD9042 UNDITHERED AND DITHERED 4k FFT OUTPUT
DITHERED
0
POWER RELATIVE TO ADC FULL SCALE - dB
POWER RELATIVE TO ADC FULL SCALE - dB
UNDITHERED
ENCODE = 41 MSPS
AIN = 19.5MHz @ -29 dBFS
NO DITHER
-20
-40
-60
2
-80
6
4
8
8
7
5
3
-100
-120
dc
4.1
8.2
12.3
16.4
20.5
0
ENCODE = 41 MSPS
AIN = 19.5MHz @ -29 dBFS
DITHER = -32.5dBm
-20
-40
-60
2
-80
4
6
8
8
7
5
3
-100
-120
dc
4.1
FREQUENCY - MHz
8.2
12.3
16.4
20.5
FREQUENCY - MHz
a
5.53
AD9042 UNDITHERED AND DITHERED SFDR
UNDITHERED
DITHERED
100
100
80
90
ENCODE = 41 MSPS
AIN = 19.5MHz
NO DITHER
WORST CASE SPURIOUS - dBc
WORST CASE SPURIOUS - dBc
90
70
60
50
40
30
20
SFDR = 80dB
REFERENCE LINE
10
0
-80
-70
-60
-50
-40
-30
-20
-10
ANALOG INPUT POWER LEVEL - dBFS
80
ENCODE = 41 MSPS
AIN = 19.5MHz
DITHER = -32.5dBm
70
60
50
40
30
20
SFDR = 80dB
REFERENCE LINE
10
0
a
0
-80
-70
-60
-50
-40
-30
-20
-10
ANALOG INPUT POWER LEVEL - dBFS
0
5.54
At lower frequencies, the FFT size must be increased from 4k to 128k (reducing the
FFT noise floor by 15dB) in order to measure the dithered SFDR. Figure 5.55 shows
the effects of dither using a 128k FFT and a 2.5MHz input signal. The SFDR with
dither is greater than 100dBFS.
48
AD9042 UNDITHERED AND DITHERED 128k FFT OUTPUTS
DITHERED
POWER RELATIVE TO ADC FULL SCALE - dB
POWER RELATIVE TO ADC FULL SCALE - dB
UNDITHERED
0
ENCODE = 41 MSPS
AIN = 2.5MHz @ -26 dBFS
NO DITHER
-20
-40
-60
-80
-100
-120
dc
4.1
8.2
12.3
FREQUENCY - MHz
16.4
20.5
a
0
ENCODE = 41 MSPS
AIN = [email protected]
DITHER = -32.5dBm
-20
-40
-60
-80
-100
-120
dc
4.1
8.2
12.3
FREQUENCY - MHz
16.4
20.5
5.55
High Speed ADC Applications in Digital Communications Systems and
Direct Broadcast Satellite (DBS) Set-Top Boxes
In a digital communications system, digital data (which can be digitized analog
signals) is formatted and transmitted serially over an appropriate medium. The
GSM cellular telephone system is an example. The ubiquitous modem
(modulator/demodulator), which PCs and FAX machines use to transmit and receive
data over the standard dial-up telephone connection, uses sophisticated modulation
techniques to place huge amounts of data in the 4kHz bandwidth telephone channel.
Most digital transmission schemes use some form of in-phase and quadrature (I and
Q) modulation to maximize the amount of data transmitted over a given channel
bandwidth. Two examples are shown in Figure 5.56 and Figure 5.57. The first is
called Quadrature Phase Shift Keying (QPSK) and is used in Direct Broadcast
Satellite systems. The diagram (constellation) shows the four possible data points,
each representing 2-bits of binary information. Each point in the constellation is
called a symbol and has a specific I and Q value. In the case of QPSK, there are two
bits of information per symbol. The symbol rate is often referred to as the baud rate.
For example, in QPSK, if the symbol (or baud) rate is 30Mbaud (1baud =
1symbol/sec), the bit rate is 60Mbits/sec. It is common practice to sample these types
of signals at twice the symbol (or baud) rate. The I and Q ADC and DSP must
identify the signal as representing one of two possible levels, and ADCs of 4, 5, or 6bits are commonly used in this application for additional noise margin and to
achieve the overall system bit-error-rate (BER) requirement.
49
QPSK MODULATION
01
Q
11
2-BITS/SYMBOL
I
10
00
I OR Q CHANNEL
SAMPLING
CLOCK
t
a
5.56
In the QPSK system, the magnitude of each symbol is equal, and only the phase is
modulated. More complex modulation schemes such as QAM (Quadrature Amplitude
Modulation), use more symbols on the constellation and thereby transmit more bits
of information per symbol (at the expense of more sensitivity to noise and more
complex digital signal processing). Figure 5.57 shows a 16-QAM constellation which
contains 4-bits of information per symbol. Note that the I and Q channel receiver
DSP must now identify the signal as representing one of the four possible levels.
Although the 16-QAM signal carries more bits per symbol, it is more sensitive to
noise, and the ADC requires more resolution (typically 8-bits) than for QPSK
modulation (typically 4, 5, or 6 bits).
50
16-QAM MODULATION
Q
1111
4-BITS/SYMBOL
I
0000
I OR Q CHANNEL
SAMPLING
CLOCK
t
a
5.57
In the digital receiver, the I and Q components are separated by a quadrature
demodulator and digitized by two ADCs operating in parallel. The ADC sampling
rate is generally twice the symbol rate. In the case of Direct Broadcast Satellite
(DBS), the symbol rate is 30Mbaud (1baud = 1symbol/sec), the bit rate 60Mbits/sec,
and the ADC sampling rate is 60MSPS. The actual signals at the ADC input are
called "eye patterns" because the intersymbol interference due to noise and limited
bandwidth smears the level transitions so that the regions where the data is valid
are located in the center of the eye opening. Figure 5.58 shows a typical I/Q
demodulator followed by a dual ADC such as the AD9066 (6-bits, 60MSPS).
51
IF SAMPLING USING AD9066 6-BIT, 60MSPS ADC
I
I
I
LPF
SIN
70MHz
IF
AD9066
ADC
QVCO
COS
Q
Q
DSP
Q
LPF
FOR DBS, SYMBOL (BAUD) RATE = 30MBAUD, QPSK
SAMPLING RATE = 60MSPS
a
5.58
A recent popular consumer application of digital communications is in Direct
Broadcast Satellite (DBS) systems. A simplified block diagram of a DBS system is
shown in Figure 5.59. The objective is to transmit up to 150 channels of video
programming to home receivers which use a small (18 inch) dish and an inexpensive
(less than $500) receiver (set-top box). The subscription costs of the services is
compatible with cable TV, but picture quality (because of digital transmission
inherent noise immunity) is generally superior over all 150 channels.
52
DIRECT BROADCAST SATELLITE (DBS)
DBS
SATELLITE
12.2 - 12.7 GHz, 150 CHANNELS
18" DISH
480MHz
UPLINK
70MHz
1GHz
LNB
DOWNCONVERTER
LO 1
VARIABLE
1ST IF
LO 2
FIXED
2ND IF
70MHz
MPEG
COMPRESSION
LPF
VIDEO
SOURCES
ADC
I
SIN
QVCO
DSP
COS
LPF
ADC
Q
MPEG
DECODE
AND
DAC
BASEBAND
VIDEO
OR
CH. 3, 4 RF
5.59
a
MPEG encoding and decoding reduces the data rates to fit the channel bandwidth.
The MPEG (Motion Picture Experts Group) standard supports various data rates
and minimizes the bandwidth used. For example, a typical 24-frame-per-second
NTSC-quality movie needs about 3Mbits/sec after encoding. A more complex and
fast-moving show, such as a soccer game, requires 5 to 6 Mbits/sec. In a DBS
system, the MPEG encoding rate is kept at a minimum value compatible with the
anticipated video signal characteristics. Multiple MPEG data streams are
multiplexed and sent through a single satellite transponder. In addition, statistical
multiplexing dynamically varies the data rate given to each source as the program
content changes.
The satellite downlink frequency is Ku-band (12.2 to 12.7GHz), and the transponder
output power is about 120W (10 to 20 times that of a typical communications
satellite which is designed for much larger receiver antennas). The LNB (Low Noise
Block Converter) converts the 12.2 to 12.7GHz untuned band down to 950MHz to
1450MHz, where the signal is easier to tune, filter, and bring into the home over
standard coaxial cable. The lower frequency signal (1GHz) incurs less loss over
standard coaxial cable from the outside antenna to the inside of the house (generally
50 feet or more) than the Ku-band signal (12GHz).
The set-top box mixes the RF (1GHz) signal down to the first fixed IF frequency of
480MHz. The LO which drives the mixer is used for channel tuning. A second fixedfrequency IF stage brings the tuned signal down to 70MHz where it is
synchronously demodulated into baseband I and Q components. The modulation
scheme is QPSK, the symbol rate is 30Mbaud, and the ADC sampling rate 60MSPS.
Figure 5.60 shows a two-chip solution to the front-end of the set-top box using the
AD6461 (quadrature demodulator and baseband filter) and the AD6462 (dual 5-bit
53
ADC and digital receiver). The input to the AD6461 is the 480MHz DBS IF signal.
The chip-set is designed to support symbol rates up to 42.5Mbaud. The AD6461
utilizes Analog Devices' XFCB process and is packaged in 28-pin SOIC dissipating
about 500mW. The AD6462 utilizes a 0.6 micron CMOS process and is packaged in
an 80-pin PQFP dissipating approximately 1.2W (operating dynamically).
NEXT-GENERATION DBS 480MHz IF SIGNAL PROCESSING
AD6461 QUADRATURE DEMOD
AND BASEBAND FILTER
MATCHED
FILTER
AD6462 DUAL 5-BIT ADC
AND DIGITAL RECEIVER
ADC
I
SIN
IF
480MHz
DEMOD
QVCO
COS
MATCHED
FILTER
AGC
ADC
VCO AND
TUNING
CONTROL
a
Q
FREQ.
SYNTH.
FORWARD
ERROR
CORRECTION
DAC
SERIAL PORT
AND
CONTROL
DATA
OUT
5.60
54
REFERENCES
1.
An Introduction to the Imaging CCD Array, Technical Note 82W-4022,
Tektronix, Inc., Beaverton, OR., 1987.
2.
Brad Brannon, Using Wide Dynamic Range Converters for Wide
Band Radios, RF Design, May 1995, pp.50-65.
3.
Joe Mitola, The Software Radio Architecture, IEEE Communications
Magazine, Vol. 33, No.5, May 1995, pp. 26-38.
4.
Jeffery Wepman, Analog-to-Digital Converters and Their Applications
in Radio Receivers, IEEE Communications Magazine,
Vol. 33, No.5, May 1995, pp. 39-45.
5.
Rupert Baines, The DSP Bottleneck, IEEE Communications
Magazine, Vol. 33, No.5, May 1995, pp. 46-54.
6.
Brad Brannon, Overcoming Converter Nonlinearities with Dither,
Application Note AN-410, Analog Devices, 1995.
7.
Chris Keate and Mark O'Brien, DBS Receiver Chip Simplifies Set-Top
Box Design, RF Design, November 1995, pp. 36-42.
8.
Bill Schweber, Direct Satellite Broadcast, EDN, December 21,
1995, pp. 53-58.
55
SECTION 6
HIGH SPEED DACs AND DDS SYSTEMS
Walt Kester
INTRODUCTION
A frequency synthesizer generates multiple frequencies from one or more frequency
references. These devices have been used for decades, especially in communications
systems. Many are based upon switching and mixing frequency outputs from a bank
of crystal oscillators. Others have been based upon well understood techniques
utilizing phase-locked loops (PLLs). This mature technology is illustrated in Figure
6.1. A fixed-frequency reference drives one input of the phase comparator. The other
phase comparator input is driven from a divide-by-N counter which is in turn driven
by a voltage-controlled-oscillator (VCO). Negative feedback forces the output of the
internal loop filter to a value which makes the VCO output frequency N-times the
reference frequency. The time constant of the loop is controlled by the loop filter.
There are many tradeoffs in designing a PLL, such a phase noise, tuning speed,
frequency resolution, etc., and there are many good references on the subject (see
References 1, 2, and 3).
FREQUENCY SYNTHESIS USING
OSCILLATORS AND PHASE-LOCKED LOOPS
OSCILLATOR BANK
PHASE-LOCKED LOOP
MIXER
XO
1
PHASE
COMPARATOR
fc
FIXED
FREQUENCY
REFERENCE
XO
2
VCO
LOOP
FILTER
f out
fout
÷ N
MIXER
XO
3
fout = N • f c
SW
XO
n
a
6.1
With the widespread use of digital techniques in instrumentation and
communications systems, a digitally-controlled method of generating multiple
frequencies from a reference frequency source has evolved called Direct Digital
Synthesis (DDS). The basic architecture is shown in Figure 6.2. In this simplified
model, a stable clock drives a programmable-read-only-memory (PROM) which
stores one or more integral number of cycles of a sinewave (or other arbitrary
1
waveform, for that matter). As the address counter steps through each memory
location, the corresponding digital amplitude of the signal at each location drives a
DAC which in turn generates the analog output signal. The spectral purity of the
final analog output signal is determined primarily by the DAC. The phase noise is
basically that of the reference clock.
The DDS system differs from the PLL in several ways. Because a DDS system is a
sampled data system, all the issues involved in sampling must be considered:
quantization noise, aliasing, filtering, etc. For instance, the higher order harmonics
of the DAC output frequencies fold back into the Nyquist bandwidth, making them
unfilterable, whereas, the higher order harmonics of the output of PLL-based
synthesizers can be filtered. There are other considerations which will be discussed
shortly.
FUNDAMENTAL DIRECT DIGITAL SYNTHESIS SYSTEM
CLOCK
fc
ADDRESS
COUNTER
SIN
LOOKUP
TABLE
N-BITS
REGISTER
N-BITS
LOOKUP TABLE CONTAINS SIN
DATA FOR INTEGRAL NUMBER
OF CYCLES
DAC
fout
LPF
a
6.2
A fundamental problem with this simple DDS system is that the final output
frequency can be changed only by changing the reference clock frequency or by
reprogramming the PROM, making it rather inflexible. A practical DDS system
implements this basic function in a much more flexible and efficient manner using
digital hardware called a Numerically Controlled Oscillator (NCO). A block diagram
of such a system is shown in Figure 6.3.
2
A FLEXIBLE DDS SYSTEM
PHASE ACCUMULATOR
n = 24-32 BITS
SERIAL
OR BYTE
LOAD
REGISTER
n
n
PARALLEL
DELTA
PHASE
REGISTER
M
n
n
PHASE n
REGISTER
SIN ROM
LOOKUP
TABLE
CLOCK
PHASE
TRUNCATION
14-16 BITS
FREQUENCY CONTROL
N-BITS
AMPLITUDE
TRUNCATION
fc
DAC
fo =
M • fc
2n
LPF
a
fo
6.3
The heart of the system is the phase accumulator whose contents is updated once
each clock cycle. Each time the phase accumulator is updated, the digital number,
M, stored in the delta phase register is added to the number in the phase
accumulator register. Assume that the number in the delta phase register is 00...01
and that the initial contents of the phase accumulator is 00...00. The phase
accumulator is updated by 00...01 on each clock cycle. If the accumulator is 32-bits
wide, 232 clock cycles (over 4 billion) are required before the phase accumulator
returns to 00...00, and the cycle repeats.
The truncated output of the phase accumulator serves as the address to a sine (or
cosine) lookup table. Each address in the lookup table corresponds to a phase point
on the sinewave from 0° to 360°. The lookup table contains the corresponding digital
amplitude information for one complete cycle of a sinewave. (Actually, only data for
90° is required because the quadrature data is contained in the two MSBs). The
lookup table therefore maps the phase information from the phase accumulator into
a digital amplitude word, which in turn drives the DAC.
Consider the case for n=32, and M=1. The phase accumulator steps through each of
232 possible outputs before it overflows. The corresponding output sinewave
frequency is equal to the clock frequency divided by 232. If M=2, then the phase
accumulator register "rolls over" twice as fast, and the output frequency is doubled.
This can be generalized as follows.
For an n-bit phase accumulator (n generally ranges from 24 to 32 in most DDS
systems), there are 2n possible phase points. The digital word in the delta phase
register, M, represents the amount the phase accumulator is incremented each clock
cycle. If fc is the clock frequency, then the frequency of the output sinewave is equal
to:
3
fo =
M ⋅ fc
.
2n
This equation is known as the DDS "tuning equation." Note that the frequency
resolution of the system is equal to fc/2n. For n=32, the resolution is greater than
one part in four billion! In a practical DDS system, all the bits out of the phase
accumulator are not passed on to the lookup table, but are truncated, leaving only
the first 13 to 15 MSBs. This reduces the size of the lookup table and does not affect
the frequency resolution. The phase truncation only adds a small but acceptable
amount of phase noise to the final output.
The resolution of the DAC is typically 2 to 4 bits less than the width of the lookup
table. Even a perfect N-bit DAC will add quantization noise to the output. Figure 6.4
shows the calculated output spectrum for a 32-bit phase accumulator, 15-bit phase
truncation, and a 12-bit DAC. The value of M was chosen so that the output
frequency was slightly offset from 0.25 times the clock frequency. Note that the
spurs caused by the phase truncation and the finite DAC resolution are all at least
90dB below the fullscale output. This performance far exceeds that of any
commercially available 12-bit DAC and is adequate for most applications.
CALCULATED OUTPUT SPECTRUM SHOWS
90dB SFDR FOR 15-BIT PHASE TRUNCATION AND
12-BIT OUTPUT DATA TRUNCATION
0
-20
MAGNITUDE - dBc
-40
-60
-80
-100
-120
0
0.05 0.1
0.15 0.2
0.25
0.3 0.35
0.4 0.45
0.5
NORMALIZED FREQUENCY - fOUT/fCLK
a
6.4
The basic DDS system described above is extremely flexible and has high resolution.
The frequency can be changed instantaneously with no phase discontinuity by
simply changing the contents of the M-register. However, practical DDS systems
first require the execution of a serial, or byte-loading sequence to get the new
frequency word into an internal buffer register which precedes the parallel-output
M-register. This is done to minimize package pin count. After the new word is loaded
4
into the buffer register, the parallel-output delta phase register is clocked, thereby
changing all the bits simultaneously. The number of clock cycles required to load the
delta-phase buffer register determines the maximum rate at which the output
frequency can be changed.
ALIASING IN DDS SYSTEMS
There is one important limitation to the range of output frequencies that can be
generated from the simple DDS system. The Nyquist Criteria states that the clock
frequency (sample rate) must be at least twice the output frequency. Practical
limitations restrict the actual highest output frequency to about 1/3 the clock
frequency. Figure 6.5 shows the output of a DAC in a DDS system where the output
frequency is 30MHz and the clock frequency is 100MHz. An antialiasing filter must
follow the reconstruction DAC to remove the lower image frequency (100–
30=70MHz) as shown in the figure.
ALIASING IN A DDS SYSTEM
LPF
sin x
x
dB
fo
30MHz
fc
100MHz
IMAGE
2
IMAGE
2
3
4
0
10
20
30
40
50
60
70
4
3
80
90
100
3
4
110
120
130
FREQUENCY (MHz)
a
6.5
Note that the amplitude response of the DAC output (before filtering) follows a
sin(x)/x response with zeros at the clock frequency and multiples thereof. The exact
equation for the normalized output amplitude, A(fo), is given by:
 πf 
sin  o 
 fc 
,
A( f o ) =
πf o
fc
where fo is the output frequency and fc is the clock frequency.
5
This rolloff is because the DAC output is not a series of zero-width impulses (as in a
perfect re-sampler), but a series of rectangular pulses whose width is equal to the
reciprocal of the update rate. The amplitude of the sin(x)/x response is down 3.92dB
at the Nyquist frequency (1/2 the DAC update rate). In practice, the transfer
function of the antialiasing filter is designed to compensate for the sin(x)/x rolloff so
that the overall frequency response is relatively flat up to the maximum output DAC
frequency (generally 1/3 the update rate).
Another important consideration is that, unlike a PLL-based system, the higher
order harmonics of the fundamental output frequency in a DDS system will fold
back into the baseband because of aliasing. These harmonics cannot be removed by
the antialiasing filter. For instance, if the clock frequency is 100MHz, and the
output frequency is 30MHz, the second harmonic of the 30MHz output signal
appears at 60MHz (out of band), but also at 100–60=40MHz (the aliased component.
Similarly, the third harmonic (90MHz) appears inband at 100–90=10MHz, and the
fourth at 120–100MHz=20MHz. Higher order harmonics also fall within the Nyquist
bandwidth (DC to fc/2). The location of the first four harmonics is shown in the
diagram.
125MSPS DDS SYSTEM (AD9850)
The AD9850 125MSPS DDS system (Figure 6.6) uses a 32-bit phase accumulator
which is truncated to 14-bits (MSBs) before being passed to the lookup table. The
final digital output is 10-bits to the internal DAC. The AD9850 allows the output
phase to be modulated using an additional register and an adder placed between the
output of the phase accumulator register and the input to the lookup table. The
AD9850 uses a 5-bit word to control the phase which allows shifting the phase in
increments of 180°, 90°, 45°, 22.5°, 11.25°, and any combination thereof. The device
also contains an internal high speed comparator which can be configured to accept
the (externally) filtered output of the DAC to generate a low-jitter output pulse
suitable for driving the sampling clock input of an ADC. The full scale output
current can be adjusted from 10 to 20mA using a single external resistor, and the
output voltage compliance is +1V. Key specifications are summarized in Figure 6.7.
6
AD9850 CMOS 125MSPS DDS/DAC SYNTHESIZER
5
5
BYTE LOAD
PHASE
CONTROL
8 BITS X 5
DATA AND
CONTROL
INPUT
REGISTER
SERIAL LOAD
32
PHASE
ACCUMULATOR
32
14
SIN
LOOKUP
TABLE
1 BIT X 40
10
10-BIT
DAC
WORD LOAD CLOCK
ANALOG
OUT
RSET
REFERENCE CLOCK IN
HIGH SPEED
COMPARATOR
+
-
a
6.6
AD9850 DDS/DAC SYNTHESIZER KEY SPECIFICATIONS
n
125MSPS Clock Rate
n
On-Chip 10-bit DAC and High Speed Comparator
n
DAC SFDR > 50dBc @ 40MHz Output
n
32-bit Frequency Tuning
n
5-bit Phase Modulation
n
Simplified Control Interface: Byte-Parallel or Serial Load
n
+5V or +3.3V Supplies
n
380mW Dissipation @ 125MSPS on +5V Supply
(30mW Power-Down Mode)
n
28-Pin Shrink Small Outline Package (SSOP)
a
6.7
The frequency tuning (delta-phase register input word) and phase modulation words
are loaded into the AD9850 via a parallel or serial loading format. The parallel load
7
format consists of five consecutive loads of an 8-bit control word (byte). The first 8bit byte controls phase modulation (5-bits), power-down enable (1-bit), and loading
format (2-bits). Bytes 2-5 comprise the 32-bit frequency tuning word. The maximum
control register update frequency is 23MHz. Serial loading of the AD9850 is
accomplished via a 40-bit serial data stream on a single pin. Maximum update rate
of the control register in the serial-load mode is 3MHz.
The AD9850 consumes only 380mW of power on a single +5V supply at a maximum
125MSPS clock rate. The device is available in a 28-pin surface mount SSOP
(Shrink Small Outline Package).
DDS SYSTEMS AS ADC CLOCK DRIVERS
DDS systems such as the AD9850 provide an excellent method of generating the
sampling clock to the ADC, especially when the ADC sampling frequency must be
under software control and locked to the system clock (see Figure 6.8). The true DAC
output current Iout, drives a 200Ω, 42MHz lowpass filter which is source and load
terminated, thereby making the equivalent load 100Ω. The filter removes spurious
frequency components above 42MHz. The filtered output drives one input of the
AD9850 internal comparator. The complementary DAC output current drives a 100Ω
load. The output of the 100kΩ resistor divider placed between the two outputs is
decoupled and generates the reference voltage for the internal comparator.
The comparator output has a 2ns rise and fall time and generates a TTL/CMOScompatible square wave. The jitter of the comparator output edges is less than 20ps
rms. True and complementary outputs are available if required.
In the circuit shown (Figure 6.8), the total output rms jitter for a 40MSPS ADC
clock is 50ps rms, and the resulting degradation in SNR must be considered in wide
dynamic range applications.
8
USING DDS SYSTEMS AS ADC CLOCK DRIVERS
I
125 MHz
AD9850
DDS/DAC
SYNTHESIZER
DAC OUTPUT
200Ω
Ω
100kΩ
Ω
470pF
200 Ω
100kΩ
Ω
Ω
100Ω
FREQ.
42MHz
200 Ω
LPF
I
CONTROL
+
-
a
6.8
AMPLITUDE MODULATION IN A DDS SYSTEM
Amplitude modulation in a DDS system can be accomplished by placing a digital
multiplier between the lookup table and the DAC input as shown in Figure 6.9.
Another method to modulate the DAC output amplitude is to vary the reference
voltage to the DAC. In the case of the AD9850, the bandwidth of the internal
reference control amplifier is approximately 1MHz. This method is useful for
relatively small output amplitude changes as long as the output signal does not
exceed the +1V compliance specification.
9
AMPLITUDE MODULATION IN A DDS SYSTEM
fc
SIN
LOOKUP
TABLE
PHASE
ACCUMULATOR
N
N
AM
REGISTER
OUTPUT
DAC
MULTIPLIER
VREF
a
6.9
THE AD9830/9831 COMPLETE DDS SYSTEMS
The AD9830/9831 CMOS DDS systems (see Figure 6.10) contain two frequency
registers and four phase registers thereby allowing both frequency and phase
modulation. The registers are loaded through a parallel microprocessor port. The
DDS chips contain a 32-bit phase accumulator register, 12-bit sin ROM lookup table,
and a 10-bit DAC. The AD9830 operates at 50MSPS and dissipates 250mW on the
+5V supply. The AD9831 operates at 25MSPS and dissipates 150mW on a +5V
supply and 35mW on +3V. Key specifications for the devices are summarized in
Figure 6.11.
AD9830/9831, 50/25MSPS COMPLETE DDS SYSTEMS
10
AVDD
DVDD
AGND
FS ADJUST
DGND
VREF
CLOCK
FullScale
Control
COMP
FSELECT
FREQ0 REG
Phase
Accumulator
(32 Bit)
MUX
Σ
12
IOUT
SIN
ROM
10-Bit DAC
IOUT
FREQ1 REG
PHASE0 REG
PHASE1 REG
MUX
PHASE2 REG
PHASE3 REG
SLEEP
Parallel Register
CONTROL REG
D0
D15
RESET
Register
Select
MPU Interface
WR
A0
CS
A1
A2
PSEL0 PSEL1
a
6.10
AD9830/9831 DDS SYSTEMS KEY SPECIFICATIONS
n
50MSPS (AD9830), 25MSPS (AD9831) Update Rate
n
Single +5V (AD9830), +5V/+3V (AD9831) Supply
n
32-bit Phase Accumulator, 12-bit Address Sine ROM
n
On Chip 10-bit DAC (70dB SFDR)
n
Two On-Chip Frequency Modulation Registers
n
Four On-Chip Phase Modulation Registers
n
On-Chip Reference
n
Power Dissipation: 250mW (AD9830),
150mW (AD9831 @ +5V), 35mW (AD9831 @ +3V)
n
48-pin TQFP
a
6.11
11
SPURIOUS FREE DYNAMIC RANGE CONSIDERATIONS IN
DDS SYSTEMS
In many DDS applications, the spectral purity of the DAC output is of primary
concern. Unfortunately, the measurement, prediction, and analysis of this
performance is complicated by a number of interacting factors.
Even an ideal N-bit DAC will produce harmonics in a DDS system. The amplitude of
these harmonics is highly dependent upon the ratio of the output frequency to the
clock frequency. This is because the spectral content of the DAC quantization noise
varies as this ratio varies, even though its theoretical rms value remains equal to
q/√12 (where q is the weight of the LSB). The assumption that the quantization
noise appears as white noise and is spread uniformly over the Nyquist bandwidth is
simply not true in a DDS system (it is more apt to be a true assumption in an ADCbased system, because the ADC adds a certain amount of noise to the signal which
tends to "dither" or randomize the quantization error. However, a certain amount of
correlation still exists). For instance, if the DAC output frequency is set to an exact
submultiple of the clock frequency, then the quantization noise will be concentrated
at multiples of the output frequency, i.e., it is highly signal dependent. If the output
frequency is slightly offset, however, the quantization noise will become more
random, thereby giving an improvement in the effective SFDR.
This is illustrated in Figure 6.12, where a 4096 point FFT is calculated based on
digitally generated data from an ideal 12-bit DAC. In the left-hand diagram, the
ratio between the clock frequency and the output frequency was chosen to be exactly
32 (128 cycles of the sinewave in the FFT record length), yielding an SFDR of about
78dBc. In the right-hand diagram, the ratio was changed to 32.25196850394 (127
cycles of the sinewave within the FFT record length), and the effective SFDR is now
increased to 92dBc. In this ideal case, we observed a change in SFDR of 14dB just
by slightly changing the frequency ratio.
12
EFFECT OF RATIO OF CLOCK TO OUTPUT FREQUENCY
ON THEORETICAL 12-BIT DAC SFDR USING 4096-POINT FFT
fc / fo = 32.25196850394
fc/ fo = 32
0
500
1000
1500
2000
0
a
500
1000
1500
2000
6.12
Best SFDR can therefore be obtained by the careful selection of the clock and output
frequencies. However, in some applications, this may not be possible. In ADC-based
systems, adding a small amount of random noise to the input tends to randomize the
quantization errors and reduce this effect. The same thing can be done in a DDS
system as shown in Figure 6.13 (Reference 5). The pseudo-random digital noise
generator output is added to the DDS sine amplitude word before being loaded into
the DAC. The amplitude of the digital noise is set to about 1/2 LSB. This
accomplishes the randomization process at the expense of a slight increase in the
overall output noise floor. In most DDS applications, however, there is enough
flexibility in selecting the various frequency ratios so that dithering is not required.
13
INJECTION OF DIGITAL DITHER IN A DDS SYSTEM TO
RANDOMIZE QUANTIZATION NOISE AND INCREASE SFDR
M
DELTA
PHASE
REGISTER
PHASE
ACCUMULATOR
fc
SINE
LOOKUP
TABLE
PSEUDORANDOM
NUMBER
GENERATOR
ADDER
DAC
q
VN = rms
2
a
6.13
A non-ideal DAC will introduce several other mechanisms of distortion. First, the
overall integral non-linearity of the DAC transfer function will introduce harmonic
distortion. This distortion behaves much like that produced by the non-linearity of
an amplifier. The distortion due to the differential non-linearity of the DAC is highly
dependent upon the nature of the differential non-linearity and is difficult to predict
mathematically. The third source of DAC distortion are code-dependent output
glitches. In a DAC there is a transient (or glitch) produced whenever the DAC input
code changes. This glitch is usually worst at midscale, where the DAC makes the
transition between the codes 1000...000 and 0111...111, and all the DAC bits must
switch. These glitches occur because of the unequal turn-on/turn-off times of the
DAC current switches. They also occur at 1/4 scale, 1/8 scale, etc., with decreasing
amplitude. Because the glitches are code-dependent (hence signal-dependent) they
produce harmonics of the fundamental output DAC frequency. For instance, each
time the sinewave crosses through mid-scale, a glitch occurs, thereby producing a
second harmonic - since the sinewave passes through midscale twice each cycle. The
harmonics produced by these code-dependent glitches fold back into the Nyquist
bandwidth due to aliasing and thereby affect the SFDR.
CONTRIBUTORS TO DDS DAC DISTORTION
n
Resolution
n
Integral Non-Linearity
n
Differential Non-Linearity
n
Code-Dependent Glitches
n
Ratio of Clock Frequency to Output Frequency
14
(Even in an Ideal DAC)
n
Mathematical Analysis is Difficult !
a
6.14
Low distortion high-speed DACs generally have a specification for the area of the
worst glitch (called glitch impulse area). In general, the smaller the glitch area, the
better the distortion-but it is difficult to mathematically relate the distortion
performance to the glitch area. The glitch impulse area for low distortion DACs is
usually less than 30pV-sec. A typical midscale glitch impulse is shown for the
AD9721 DAC in Figure 6.15.
AD9720/AD9721 DAC MIDSCALE GLITCH SHOWS 1.34pV-s
NET IMPULSE AREA AND SETTLING TIME OF 4.5ns
AD9720
IOUT
100MHz
LPF
TEST CIRCUIT
a
5ns/DIVISION
50Ω
6.15
The best way to measure DAC performance is with a spectrum analyzer, with a
DDS system used to drive the DAC (Figure 6.16). Because there are nearly an
infinite combination of possible clock and output frequencies, SFDR is generally
specified for only a few selected combinations. One method is to plot the SFDR as a
function of clock frequency for the output frequency slightly offset from 1/3 or 1/4 the
clock frequency. The small frequency offset randomizes the quantization noise and
also allows the distortion products to be easily observed.
15
TEST SETUP FOR MEASURING DAC SFDR
PARALLEL OR
SERIAL PORT
N
PC
DDS
N
LATCH
DAC
fo
STABLE
FREQUENCY
REFERENCE
fc
SPECTRUM
ANALYZER
a
6.16
Note that for the output slightly offset from fc/3, the even harmonics will be aliased
very close to the output signal as shown in Figure 6.17. Similarly, for the output
slightly offset from fc/4, the odd harmonics will fall close to the output frequency
(Figure 6.18).The SFDR at fc/3 is usually considered a worse case condition and is
often plotted as a function of clock frequency as shown in Figure 6.19 for the
AD9721 10-bit, 100MSPS TTL-compatible DAC.
16
LOCATION OF EVEN HARMONICS FOR
fo = fc / 3 - ∆f
fo
∆f
10
4
2
8
fc
3
a
6.17
LOCATION OF ODD HARMONICS FOR
fo = fc / 4 - ∆ f
fo
∆f
9
5
3
7
fc
4
a
6.18
17
SFDR OF AD9721 10-BIT DAC FOR
fo ~ fc / 3 (BANDWIDTH: DC TO fc / 2)
80
SFDR
(dBc)
70
60
50
40
0
20
40
60
80
100
CLOCK FREQUENCY (MHz)
a
6.19
HIGH SPEED LOW DISTORTION DAC ARCHITECTURES
Because of the emphasis in communications systems for DDS DACs with high
SFDR, much effort has been placed on determining optimum DAC architectures.
Practically all low distortion high speed DACs make use of some form of nonsaturating current-mode switching. A straight binary DAC with one current switch
per bit produces code-dependent glitches as discussed above and is certainly not the
most optimum architecture (Figure 6.20). A DAC with one current source per code
level can be shown not to have code-dependent glitches, but it is not practical to
implement for high resolutions. However, this performance can be approached by
decoding the first few MSBs into a "thermometer" code and have one current switch
per level. For example, a 5-bit thermometer DAC would have an architecture similar
to that shown in Figure 6.21.
18
5-BIT BINARY DAC ARCHITECTURES
I
I
I
I
I
I
MSB
I/2
I/4
I/8
I/16
MSB
R
OUTPUT
R
2R
R
2R
R
2R
R
OUTPUT
R (CAN BE EXTERNAL)
a
6.20
5-BIT "THERMOMETER" OR "FULLY-DECODED" DAC
MSB
5-BIT
LATCH
5-TO-31
DECODE
LOGIC
31-BIT
LATCH
31
LINES
LSB
31
CURRENT
EQUAL
31
OUTPUT
CURRENT
LINES
SWITCHES
CLOCK
a
6.21
The input binary word is latched and then decoded into 31 outputs which drive a
second latch. The output of the second latch drives 31 equally weighted current
switches whose outputs are summed together. This scheme effectively removes
nearly all the code-dependence of the output glitch. The residual glitch that does
occur at the output is equal regardless of the output code change and can be filtered.
19
The distortion mechanisms associated with the full-decoded architecture are
primarily asymmetrical output slewing, finite switch turn-on and turn-off times, and
integral nonlinearity.
The obvious disadvantage of this type of thermometer DAC is the large number of
latches and switches required to make a 12, 10, or even 8-bit DAC. However, if this
technique is used on the 5 MSBs of an 8, 10, or 12-bit DAC, a significant reduction
in the code-dependent glitch is possible. This process is called segmentation and is
quite common in low distortion DACs.
Figure 6.22 shows a scheme whereby the first 5 bits of a 10-bit DAC are decoded as
described above and drive 31 equally weighted switches. The last 5 bits are derived
from binarily weighted current sources. Equally weighted current sources driving an
R/2R resistor ladder could be used to derive the LSBs, however, this approach
requires thin film resistors which are not generally available on a low-cost CMOS
process. Also, the use of R/2R networks lowers the DAC output impedance, thereby
requiring more drive current to develop the same voltage across a fixed load
resistance.
10-BIT SEGMENTED DAC
5
31
31
MSB
DECODE
10
10-BIT
LATCH
36-BIT
LATCH
5
FULLY
DECODED
MSB
DAC
CURRENT
OUTPUT
5
BINARY
LSB
DAC
CLOCK
a
6.22
The AD9850 internal 10-bit DAC uses two major stages of segmentation as shown
in Figure 6.23. The first 5 bits (MSBs) are fully decoded and drive 31 equally
weighted current switches (320µA each). The next 4 bits are decoded into 15 lines
which drive 15 current switches, each supplying 20µA (1/16 the current supplied by
each MSB switch). The LSB is latched and drives a single current switch which
supplies 10µA (1/32 the current supplied by each MSB switch). A total of 47 current
switches and latches are required to implement this architecture.
20
AD9850 10-BIT CMOS CURRENT SWITCH DAC CORE
5
31
31
31
CURRENT
SWITCHES
BITS 1-5
DECODE
10
10-BIT
LATCH
5-TO-31
4
47-BIT
LATCH
15
320µ
µA
15
BITS 6-9
DECODE
4-TO-15
LSB
1
1
CLOCK
a
15
CURRENT
SWITCHES
µA
20µ
1
CURRENT
SWITCH
10µ
µA
CURRENT
OUTPUT
FS =
10.23mA
6.23
The basic current switching cell is made up of a differential PMOS transistor pair as
shown in Figure 6.24. The differential pairs are driven with low-level logic to
minimize switching transients and time skew. The DAC outputs are symmetrical
differential currents which help to minimize even-order distortion products
(especially which driving a differential output such as a transformer or an op amp
differential I/V converter).
The overall architecture of the AD9850 is an excellent tradeoff between
power/performance and allows the entire DDS function to be implemented on a
standard CMOS process with no thin film resistors. Single-supply operation on
+3.3V or +5V makes the device extremely attractive for portable and low power
applications. The SFDR performance is typically 60, 55, and 45dBc for output
frequencies of 1, 20, and 40MHz, respectively (clock frequency = 125MSPS).
21
PMOS TRANSISTOR CURRENT SWITCHES
+VS
RL
RL
6.24
a
The AD9760 (10-bit), AD9762 (12-bit) and AD9764 (14-bit) 100MSPS DACs utilize
the same basic switching core as the AD9850. This family of DACs is pin-compatible,
and offers exceptional AC and DC performance. They operate on single +5V or +3V
supplies and contain on-chip latches, reference, and are ideal for the transmit
channel in wireless basestations, ADSL/HFC modems, and DDS applications. Key
specifications for the family are summarized in Figure 6.25.
AD9760/9762/9764 FAMILY OF 100MSPS DACs
n
Pin-Compatible 10-bit (AD9760), 12-bit (AD9762),
and 14-bit (AD9764)
n
SFDR for 15MHz Output: -60dBc
n
Low Glitch Impulse: 5pVsec
n
On-Chip Reference
n
Single +5V or +3V Supplies
n
Power Dissipation: 175mW @ 5V
n
Power-Down Mode: 30mW
a
6.25
22
IMPROVING SFDR USING SAMPLE-AND-HOLD
DEGLITCHERS
High-speed sample-and-hold amplifiers (such as the AD9100 and AD9101) can be
used to deglitch DAC outputs as shown in Figure 6.26. Just prior to latching new
data into the DAC, the SHA is put into the hold mode so that the DAC switching
glitches are isolated from the output. The switching transients produced by the SHA
are code-independent and occur at the clock frequency and hence are easily filtered.
However, great care must be taken so that the relative timing between the SHA
clock and the DAC update clock is optimum. In addition, the distortion performance
of the SHA must be at least 6 to 10dB better than the DAC, or no improvement in
SFDR will be realized. Achieving good results using an external SHA deglitcher
becomes increasingly more difficult as clock frequencies approach 100MSPS.
SAMPLE-AND-HOLD (SHA) USED AS DAC DEGLITCHER
DAC
SHA
OUTPUT
MODE CONTROL
DAC
INPUT
DAC
OUTPUT
SHA
MODE
1000 . . . 0
0111 . . . 1
T
H
H
H
T
T
T
SHA
OUTPUT
a
6.26
The AD6742 is a 12-bit, 65MSPS low distortion DAC with on-chip SHA deglitcher
designed for communications applications. This DAC is fabricated on the XFCB
process and provides 75dB SFDR for a 20MHz output. A functional diagram is
shown in Figure 6.27, and key specifications in Figure 6.28.
AD6742 12-BIT, 65MSPS DEGLITCHED DAC
23
V R E F REFIN B Y P A S S
DACREF
RL
- 2.4V
RE F.
IR E F
VCC
GND
RL
R-2R
NETW ORK S
DA TA
R E F.
I(D4-D0)
VEE
CURRE NT
V
T /H
AMP
VOUT
V
RSET
DREF
VEE
I(D 1 1- D5 )
S OU R C E S / S W I TC H E S
................. .
IN T E R N A L
TIM ING
DIGITAL INPUT STAGES AND LATCHES
(M S B )
(L SB )
D11 D 1 0
D9
D8
D7
D6
D5
D4
D3
D2
D1
D0
C L O C K CL OC K
a
6.27
AD6742 12-BIT, 65MSPS DAC KEY SPECIFICATIONS
n
12-bit, 65MSPS Communications DAC
n
Ideal for Wideband Multichannel Transmit Path
n
High SFDR: 78dB (typ) @ 20MHz Output, 65MSPS Update
n
Fabricated on XFCB process
n
On-Chip Reference
n
Dual 5V Supplies, 900mW power dissipation
a
6.28
HIGH SPEED INTERPOLATING DACS
Consider a DDS system which operates at a clock frequency of 100MSPS and
outputs a 30MHz sinewave (see Figure 6.29). The first aliased (or image) frequency
occurs at 100–30 = 70MHz. Assume we wish the antialiasing filter to attenuate this
image frequency component by 60dB. The filter must go from a passband of 30MHz
to 60dB stopband attenuation over the transition band lying between 30 and 70MHz
(approximately one octave). A Butterworth filter design gives 6dB attenuation per
octave for each pole. Therefore, a minimum of 10 poles is required to provide the
desired attenuation. Filters become even more complex as the transition band
becomes narrower.
24
LPF REQUIRED TO REJECT IMAGE FREQUENCY
TRANSITION
BAND
LPF
0
dB
fo
30MHz
fc
100MHz
IMAGE
IMAGE
-60
0
10
20
30
40
50
60
70
80
90
100
110
120
130
FREQUENCY (MHz)
a
6.29
In ADC-based systems, oversampling can ease the requirements on the antialiasing
filter, and a sigma-delta ADC has this inherent advantage. In a DAC-based system
(such as DDS), the concept of interpolation can be used in a similar manner. This
concept is common in digital audio CD players, where the basic update rate of the
data from the CD is about 44kSPS. "Zeros" are inserted into the parallel data,
thereby increasing the effective update rate to 4-times, 8-times, or 16-times the
fundamental throughput rate. The 4x, 8x, or 16x data stream is passed through a
digital interpolation filter which generates the extra data points. The high
oversampling rate moves the image frequencies higher, thereby allowing a less
complex filter with a wider transition band.
The same concept can be applied to a high speed DDS DAC. Assume a traditional
DAC is driven at an input word rate of 30MSPS (see Figure 6.30). The maximum
realizable DAC output frequency is about 10MHz. The image frequency component
at 30–10 = 20MHz must be attenuated by the analog antialiasing filter, and the
transition band of the filter is 10 to 20MHz.
Assume that we increase the update rate to 60MSPS by inserting a "zero" between
each original data sample. The parallel data stream is now 60MSPS and is passed
through the digital interpolation filter which computes the additional data points.
The response of the digital filter relative to the 2-times oversampling frequency is
shown in Figure 6.30. The analog antialiasing filter transition zone is now 10 to
50MHz (the first image occurs at 2fc–fo=60–10=50MHz).
25
fc = 30MSPS AND fc = 60MSPS
ANALOG LPF
fCLOCK = 30MSPS
dB
fo
IMAGE
IMAGE
10
20
30
40
IMAGE
50
IMAGE
60
70
80
FREQUENCY (MHz)
fCLOCK = 60MSPS
dB
fo
ANALOG
LPF
IMAGE
IMAGE
10
20
30
40
a
50
60
70
80
6.30
The AD977x is a 4-times oversampling interpolating 10-bit DAC, and a simplified
block diagram is shown in Figure 6.31. The device is designed to handle 10-bit input
word rates up to about 30MSPS. The internal digital filter consists of a 15-tap filter
operating at 2fc followed by a 7-tap filter operating at 4fc. The output word rate is
120MSPS, putting the image frequency at 4fc–fo=120–10=110MHz. SFDR of the
DAC for a 10MHz output is approximately 60dBc.
26
INCREASING THE DAC THROUGHPUT RATE BY "K"
USING A PLL AND A DIGITAL INTERPOLATION FILTER
(INTERPOLATING DAC)
10
fc
a
LATCH
10
DIGITAL
10
INTERPOLATION
FILTER
10
LATCH
K•fc
PLL
DAC
LPF
TYPICAL APPLICATION: fc = 33MSPS
fo = 10MHz
K = 4 OR 8
fo
6.31
QPSK SIGNAL GENERATION USING DDS (AD9853)
The AD9853 is a digital Quadrature Phase Shift Keying (QPSK) modulator useful in
the 5 to 40MHz return path transmitter in a hybrid fiber coax (HFC) CATV cable
modem application (see Figure 6.32). This allows asynchronous data transfer over
the HFC cable plant. The device takes the serial QPSK data input, splits it into an
in-phase (I) and quadrature (Q) signal. The I and Q channel data is then filtered and
passed through a digital quadrature modulator. The quadrature modulators are
driven by the sine and cosine outputs from the DDS section. The modulator outputs
are then recombined digitally and then converted into analog by an internal 10-bit
DAC. The resulting QPSK constellation is shown in Figure 6.33. This scheme of
modulation is quite common, and results in relatively high noise immunity. Key
specifications for the AD9853 are given in Figure 6.34.
27
AD9853 DIGITAL QPSK MODULATOR
I
FIR
FILTER
INTERPOLATING
FILTER
SERIAL
10
MUX
QPSK
INPUT
10-BIT
DAC
QPSK
FORMAT
FIR
FILTER
Q
INTERPOLATING
FILTER
2
COS
SIN
CLOCKS
140MSPS DDS AND CONTROL FUNCTIONS
X4 PLL
REF CLOCK INPUT
BURST
MODE
CONTROL
MASTER
RESET
FREQ.
UPDATE
WORD
LOAD
CLOCK
32-BIT FREQ.
TUNING AND
CONTROL WORD
a
6.32
QPSK CONSTELLATION
Q
11
01
I
00
10
a
6.33
AD9853 DIGITAL QPSK MODULATOR KEY SPECIFICATIONS
28
n
Performs Transmit Function for QPSK 5-40MHz
Hybrid Fiber Coax (HFC) Return Path
n
Includes Raised Cosine Pulse-Shaping Filter
(Alpha = 0.5) and Interpolation Filters
n
140MSPS Clock Frequency
n
46dBc SFDR @ 40MHz Output
n
+5V or +3.3V Operation
n
300mW Dissipation @ 125MSPS Clock Frequency
(30mW Power-Down Mode)
n
28-Pin SSOP Surface-Mount Package
a
6.34
29
REFERENCES
1.
R.E. Best, Phase-Locked Loops, McGraw-Hill, New York, 1984.
2.
F.M. Gardner, Phaselock Techniques, 2nd Edition, John Wiley,
New York, 1979.
3.
Phase-Locked Loop Design Fundamentals, Applications Note AN-535,
Motorola, Inc.
4.
The ARRL Handbook for Radio Amateurs, American Radio
Relay League, Newington, CT, 1992.
5.
Richard J. Kerr and Lindsay A. Weaver, Pseudorandom Dither for
Frequency Synthesis Noise, United States Patent Number 4,901,265,
February 13, 1990.
6.
Henry T. Nicholas, III and Henry Samueli, An Analysis of the Output
Spectrum of Direct Digital Frequency Synthesizers in the Presence of
Phase-Accumulator Truncation, IEEE 41st Annual Frequency Control
Symposium Digest of Papers, 1987, pp. 495-502, IEEE Publication No.
CH2427-3/87/0000-495.
7.
Henry T. Nicholas, III and Henry Samueli, The Optimization of Direct
Digital Frequency Synthesizer Performance in the Presence of Finite Word
Length Effects, IEEE 42nd Annual Frequency Control Symposium
Digest of Papers, 1988, pp. 357-363, IEEE Publication No. CH25882/88/0000-357.
30
SECTION 7
HIGH SPEED HARDWARE DESIGN
TECHNIQUES
Walt Kester, James Bryant, Walt Jung, Adolfo
Garcia, John McDonald, Joe Buxton
ANALOG CIRCUIT SIMULATION
Walt Kester, Joe Buxton
In recent years there has been much pressure placed on system designers to verify
their designs with computer simulations before committing to actual printed circuit
board layouts and hardware. Simulating complex digital designs is extremely
beneficial, and very often, the prototype phase can be eliminated entirely. However,
bypassing the prototype phase in high-speed/high-performance analog or mixedsignal circuit designs can be risky for a number of reasons.
For the purposes of this discussion, an analog circuit is any circuit which uses ICs
such as op amps, instrumentation amps, programmable gain amps (PGAs), voltage
controlled amps (VCAs), log amps, mixers, analog multipliers, etc. A mixed-signal
circuit is an A/D converter (ADC), D/A converter (DAC), or combinations of these in
conjunction with some amount of digital signal processing which may or may not be
on the same IC as the converters.
Consider a typical IC operational amplifier. It may contain some 20-40 transistors,
almost as many resistors, and a few capacitors. A complete SPICE (Simulation
Program with Integrated Circuit Emphasis, see Reference 1) model will contain all
these components, and probably a few of the more important parasitic capacitances
and spurious diodes formed by the various junctions in the op-amp chip. For highspeed ICs, the package and wirebond parasitics may also be included. This is the
type of model that the IC designer uses to optimize the device during the design
phase and is typically run on a CAD workstation. Because it is a detailed model, it
will be referred to as a micromodel. In simulations, such a model will behave very
much like the actual op-amp, but not exactly.
The IC designer uses transistor and other device models based on the actual process
upon which the component is fabricated. Semiconductor manufacturers invest
considerable time and money developing and refining these device models so that
the IC designers can have a high degree of confidence that the first silicon will work
and that mask changes (costing additional time and money) required for the final
manufactured product are minimized.
However, these device models are not published, neither are the IC micromodels, as
they contain proprietary information which would be of use to other semiconductor
companies who might wish to copy or improve on the design. It would also take far
too long for a simulation of a system containing several ICs (each represented by its
own micromodel) to reach a useful result. SPICE micromodels of analog ICs often
1
fail to converge (especially under transient conditions), and multiple IC circuits
make this a greater possibility.
For these reasons, the SPICE models of analog circuits published by manufacturers
or software companies are macromodels (as opposed to micromodels), which simulate
the major features of the component, but lack fine detail. Most manufacturers of
linear ICs (including Analog Devices) provide these macromodels for components
such as operational amplifiers, analog multipliers, references, etc. (Reference 2 and
3). These models represent approximations to the actual circuit, and parasitic effects
such as package capacitance and inductance and PC board layout are rarely
included. The models are designed to work with various versions of SPICE
simulation programs such as PSpice® (Reference 4) and run on workstations or
personal computers. The models are simple enough so that circuits using multiple
ICs can be simulated in a reasonable amount of computation time and with good
certainty of convergence. Consequently, SPICE modeling does not always reproduce
the exact performance of a circuit and should always be verified experimentally
using a carefully built prototype.
Finally, there are mixed-signal ICs such as A/D and D/A converters which have no
SPICE models, or if they exist, the models do not simulate dynamic performance
(Signal-to-noise, effective bits, etc.), and prototypes of circuits using them should
always be built.
SPICE SIMULATIONS:
MACROMODEL OR MICROMODEL?
MACROMODEL
MICROMODEL
METHODOLOGY
ADVANTAGES
DISADVANTAGES
Ideal Elements
Model the Device
Behavior
Fully
Characterized
Transistor Level
Fast Simulation
Time, Easy to
Modify
Most Complete
Model
May Not Model All
Characteristics
Slow Simulation,
Difficulty in
Convergence
Not Available to
Customers
a
7.1
The ADSpice Model
The ADSpice model was developed to advance the state-of-the-art in op amp
macromodelling and provide a tool for designers to simulate accurately their circuits.
Previously, the dominant model architecture was the Boyle model (Reference 3).
However, this model was developed over 20 years ago and does not accurately model
many of today's higher speed amplifiers. The primary reason for this is that the
Boyle model has only two frequency shaping poles and no zeroes. In contrast, the
2
ADSpice model has an open architecture that allows for unlimited poles and zeroes,
leading to much more accurate AC and transient responses.
The ADSpice model is comprised of three main portions: the input and gain stage,
the pole/zero stages, and the output stage. The input stage shown in Figure 7.2 uses
the only two transistors in the entire model. These are needed to model properly an
op amp's differential input stage characteristics. Although the example here uses
NPN transistors, the input stage can easily be modified to include PNP, JFET, or
CMOS devices. The rest of the input stage uses simple SPICE elements such as
resistors, capacitors, and controlled sources.
ADSpice INPUT AND GAIN STAGE MODEL
V+
OPEN LOOP GAIN = gm1 · R7
R3
+
IN-
1
R4
C2
VD
Q2
Q1
CIN
gm1
R7
R1
C3
Vd
R5
IOS
IN+
2
R6
R2
- +
VCS
3
+
-
EREF
I
V-
a
7.2
An example of a controlled source is gm1 in the gain stage, which is a voltage
controlled current source. It senses the differential collector voltage from the input
stage and converts that to a current. When the current flows through R7, a singleended voltage is produced. By making the product of gm1 and R7 equal to the open
loop gain, the entire open-loop gain is produced in the gain stage, which means that
all other stages are set to unity gain. This leads to significant flexibility in adding
and deleting stages.
Following the gain stage are an unlimited number of pole / zero stages and their
combinations. The typical topology of these stages is shown in Figure 7.3, which is
similar to the gain stage. The main difference is that now the product of gm2 times
R8 is equal to unity. The pole or zero frequency is set by the parallel combination of
the resistor and capacitor, R8-C4 for the pole and Rg-C5 for the zero. Because these
stages are unity gain, any number of them can be added or deleted without affecting
the low frequency response of the model. Instead, the high frequency gain and phase
response can be tailored to match accurately the actual amplifier's response. The
benefits are especially apparent in closed loop pulse response and stability analysis.
3
POLE AND ZERO STAGE
C5
Rg
gm2
R8
C4
+
-
E1
+
-
EREF
R10
gm2 R8 = 1
+
-
EREF
POLE
R10
E1 Rg + R10 = 1
ZERO
a
7.3
The output stage in Figure 7.4 not only models the open loop output impedance at
DC but with the inclusion of an inductor also models the rise in impedance at high
frequencies. Additionally, the output current is correctly reflected in the supply
currents. This is a significant improvement over the Boyle model because now the
power consumption of the circuit under load can be analyzed accurately.
Furthermore, circuits that use the supply currents for feedback can also be
simulated.
4
OUTPUT STAGE
V+
Rg
gm5
R13
D5
ISY
V5
+LO
VX
VO
D6
R14
R11
D8
D7
gm3
D10
D7
V6
-+
gm6
R12
gm4
VOUTPUT IMPEDANCE =
a
R11 + R12
2
+sLO
7.4
As an illustration of using the ADSpice model to predict circuit performance, the
AD847 op amp (50MHz unity gain-bandwidth product) output was loaded in a 65pF
capacitor and the response measured (both in ADSpice and in the circuit). The
results shown in Figure 7.5 illustrate good correlation between the simulated and
the actual response. As an additional example, extra parasitic capacitances were
added as shown in Figure 7.6, and the simulated and actual responses compared.
Again, note the excellent general agreement.
5
AD847 PULSE RESPONSE
20 0 mV
845Ω
Ω
1 00 mV
0 mV
AD847
+
65pF
-1 00 mV
50Ω
Ω
-2 00 mV
0ns
100 ns
300n s
200 ns
T ime
(A)
(C)
(B)
Properly laid out PC board and simulation agree closely
a
7.5
PC BOARD PARASITICS WILL ALTER THE RESULTS
2pF
2 00 mV
845Ω
1 00 mV
8pF
AD847
+
0 mV
65pF
-1 00 mV
50Ω
8pF
-2 00 mV
0ns
100ns
300ns
200 ns
Time
(A)
(B)
(C)
• Parasitic capacitances worsen the circuit's response
• Properly modelling the parasitics in SPICE yields good results
a
7.6
Other Features of ADSpice Models
In addition to offering models of op amps (both voltage and current feedback), which
allow simulation of AC and DC performance, Analog Devices has included noise in
many of its amplifier models. The capability to model a circuit's noise performance in
SPICE can be appreciated by anyone who has tried to analyze noise by hand. A
6
complete analysis is a very involved and tedious task which requires calculating all
the individual noise contributors and reflecting them to the input or output. The
procedure is further complicated by the fact that noise gain is generally a function of
frequency and can significantly affect results if not carefully considered.
To greatly simplify this task, the ADSpice model was enhanced to include noise
generators which accurately predict the broadband and 1/f noise of the actual
amplifier. Noise is currently modeled in a number of ADI op amps, variable gain
amplifiers, and voltage references. For further discussion on the noise model details,
see Reference 2.
In addition to amplifiers, ADSpice models exist for instrumentation amplifiers,
analog multipliers, voltage references, analog switches, multiplexers, matched
transistors, and buffers. A complete set of ADSpice models is available from Analog
Devices upon request.
ADSpice will give good approximations to actual performance, if used correctly.
However, the user must include the external components and parasitics which may
affect the device performance in the circuit. This becomes a difficult task at
frequencies much above 100MHz, and caution must be used in interpreting the
simulation results. There is no substitute for prototyping at these frequencies.
While pulse and frequency response can be successfully simulated using the
ADSpice models, distortion performance cannot be predicted since non-linear effects
are not included in the models. As mentioned previously, models for ADCs and DACs
are not available due to the difficulty in modeling their AC performance.
SUMMARY: ADSpice FEATURES
n
Transistor-Level Input Stage Model
n
Unlimited Poles and Zeros
n
Noise is Included in Some Models
n
Distortion is not Modeled
n
Over 500 Models Exist for:
u
u
u
u
u
u
n
Amplifiers
Instrumentation Amplifiers
Analog Multipliers
Voltage References
VCAs
Multiplexers and Switches
But There is no Substitute for a Good Prototype!!
a
7.7
7
PROTOTYPING TECHNIQUES
James Bryant, Walt Kester
The basic principle of a breadboard or prototype is that it is a temporary structure,
designed to test the performance of a circuit or system, and must therefore be easy
to modify.
There are many commercial prototyping systems, but almost all of them are
designed to facilitate the prototyping of digital systems, where noise immunities are
hundreds of millivolts or more. Non copper-clad Matrix board, Vectorboard,
wire-wrap, and plug-in breadboard systems are, without exception, unsuitable for
high performance or high frequency analog prototyping because their resistance,
inductance, and capacitance are too high. Even the use of standard IC sockets is
inadvisable in many prototyping applications.
An important consideration in selecting a prototyping method is the requirement for
a large-area ground plane. This is required for high frequency circuits as well as low
speed precision circuits, especially when prototyping circuits involving ADCs or
DACs. The differentiation between high-speed and high-precision mixed-signal
circuits is difficult to make. For example, 16+ bit ADCs (and DACs) may operate on
high speed clocks (>10MHz) with rise and fall times of less than a few nanoseconds,
while the effective throughput rate of the converters may be less than 100kSPS.
Successful prototyping of these circuits requires that equal attention be given to
good high-speed and high-precision circuit techniques.
The simplest technique for analog prototyping uses a solid copper-clad board as a
ground plane (Reference 5 and 6). The ground pins of the ICs are soldered directly to
the plane, and the other components are wired together above it. This allows HF
decoupling paths to be very short indeed. All lead lengths should be as short as
possible, and signal routing should separate high-level and low-level signals.
Connection wires should be located close to the surface of the board to minimize the
possibility of stray inductive coupling. In most cases, 18-gauge or larger insulated
wire should be used. Parallel runs should not be "bundled" because of possible
coupling. Ideally the layout (at least the relative placement of the components on the
board) should be similar to the layout to be used on the final PCB. This approach is
often referred to as deadbug prototyping because the ICs are often mounted upside
down with their leads up in the air (with the exception of the ground pins, which are
bent over and soldered directly to the ground plane). The upside-down ICs look like
deceased insects, hence the name.
Figure 7.8 shows a hand-wired breadboard using two high speed op amps which
gives excellent performance in spite of its lack of esthetic appeal. The IC op amps
are mounted upside down on the copper board with the leads bent over. The signals
are connected with short point-to-point wiring. The characteristic impedance of a
wire over a ground plane is about 120Ω, although this may vary as much as ±40%
depending on the distance from the plane. The decoupling capacitors are connected
directly from the op amp power pins to the copper-clad ground plane. When working
at frequencies of several hundred MHz, it is a good idea to use only one side of the
board for ground. Many people drill holes in the board and connect both sides
together with short pieces of wire soldered to both sides of the board. If care is not
8
taken, however, this may result in unexpected ground loops between the two sides of
the board, especially at RF frequencies.
"DEADBUG" PROTOTYPE
a
7.8
Pieces of copper-clad board may be soldered at right angles to the main ground
plane to provide screening, or circuitry may be constructed on both sides of the board
(with connections through holes) with the board itself providing screening. In this
case, the board will need standoffs at the corners to protect the components on the
underside from being crushed.
When the components of a breadboard of this type are wired point-to-point in the air
(a type of construction strongly advocated by Robert A. Pease of National
Semiconductor (Reference 6) and sometimes known as "bird's nest" construction)
there is always the risk of the circuitry being crushed and resulting short-circuits.
Also, if the circuitry rises high above the ground plane, the screening effect of the
ground plane is diminished, and interaction between different parts of the circuit is
more likely. Nevertheless, the technique is very practical and widely used because
the circuit may easily be modified (assuming the person doing the modifications is
adept at using a soldering iron, solder-wick, and a solder-sucker).
Another prototype breadboard is shown in Figure 7.9. The single-sided copper-clad
board has pre-drilled holes on 0.1" centers (Reference 7). Power busses are at the top
and bottom of the board. The decoupling capacitors are used on the power pins of
each IC. Because of the loss of copper area due to the pre-drilled holes, this
technique does not provide as low a ground impedance as a completely covered
copper-clad board.
9
"DEADBUG" PROTOTYPE USING PRE-DRILLED
SINGLE-SIDED COPPER-CLAD BOARD
a
7.9
In a variation of this technique, the ICs and other components are mounted on the
non-copper-clad side of the board. The holes are used as vias, and the point-to-point
wiring is done on the copper-clad side of the board. The copper surrounding each
hole used for a via must be drilled out to prevent shorting. This approach requires
that all IC pins be on 0.1" centers. Low profile sockets can be used for low frequency
circuits, and the socket pins allow easy point-to-point wiring.
There is a commercial breadboarding system which has most of the advantages of
the above techniques (robust ground, screening, ease of circuit alteration, low
capacitance and low inductance) and several additional advantages: it is rigid,
components are close to the ground plane, and where necessary, node capacitances
and line impedances can be calculated easily. This system is made by Wainwright
Instruments and is available in Europe as "Mini-Mount" and in the USA (where the
trademark "Mini-Mount" is the property of another company) as "Solder-Mount"
(Reference 8).
Solder-Mount consists of small pieces of PCB with etched patterns on one side and
contact adhesive on the other. These pieces are stuck to the ground plane, and
components are soldered to them. They are available in a wide variety of patterns,
including ready-made pads for IC packages of all sizes from 8-pin SOICs to 64-pin
DILs, strips with solder pads at intervals (which intervals range from 0.040" to
0.25", the range includes strips with 0.1" pad spacing which may be used to mount
DIL devices), strips with conductors of the correct width to form microstrip
transmission lines (50Ω, 60Ω, 75Ω or 100Ω) when mounted on the ground plane, and
a variety of pads for mounting various other components. Self-adhesive tinned
10
copper strips and rectangles (LO-PADS) are also available as tie-points for
connections. They have a relatively high capacitance to ground and therefore serve
as low-inductance decoupling capacitors. They come in sheet form and may be cut
with a knife or scissors. A few of the many types of Solder-Mount building-block
components are shown in Figure 7.10.
SAMPLES OF "SOLDER-MOUNT" COMPONENTS
a
7.10
The main advantage of Solder-Mount construction over "bird's nest" or "deadbug" is
that the resulting circuit is far more rigid, and, if desired, may be made far smaller
(the latest Solder-Mounts are for surface-mount devices and allow the construction
of breadboards scarcely larger than the final PC board, although it is generally more
convenient if the prototype is somewhat larger). Solder-Mount is sufficiently durable
that it may be used for small quantity production as well as prototyping.
Figure 7.11 shows an example of a 2.5GHz phase-locked-loop prototype built with
Solder-Mount. This is a high speed circuit, but the technique is equally suitable for
the construction of high resolution low frequency analog circuitry. A particularly
convenient feature of Solder-Mount at VHF is the ease with which it is possible to
make a transmission line.
"SOLDER-MOUNT" PROTOTYPE
11
a
7.11
If a conductor runs over a ground plane, it forms a microstrip transmission line. The
Solder-Mount components include strips which form microstrip lines when mounted
on a ground plane (they are available with impedances of 50Ω, 60Ω, 75Ω, and
100Ω). These strips may be used as transmission lines, for impedance matching, or
simply as power buses. (Glass fiber/epoxy PCB is somewhat lossy at VHF and UHF,
but the losses will probably be tolerable if microstrip runs are short.)
Both the "deadbug" and the "Solder-Mount" prototyping techniques become
somewhat tedious for complex analog or mixed-signal circuits. Larger circuits are
often better prototyped using more formal layout techniques.
An approach to prototyping more complex analog circuits is to actually lay out a
double-sided board using CAD techniques. PC-based software layout packages offer
ease of layout as well as schematic capture to verify connections (Reference 9).
Although most layout software has some amount of auto-routing capability, this
feature is best left to digital designs. After the components are placed in their
desired positions, the interconnections should be routed manually following good
analog layout guidelines. After the layout is complete, the software verifies the
connections per the schematic diagram net list.
Many design engineers find that they can use CAD techniques to lay out simple
boards themselves, or work closely with a layout person who has experience in
analog circuit boards. The result is a pattern-generation tape (or Gerber file) which
would normally be sent to a PCB manufacturing facility where the final board is
made. Rather than use a PC board manufacturer, however, automatic drilling and
milling machines are available which accept the PG tape directly (Reference 10).
These systems produce single and double-sided circuit boards directly by drilling all
holes and using a milling technique to remove copper and create insulation paths
and finally, the finished board. The result is a board very similar to the final
manufactured double-sided PC board, the chief exception being that there is no
12
"plated-through" hole capability, and any "vias" between the two layers of the board
must be wired and soldered on both sides. Minimum trace widths of 25 mils (1 mil =
0.001") and 12 mil spacing between traces are standard, although smaller trace
widths can be achieved with care. The minimum spacing between lines is dictated
by the size of the milling bit, typically 10 to 12 mils. An example of such a prototype
board is shown in Figure 7.12 (top view) and Figure 7.13 (bottom view).
"MILLED" PROTOTYPE - TOP VIEW
a
7.12
"MILLED" PROTOTYPE - BOTTOM VIEW
13
a
7.13
IC sockets can degrade the performance of high speed or high precision analog ICs.
Although they make prototyping easier, even low-profile sockets often introduce
enough parasitic capacitance and inductance to degrade the performance of the
circuit. If sockets must be used in high speed circuits, an IC socket made of
individual pin sockets (sometimes called cage jacks) mounted in the ground plane
board may be acceptable (clear the copper, on both sides of the board, for about
0.5mm around each ungrounded pin socket and solder the grounded ones to ground
on both sides of the board). Both capped and uncapped versions of these pin sockets
are available (AMP part numbers 5-330808-3, and 5-330808-6, respectively). The
pin sockets protrude through the board far enough to allow point-to-point wiring
interconnections between them (see Figure 7.14).
The spring-loaded gold-plated contacts within the pin socket makes good electrical
and mechanical connection to the IC pins. Multiple insertions, however, may
degrade the performance of the pin socket. The uncapped versions allow the IC pins
to extend out the bottom of the socket. After the prototype is functional and no
further changes are to be made, the IC pins can be soldered directly to the bottom of
the socket, thereby making a permanent and rugged connection.
PIN SOCKETS (CAGE JACKS) HAVE MINIMUM PARASITIC
RESISTANCE, INDUCTANCE, AND CAPACITANCE
COPPER
SOLDER
SPRING
CONTACTS
SOLDER
PCB DIELECTRIC
PCB DIELECTRIC
SOLDER
SOLDER
CAPPED OR UNCAPPED
VERSIONS AVAILABLE
a
7.14
The prototyping techniques discussed so far have been limited to single or doublesided PC boards. Multilayer PC boards do not easily lend themselves to standard
prototyping techniques. If multilayer board prototyping is required, one side of a
double-sided board can be used for ground and the other side for power and signals.
Point-to-point wiring can be used for additional runs which would normally be
placed on the additional layers provided by a multi-layer board. However, it is
difficult to control the impedance of the point-to-point wiring runs, and the high
14
frequency performance of a circuit prototyped in this manner may differ significantly
from the final multilayer board.
Other difficulties in prototyping may occur with op amps or other linear devices
having bandwidths greater than a few hundred megahertz. Small variations in
parasitic capacitance (<1pF) between the prototype and the final board may cause
subtle differences in bandwidth and settling time. Oftentimes prototyping is done
with DIP packages, when the final production package is an SOIC. This can account
for differences between prototype and final PC board performance.
EVALUATION BOARDS
Walt Kester
Most manufacturers of analog ICs provide evaluation boards (usually at a nominal
cost) which allow customers to evaluate products without constructing their own
prototypes. Regardless of the product, the manufacturer has taken proper
precautions regarding grounding, layout, and decoupling to ensure optimum device
performance. The artwork or CAD file is usually made available free of charge,
should the customer wish to copy the layout directly or make modifications to suit
the application.
Figure 7.15 shows the schematic for the AD8001 (SOIC package) 800MHz op amp
evaluation board. Figures 7.16 and 7.17, respectively, show the top and bottom side
of the PCB. The amplifier is connected in the non-inverting mode. The top side
(Figure 7.16) shows the top side of the SOIC package along with input and output
SMA connectors. Notice that the ground plane is cut away around the SOIC in order
to minimize parasitic capacitance. The bottom side of the board (Figure 7.17) shows
the surface mount resistors and capacitors which comprise the op amp gain-setting
and power supply decoupling circuits, respectively.
15
AD8001AR (SOIC) 800MHz OP AMP: NON-INVERTING
MODE EVALUATION BOARD SCHEMATIC
681Ω
RF
+VS
+
10µF
0.01µF
1000pF
RG
681Ω
RO
OUT
AD8001
49.9Ω
+
IN
RT
49.9Ω
1000pF
0.01µF
+
10µF
-VS
a
7.15
AD8001AR (SOIC) EVALUATION BOARD - TOP VIEW
a
7.16
AD8001AR (SOIC) EVALUATION BOARD - BOTTOM VIEW
16
a
7.17
In high speed/high precision ICs, special attention must be given to power supply
decoupling. For example, fast slewing signals into relatively low impedance loads
produce high speed transient currents at the power supply pins of an op amp. The
transient currents, in turn, produce corresponding voltages across any parasitic
impedance which may exist in the power supply traces. These voltages, in turn, may
couple to the amplifier output because of the op amp's finite power supply rejection
at high frequencies.
A three-capacitor decoupling scheme was chosen for the AD8001 evaluation board to
ensure a low impedance path to ground at all transient frequencies. The highest
frequency transients are shunted to ground by the 1000pF and the 0.01µF ceramic
capacitors. These are located as close to the power supply pins as possible to
minimize any series inductance and resistance. Because the devices are surface
mount, there is minimum stray inductance and resistance in the path to the ground
plane. The lower frequency transient currents are shunted to ground by the 10µF
tantalum capacitors.
The input and output signal traces are of the AD8001 evaluation board are 50Ω
microstrip transmission lines. Notice that there is considerable continuous ground
plane area on both sides of the PCB. Plated-through holes connect the top and
bottom side ground planes at several points in order to maintain low impedance
ground continuity at high frequencies.
Evaluation boards can range from relatively simple ones (op amps, for example) to
rather complex ones for mixed-signal ICs such as A/D converters. ADC evaluation
boards often have on-board memory and DSPs for analyzing the ADC performance.
Software is often provided with these more complex evaluation boards so that they
can interface with a personal computer to perform complex signal analysis such as
histogram and FFT testing.
17
Complete evaluations of ADCs requires the use of FFTs to fully characterize the
devices AC performance. A typical test setup is shown in Figure 7.18. The
manufacturer's evaluation board is used as a means for interfacing to the ADC. The
evaluation board is designed to allow easy access to the ADC inputs and outputs
while also providing a good layout (including all necessary references, buffer
amplifiers, and decoupling). The evaluation board allows the ADC output data to be
captured on a parallel output connector. Most ADC evaluation boards contain an onboard DAC which can be used to check the functionality of the ADC, but is
somewhat limited in performing meaningful AC testing. A block diagram of the
AD9042 (12-bits, 41MSPS) evaluation board is shown in Figure 7.19, and a photo in
Figure 7.20.
TEST SETUP REQUIRED TO EVALUATE HIGH SPEED ADCs
POWER
SUPPLIES
LOW PHASE
JITTER
SINEWAVE
SOURCE
BANDPASS
FILTER
MEMORY
OR LOGIC
ANALYZER
ADC ON
EVALUATION
BOARD
CLOCK
LOW PHASE
JITTER
SAMPLING
CLOCK
SOURCE
fs
PC
PARALLEL OR
SERIAL PORT
a
MONI TOR
18
7.18
AD9042 12-BIT, 41MSPS ADC EVALUATION BOARD
FUNCTIONAL DIAGRAM
74AS00
74AS00
XTAL
OSC
40.96MHz
499Ω
Ω
100 Ω
AD9042
ENCODE
EXTERNAL
SAMPLING
CLOCK
50 Ω
CK
74AC574
REGISTERS
(2)
ENCODE
T1 - 1T
MINICIRCUITS
499Ω
Ω
µF
0.1µ
ANALOG
INPUT
60.4Ω
Ω
a
7.19
AD9042 EVALUATION BOARD - TOP VIEW
a
7.20
19
The most complex part of the problem is usually designing the buffer memory
module. A high speed logic analyzer is one method of capturing the ADC output
data, and interfaces easily to the ADC evaluation board. Data from the logic
analyzer can be loaded into a PC through either parallel or serial ports. Once the
ADC data is inside the PC, software packages such as Mathcad can be used to
perform the actual FFT.
Another alternative is to use a commercially-available data acquisition module that
plugs directly into a card slot of the PC. These modules come complete with FFT and
other ADC test software, but are not easily portable from one PC to another and are
generally difficult to interface with laptop computers.
Although fast and relatively low power memories (FIFOs) are available
commercially, designing a buffer memory, the interfaces to the ADC and the PC,
and the necessary software can be a time-consuming project. Analog Devices has
designed a simple 16-bit by 16k deep 100MHz memory board (3 x 4 inches) and the
necessary software to allow high speed ADC evaluation boards to interface directly
with the parallel printer port of most PCs. The core of the memory design is the
IDT72265 16k by 18-bit wide FIFO or alternately, the IDT72255 is an 8k pin
compatible device which may be substituted if the deeper memory is not required.
This FIFO chip features fully independent I/O ports that allow data to be loaded at
up to 100MSPS and downloaded at the rate of a parallel printer port. Since the ports
are independent, both can operate simultaneously, i.e., data may be read out while
new data is being written. The chip takes care of all addressing, overhead and much
of the hand-shaking for these operations. Included is circuitry that prevents unread
data from being overwritten, eliminating the need for extensive write control
circuitry.
A photograph of the Fifo Memory board is shown in Figure 7.21, and Figure 7.22
shows it connected to the AD9042 evaluation board.
BUFFER MEMORY FIFO BOARD - TOP VIEW
20
a
7.21
MEMORY BOARD / AD9042 EVALUATION BOARD
a
7.22
Using this hardware and Windows-based software to capture the ADC data, many
testing possibilities exist. Figure 7.23 shown a time-domain plot of data captured
using the fifo memory. Once the data is captured, FFT analysis (Figure 7.24) or
DNL histograms (Figure 7.25) are easily generated.
DATA CAPTURE PC OUTPUT DISPLAY
a
7.23
21
FFT OUTPUT
a
7.24
DNL HISTOGRAM
22
a
7.25
In summary, good analog designers utilize as many tools as possible to ensure that
the final system design performs correctly. The first step is the intelligent use of IC
macromodels, where available, to simulate the circuit. The second step is the
construction of a prototype board to further verify the design and the simulation.
The final PCB layout should be then be based on the prototype layout as much as
possible.
Finally, evaluation boards can be extremely useful in evaluating new analog ICs,
and allow designers to verify the IC performance with a minimum amount of effort.
The layout of the components on the evaluation board can serve as a guide to both
the prototype and the final PC board layout. Gerber files are generally available for
all evaluation board layouts and may be obtained at no charge.
23
REFERENCES: SIMULATION, PROTOTYPING, AND
EVALUATION BOARDS
1.
Paolo Antognetti and Guiseppe Massobrio, Ed, Semiconductor Device
Modeling with SPICE, McGraw Hill, 1988.
2.
Amplifier Applications Guide, Section 13, Analog Devices, Inc.,
Norwood, MA, 1992.
3.
Boyle, et al, Macromodelling of Integrated Circuit
Operational Amplifiers, IEEE Journal of Solid State Circuits,
Vol. SC-9, no.6, December 1974.
4.
PSpice® Simulation software.
MicroSim Corporation, 20 Fairbanks, Irvine, CA 92718, 714-770-3022
5.
Jim Williams, High Speed Amplifier Techniques, Linear Technology
Application Note 47, August, 1991.
6.
Robert A. Pease, Troubleshooting Analog Circuits, ButterworthHeinemann, 1991.
7.
Vector Electronic Company, 12460 Gladstone Ave., Sylmar, CA 91342,
Tel. 818-365-9661, Fax. 818-365-5718.
8.
Wainwright Instruments Inc., 69 Madison Ave., Telford, PA,
18969-1829, Tel. 215-723-4333, Fax. 215-723-4620.
Wainwright Instruments GmbH, Widdersberger Strasse 14,
DW-8138 Andechs-Frieding, Germany. Tel: +49-8152-3162,
Fax: +49-8152-40525.
9.
Schematic Capture and Layout Software:
PADS Software, INC, 165 Forest St., Marlboro, MA, 01752 and
ACCEL Technologies, Inc., 6825 Flanders Dr., San Diego, CA,
92121
10.
Prototype Board Cutters:
LPKF CAD/CAM Systems, Inc., 6190 Artic Dr, PO Box 6209,
Beaverton, OR, 97005 and
T-Tech, Inc., 5591-B New Peachtree Road, Atlanta, GA,
34341
11.
Howard W. Johnson and Martin Graham, High-Speed Digital Design,
PTR Prentice Hall, 1993.
12.
Practical Analog Design Techniques, Analog Devices, 1995.
24
25
GROUNDING IN HIGH SPEED SYSTEMS
Walt Kester, James Bryant
The importance of maintaining a low impedance large area ground plane is critical
to practically all analog circuits today, especially at high speeds. The ground plane
not only acts as a low impedance return path for high frequency currents but also
minimizes EMI/RFI emissions. Because of the shielding action of the ground plane,
the circuits susceptibility to external EMI/RFI is also reduced.
All IC ground pins should be soldered directly to the ground plane to minimize series
inductance. Power supply pins should be decoupled to the ground plane using low
inductance ceramic surface mount capacitors. If through-hole mounted ceramic
capacitors must be used, their leads should be less than 1mm. Ferrite beads may be
also required.
The ground plane allows the impedance of PCB traces to be controlled, and high
frequency signals can be terminated in the characteristic impedance of the trace to
minimize reflections when necessary.
Each PCB in the system should have at least one complete layer dedicated to the
ground plane. Ideally, a double-sided board should have one side dedicated to
ground and the other side for interconnections. In practice, this is not possible, since
some of the ground plane will certainly have to be removed to allow for signal and
power crossovers and vias. Nevertheless, as much area as possible should be
preserved, and at least 75% should remain. After completing an initial layout, the
ground layer should be checked carefully to make sure there are no isolated ground
"islands." IC ground pins located in a ground "island" have no current return path to
the ground plane.
The best way of minimizing ground impedance in a multicard system is to use
another PCB as a backplane for interconnections between cards, thus providing a
continuous ground plane to the mother card. The PCB connector should have at
least 30-40% of its pins devoted to ground, and these pins should be connected to the
ground plane on the backplane mother card. To complete the overall system
grounding scheme there are two possibilities: (1) The backplane ground plane can be
connected to chassis ground at numerous points, thereby diffusing the various
ground current return paths. (2) The ground plane can be connected to a single
system "star ground" point (generally at the power supply).
The first approach is often used at very high frequencies and where the return
currents are relatively constant. The low ground impedance is maintained all the
way through the PC boards, the backplane, and ultimately the chassis. It is critical
that good electrical contact be made where the grounds are connected to the sheet
metal chassis. This requires self-tapping sheet metal screws or "biting" washers.
Special care must be taken where anodized aluminum is used for the chassis
material, since its surface acts as an insulator.
In other systems, especially high speed ones with large amounts of digital circuitry,
it is highly desirable to physically separate sensitive analog components from noisy
digital components. It is usually desirable to use separate ground planes for the
analog and the digital circuitry. On PCBs which have both analog and digital
1
circuits, there are two separate ground planes. These planes should not overlap in
order to minimize capacitive coupling between the two. The separate analog and
digital ground planes are continued on the backplane using either motherboard
ground planes or "ground screens" which are made up of a series of wired
interconnections between the connector ground pins. The arrangement shown in
Figure 7.26 illustrates that the two planes are kept separate all the way back to a
common system "star" ground, generally located at the power supplies. The
connections between the ground planes, the power supplies, and the "star" should be
made up of multiple bus bars or wide copper brads for minimum resistance and
inductance. The back-to-back Schottky diodes on each PCB are inserted to prevent
accidental DC voltage from developing between the two ground systems when cards
are plugged and unplugged.
SEPARATING ANALOG AND DIGITAL GROUNDS
VA
PCB
ANALOG
GROUND
PLANE
A
VD
VA
ANALOG
GROUND
PLANE
DIGITAL
GROUND
PLANE
D
VD
DIGITAL
GROUND
PLANE
A
PCB
D
DIGITAL GND PLANE
BACKPLANE
ANALOG GND PLANE
VA
POWER
SUPPLIES
SYSTEM STAR
GROUND
a
VD
7.26
Sensitive analog components such as amplifiers and voltage references are
referenced and decoupled to the analog ground plane. The ADCs and DACs (and
even some mixed-signal ICs) should be treated as analog components and also
grounded and decoupled to the analog ground plane. At first glance, this may seem
somewhat contradictory, since a converter has an analog and digital interface and
usually pins designated as analog ground (AGND) and digital ground (DGND). The
diagram shown in Figure 7.27 will help to explain this seeming dilemma.
2
PROPER GROUNDING OF ADCs, DACs,
AND OTHER MIXED-SIGNAL ICs
VA
A
ADC,
OR
DAC
ANALOG
IN/OUT
D
VA
VD
D
A
CSTRAY
ANALOG
CIRCUITS
DIGITAL
CIRCUITS
"QUIET"
DIGITAL
BUFFER
LATCH
NOISY
DATA BUS
B
A
CSTRAY
IA
ID
AGND
DGND
A
A
A
∆V
D
= DIGITAL
D GROUND PLANE
= ANALOG
A GROUND PLANE
a
7.27
Inside an IC that has both analog and digital circuits, such as an ADC or a DAC, the
grounds are usually kept separate to avoid coupling digital signals into the analog
circuits. Figure 7.27 shows a simple model of a converter. There is nothing the IC
designer can do about the wirebond inductance and resistance associated with
connecting the pads on the chip to the package pins except to realize it's there. The
rapidly changing digital currents produce a voltage at point B which will inevitably
couple into point A of the analog circuits through the stray capacitance, CSTRAY. In
addition, there is approximately 0.2pF unavoidable stray capacitance between every
pin of the IC package! It's the IC designer's job to make the chip work in spite of
this. However, in order to prevent further coupling, the AGND and DGND pins
should be joined together externally to the analog ground plane with minimum lead
lengths. Any extra impedance in the DGND connection will cause more digital noise
to be developed at point B; it will, in turn, couple more digital noise into the analog
circuit through the stray capacitance.
The name "DGND" on an IC tells us that this pin connects to the digital ground of the
IC. This does not imply that this pin must be connected to the digital ground of the
system.
It is true that this arrangement will inject a small amount of digital noise on the
analog ground plane. These currents should be quite small, and can be minimized by
ensuring that the converter input/or output does not drive a large fanout (they
normally can't by design). Minimizing the fanout on the converter's digital port will
also keep the converter logic transitions relatively free from ringing, and thereby
minimize any potential coupling into the analog port of the converter. The logic
3
supply pin (VD) can be further isolated from the analog supply by the insertion of a
small lossy ferrite bead as shown in Figure 7.27. The internal digital currents of the
converter will return to ground through the VD pin decoupling capacitor (mounted
as close to the converter as possible) and will not appear in the external ground
circuit. It is always a good idea (as shown in Figure 7.27) to place a buffer latch
adjacent to the converter to isolate the converter's digital lines from any noise which
may be on the data bus. Even though a few high speed converters have three-state
outputs/inputs, this isolation latch represents good design practice.
The buffer latch and other digital circuits should be grounded and decoupled to the
digital ground plane of the PC board. Notice that any noise between the analog and
digital ground plane reduces the noise margin at the converter digital interface.
Since digital noise immunity is of the orders of hundreds or thousands of millivolts,
this is unlikely to matter.
POWER SUPPLY, GROUNDING, AND DECOUPLING POINTS
VA
VA
A
A
VD
VA
VD
A
D
BUFFER
LATCH
ADC
OR
DAC
A
A
TO OTHER
DIGITAL
CIRCUITS
AGND DGND
VA
A
A
VOLTAGE
REFERENCE
A
a
D
SAMPLING
CLOCK
GENERATOR
A
VA
A
A
D
= ANALOG GROUND PLANE
= DIGITAL GROUND PLANE
7.28
Separate power supplies for analog and digital circuits are also highly desirable. The
analog supply should be used to power the converter. If the converter has a pin
designated as a digital supply pin (VD), it should either be powered from a separate
analog supply, or filtered as shown in the diagram. All converter power pins should
be decoupled to the analog ground plane, and all logic circuit power pins should be
decoupled to the digital ground plane. If the digital power supply is relatively quiet,
it may be possible to use it to supply analog circuits as well, but be very cautious.
The sampling clock generation circuitry should also be grounded and heavilydecoupled to the analog ground plane. As previously discussed, phase noise on the
sampling clock produces degradation in system SNR.
A low phase-noise crystal oscillator should be used to generate the ADC sampling
clock, because sampling clock jitter modulates the input signal and raises the noise
4
and distortion floor. The sampling clock generator should be isolated from noisy
digital circuits and grounded and decoupled to the analog ground plane, as is true
for the op amp and the ADC.
Ideally, the sampling clock generator should be referenced to the analog ground
plane in a split-ground system. However, this is not always possible because of
system constraints. In many cases, the sampling clock must be derived from a
higher frequency multi-purpose system clock which is generated on the digital
ground plane. If it is passed between its origin on the digital ground plane to the
ADC on the analog ground plane, the ground noise between the two planes adds
directly to the clock and will produce excess jitter. The jitter can cause degradation
in the signal-to-noise ratio and also produce unwanted harmonics. This can be
remedied somewhat by transmitting the sampling clock signal as a differential one
using either a small RF transformer or a high speed differential driver and receiver
as shown in Figure 7.29. The driver and receiver should be ECL to minimize phase
jitter. In either case, the original master system clock should be generated from a
low phase noise crystal oscillator.
SAMPLING CLOCK DISTRIBUTION FROM
DIGITAL TO ANALOG GROUND PLANES
+VD
+VD
LOW PHASE
NOISE
MASTER CLOCK
DIGITAL
GROUND PLANE
(D)
ANALOG
GROUND PLANE
(A)
SYSTEM CLOCK
GENERATORS
SAMPLING
CLOCK
METHOD 1
D
D
A
D
+VD
+VD
+VA
SAMPLING
CLOCK
DSP
METHOD 2
D
D
a
A
7.29
It is evident that noise can be minimized by paying attention to the system layout
and preventing different signals from interfering with each other. High level analog
signals should be separated from low level analog signals, and both should be kept
away from digital signals. We have seen elsewhere that in waveform sampling and
reconstruction systems the sampling clock (which is a digital signal) is as vulnerable
to noise as any analog signal, but is as liable to cause noise as any digital signal,
and so must be kept isolated from both analog and digital systems.
If a ground plane is used, as it should in be most cases, it can act as a shield where
sensitive signals cross. Figure 7.30 shows a good layout for a data acquisition board
5
where all sensitive areas are isolated from each other and signal paths are kept as
short as possible. While real life is rarely as tidy as this, the principle remains a
valid one.
A PC BOARD LAYOUT SHOWING GOOD SIGNAL ROUTING
SAMPLING CLOCK
GENERATOR
TIMING
CIRCUITS
CONTROL
LOGIC
ADC
BUFFER
LATCH
DEMULTIPLEXER
REF
FILTER
DSP
BUFFER
MEMORY
AMPLIFIER
POWER
MULTIPLE
GROUNDS
ANALOG
INPUT
ADDRESS
BUS
DATA
BUS
a
MULTIPLE
GROUNDS
7.30
There are a number of important points to be considered when making signal and
power connections. First of all a connector is one of the few places in the system
where all signal conductors must run parallel - it is therefore a good idea to separate
them with ground pins (creating a faraday shield) to reduce coupling between them.
Multiple ground pins are important for another reason: they keep down the ground
impedance at the junction between the board and the backplane. The contact
resistance of a single pin of a PCB connector is quite low (of the order of 10 mOhms)
when the board is new - as the board gets older the contact resistance is likely to
rise, and the board's performance may be compromised. It is therefore well
worthwhile to afford extra PCB connector pins so that there are many ground
connections (perhaps 30-40% of all the pins on the PCB connector should be ground
pins). For similar reasons there should be several pins for each power connection,
although there is no need to have as many as there are ground pins.
6
POWER SUPPLY NOISE REDUCTION AND
FILTERING
Walt Jung and John McDonald
Precision analog circuitry has traditionally been powered from well regulated, low
noise linear power supplies. During the last decade however, switching power
supplies have become much more common in electronic systems. As a consequence,
they also are being used for analog supplies. Good reasons for the general popularity
include their high efficiency, low temperature rise, small size, and light weight.
In spite of these benefits, switchers do have drawbacks, most notably high output
noise. This noise generally extends over a broad band of frequencies, resulting in
both conducted and radiated noise, as well as unwanted electric and magnetic fields.
Voltage output noise of switching supplies are short-duration voltage transients, or
spikes. Although the fundamental switching frequency can range from 20kHz to
1MHz, the spikes can contain frequency components extending to 100MHz or more.
While specifying switching supplies in terms of RMS noise is common vendor
practice, as a user you should also specify the peak (or p-p) amplitudes of the
switching spikes, with the output loading of your system.
The following section discusses filter techniques for rendering a noisy switcher
output analog ready, that is sufficiently quiet to power precision analog circuitry
with relatively small loss of DC terminal voltage. The filter solutions presented are
generally applicable to all power supply types incorporating switching element(s) in
their energy path. This includes various DC-DC converters as well as popular 5V
(PC type) supplies.
An understanding of the EMI process is necessary to understand the effects of
supply noise on analog circuits and systems. Every interference problem has a
source, a path, and a receptor [Reference 1]. In general, there are three methods for
dealing with interference. First, source emissions can be minimized by proper
layout, pulse-edge rise time control/reduction, filtering, and proper grounding.
Second, radiation and conduction paths should be reduced through shielding and
physical separation. Third, receptor immunity to interference can be improved, via
supply and signal line filtering, impedance level control, impedance balancing, and
utilizing differential techniques to reject undesired common-mode signals. This
section focuses on reducing switching power supply noise with external post filters.
Tools useful for combating high frequency switcher noise are shown by Figure 7.31.
They differ in electrical characteristics as well as practicality towards noise
reduction, and are listed roughly in an order of priorities. Of these tools, L and C are
the most powerful filter elements, and are the most cost-effective, as well as small
sized.
NOISE REDUCTION TOOLS
n
Capacitors
n
Inductors
7
n
Ferrites
n
Resistors
n
Linear Post Regulation
n
PHYSICAL SEPARATION FROM SENSITIVE
ANALOG CIRCUITS !!
a
7.31
Capacitors are probably the single most important filter component for switchers.
There are many different types of capacitors, and an understanding of their
individual characteristics is absolutely mandatory to the design of effective practical
supply filters. There are generally three classes of capacitors useful in 10kHz100MHz filters, broadly distinguished as the generic dielectric types; electrolytic,
film, and ceramic. These can in turn can be further sub-divided. A thumbnail sketch
of capacitor characteristics is shown in the chart of Figure 7.32.
CAPACITOR SELECTION
Aluminum
Electrolytic
Aluminum
Electrolytic
(General
Purpose)
(Switching
Size
100 µF (1)
120 µF (1)
100 µF (1)
1 µF
0.1 µF
Rated
Voltage
25 V
25 V
20 V
400 V
50 V
ESR
0.6 Ω @
0.18 Ω @
0.12 Ω @
0.11 Ω @
0.12 Ω @
100 kHz
100 kHz
100 kHz
1 MHz
1 MHz
≅ 100 kHz
≅ 500 kHz
≅ 1 MHz
≅ 10 MHz
≅ 1 GHz
Operating
Frequenc
y (2)
Tantalum
Electrolytic
Polyester
(Stacked
Ceramic
(Multilayer)
Film)
Type)
(1) Types shown in Figure 7.33 data
(2) Upper frequency limit is strongly size and package dependent
a
7.32
With any dielectric, a major potential filter loss element is ESR (equivalent series
resistance), the net parasitic resistance of the capacitor. ESR provides an ultimate
limit to filter performance, and requires more than casual consideration, because it
8
can vary both with frequency and temperature in some types. Another capacitor loss
element is ESL (equivalent series inductance). ESL determines the frequency where
the net impedance characteristic switches from capacitive to inductive. This varies
from as low as 10kHz in some electrolytics to as high as 100MHz or more in chip
ceramic types. Both ESR and ESL are minimized when a leadless package is used.
All capacitor types mentioned are available in surface mount packages, preferable
for high speed uses.
The electrolytic family provides an excellent, cost-effective low-frequency filter
component, because of the wide range of values, a high capacitance-to-volume ratio,
and a broad range of working voltages. It includes general purpose aluminum
electrolytic types, available in working voltages from below 10V up to about 500V,
and in size from 1 to several thousand µF (with proportional case sizes). All
electrolytic capacitors are polarized, and thus cannot withstand more than a volt or
so of reverse bias without damage. They also have relatively high leakage currents
(up to tens of µA, and strongly dependent upon design specifics).
A subset of the general electrolytic family includes tantalum types, generally limited
to voltages of 100V or less, with capacitance of 500µF or less[Reference 3]. In a given
size, tantalums exhibit a higher capacitance-to-volume ratios than do general
purpose electrolytics, and have both a higher frequency range and lower ESR. They
are generally more expensive than standard electrolytics, and must be carefully
applied with respect to surge and ripple currents.
A subset of aluminum electrolytic capacitors is the switching type, designed for
handling high pulse currents at frequencies up to several hundred kHz with low
losses [Reference 4]. This capacitor type competes directly with tantalums in high
frequency filtering applications, with the advantage of a broader range of values.
A more specialized high performance aluminum electrolytic capacitor type uses an
organic semiconductor electrolyte [Reference 5]. The OS-CON capacitors feature
appreciably lower ESR and higher frequency range than do other electrolytic types,
with an additional feature of low low-temperature ESR degradation.
Film capacitors are available in very broad value ranges and an array of dielectrics,
including polyester, polycarbonate, polypropylene, and polystyrene. Because of the
low dielectric constant of these films, their volumetric efficiency is quite low, and a
10µF/50V polyester capacitor (for example) is actually a handful. Metalized (as
opposed to foil) electrodes does help to reduce size, but even the highest dielectric
constant units among film types (polyester, polycarbonate) are still larger than any
electrolytic, even using the thinnest films with the lowest voltage ratings (50V).
Where film types excel is in their low dielectric losses, a factor which may not
necessarily be a practical advantage for filtering switchers. For example, ESR in
film capacitors can be as low as 10mΩ or less, and the behavior of films generally is
very high in terms of Q. In fact, this can cause problems of spurious resonance in
filters, requiring damping components.
Typically using a wound layer-type construction, film capacitors can be inductive,
which can limit their effectiveness for high frequency filtering. Obviously, only noninductively made film caps are useful for switching regulator filters. One specific
style which is non-inductive is the stacked-film type, where the capacitor plates are
cut as small overlapping linear sheet sections from a much larger wound drum of
9
dielectric/plate material. This technique offers the low inductance attractiveness of a
plate sheet style capacitor with conventional leads [see References 4, 5, 6].
Obviously, minimal lead length should be used for best high frequency effectiveness.
Very high current polycarbonate film types are also available, specifically designed
for switching power supplies, with a variety of low inductance terminations to
minimize ESL [Reference 7].
Dependent upon their electrical and physical size, film capacitors can be useful at
frequencies to well above 10MHz. At the highest frequencies, only stacked film types
should be considered. Some manufacturers are now supplying film types in leadless
surface mount packages, which eliminates the lead length inductance.
Ceramic is often the capacitor material of choice above a few MHz, due to its
compact size, low loss, and availability up to several µF in the high-K dielectric
formulations (X7R and Z5U), at voltage ratings up to 200V [see ceramic families of
Reference 3]. NP0 (also called COG) types use a lower dielectric constant
formulation, and have nominally zero TC, plus a low voltage coefficient (unlike the
less stable high-K types). NP0 types are limited to values of 0.1µF or less, with
0.01µF representing a more practical upper limit.
Multilayer ceramic “chip caps” are very popular for bypassing/ filtering at 10MHz or
more, simply because their very low inductance design allows near optimum RF
bypassing. For smaller values, ceramic chip caps have an operating frequency range
to 1GHz. For high frequency applications, a useful selection can be ensured by
selecting a value which has a self-resonant frequency above the highest frequency of
interest.
All capacitors have some finite ESR. In some cases, the ESR may actually be helpful
in reducing resonance peaks in filters, by supplying “free” damping. For example, in
most electrolytic types, a nominally flat broad series resonance region can be noted
in an impedance vs. frequency plot. This occurs where |Z| falls to a minimum level,
nominally equal to the capacitor’s ESR at that frequency. This low Q resonance can
generally be noted to cover a relatively wide frequency range of several octaves.
Contrasted to the very high Q sharp resonances of film and ceramic caps, the low Q
behavior of electrolytics can be useful in controlling resonant peaks.
In most electrolytic capacitors, ESR degrades noticeably at low temperature, by as
much as a factor of 4-6 times at –55°C vs. the room temperature value. For circuits
where ESR is critical to performance, this can lead to problems. Some specific
electrolytic types do address this problem, for example within the HFQ switching
types, the –10°C ESR at 100kHz is no more than 2× that at room temperature. The
OSCON electrolytics have a ESR vs. temperature characteristic which is relatively
flat.
Figure 7.33 illustrates the high frequency impedance characteristics of a number of
electrolytic capacitor types, using nominal 100µF/20V samples. In these plots, the
impedance, |Z|, vs. frequency over the 20Hz-200kHz range is displayed using a
high resolution 4-terminal setup [Reference 8]. Shown in this display are
performance samples for a 100µF/25V general purpose aluminum unit (top curve @
right), a 120µF/25V HFQ unit (next curve down @ right), a 100µF/20V tantalum
bead type (next curve down @ right), and a 100µF/20V OS-CON unit (lowest curve @
right). While the HFQ and tantalum samples are close in 100kHz impedance, the
10
general purpose unit is about 4 times worse. The OS-CON unit is nearly an order of
magnitude lower in 100kHz impedance than the tantalum and switching electrolytic
types.
IMPEDANCE Z(Ω ) VS. FREQUENCY FOR 100 µ F
ELECTROLYTIC CAPACITORS (AC CURRENT = 50mA RMS)
100
10
Z(Ω)
1
GEN. PURPOSE AL
100µF, 25V
0.1
"HFQ" 120µ F, 25V
TANTALUM BEAD
100µ F, 20V
10m
OS-CON AL
100µF, 20V
1m
20
100
1k
10k
100k
200k
FREQUENCY (Hz)
a
7.33
As noted, all real capacitors have parasitic elements which limit their performance.
The equivalent electrical network representing a real capacitor models both ESR
and ESL as well as the basic capacitance, plus some shunt resistance. In such a
practical capacitor, at low frequencies the net impedance is almost purely capacitive
(noted in Figure 7.33 by the 100Hz impedance). At intermediate frequencies, the net
impedance is determined by ESR, for example about 0.12Ω to 0.4Ω at 125kHz, for
several types. Above about 1MHz these capacitor types become inductive, with
impedance dominated by the effect of ESL (not shown). All electrolytics will display
impedance curves similar in general shape. The minimum impedance will vary with
the ESR, and the inductive region will vary with ESL (which in turn is strongly
effected by package style).
Regarding inductors, Ferrites (non-conductive ceramics manufactured from the
oxides of nickel, zinc, manganese, or other compounds) are extremely useful in
power supply filters [Reference 9]. At low frequencies (<100kHz), ferrites are
inductive; thus they are useful in low-pass LC filters. Above 100kHz, ferrites become
resistive, an important characteristic in high-frequency filter designs. Ferrite
impedance is a function of material, operating frequency range, DC bias current,
number of turns, size, shape, and temperature. Figure 7.34 summarize a number
ferrite characteristics.
CHARACTERISTICS OF FERRITES
n
Good for frequencies above 25kHz
11
n
Many sizes and shapes available including
leaded "resistor style"
n
Ferrite impedance at high frequencies is
primarily resistive -- Ideal for HF filtering
n
Low DC loss: Resistance of wire passing through
ferrite is very low
n
High saturation current
n
Low cost
a
7.34
Several ferrite manufacturers offer a wide selection of ferrite materials from which
to choose, as well as a variety of packaging styles for the finished network (see
References 10 and 11). A simple form is the bead of ferrite material, a cylinder of the
ferrite which is simply slipped over the power supply lead to the decoupled stage.
Alternately, the leaded ferrite bead is the same bead, pre-mounted on a length of
wire and used as a component (see Reference 11). More complex beads offer multiple
holes through the cylinder for increased decoupling, plus other variations. Surface
mount beads are also available.
PSpice ferrite models for Fair-Rite materials are available, and allow ferrite
impedance to be estimated [see Reference 12]. These models have been designed to
match measured impedances rather than theoretical impedances.
A ferrite’s impedance is dependent upon a number of inter-dependent variables, and
is difficult to quantify analytically, thus selecting the proper ferrite is not
straightforward. However, knowing the following system characteristics will make
selection easier. First, determine the frequency range of the noise to be filtered.
Second, the expected temperature range of the filter should be known, as ferrite
impedance varies with temperature. Third, the DC current flowing through the
ferrite must be known, to ensure that the ferrite does not saturate. Although models
and other analytical tools may prove useful, the general guidelines given above,
coupled with some experimentation with the actual filter connected to the supply
output under system load conditions, should lead to a proper ferrite selection.
CHOOSING THE RIGHT FERRITE DEPENDS ON
n
Source of Interference
n
Interference Frequency Range
n
Impedance Required at Interference Frequency
n
Environmental Conditions:
Temperature, AC and DC Field Strength,
12
Size / Space Available
n
Don't fail to Test the Design ------EXPERIMENT! EXPERIMENT!
a
7.35
Using proper component selection, low and high frequency band filters can be
designed to smooth a noisy switcher’s DC output so as to produce an analog ready
5V supply. It is most practical to do this over two (and sometimes more) stages, each
stage optimized for a range of frequencies. A basic stage can be used to carry all of
the DC load current, and filter noise by 60dB or more up to a 1-10MHz range. This
larger filter is used as a card entry filter providing broadband filtering for all power
entering a PC card. Smaller, more simple local filter stages are also used to provide
higher frequency decoupling right at the power pins of individual stages.
Figure 7.36 illustrates a card entry filter suitable for use with switching supplies.
With a low rolloff point of 1.5kHz and mV level DC errors, it is effective for a wide
variety of filter applications just as shown. This filter is a single stage LC low-pass
filter covering the 1kHz to 1MHz range, using carefully chosen parts. Because of
component losses, it begins to lose effectiveness above a few MHz, but is still able to
achieve an attenuation approaching 60dB at 1MHz.
"CARD-ENTRY" SWITCHING SUPPLY FILTER
+
L1
+
100µH
5V INPUT
FROM NOISY
SWITCHING
SUPPLY OR DC
TO DC
CONVERTER
C1
100µF,20V
TANTALUM
+
C2
1µF
CERAMIC
OUTPUT TO
300mA LOAD
ANALOG STAGE
R1
1Ω
-
-
a
7.36
The key to low DC losses is the use of input choke, L1, a ferrite-core unit selected for
a low DC resistance (DCR) of <0.25Ω at the 100µH inductance (either an axial lead
13
type 5250 or a radial style 6000-101K choke should give comparable results)
[Reference 13]. These chokes have low inductance shift with a 300mA load current,
and the low DCR allows the 300mA to be passed with no more than 75mV of DC
error. Alternately, resistive filtering might be used in place of L1, but a basic
tradeoff here is that load current capacity will be compromised for comparable DC
errors. C1, a 100µF/20V tantalum type, provides the bulk of the capacitive filtering,
shunted by a 1µF multilayer ceramic.
Figure 7.37 shows the frequency response of this filter in terms of SPICE simulation
and lab measurements, with good agreement between the simulation and the
measurements below 1MHz.
OUTPUT RESPONSE OF "CARD-ENTRY" FILTER
LAB VS. SIMULATION
0
X
X
X
X
X
-20
X
Z(Ω)
X
= SPICE SIMULATION
X = LAB RESULTS
-40
X
X
X
-60
X
X
X
-80
-100
10
a
100
1.0k
10k
100k
1.0M
10M
100M
FREQUENCY (Hz)
7.37
This type of filter does have some potential pitfalls, and one of them is the control of
resonances. If the LCR circuit formed does not have sufficiently high resistance at
the resonant frequency, amplitude peaking will result. This peaking can be
minimized with resistance at two locations: in series with L1, or in series with
C1+C2. Obviously, limited resistance is usable in series with L1, as this increases
the DC errors.
In the filter, R1 is a damping resistor, used to control resonant peaks, and it should
not be eliminated. A 1Ω value provides a slightly underdamped response, with
peaking on the order of 1dB. Alternately, 1.5Ω can be used for less peaking, with a
tradeoff of less attenuation below 1MHz. Note that for wide temperature range
applications, all temperature sensitive filter components will need consideration.
14
A local high frequency filter useful with the card entry filter is shown in Figure 7.38.
This simple filter can be considered an option, one which is exercised dependent
upon the high frequency characteristics of the associated IC and the relative
attenuation desired. It uses Z1, a leaded ferrite bead such as the Panasonic
EXCELSA39, providing a resistance of more than 80Ω at 10MHz, increasing to over
100Ω at 100MHz. The ferrite bead is best used with a local high frequency
decoupling cap right at the IC power pins, such as a 0.1µF ceramic unit shown.
HIGH FREQUENCY LOCALIZED DECOUPLING
+5V
FROM
CARD-ENTRY
FILTER
ANALOG
IC
TO ADDITIONAL
STAGES
Z1
LEADED FERRITE BEAD
PANASONIC EXCELSA39
(OPTIONAL)
U1
U1
0.1µF
CERAMIC
a
7.38
Both the card entry filter and the local high frequency decoupling filters are
designed to filter differential-mode noise only, and use common, off the shelf
components [Reference 14].
The following list summarizes the switching power supply filter layout/construction
guidelines which will help ensure that the filter does the best possible job:
(1) Pick the highest electrical value and voltage rating for filter capacitors which is
consistent with budget and space limits. This minimizes ESR, and maximizes filter
performance. Pick chokes for low ∆L at the rated DC current, as well as low DCR.
(2) Use short and wide PCB tracks to decrease voltage drops and minimize
inductance. Make track widths at least 200 mils for every inch of track length for
lowest DCR, and use 1 oz or 2 oz copper PCB traces to further reduce IR drops and
inductance.
(3) Use short leads or better yet, leadless components, to minimize lead inductance.
This minimizes the tendency to add excessive ESL and/or ESR. Surface mount
packages are preferred.
15
(4) Use a large-area ground plane for minimum impedance.
(5) Know what your components do over frequency, current and temperature
variations! Make use of vendor component models for the simulation of prototype
designs, and make sure that lab measurements correspond reasonably with the
simulation. While simulation is not absolutely necessary, it does instill confidence in
a design when correlation is achieved(see Reference 15).
The discussion above assumes that the incoming AC power is relatively clean, an
assumption not always valid. The AC power line can also be an EMI entry/exit path!
To remove this noise path and reduce emissions caused by the switching power
supply or other circuits, a power line filter is required.
It is important to remember that AC line power can potentially be lethal! Do not
experiment without proper equipment and training! All components used in power
line filters should be UL approved, and the best way to provide this is to specify a
packaged UL approved filter. It should be installed in such a manner that it is the
first thing the AC line sees upon entering the equipment (see Figure 7.39). Standard
three wire IEC style line cords are designed to mate with three terminal male
connectors integral to many line filters. This is the best way to achieve this
function, as it automatically grounds the third wire to the shell of the filter and
equipment chassis via a low inductance path.
POWER LINE FILTERING IS ALSO IMPORTANT
POWER
LINE
POWER
LINE
FILTER
SWITCHING
POWER
SUPPLY
SWITCHING
POWER
SUPPLY
FILTER
Power Line Filter Blocks EMI from Entering
or Exiting Box Via Power Lines
a
7.39
Commercial power line filters can be quite effective in reducing AC power-line noise.
This noise generally has both common-mode and differential-mode components.
Common-mode noise is noise that is found on any two of the three power connections
16
(black, white, or green) with the same amplitude and polarity. In contrast,
differential-mode noise is noise found only between two lines. By design, most
commercially available filters address both noise modes (see Reference 16).
17
REFERENCES: NOISE REDUCTION AND FILTERING
1.
EMC Design Workshop Notes, Kimmel-Gerke Associates, Ltd.,
St. Paul, MN. 55108, (612) 330-3728.
2.
Walt Jung, Dick Marsh, Picking Capacitors, Parts 1 & 2, Audio,
February, March, 1980.
3.
Tantalum Electrolytic and Ceramic Capacitor Families, Kemet
Electronics, Box 5928, Greenville, SC, 29606, (803) 963-6300.
4.
Type HFQ Aluminum Electrolytic Capacitor and type V Stacked
Polyester Film Capacitor, Panasonic, 2 Panasonic Way, Secaucus,
NJ, 07094, (201) 348-7000.
5.
OS-CON Aluminum Electrolytic Capacitor 93/94 Technical Book,
Sanyo, 3333 Sanyo Road, Forrest City, AK, 72335, (501) 633-6634.
6.
Ian Clelland, Metalized Polyester Film Capacitor Fills High Frequency
Switcher Needs, PCIM, June 1992.
7.
Type 5MC Metallized Polycarbonate Capacitor, Electronic Concepts, Inc.,
Box 1278, Eatontown, NJ, 07724, (908) 542-7880.
8.
Walt Jung, Regulators for High-Performance Audio, Parts 1 and 2,
The Audio Amateur, issues 1 and 2, 1995.
9.
Henry Ott, Noise Reduction Techniques in Electronic Systems,
2d Ed., 1988, Wiley.
10.
Fair-Rite Linear Ferrites Catalog, Fair-Rite Products, Box J, Wallkill,
NY, 12886, (914) 895-2055.
11.
Type EXCEL leaded ferrite bead EMI filter, and type EXC L leadless
ferrite bead, Panasonic, 2 Panasonic Way, Secaucus, NJ, 07094,
(201) 348-7000.
12.
Steve Hageman, Use Ferrite Bead Models to Analyze EMI Suppression,
The Design Center Source, MicroSim Newsletter, January, 1995.
13.
Type 5250 and 6000-101K chokes, J. W. Miller, 306 E. Alondra Blvd.,
Gardena, CA, 90247, (310) 515-1720.
14.
DIGI-KEY, PO Box 677, Thief River Falls, MN, 56701-0677,
(800) 344-4539.
15.
Tantalum Electrolytic Capacitor SPICE Models, Kemet Electronics,
Box 5928, Greenville, SC, 29606, (803) 963-6300.
16.
Eichhoff Electronics, Inc., 205 Hallene Road, Warwick, RI., 02886,
(401) 738-1440.
18
POWER SUPPLY REGULATION/CONDITIONING
Walt Jung
Many analog circuits require stable regulated voltages relatively close in potential
to an unregulated source. An example would be a linear post regulator for a
switching power supply, where voltage loss (dropout) is critical. This low dropout
type of regulator is readily implemented with a rail-rail output op amp. The wide
output swing and low saturation voltage enables outputs to come within a fraction of
a volt of the source for medium current (<30mA) loads, such as reference
applications. For higher output currents, the rail-rail voltage swing feature allows
direct drive to low saturation voltage pass devices, such as power PNPs or P-channel
MOSFETs. Op amps working from 3V up with the rail-rail features are most
suitable here, providing power economy and maximum flexibility.
LOW DROPOUT REFERENCES
Basic references
Among the many problems in making stable DC voltage references work from 5V
and lower supplies are quiescent power consumption, overall efficiency, the ability to
operate down to 3V, low input/output (dropout) capability, and minimum noise
output. Because low voltage supplies can't support zeners of ≅6V, low voltage references must necessarily be bandgap based-- a basic ≅1.2V potential. With low voltage
systems, power conservation can be a critical issue with references, as can output
DC precision.
For many applications, simple one-package fixed (or variable) voltage references
with minimal external circuitry and high accuracy are attractive. Two unique
features of the three terminal REF19X bandgap reference family are low power, and
shutdown capability. The series allows fixed outputs from 2.048-5V to be controlled
between ON and OFF, via a TTL/CMOS power control input. It provides precision
reference quality for those popular voltages shown in Figure 7.40.
19
30mA REFERENCE FAMILY WITH OPTIONAL SHUTDOWN
VS
VS > VOUT + 0.5V
TO +15V
VC
POWER
CONTROL
TTL/CMOS
LEVELS
6
2
3
+
U1
REF19X
+
4
C1
1µF
(TANTALUM)
VOUT
C2
10µF
HIGH (OR OPEN) =
ON
LOW = OFF
U1
VOUT (V)
REF191
REF192
REF193
REF196
REF198
REF194
REF195
2.048
2.5
3.0
3.3
4.096
4.5
5.0
a
7.40
The REF19X family can be used as a simple three terminal fixed reference as per
the table by tying pins 2 and 3 together, or as an ON/OFF controlled device, by
programming pin 3 as noted. In addition to the shutdown capacity, the
distinguishing functional features are a low dropout of 0.5V at 10mA, and a low
current drain for both quiescent and shutdown states, 45 and 15µA (max.),
respectively. For example, working from inputs in the range of 6.3 to 15V, a REF195
used as shown drives 5V loads at up to 30mA, with grade dependent tolerances of ±2
to ±5mV, and max TCs of 5 to 25ppm/°C. Other devices in the series provide
comparable accuracy specifications, and all have low dropout features.
To maximize DC accuracy in this circuit, the output of U1 should be connected
directly to the load with short heavy traces, to minimize IR drops. The common
terminal (pin 4) is less critical due to lower current in this leg.
Scaled References
Another approach, one with the advantage of voltage flexibility, is to buffer/scale a
low voltage reference diode. With this approach, one difficulty is getting an amplifier
to work well at 3V. A workhorse solution is the low power reference and scaling
buffer shown in Figure 7.41. Here a low current 1.2V, two-terminal reference diode
is used for D1, either the 1.235V AD589 or the 1.225V AD1580. Resistor R1 sets the
diode current, chosen for 50µA at a minimum supply of 2.7V. Obviously, loading on
the unbuffered diode must be minimized at the VREF node.
20
RAIL-TO-RAIL OUTPUT OP AMPS ALLOW GREATEST
FLEXIBILITY IN LOW DROPOUT REGULATORS
+3V OR MORE
C1
R1
27.4kΩ
Ω
0.1µ
µF
+
VOUT = VREF
OR
VOUT = VREF× (1 + R2/R3)
U1
D1
AD589
+1.235V
AD1580
+1.225V
U1: SEE TEXT
R2
R3
VREF
(UNBUFFERED)
a
7.41
Amplifier U1 both buffers and optionally scales up the nominal 1.2V reference,
allowing much higher source/sink currents. A higher op amp quiescent current is
expended in doing this, but this is a basic tradeoff of the approach. Quiescent
current is amplifier dependent, ranging from 45µA/channel with the OP196/296/496
series to 1000-2000µA/channel with the OP284 and OP279. The former series is
most useful for very light loads (<2mA), while the latter series provide device
dependent outputs up to 50mA. Various devices can be used in the circuit as shown,
and their key specs are summarized in Figure 7.42.
OP AMPS USEFUL IN LOW VOLTAGE RAIL-RAIL REFERENCES
AND REGULATORS
Device*
Iq/channel
Vsat(+),
Vsat(-), V
Isc, mA
mA
V(min @ mA)
(max @ mA)
(min)
OP193/293/493
0.017
4.20 @ 1
0.280 @ 1
(typ)
±8
OP196/296/496
0.045
4.30 @ 1
0.430 @ 1
±4
OP295/495
0.150 (max)
4.50 @ 1
0.110 @ 1
± 11
OP191/291/491
0.300
4.80 @ 2.5
0.075 @ 2.5
± 8.75
AD820/822/824
0.620
4.89 @ 2
0.055 @ 2
± 15
OP184/284/484
1.250 (max)
4.85 @ 2.5
0.125 @ 2.5
± 7.5
21
OP279
2.000
4.80 @ 10
± 45
0.075 @ 10
*Typical device specifications @ Vs = +5V, TA = 25°°C, unless otherwise noted.
a
7.42
In Figure 7.41, without gain scaling resistors R2-R3, VOUT is simply equal to VREF.
With the scaling resistors, VOUT can be set anywhere between VREF and the
positive rail, due to the op amp’s rail-rail output swing. Also, this buffered reference
is inherently low dropout, allowing a +4.5V reference output on a +5V supply, for
example. The general expression for VOUT is shown in the figure, where VREF is
the reference voltage.
Amplifier standby current can be further reduced below 20µA, if an amplifier from
the OP193/293/493 series is used. This will be at the expense of current drive and
positive rail saturation, but does provide the lowest possible quiescent current if
necessary. All devices in Figure 7.42 operate from voltages down to 3V (except the
OP279, which operates at 5V).
Low Dropout Regulators
By adding a boost transistor to the basic rail-rail output low dropout reference of
Figure 7.41, output currents of 100mA or more are possible, still retaining features
of low standby current and low dropout voltage. Figure 7.43 shows a low dropout
regulator with 800µA standby current, suitable for a variety of outputs at current
levels of 100mA.
100mA LOW NOISE, LOW DROPOUT REGULATOR
Q1
MJE170
+VIN
VOUT
3-6V (TABLE)
OUTPUT TABLE
R4
39.2kΩ
Ω
C1
100µ
µF/25V
(LOW ESR)
U1
6 AD820
4
7
+
3
2
+
R5
100kΩ
Ω
D1
AD589
1.235V
R3
2kΩ
Ω
C5
0.1µ
µF
C3
1µ
µF
FILM
R1
(TABLE)
VOUT
R1
VIN(min)
6V
5V
4V
3.3V
3.0V
383kΩ
Ω
301kΩ
Ω
226kΩ
Ω
169kΩ
Ω
143kΩ
Ω
6.2V
5.2V
4.2V
3.5V
3.2V
C4
0.01µ
µF
+
C2
100µ
µF/25V
(LOW ESR)
R2
100kΩ
Ω
COMMON
a
7.43
22
The 100mA output is achieved with a controlled gain bipolar power transistor for
pass device Q1, an MJE170. Maximum output current control is provided by
limiting base drive to Q1 via series resistor R3. This limits the base current to about
2mA, so the max HFE of Q1 then allows no more than 500mA. This limits Q1’s short
circuit power dissipation to safe levels.
Overall, the circuit operates as a follower with gain, as was true in the case of
Figure 7.41, so VOUT has a similar output expression. The circuit is adapted for
different voltages simply by programming R1 via the table. Dropout with a 100mA
load is about 200mV, thus a 5V output is maintained for inputs above 5.2V (see
table), and VOUT levels down to 3V are possible. Step load response of this circuit is
quite good, and transient error is only a few mVp-p for a 30-100mA load change.
This is achieved with low ESR switching type capacitors at C1-C2, but the circuit
also works with conventional electrolytics (with higher transient errors).
If desired, lowest output noise with the AD820 is reached by including the optional
reference noise filter, R5-C3. Lower current op amps can also be used for lower
standby current, but with larger transient errors due to reduced bandwidth.
While the 30mA rated output current of the REF19X series is higher than most
reference ICs, it can be boosted to much higher levels if desired, with the addition of
a PNP transistor, as shown in Figure 7.44. This circuit uses full time current
limiting for protection of pass transistor shorts.
150 mA BOOSTED OUTPUT REGULATOR/REFERENCE
WITH CURRENT LIMITING
Q2
TIP32A
R4
2Ω
+VS = 6 TO 9V
(SEE TEXT)
(SEE TEXT)
Q2
2N3906
C2
100µF/25V
R2
1.5k
+
D1
VC
OUTPUT TABLE
R1
1k
(SEE TEXT
ON SLEEP)
4
R3
1.82k
VS
COMMON
C3
0.1µF
VOUT(V)
REF192
REF193
REF196
REF194
REF195
2.5
3.0
3.3
4.5
5.0
F
S
C1
10µF/25V
(TANTALUM)
+
1N4148
2
U1
3 REF196 6
(SEE TABLE)
U
1
+VOUT
3.3V
@ 150mA
RL
S
F
VOUT
COMMON
7.44
a
In this circuit the supply current of reference U1 flows in R1-R2, developing a base
drive for pass device Q1, whose collector provides the bulk of the output current.
With a typical gain of 100 in Q1 for 100-200mA loads, U1 is never required to
furnish more than a few mA, and this factor minimizes temperature related drift.
23
Short circuit protection is provided by Q2, which clamps drive to Q1 at about 300mA
of load current. With separation of control/power functions, DC stability is optimum,
allowing best advantage of premium grade REF19X devices for U1. Of course, load
management should still be exercised. A short, heavy, low resistance conductor
should be used from U1-6 to the VOUT sense point “S”, where the collector of Q1
connects to the load.
Because of the current limiting, dropout voltage is raised about 1.1V over that of the
REF19X devices. However, overall dropout typically is still low enough to allow
operation of a 5 to 3.3V regulator/reference using the 3.3V REF-196 for U1, with a
Vs of 4.5V and a load current of 150mA.
The heat sink requirements of Q1 depend upon the maximum power. With Vs = 5V
and a 300mA current limit, the worst case dissipation of Q1 is 1.5W, less than the
TO-220 package 2W limit. If TO-39 or TO-5 packaged devices such as the 2N4033
are used, the current limit should be reduced to keep maximum dissipation below
the package rating, by raising R4. A tantalum output capacitor is used at C1 for its
low ESR, and the higher value is required for stability. Capacitor C2 provides input
bypassing, and can be an ordinary electrolytic.
Shutdown control of the booster stage is shown as an option, and when used, some
cautions are in order. To enable shutdown control, the connection to U1-2 and U1-3
is broken at “X”, and diode D1 allows a CMOS control source to drive U1-3 for
ON/OFF control. Startup from shutdown is not as clean under heavy load as it is
with the basic REF19X series stand-alone, and can require several milliseconds
under load. Nevertheless, it is still effective, and can fully control 150mA loads.
When shutdown control is used, heavy capacitive loads should be minimized.
Dedicated low dropout linear IC regulators offer all the virtues of the discrete
approaches, but in a easier-to-use compact format. The ADP3367 is such a device,
providing either a fixed output of 5V ±2%, or adjustable outputs over a range of 1.3
to 16.5V, with current outputs up to 300mA. Using a CMOS architecture with a
PNP pass transistor, it has a quiescent current of 25µA (max., unloaded), and a
dropout voltage of 175mV (max.) with a 100mA output.
Figure 7.45 shows the basic hookup for the ADP3367, which uses the "thermal
coastline" 8 pin SOIC package, which is designed for power dissipation up to
960mW. For fixed 5V outputs, R1 and R2 aren’t used, and the SET pin is grounded
as shown. With the SHDN pin also grounded, this simple hookup provides a
constant 5V at VOUT, with the low dropout features mentioned.
24
300mA LOW DROPOUT FIXED/VARIABLE
REGULATOR WITH OPTIONAL SHUTDOWN
VIN
IN
+5.15V
TO
+16.5V
VOUT
+5V
OUT
R2
162k Ω
ADP3367
+
C1
0.1µ
µF
SHDN
*
C2
10 µF
SET
GND
R1
100kΩ
Ω
VSHDN
* WITH R1, R2:
R2
VOUT = V REF 1 + R1
WHERE VREF = +1.255V,
(
)
VOUT = +3.3V (VALUES SHOWN)
a
7.45
The ADP3367’s useful output current capacity will be dependent upon the VIN-VOUT
differential, such that the resulting power it dissipates is contained to 960 mW or
less. For example, at low input-output differences of 2.5V, up to 300mA is available.
For higher input-output differences, the allowable current is reduced according to
the curves shown in Figure 7.46. The upper shaded curve corresponds to the output
current which is consistent with the ADP3367’s package limitations. Note that the
allowable output current is appreciably higher than that of a standard SO package,
shown in the lower shaded curve.
25
ADP3367 LOAD CURRENT VS. INPUT - OUTPUT VOLTAGE
400
LOAD CURRENT - mA
TA = +50°C
300
GUARANTEED 300mA
200
ADP3367
DISSIPATION LIMIT
100
STANDARD
SO PACKAGE
DISSIPATION LIMIT
0
0
5
10
15
VIN-VOUT - V
a
7.46
The ADP3367 can be placed in a shutdown mode, which reduces the output voltage
to zero and drops the standby current to less than 1µA. When implemented,
shutdown is accomplished by applying a control voltage of more than 1.5V to VSHDN.
Otherwise, this pin should be tied to ground as shown. The SET pin has a dual
function, and can be used either to select an internal divider (which provides the
fixed 5V output), or it can be used with an external divider, R1-R2. When the SET
pin is grounded, the internal divider is active, and the 5 V output results. When the
SET pin is used with the external divider, VOUT is programmed as:
R2 

VOUT = V REF ∗ 1 +


R1 
where VREF is 1.255V, the internal reference voltage of the ADP3367. The divider’s
absolute resistance values are not critical, since the input current at the SET pin is
low, typically 10pA. This allows resistances of 100k - 1meg, consistent with the
overall low standby power objectives. The example 1% values shown provide a 3.3V
output. They can be further increased, if it is desired to lower standby current
consumption below the ≅12µA resulting with the values shown.
C2, the output capacitor, is a 10µF type, and is required for regulator stability.
Larger sizes are permissible, and will help improve transient response. An input
bypass is also recommended, C1.
To achieve the full power capability inherent to the design, the ADP3367 should be
mounted on a PCB in such as way that internally-generated heat can flow outward
easily from the die to the PCB. Large area PCB copper traces should be used
beneath and around the IC, and mounting should be such that the part is exposed to
unrestricted air flow [see Reference 5].
26
27
REFERENCES: POWER SUPPLY
REGULATION/CONDITIONING
1.
Walt Jung, Build an Ultra-Low-Noise Voltage Reference,
Electronic Design Analog Applications Issue, June 24, 1993.
2.
Walt Jung, Getting the Most from IC Voltage References, Analog
Dialogue 28-1, 1994.
3.
Walt Jung, The Ins and Outs of ‘Green’ Regulators/References ,
Electronic Design Analog Applications Issue, June 27, 1994.
4.
Walt Jung, Very-Low-Noise 5-V Regulator, Electronic Design,
July 25, 1994.
5.
“Power Dissipation” Discussions, ADP3367 Data Sheet, Analog Devices.
28
THERMAL MANAGEMENT
Walt Jung
For reliability reasons, modern semiconductor based systems are increasingly called
upon to observe some form of thermal management. All semiconductors have some
specified safe upper limit to junction temperature (TJ), usually on the order of 150°C
(but sometimes 175°). Like maximum power supply potentials, maximum junction
temperature is a worst case limitation which shouldn’t be exceeded. In conservative
designs, it won’t be approached by less than an ample safety margin. This is a
critical point, since the lifetime of all semiconductors is inversely related to their
operating junction temperature. The cooler semiconductors can be kept during
operation, the more closely they will approach maximum useful life.
Thermal basics
The general symbol θ is used for thermal resistance, that is:
θ = thermal resistance, in units of °C/watt (or, °C/W).
θJA and θJC are two more specific terms used in dealing with semiconductor
thermal issues, which are further explained below.
In general, a device with a thermal resistance θ equal to 100°C/W will exhibit a
temperature differential of 100°C for a power dissipation of 1W, as measured
between two reference points. Note that this is a linear relation, so a 500mW
dissipation in the same part will produce a 50°C differential, and so forth. For any
power P (in watts), calculate the effective temperature differential (∆T) in °C as:
∆T = P × θ
where θ is the total applicable thermal resistance. Figure 7.47 summarizes these
thermal relationships.
THERMAL BASICS
n
θ = Thermal Resistance (°C/W)
n
∆T = P × θ
n
θJA = Junction - to - Ambient Thermal Resistance
n
θJC = Junction - to - Case Thermal Resistance
n
θCA = Case - to - Ambient Thermal Resistance
n
θJA = θJC + θCA
n
TJ = TA + (P × θJA), P = Total Device Power Dissipation
n
TJ(Max) = 150°C
(Sometimes 175°C)
29
a
7.47
A real example illustrating this relationship is shown by Figure 7.48. These curves
indicate the maximum power dissipation vs. temperature characteristic for a device
using 8 pin DIP and SOIC packaging. For a TJ(max) of 150°C, the upper curve
shows the allowable power in a DIP package. This corresponds to a θ which can be
calculated by dividing the ∆T by P at any point. For example, 1W of power is allowed
at a TA of 60°C, so the ∆T is 150°C – 60°C = 90°C. Dividing by 1W gives this DIP
package’s θ of 90°C/W. Similarly, the SOIC package yields 160°C/W. These figures
are in fact the θJA for the AD823 op amp, but they also happen to be quite similar
to other 8 pin devices. Given such data as these curves, the θJA for a given device
can be readily determined, as above.
MAXIMUM POWER DISSIPATION VS. TEMPERATURE
FOR 8-PIN MINI-DIP AND 8-PIN SOIC PACKAGES
MAXIMUM POWER DISSIPATION - Watts
2.0
8-PIN MINI-DIP PACKAGE
TJ = +150°C
1.5
1.0
8-PIN SOIC PACKAGE
0.5
0
-50 -40 -30 -20 -10 0 10 20 30 40 50 60
AMBIENT TEMPERATURE - °C
70 80 90
7.48
As the relationship signifies, to maintain a low TJ, either θ or the power dissipated
(or both) must be kept low. A low ∆T is the key to extending semiconductor lifetimes,
as it leads to low maximum junction temperatures.
In semiconductors, one temperature reference point is always the device junction,
taken to mean the hottest spot inside the chip operating within a given package.
The other relevant reference point will be either the case of the device, or the
ambient temperature, TA, that of the surrounding air. This then leads in turn to the
above mentioned individual thermal resistances, θJA and θJC.
30
Taking the more simple case first, θJA is the thermal resistance of a given device
measured between its junction and the ambient air. This thermal resistance is most
often used with small, relatively low power ICs which do not dissipate serious
amounts of power, that is 1W or less. θJA figures typical of op amps and other small
devices are on the order of 90-100°C/W for a plastic 8 pin DIP package. It must be
understood that thermal resistances are highly package dependent, as different
materials have differing degrees of thermal conductivity. As a general rule of thumb,
thermal resistance for the conductors within packaging materials is closely
analogous to electrical resistances, that is copper is the best, followed by aluminum,
steel, and so on. Thus copper lead frame packages offer the highest performance
(lowest θ).
A summary of the thermal resistances of various IC packages is shown in Figures
7.49 and 7.50. In general, most of these packages do not lend themselves to easy
heat sink attachment (with notable exceptions, such as the older round metal can
types). Devices which are amenable to heat sink attachment will often be noted by a
θJC dramatically lower than the θJA. See for example the 15 pin SIP package (used
by the AD815).
STANDARD PACKAGE THERMAL RESISTANCES - 1
Package
ADI
designation
θJA
(°°C/W)
θJC
(°°C/W)
8 pin plastic DIP
8 pin ceramic DIP
8 pin SOIC
8 pin SOIC
8 pin metal can
N-8
D-8
R-8
R-8
H-08A (TO-99)
90
110
160
90
150
22
60
60
45
10 pin metal can
H-10A (TO-100)
150
25
AD582
12 pin metal can
H-12A (TO-8)
100
30
AD841
14 pin plastic DIP
14 pin ceramic DIP
14 pin SOIC
N-14
D-14
R-14
150
110
120
30
AD713
AD585
AD813
15 pin SIP
Y-15
41
2
AD815 Through-Hole
16 pin plastic DIP
16 pin ceramic DIP
16 pin SOIC
N-16
D-16
R-16
120
95
85
40
22
Comment
AD823
AD712
ADP3367 Thermal Coastline
OP07
AD524
AD811
a
7.49
STANDARD PACKAGE THERMAL RESISTANCES - 2
Package
ADI
designation
θJA
(°°C/W)
31
θJC
(°°C/W)
Comment
18 pin ceramic DIP
D-18
120
35
20 pin plastic DIP
20 pin ceramic DIP
20 pin SOIC
N-20
D-20
R-20
102
70
74
31
10
24
24 pin plastic DIP
24 pin ceramic DIP
N-24
D-24
105
120
35
35
28 pin plastic DIP
28 pin ceramic DIP
28 pin SOIC
N-28
D-28
R-28
74
51
71
24
8
23
AD7575
AD7547
a
7.50
θJC is the thermal resistance of a given device as measured between its junction and
the device case. This form is most often used with larger power semiconductors
which do dissipate significant amounts of power, that is typically more than 1W. The
reason for this is that a heat sink generally must be used with such devices, to
maintain a sufficiently low internal junction temperature. A heat sink is simply an
additional low thermal resistance device attached externally to a semiconductor part
to aid in heat removal. It will have some additional thermal resistance of its own,
also rated in °C/W.
Rather than just a single number, θ in this case will be composed of more than one
component, i.e., θ1, θ2, etc. Like series resistors, thermal impedances add, making a
net calculation relatively simple. For example, to compute a net θJA given a
relevant θJC, the thermal resistance of the heat sink, θCA, or case to ambient is
added to the θJC as:
θJA = θJC + θCA
and the result is the θJA for that specific circumstance.
A second form of the general overall relationship between TJ, TA, P and θ is:
TJ = TA + (P × θ)
To take a real world example, the AD815AVR power-tab packaged op amp has a
θJA of 41°C/W with no additional heat sinking (the device simply operating in still
air). Using it just as this would allow a power of:
P = (TJ – TA)/ θJA
or, (150°C – 70°C)/41°C /W, which results in an allowable power of about 2W.
32
However, such a mode of operation falls short of the device’s full power handling
capacity. The AD815AVR’s θJC is quite low at about 2°C/W, and if a heat sink of
significantly less than 38°C/W is used with it, then it can dissipate much more
power for a given junction temperature. A 20°C/W heat sink will allow almost twice
the power to be dissipated by the same device, simply because of the lower net θJA
only 22°C/W. This can be accomplished by a double-sided PCB copper plane area of
1k mm2 [see Reference 1].
To illustrate, the general relationship of the AD815AVR and PCB heat sink net θJA
is shown by Figure 7.51. In the first example cited above, full advantage of PCB
heat sink area was not taken, and as the graph shows, the net θJA can be reduced
to as low as ≅17°C/W by increasing the heat sink area further. The tradeoff is simply
one of board area, and with a 2k mm2 heat sink area, nearly 5W of power can be
handled by the same device, assuming the same ∆T and max TJ. Of course, for the
AD815 (and other devices) even more conservative operation is optionally possible
by holding to a lower maximum TJ.
Note that for the data of Figure 7.51, these data assume that the AD815AVR is
soldered directly to one of the dual copper PCB planes.
The power tab style package used with the AD815AVR can also be used with
conventional PC mounted heat sinks, with θJC of 20°C/W and less. See Reference 2.
AD815AVR AND PCB HEAT SINK θ JA VS.
PCB HEAT SINK AREA
35
30
θJA - ºC/W
AD815AVR, AY (θJC = 2°C/W)
25
20
15
10
0
0.5k
1k
1.5k
2k
COPPER HEAT SINK AREA (TOP AND BOTTOM) - mm2
2.5k
7.51
33
Calculating Power In Various Devices
In all instances of thermal calculations, a basic assumption is that the power is the
total for a given package. With many modern devices now using more than one
supply, the net total power dissipated will be the sum of all individual supply
quiescent powers, plus any load dependent power. For many low output current op
amps for example, total power will then be essentially the same as the quiescent. As
long as this is safely less than the package can support, there is little worry.
However, with some devices operable over a wide range of supply voltages, there are
instances where high supply voltages and a medium to high quiescent current plus
load current can be a problem.
The AD811 is such an example, being capable of operation from ±5V to ±15V, with a
quiescent current of about 16mA. If operated at ±15V, the quiescent dissipation is
nearly 500mW, which with a 90°C/W θJA, will push TJ to about 115°C in a 70°C
ambient, high enough for concern. If the signal voltage output for such an amplifier
doesn’t require the ±15V supplies, then reducing the supplies will lower the
quiescent power, and TJ.
To illustrate a general relationship of the power dissipated in an op amp and the
power in a load for family of supply voltages, Figure 7.52 was prepared. This is a
test simulation of a standard gain-of-2 non inverting amplifier driving a 150Ω load,
with 1kΩ gain and feedback resistors. Assuming an input voltage of 1V DC, the 2V
output across the net resistor load of 150Ω||2kΩ=140Ω will produce a power Pr of
about 29mW. The AD817 amplifier operates over a supply range of ±5V to ±15V,
which is the Vs sweep range for the test circuit. The op amp quiescent power Pq
increases to 210mW at ±15V, while the signal power Ps dissipated by the op amp
increases to 187mW at ±15V. The total power in the op amp is their sum, 397mW at
±15V. Clearly, operating relatively high current and low voltage loads from an op
amp does waste considerable power, and lower voltage supplies will be much more
efficient, where allowable.
34
AD817 OP AMP POWER DISSIPATION VS. SUPPLY VOLTAGE
400
Ps + Pq = TOTAL
OP AMP POWER
+VS
+1V
POWER
(mW)
+2V
+
AD817
300
150 Ω
-VS
Pq = QUIESCENT POWER
200
Ps = SIGNAL POWER
1k Ω
1k Ω
100
Pr = LOAD POWER
0
0
5
10
15
±VS, VOLTS
a
7.52
Where appropriate, a clip on DIP compatible heat sink such as the AAVID 580100
can be used [Reference 3]. This series has sinks compatible with ICs of 8 through 40
pin sizes, using a staggered fin design. Performance of these (and all) heat sinks is
enhanced by air movement, either through forced convection, or as a minimum, by
arranging PCB cards vertically to enhance natural convection.
A/D converters can consume considerable power, although the trend is towards
lower voltage and lower power dissipation. Like op amps, they are generally
analyzed by adding up the total power in the package, which can then be used with
the package’s θJA to compute junction temperature. In adding various power totals,
some care should be made to ascertain if any power is clock dependent. In some
CMOS based designs, there can be appreciable differences in power as a function
high/low clock speed as shown in Figure 7.53 for the AD9220 12-bit, 10MSPS ADC.
35
AD9220 12-BIT, 10MSPS CMOS ADC POWER DISSIPATION
VS. SAMPLING CLOCK FREQUENCY
300
280
POWER - mW
INPUT = 5Vp-p
260
INPUT = 2Vp-p
240
220
200
0
2
4
6
8
10
12
14
CLOCK FREQUENCY - MHz
a
7.53
For example, the AD9042 12 bit A/D consumes about 600mW total on two 5V
supplies, and its 28 pin DIP package has a θJA of 34°C/W. What will be the max TJ
for this part in a TA of 70°C? You should get a TJ of 90.4°C (∆T = 0.6W × 34°C/W =
20.4°C, so TJ for TA of 70°C = 70°C + 20.4°C). This particular part is therefore in
good shape for this TA, assuming that there are no adjacent “hot spot” sources to
increase the device’s effective TA.
Airflow Control
For large power dissipations and/or to maintain low TJ’s, forced air movement can
be used to increase air flow and aid in heat removal. In its most simple form this can
consist of a continuously or thermostatically operated fan, directed across high
temperature, high wattage dissipation devices such as CPUs, DSP chips, etc.
Quite often however, more sophisticated temperature control is necessary. Recent
temperature monitoring and control ICs such as the TMP12, an airflow temperature
sensor IC, lend themselves to such applications.
The TMP12 includes on chip two comparators, a voltage reference, a temperature
sensor and a heater. The heater is used to force a predictable internal temperature
rise, to match a power IC such as a microprocessor. The temperature sensing and
control portions of the IC can then be programmed to respond to the temperature
changes and control an external fan, so as to maintain some range of temperature.
Compared to a simple thermostat, this allows infinite resolution of user control for
control points and ON/OFF hysteresis.
The device is placed in an airstream near the power IC, such that both see the same
stream of air, and will thus have like temperature profiles, assuming proper control
of the stream. This is shown in basic form by the layout diagram of Figure 7.54.
36
SYSTEM USE OF TMP12 AIRFLOW SENSOR
PGA
PACKAGE
AIR FLOW
PGA
SOCKET
POWER I.C.
PC BOARD
TMP12
7.54
a
With the TMP12’s internal 250mW heater ON and no airflow, the TMP12 thermal
profile will look like the curve “A” of Figure 7.55, and will show a 20°C rise above
TA. When airflow is provided, this same dissipation results in a lower temperature,
“D”. In programming the device for airspeed control, the designer can set up to two
switch points, shown here symbolically by “B” and “C”, which are HIGH and LOW
setpoints, respectively. The basic idea is that when the IC substrate reaches point B
in temperature, the external fan will be turned on to create the airstream, and lower
the temperature. If the overall system setup is reasonable in terms of thermal
profiling, this small IC can thus be used to indirectly control another larger and
independent power source with regard to its temperature. Note that the dual mode
control need not necessarily be used, in all applications. An unused comparator is
simply wired high or low.
37
TMP12 TEMPERATURE RELATIONSHIPS
65
a
DIE TEMPERATURE (ºC)
60
b
55
c
50
d
45
a. TMP12 DIE TEMP NO AIR FLOW
b. HIGH SET POINT
c. LOW SET POINT
d. TMP12 DIE TEMP MAX AIR FLOW
e. SYSTEM AMBIENT TEMPERATURE
40
e
35
0
50
100
150
200
250
TMP12 PD (mW)
7.55
a
Figure 7.56 shows a circuit diagram using the TMP12 as a general purpose
controller. The device is connected to a 5V supply, which is also used to power a
control relay and the TMP12’s internal heater at pin 5. Setpoint programming of the
TMP12 is accomplished by the resistor string at pins 4 through 1, R1 - R3. These
resistors establish a current drain from the internal reference source at pin 4, which
sets up a reference current, IREF, which is set as:
IREF = (5µA/°C × THYS)+ 7µA
In this expression, THYS is the hysteresis temperature swing desired about the
setpoint, in °C, and the 7µA is recommended minimum loading of the reference. For
a 2°C hysteresis for example, IREF is 17µA; for 5°C, it would be 32µA.
Given a desired setpoint temperature in °C, the setpoint can be converted to a
corresponding voltage. Although not available externally, the internal temperature
dependent voltage of the TMP12 is scaled at 5mV/°C, and is equal to 1.49V at 25°C.
To convert a setpoint temperature to a voltage VSETPOINT,
VSETPOINT = 1.49V + [ 5mV/°C × (TSETPOINT – T25) ]
where TSETPOINT is the desired setpoint temperature, and T25 is 25°C. For a
50°C high setpoint, this works out to be VSETPOINT(HI) = 1.615V. For a lower
setpoint of 35°C, the voltage VSETPOINT(LO) would be 1.59V.
38
The divider resistors are then chosen to draw the required current IREF while
setting the two tap voltages corresponding to VSETPOINT(HI) and
VSETPOINT(LO).
RTOTAL = VREF / IREF
= 2.5V / IREF
R1 = [ VREF – VSETPOINT(HI) ] / IREF
= [ 2.5V – VSETPOINT(HI) ] / IREF
R2= [ VSETPOINT(HI) – VSETPOINT(LO) ] / IREF
R3 = VSETPOINT(LO) / IREF
In the example of the figure, the resulting standard values for R1 - R3 correspond to
the temperature/voltage setpoint examples noted above. Ideal 1% values shown give
resistor related errors of only 0.1°C from ideal. Note that this is error is independent
of the TMP12 temperature errors, which are ±2°C.
As noted above, both comparators of the device need not always be used, and in this
case the lower comparator output is not used. For a single point 50°C controller, the
35°C setpoint is superfluous. One resistor can be eliminated by making R2 + R3 a
single value of 95.3kΩ and connecting pin 3 to GND. Pin 6 should be left as a noconnect. If a greater hysteresis is desired, the resistor values will be proportionally
lowered.
It is also important to minimize potential parasitic temperature errors associated
with the TMP12. Although the open-collector outputs can sink up to 20mA, it is
advised that currents be kept low at this node, to limit any additional temperature
rise. The Q1 - Q2 transistor buffer shown in the figure raises the current drive to
100mA, allowing a 50Ω/5V coil to be driven. The relay type shown is general
purpose, and many other power interfaces are possible with the TMP12. If used as
shown, the relay contacts would be used to turn on a fan for airflow when the active
low output at pin 7 changes, indicating the upper setpoint threshold.
A basic assumption of the TMP12’s operation is that it will “mimic” another device
in temperature rise. Therefore, a practical working system must be arranged and
tested for proper airflow channeling, minimal disturbances from adjacent devices,
etc. Some experimentation should be expected before a final setup will result.
39
TMP12 50º SETPOINT CONTROLLER
+5V
TMP12
1
TEMPERATURE
SENSOR AND
VOLTAGE
REFERENCE
VREF
R1
2
µF
0.1µ
VPTAT
-
D1
IN4002
R5
Ω
390Ω
8
R4
10k Ω
7
Q2
Q1
TO FAN OR
COOLING DEVICE
SPDT RELAY
5V COIL, 50Ω
Ω MIN
OMRON G2R-14-DC5
Q1, Q2 = 2N2222
+
R2
3
R3
6
+
NC
FOR THYS = 2ºC, IREF = 17µ
µA
SETPOINT (HI) = 50ºC
SETPOINT (LO) = 35ºC (IF USED)
R1 = 52.3kΩ
Ω
R2 = 4.42kΩ
Ω OR 95.3kΩ
Ω
R3 = 90.9kΩ
Ω
-
4
5
HYSTERESIS
GENERATOR
IREF =
Ω
100Ω
VREF
R1 + R2 + R3
a
7.56
40
REFERENCES: THERMAL MANAGEMENT
1.
Power Consideration Discussions, AD815 Data Sheet, Analog Devices.
2.
Heat Sinks for Multiwatt Packages, AAVID Engineering, Inc., One Kool
Path, Laconia, NH, 03246, (603) 528-3400.
3.
General Catalog, AAVID Engineering, Inc., One Kool Path, Laconia, NH,
03246, (603) 528-3400.
41
EMI/RFI CONSIDERATIONS
Adolfo A. Garcia
Electromagnetic interference (EMI) has become a hot topic in the last few years
among circuit designers and systems engineers. Although the subject matter and
prior art have been in existence for over the last 50 years or so, the advent of
portable and high-frequency industrial and consumer electronics has provided a
comfortable standard of living for many EMI testing engineers, consultants, and
publishers. With the help of EDN Magazine and Kimmel Gerke Associates, this
section will highlight general issues of EMC (electromagnetic compatibility) to
familiarize the system/circuit designer with this subject and to illustrate proven
techniques for protection against EMI.
A PRIMER ON EMI REGULATIONS
The intent of this section is to summarize the different types of electromagnetic
compatibility (EMC) regulations imposed on equipment manufacturers, both
voluntary and mandatory. Published EMC regulations apply at this time only to
equipment and systems, and not to components. Thus, EMI hardened equipment
does not necessarily imply that each of the components used (integrated circuits,
especially) in the equipment must also be EMI hardened.
Commercial Equipment
The two driving forces behind commercial EMI regulations are the FCC (Federal
Communications Commission) in the U. S. and the VDE (Verband Deutscher
Electrotechniker) in Germany. VDE regulations are more restrictive than the FCC’s
with regard to emissions and radiation, but the European Community will be
adding immunity to RF, electrostatic discharge, and power-line disturbances to the
VDE regulations, and now requires mandatory compliance. In Japan, commercial
EMC regulations are covered under the VCCI (Voluntary Control Council for
Interference) standards and, implied by the name, are much looser than their FCC
and VDE counterparts.
All commercial EMI regulations primarily focus on radiated emissions, specifically to
protect nearby radio and television receivers, although both FCC and VDE
standards are less stringent with respect to conducted interference (by a factor of 10
over radiated levels). The FCC Part 15 and VDE 0871 regulations group commercial
equipment into two classes: Class A, for all products intended for business
environments; and Class B, for all products used in residential applications. For
example, Table 7.1 illustrates the electric-field emission limits of commercial
computer equipment for both FCC Part 15 and VDE 0871 compliance.
Radiated Emission Limits for Commercial Computer Equipment
Frequency (MHz)
Class A
42
Class B
30 - 88
88 - 216
216 - 1000
( at 3 m)
300 µV/m
500 µV/m
700 µV/m
(at 3 m)
100 µV/m
150 µV/m
200 µV/m
Reprinted from EDN Magazine (January 20, 1994), © CAHNERS PUBLISHING
COMPANY 1995, A Division of Reed Publishing USA.
Table 7.1
In addition to the already stringent VDE emission limits, the European Community
EMC standards (IEC and IEEE) now requires mandatory compliance to these
additional EMI threats: Immunity to RF fields, electrostatic discharge, and powerline disturbances. All equipment/systems marketed in Europe must exhibit an
immunity to RF field strengths of 1-10V/m (IEC standard 801-3), electrostatic
discharge (generated by human contact or through material movement) in the range
of 10-15kV (IEC standard 801-2), and power-line disturbances of 4kV EFTs
(extremely fast transients, IEC standard 801-4) and 6kV lightning surges (IEEE
standard C62.41).
Military Equipment
The defining EMC specification for military equipment is MIL-STD-461 which
applies to radiated equipment emissions and equipment susceptibility to
interference. Radiated emission limits are very typically 10 to 100 times more
stringent than the levels shown in Table 7.1. Required limits on immunity to RF
fields are typically 200 times more stringent (RF field strengths of 5-50mV/m) than
the limits for commercial equipment.
Medical Equipment
Although not yet mandatory, EMC regulations for medical equipment are presently
being defined by the FDA (Food and Drug Administration) in the USA and the
European Community. The primary focus of these EMC regulations will be on
immunity to RF fields, electrostatic discharge, and power-line disturbances, and
may very well be more stringent than the limits spelled out in MIL-STD-461. The
primary objective of the medical EMC regulations is to guarantee safety to humans.
Industrial- and Process-Control Equipment
Presently, equipment designed and marketed for industrial- and process-control
applications are not required to meet pre-existing mandatory EMC regulations. In
fact, manufacturers are exempt from complying to any standard in the USA.
However, since industrial environments are very much electrically hostile, all
equipment manufacturers will be required to comply with all European Community
EMC regulations in 1996.
Automotive Equipment
Perhaps the most difficult and hostile environment in which electrical circuits and
systems must operate is that found in the automobile. All of the key EMI threats to
43
electrical systems exist here. In addition, operating temperature extremes, moisture,
dirt, and toxic chemicals further exacerbate the problem. To complicate matters
further, standard techniques (ferrite beads, feed-through capacitors, inductors,
resistors, shielded cables, wires, and connectors) used in other systems are not
generally used in automotive applications because of the cost of the additional
components.
Presently, automotive EMC regulations, defined by the very comprehensive SAE
Standards J551 and J1113, are not yet mandatory. They are, however, very
rigorous. SAE standard J551 applies to vehicle-level EMC specifications, and
standard J1113 (functionally similar to MIL-STD-461) applies to all automotive
electronic modules. For example, the J1113 specification requires that electronic
modules cannot radiate electric fields greater than 300nV/m at a distance of 3
meters. This is roughly 1000 times more stringent than the FCC Part 15 Class A
specification. In many applications, automotive manufacturers are imposing J1113
RF field immunity limits on each of the active components used in these modules.
Thus, in the very near future, automotive manufacturers will require that IC
products comply with existing EMC standards and regulations.
EMC Regulations’ Impact on Design
In all these applications and many more, complying with mandatory EMC
regulations will require careful design of individual circuits, modules, and systems
using established techniques for cable shielding, signal and power-line filtering
against both small- and large-scale disturbances, and sound multi-layer PCB
layouts. The key to success is to incorporate sound EMC principles early in the
design phase to avoid time-consuming and expensive redesign efforts.
A DIAGNOSTIC FRAMEWORK FOR EMI/RFI PROBLEM
SOLVING
With any problem, a strategy should be developed before any effort is expended
trying to solve it. This approach is similar to the scientific method: initial circuit
misbehavior is noted, theories are postulated, experiments designed to test the
theories are conducted, and results are again noted. This process continues until all
theories have been tested and expected results achieved and recorded. With respect
to EMI, a problem solving framework has been developed. As shown in Figure 7.57,
the model suggested by Kimmel-Gerke in [Reference 1] illustrates that all three
elements (a source, a receptor or victim, and a path between the two) must exist in
order to be considered an EMI problem. The sources of electromagnetic interference
can take on many forms, and the ever-increasing number of portable
instrumentation and personal communications/computation equipment only adds
the number of possible sources and receptors.
A DIAGNOSTIC FRAMEWORK FOR EMI
Reprinted from EDN Magazine (January 20,1994), © CAHNERS PUBLISHING COMPANY
1995, A Division of Reed Publishing USA
ANY INTERFERENCE PROBLEM CAN BE BROKEN DOWN INTO:
n
The SOURCE of interference
44
n
The RECEPTOR of interference
n
The PATH coupling the source to the receptor
SOURCES
Microcontroller
u Analog
u Digital
ESD
Communications
Transmitters
Power Disturbances
Lightning
PATHS
Radiated
u EM Fields
u Crosstalk
Capacitive
Inductive
Conducted
u Signal
u Power
u Ground
a
RECEPTORS
Microcontroller
u Analog
u Digital
Communications
u Receivers
Other Electronic
Systems
7.57
Interfering signals reach the receptor by conduction (the circuit or system
interconnections) or radiation (parasitic mutual inductance and/or parasitic
capacitance). In general, if the frequencies of the interference are less than 30MHz,
the primary means by which interference is coupled is through the interconnects.
Between 30MHz and 300MHz, the primary coupling mechanism is cable radiation
and connector leakage. At frequencies greater than 300MHz, the primary
mechanism is slot and board radiation. There are many cases where the
interference is broadband, and the coupling mechanisms are combinations of the
above.
When all three elements exist together, a framework for solving any EMI problem
can be drawn from Figure 7.58. There are three types of interference with which the
circuit or system designer must contend. The first type of interference is that
generated by and emitted from an instrument; this is known as circuit/system
emission and can be either conducted or radiated. An example of this would be the
personal computer. Portable and desktop computers must pass the stringent FCC
Part 15 specifications prior to general use.
45
THREE TYPES OF INTERFERENCE
EMISSIONS - IMMUNITY - INTERNAL
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
HANDHELD
TRANSMITTER
RADIO
TRANSMITTER
LIGHTNING
RADIATED
EMISSIONS
INTERNAL
ELECTRONICS
CONDUCTED
EMISSIONS
HUMAN ESD
POWER
DISTURBANCE
a
7.58
The second type of interference is circuit or system immunity. This describes the
behavior of an instrument when it is exposed to large electromagnetic fields,
primarily electric fields with an intensity in the range of 1 to 10V/m at a distance of
3 meters. Another term for immunity is susceptibility, and it describes circuit/system
behavior against radiated or conducted interference.
The third type of interference is internal. Although not directly shown on the figure,
internal interference can be high-speed digital circuitry within the equipment which
affects sensitive analog (or other digital circuitry), or noisy power supplies which can
contaminate both analog and digital circuits. Internal interference often occurs
between digital and analog circuits, or between motors or relays and digital circuits.
In mixed signal environments, the digital portion of the system often interferes with
analog circuitry. In some systems, the internal interference reaches such high levels
that even very high-speed digital circuitry can affect other low-speed digital circuitry
as well as analog circuits.
In addition to the source-path-receptor model for analyzing EMI-related problems,
Kimmel Gerke Associates have also introduced the FAT-ID concept [Reference 1].
FAT-ID is an acronym that describes the five key elements inherent in any EMI
problem. These five key parameters are: frequency, amplitude, time, impedance, and
distance.
The frequency of the offending signal suggests its path. For example, the path of lowfrequency interference is often the circuit conductors. As the interference frequency
increases, it will take the path of least impedance, usually stray capacitance. In this
case, the coupling mechanism is radiation.
Time and frequency in EMI problems are interchangeable. In fact, the physics of
EMI have shows that the time response of signals contains all the necessary
46
information to construct the spectral response of the interference. In digital systems,
both the signal rise time and pulse repetition rate produce spectral components
according to the following relationship:
f EMI =
1
π ⋅ t rise
Eq. 7.1
For example, a pulse having a 1ns rise time is equivalent to an EMI frequency of
over 300MHz. This time-frequency relationship can also be applied to high-speed
analog circuits, where slew rates in excess of 1000V/µs and gain-bandwidth products
greater than 500MHz are not uncommon.
When this concept is applied to instruments and systems, EMI emissions are again
functions of signal rise time and pulse repetition rates. Spectrum analyzers and high
speed oscilloscopes used with voltage and current probes are very useful tools in
quantifying the effects of EMI on circuits and systems.
Another important parameter in the analysis of EMI problems is the physical
dimensions of cables, wires, and enclosures. Cables can behave as either passive
antennas (receptors) or very efficient transmitters (sources) of interference. Their
physical length and their shield must be carefully examined where EMI is a concern.
As previously mentioned, the behavior of simple conductors is a function of length,
cross-sectional area, and frequency. Openings in equipment enclosures can behave
as slot antennas, thereby allowing EMI energy to affect the internal electronics.
PASSIVE COMPONENTS: YOUR ARSENAL AGAINST EMI
Minimizing the effects of EMI requires that the circuit/system designer be
completely aware of the primary arsenal in the battle against interference: passive
components. To use successfully these components, the designer must understand
their non-ideal behavior. For example, Figure 7.59 illustrates the real behavior of
the passive components used in circuit design. At very high frequencies, wires
become transmission lines, capacitors become inductors, inductors become
capacitors, and resistors behave as resonant circuits.
47
ALL PASSIVE COMPONENTS EXHIBIT
"NON IDEAL" BEHAVIOR
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
COMPONENT
LF BEHAVIOR
HF BEHAVIOR
RESPONSE
Z
WIRE
f
Z
f
CAPACITOR
Z
INDUCTOR
f
Z
RESISTOR
f
a
7.59
A specific case in point is the frequency response of a simple wire compared to that
of a ground plane. In many circuits, wires are used as either power or signal returns,
and there is no ground plane. A wire will behave as a very low resistance (less than
0.02Ω/ft for 22-gauge wire) at low frequencies, but because of its parasitic
inductance of approximately 20nH/inch, it becomes inductive at frequencies above
13kHz. Furthermore, depending on size and routing of the wire and the frequencies
involved, it ultimately becomes a transmission line with an uncontrolled impedance.
From our knowledge of RF, unterminated transmission lines become antennas with
gain. On the other hand, large area ground planes are much more well-behaved, and
maintain a low impedance over a wide range of frequencies. With a good
understanding of the behavior of real components, a strategy can now be developed
to find solutions to most EMI problems.
RADIO FREQUENCY INTERFERENCE
The world is rich in radio transmitters: radio and TV stations, mobile radios,
computers, electric motors, garage door openers, electric jackhammers, and
countless others. All this electrical activity can affect circuit/system performance
and, in extreme cases, may render it inoperable. Regardless of the location and
magnitude of the interference, circuits/systems must have a minimum level of
immunity to radio frequency interference (RFI). The next section will cover two
general means by which RFI can disrupt normal instrument operation: the direct
effects of RFI sensitive analog circuits, and the effects of RFI on shielded cables.
Two terms are typically used in describing the sensitivity of an electronic system to
RF fields. In communications, radio engineers define immunity to be an
instrument’s susceptibility to the applied RFI power density at the unit. In more
general EMI analysis, the electric-field intensity is used to describe RFI stimulus.
48
For comparative purposes, Equation 7.2 can be used to convert electric-field
intensity to power density and vice-versa:
r  V
 mW 

E   = 61.4 PT 
 m
 cm 2 
Eq. 7.2
where E = Electric Field Strength, in volts per meter, and
PT = Transmitted power, in milliwatts per cm2.
From the standpoint of the source-path-receptor model, the strength of the electric
field, E, surrounding the receptor is a function of transmitted power, antenna gain,
and distance from the source of the disturbance. An approximation for the electricfield intensity (for both near- and far-field sources) in these terms is given by
Equation 7.3:
 PT ⋅ G A
r  V
E   = 5.5 
 m
d





Eq. 7.3
where E = Electric field intensity, in V/m;
PT = Transmitted power, in mW/cm2;
GA = Antenna gain (numerical); and
d = distance from source, in meters
For example, a 1W hand-held radio at a distance of 1 meter can generate an electricfield of 5.5V/m, whereas a 10kW radio transmission station located 1km away
generates a field smaller than 0.6V/m.
Analog circuits are generally more sensitive to RF fields than digital circuits because
analog circuits, operating at high gains, must be able to resolve signals in the
microvolt/millivolt region. Digital circuits, on the other hand, are more immune to
RF fields because of their larger signal swings and noise margins. As shown in
Figure 7.60, RF fields can use inductive and/or capacitive coupling paths to generate
noise currents and voltages which are amplified by high-impedance analog
instrumentation. In many cases, out-of-band noise signals are detected and rectified
by these circuits. The result of the RFI rectification is usually unexplained offset
voltage shifts in the circuit or in the system.
49
RFI CAN CAUSE RECTIFICATION IN
SENSITIVE ANALOG CIRCUITS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
INPUTS PICK UP HIGH FREQUENCY ENERGY ON
SIGNAL LINE, WHICH IS DETECTED BY THE AMPLIFIER
VCC
OUTPUT DRIVERS CAN BE JAMMED, TOO: ENERGY
COUPLES BACK TO INPUT VIA VCC OR SIGNAL LINE
AND THEN IS DETECTED OR AMPLIFIED
a
7.60
There are techniques that can be used to protect analog circuits against interference
from RF fields (see Figure 7.61). The three general points of RFI coupling are signal
inputs, signal outputs, and power supplies. At a minimum, all power supply pin
connections on analog and digital ICs should be decoupled with 0.1µF ceramic
capacitors. As was shown in Reference 3, low-pass filters, whose cutoff frequencies
are set no higher than 10 to 100 times the signal bandwidth, can be used at the
inputs and the outputs of signal conditioning circuitry to filter noise.
50
KEEPING RFI AWAY FROM ANALOG CIRCUITS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
LOCAL
VPOS
REMOTE
VNEG
Decouple all voltage supplies to analog chip with high-frequency capacitors
Use high-frequency filters on all lines that leave the board
Use high-frequency filters on the voltage reference if it is not grounded
a
7.61
Care must be taken to ensure that the low pass filters (LPFs) are effective at the
highest RF interference frequency expected. As illustrated in Figure 7.62, real lowpass filters may exhibit leakage at high frequencies. Their inductors can lose their
effectiveness due to parasitic capacitance, and capacitors can lose their effectiveness
due to parasitic inductance. A rule of thumb is that a conventional low-pass filter
(made up of a single capacitor and inductor) can begin to leak when the applied
signal frequency is 100 to 1000 higher than the filter’s cutoff frequency. For
example, a 10kHz LPF would not be considered very efficient at filtering frequencies
above 1MHz.
51
A SINGLE LOW PASS FILTER LOSES EFFECTIVENESS
AT 100 - 1000 f3dB
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
TYPICALLY 100 - 1000 f3dB
FILTER
ATTENUATION
f3dB
FREQUENCY
a
7.62
Rather than use one LPF stage, it is recommended that the interference frequency
bands be separated into low-band, mid-band, and high-band, and then use
individual filters for each band. Kimmel Gerke Associates use the stereo speaker
analogy of woofer-midrange-tweeter for RFI low-pass filter design illustrated in
Figure 7.63. In this approach, low frequencies are grouped from 10kHz to 1MHz,
mid-band frequencies are grouped from 1MHz to 100MHz, and high frequencies
grouped from 100MHz to 1GHz. In the case of a shielded cable input/output, the
high frequency section should be located close to the shield to prevent highfrequency leakage at the shield boundary. This is commonly referred to as feedthrough protection. For applications where shields are not required at the
inputs/outputs, then the preferred method is to locate the high frequency filter
section as close the analog circuit as possible. This is to prevent the possibility of
pickup from other parts of the circuit.
52
MULTISTAGE FILTERS ARE MORE EFFECTIVE
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
FEEDTHROUGH
CAPACITOR
.01µF
TWEETER
FERRITE
BEAD
.01µF
MIDRANGE
IRON
CORE
1µF
WOOFER
STEREO SPEAKER ANALOGY
a
7.63
Another cause of filter failure is illustrated in Figure 7.64. If there is any impedance
in the ground connection (for example, a long wire or narrow trace connected to the
ground plane), then the high-frequency noise uses this impedance path to bypass the
filter completely. Filter grounds must be broadband and tied to low-impedance
points or planes for optimum performance. High frequency capacitor leads should be
kept as short as possible, and low-inductance surface-mounted ceramic chip
capacitors are preferable.
53
NON-ZERO (INDUCTIVE AND/OR RESISTIVE) FILTER
GROUND REDUCES EFFECTIVENESS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
FILTER
HF
ENERGY
HF
ENERGY
BOND IMPEDANCE
a
7.64
SOLUTIONS FOR POWER-LINE DISTURBANCES
The goal of this next section is not to describe in detail all the circuit/system failure
mechanisms which can result from power-line disturbances or faults. Nor is it the
intent of this section to describe methods by which power-line disturbances can be
prevented. Instead, this section will describe techniques that allow circuits and
systems to accommodate transient power-line disturbances.
Figure 7.65 is an example of a hybrid power transient protection network commonly
used in many applications where lightning transients or other power-line
disturbances are prevalent. These networks can be designed to provide protection
against transients as high as 10kV and as fast as 10ns. Gas discharge tubes
(crowbars) and large geometry zener diodes (clamps) are used to provide both
differential and common-mode protection. Metal-oxide varistors (MOVs) can be
substituted for the zener diodes in less critical, or in more compact designs. Chokes
are used to limit the surge current until the gas discharge tubes fire.
54
POWER LINE DISTURBANCES CAN GENERATE EMI
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
GAS DISCHARGE
TUBES
"CROWBARS"
CHOKES
TRANSIENT
SUPPRESSORS
BIG ZENERS
OR MOVs
V
LINE
LOAD
N
G
COMMON-MODE AND DIFFERENTIAL MODE PROTECTION
a
7.65
Commercial EMI filters, as illustrated in Figure 7.66, can be used to filter less
catastrophic transients or high-frequency interference. These EMI filters provide
both common-mode and differential mode filtering as in Figure 7.66. An optional
choke in the safety ground can provide additional protection against common-mode
noise. The value of this choke cannot be too large, however, because its resistance
may affect power-line fault clearing. These filters work in both directions: they are
not only protect the equipment from surges on the power line but also prevent
transients from the internal switching power supplies from corrupting the power
line.
55
SCHEMATIC FOR A COMMERCIAL POWER LINE FILTER
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
HOT
HOT
LINE
LOAD
NEU
NEU
GND
OPTIONAL
NOTE: OPTIONAL CHOKE ADDED FOR COMMON-MODE PROTECTION
a
7.66
Transformers provide the best common-mode power line isolation. They provide good
protection at low frequencies (<1MHz), or for transients with rise and fall times
greater than 300ns. Most motor noise and lightning transients are in this range, so
isolation transformers work well for these types of disturbances. Although the
isolation between input and output is galvanic, isolation transformers do not provide
sufficient protection against extremely fast transients (<10ns) or those caused by
high-amplitude electrostatic discharge (1 to 3ns). As illustrated in Figure 7.67,
isolation transformers can be designed for various levels of differential- or commonmode protection. For differential-mode noise rejection, the Faraday shield is
connected to the neutral, and for common-mode noise rejection, the shield is
connected to the safety ground.
56
FARADAY SHIELDS IN ISOLATION TRANSFORMERS
PROVIDE INCREASING LEVELS OF PROTECTION
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
STANDARD TRANSFORMER - NO SHIELD
NOTE CONNECTION FROM SECONDARY
TO SAFETY GROUND TO ELIMINATE
GROUND-TO-NEUTRAL VOLTAGE
SINGLE FARADAY SHIELD
CONNECT TO SAFETY GROUND FOR
COMMON-MODE PROTECTION
SINGLE FARADAY SHIELD
CONNECT TO NOISY-SIDE NEUTRAL
WIRE FOR DIFFERENTIAL-MODE
PROTECTION
TRIPLE FARADAY SHIELD
a
CONNECT TO SAFETY GROUND FOR
COMMON MODE
CONNECT TO NEUTRALS FOR
DIFFERENTIAL MODE
7.67
PRINTED CIRCUIT BOARD DESIGN FOR EMI
PROTECTION
This section will summarize general points regarding the most critical portion of the
design phase: the printed circuit board layout. It is at this stage where the
performance of the system is most often compromised. This is not only true for
signal-path performance, but also for the system’s susceptibility to electromagnetic
interference and the amount of electromagnetic energy radiated by the system.
Failure to implement sound PCB layout techniques will very likely lead to
system/instrument EMC failures.
Figure 7.68 is a real-world printed circuit board layout which shows all the paths
through which high-frequency noise can couple/radiate into/out of the circuit.
Although the diagram shows digital circuitry, the same points are applicable to
precision analog, high-speed analog, or mixed analog/digital circuits. Identifying
critical circuits and paths helps in designing the PCB layout for both low emissions
and susceptibility to radiated and conducted external and internal noise sources.
57
METHODS BY WHICH HIGH FREQUENCY ENERGY
COUPLES AND RADIATES INTO CIRCUITRY VIA PLACEMENT
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
COUPLING TO I/O VIA
CROSSTALK OR RADIATION
RADIATION FROM
POWER WIRING
COUPLING VIA COMMON
POWER IMPEDANCE
COUPLING VIA COMMON
GROUND IMPEDANCE
a
RADIATION
FROM I/O
WIRING
7.68
A key point in minimizing noise problems in a design is to choose devices no faster
than actually required by the application. Many designers assume that faster is
better: fast logic is better than slow, high bandwidth amplifiers are clearly better
than low bandwidth ones, and fast DACs and ADCs are better, even if the speed is
not required by the system. Unfortunately, faster is not better, but worse where
EMI is concerned.
Many fast DACs and ADCs have digital inputs and outputs with rise and fall times
in the nanosecond region. Because of their wide bandwidth, the sampling clock and
the digital inputs and can respond to any form of high frequency noise, even glitches
as narrow as 1 to 3ns. These high speed data converters and amplifiers are easy
prey for the high frequency noise of microprocessors, digital signal processors,
motors, switching regulators, hand-held radios, electric jackhammers, etc. With
some of these high-speed devices, a small amount of input/output filtering may be
required to desensitize the circuit from its EMI/RFI environment. Adding a small
ferrite bead just before the decoupling capacitor as shown in Figure 7.69 is very
effective in filtering high frequency noise on the supply lines. For those circuits that
require bipolar supplies, this technique should be applied to both positive and
negative supply lines.
To help reduce the emissions generated by extremely fast moving digital signals at
DAC inputs or ADC outputs, a small resistor or ferrite bead may be required at each
digital input/output.
58
POWER SUPPLY FILTERING AND SIGNAL LINE
SNUBBING GREATLY REDUCES EMI EMISSIONS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
FERRITE
BEAD
FERRITE BEAD OR
10 - 33Ω RESISTOR
VCC
GND
MICROPROCESSOR
OR OTHER HIGH-SPEED
CLOCKED CIRCUIT
a
7.69
Once the system’s critical paths and circuits have been identified, the next step in
implementing sound PCB layout is to partition the printed circuit board according to
circuit function. This involves the appropriate use of power, ground, and signal
planes. Good PCB layouts also isolate critical analog paths from sources of high
interference (I/O lines and connectors, for example). High frequency circuits (analog
and digital) should be separated from low frequency ones. Furthermore, automatic
signal routing CAD layout software should be used with extreme caution, and
critical paths routed by hand.
Properly designed multilayer printed circuit boards can reduce EMI emissions and
increase immunity to RF fields by a factor of 10 or more compared to double-sided
boards. A multilayer board allows a complete layer to be used for the ground plane,
whereas the ground plane side of a double-sided board is often disrupted with signal
crossovers, etc. If the system has separate analog and digital ground and power
planes, the analog ground plane should be underneath the analog power plane, and
similarly, the digital ground plane should be underneath the digital power plane.
There should be no overlap between analog and digital ground planes nor analog
and digital power planes.
The preferred multi-layer board arrangement is to embed the signal traces between
the power and ground planes, as shown in Figure 7.70. These low-impedance planes
form very high-frequency stripline transmission lines with the signal traces. The
return current path for a high frequency signal on a trace is located directly above
and below the trace on the ground/power planes. The high frequency signal is thus
contained inside the PCB, thereby minimizing emissions. The embedded signal trace
approach has an obvious disadvantage: debugging circuit traces that are hidden
from plain view is difficult.
59
"TO EMBED OR NOT TO EMBED"
THAT IS THE QUESTION
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
BEFORE
AFTER
Route
Power
Power
Route
Ground
Route
Route
Ground
Advantages of Embedding
Lower impedances, therefore lower emissions and crosstalk
Reduction in emissions and crosstalk is significant above 50MHz
Traces are protected
Disadvantages of Embedding
Lower interboard capacitance, harder to decouple
Impedances may be too low for matching
Hard to prototype and troubleshoot buried traces
a
7.70
Much has been written about terminating printed circuit board traces in their
characteristic impedance to avoid reflections. A good rule-of-thumb to determine
when this is necessary is as follows: Terminate the line in its characteristic
impedance when the one-way propagation delay of the PCB track is equal to or
greater than one-half the applied signal rise/fall time (whichever edge is faster). A
conservative approach is to use a 2 inch (PCB track length)/nanosecond (rise-, falltime) criterion. For example, PCB tracks for high-speed logic with rise/fall time of
5ns should be terminated in their characteristic impedance and if the track length is
equal to or greater than 10 inches (including any meanders). The 2 inch/nanosecond
track length criterion is summarized in Figure 7.71 for a number of logic families.
LINE TERMINATION SHOULD BE USED WHEN
LENGTH OF PCB TRACK EXCEEDS 2 inches / ns
Reprinted from EDN Magazine (January 20,1994), © CAHNERS PUBLISHING
COMPANY 1995, A Division of Reed Publishing USA
DIGITAL IC
FAMILY
GaAs
ECL
Schottky
FAST
AS
AC
ALS
LS
TTL
tr, tf
(ns)
0.1
0.75
3
3
3
4
6
8
10
PCB TRACK LENGTH
(inches)
0.2
1.5
6
6
6
8
12
16
20
60
PCB TRACK LENGTH
(cm)
0.5
3.8
15
15
15
20
30
40
50
HC
18
36
90
tr = rise time of signal in ns
tf = fall time of signal in ns
n
For analog signals @ fmax, calculate tr = tf = 0.35 / fmax
a
7.71
This same 2 inch/nanosecond rule of thumb should be used with analog circuits in
determining the need for transmission line techniques. For instance, if an amplifier
must output a maximum frequency of fmax, then the equivalent risetime, tr, can be
calculated using the equation tr = 0.35/fmax. The maximum PCB track length is
then calculated by multiplying the risetime by 2 inch/nanosecond. For example, a
maximum output frequency of 100MHz corresponds to a risetime of 3.5ns, and a
track carrying this signal greater than 7 inches should be treated as a transmission
line.
Equation 7.4 can be used to determine the characteristic impedance of a PCB track
separated from a power/ground plane by the board’s dielectric (microstrip
transmission line):
Z o (Ω) =
87
 5.98d 
ln 
 Eq. 7.4
ε r + 1.41
 0.89w + t 
where εr = dielectric constant of printed circuit board material;
d = thickness of the board between metal layers, in mils;
w = width of metal trace, in mils; and
t = thickness of metal trace, in mils.
The one-way transit time for a single metal trace over a power/ground plane can be
determined from Eq. 7.5:
t pd ( ns / ft) = 1.017 0.475ε r + 0.67 Eq. 7.5
For example, a standard 4-layer PCB board might use 8-mil wide, 1 ounce (1.4 mils)
copper traces separated by 0.021" FR-4 (εr=4.7) dielectric material. The
characteristic impedance and one-way transit time of such a signal trace would be
88Ω and 1.7ns/ft (7"/ns), respectively. Transmission lines can be effectively
terminated in several ways depending on the application, as described in Section 2
of this book.
61
REFERENCES ON EMI/RFI
1.
EDN’s Designer’s Guide to Electromagnetic Compatibility, EDN,
January, 20, 1994, material reprinted by permission of Cahners Publishing
Company, 1995.
2.
Designing for EMC (Workshop Notes), Kimmel Gerke Associates, Ltd., 1994.
3.
Systems Application Guide, Chapter 1, pg. 21-55, Analog Devices,
Incorporated, Norwood, MA, 1994.
4.
Henry Ott, Noise Reduction Techniques In Electronic Systems,
Second Edition, New York, John Wiley & Sons, 1988.
5.
Ralph Morrison, Grounding And Shielding Techniques In
Instrumentation, Third Edition, New York, John Wiley & Sons, 1986.
6.
Amplifier Applications Guide, Chapter XI, pg. 61, Analog Devices,
Incorporated, Norwood, MA, 1992.
7.
B.Slattery and J.Wynne, Design and Layout of a Video Graphics
System for Reduced EMI, Analog Devices Application Note AN-333.
8.
Paul Brokaw, An IC Amplifier User Guide To Decoupling, Grounding,
And Making Things Go Right For A Change, Analog Devices
Application Note, Order Number E1393-5-590.
9.
A. Rich, Understanding Interference-Type Noise, Analog Dialogue, 16-3,
1982, pp. 16-19.
10.
A. Rich, Shielding and Guarding, Analog Dialogue, 17-1, 1983, pp. 8-13.
11.
EMC Test & Design, Cardiff Publishing Company, Englewood, CO.
An excellent, general purpose trade journal on issues of EMI and EMC.
62
SHIELDING CONCEPTS
Adolfo Garcia, John McDonald
The concepts of shielding effectiveness presented next are background material.
Interested readers should consult References 1,2, and 6 cited at the end of the
section for more detailed information.
Applying the concepts of shielding requires an understanding of the source of the
interference, the environment surrounding the source, and the distance between the
source and point of observation (the receptor or victim). If the circuit is operating
close to the source (in the near-, or induction-field), then the field characteristics are
determined by the source. If the circuit is remotely located (in the far-, or radiationfield), then the field characteristics are determined by the transmission medium.
A circuit operates in a near-field if its distance from the source of the interference is
less than the wavelength (λ) of the interference divided by 2π, or λ/2π. If the distance
between the circuit and the source of the interference is larger than this quantity,
then the circuit operates in the far field. For instance, the interference caused by a
1ns pulse edge has an upper bandwidth of approximately 350MHz. The wavelength
of a 350MHz signal is approximately 32 inches (the speed of light is approximately
12"/ns). Dividing the wavelength by 2π yields a distance of approximately 5 inches,
the boundary between near- and far-field. If a circuit is within 5 inches of a 350MHz
interference source, then the circuit operates in the near-field of the interference. If
the distance is greater than 5 inches, the circuit operates in the far-field of the
interference.
Regardless of the type of interference, there is a characteristic impedance associated
with it. The characteristic, or wave impedance of a field is determined by the ratio of
its electric (or E-) field to its magnetic (or H-) field. In the far field, the ratio of the
electric field to the magnetic field is the characteristic (wave impedance) of free
space, given by Zo = 377Ω. In the near field, the wave-impedance is determined by
the nature of the interference and its distance from the source. If the interference
source is high-current and low-voltage (for example, a loop antenna or a power-line
transformer), the field is predominately magnetic and exhibits a wave impedance
which is less than 377Ω. If the source is low-current and high-voltage (for example, a
rod antenna or a high-speed digital switching circuit), then the field is
predominately electric and exhibits a wave impedance which is greater than 377Ω.
Conductive enclosures can be used to shield sensitive circuits from the effects of
these external fields. These materials present an impedance mismatch to the
incident interference because the impedance of the shield is lower than the wave
impedance of the incident field. The effectiveness of the conductive shield depends on
two things: First is the loss due to the reflection of the incident wave off the
shielding material. Second is the loss due to the absorption of the transmitted wave
within the shielding material. Both concepts are illustrated in Figure 7.72. The
amount of reflection loss depends upon the type of interference and its wave
impedance. The amount of absorption loss, however, is independent of the type of
interference. It is the same for near- and far-field radiation, as well as for electric or
magnetic fields.
63
REFLECTION AND ABSORPTION ARE THE TWO
PRINCIPAL SHIELDING MECHANISMS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
INCIDENT RAY
REFLECTED RAY
TRANSMITTED
RAY
SHIELD
MATERIAL
a
ABSORPTIVE
REGION
7.72
Reflection loss at the interface between two media depends on the difference in the
characteristic impedances of the two media. For electric fields, reflection loss
depends on the frequency of the interference and the shielding material. This loss
can be expressed in dB, and is given by:
 σ

r  Eq. 7.6
R e (dB) = 322 + 10log10 
µ f 3 r2 
 r

where σr = relative conductivity of the shielding material, in Siemens per meter;
µr = relative permeability of the shielding material, in Henries per meter;
f = frequency of the interference, and
r = distance from source of the interference, in meters
For magnetic fields, the loss depends also on the shielding material and the
frequency of the interference. Reflection loss for magnetic fields is given by:
 f r2 σ 
r  Eq. 7.7
R m (dB) = 14.6 + 10log10 
 µr 


and, for plane waves ( r > λ/2π), the reflection loss is given by:
σ 
R pw (dB) = 168 + 10log10  r 
µr f 
Eq. 7.8
64
Absorption is the second loss mechanism in shielding materials. Wave attenuation
due to absorption is given by:
A (dB) = 3.34 t σ r µ r f
Eq. 7.9
where t = thickness of the shield material, in inches. This expression is valid for
plane waves, electric and magnetic fields. Since the intensity of a transmitted field
decreases exponentially relative to the thickness of the shielding material, the
absorption loss in a shield one skin-depth (δ) thick is 9dB. Since absorption loss is
proportional to thickness and inversely proportional to skin depth, increasing the
thickness of the shielding material improves shielding effectiveness at high
frequencies.
Reflection loss for plane waves in the far field decreases with increasing frequency
because the shield impedance, Zs, increases with frequency. Absorption loss, on the
other hand, increases with frequency because skin depth decreases. For electric
fields and plane waves, the primary shielding mechanism is reflection loss, and at
high frequencies, the mechanism is absorption loss. For these types of interference,
high conductivity materials, such as copper or aluminum, provide adequate
shielding. At low frequencies, both reflection and absorption loss to magnetic fields
is low; thus, it is very difficult to shield circuits from low-frequency magnetic fields.
In these applications, high-permeability materials that exhibit low-reluctance
provide the best protection. These low-reluctance materials provide a magnetic
shunt path that diverts the magnetic field away from the protected circuit. Some
characteristics of metallic materials commonly used for shielded enclosures are
shown in Figure 7.73.
IMPEDANCE AND SKIN DEPTHS
FOR VARIOUS SHIELDING MATERIALS
Material
Cu
Al
Conductivity
Permeability
Shield Impedance
Skin Depth
σr
µr
|Zs|
δ (inch)
1
1
3.68E- 7 ⋅ f
2.6
4.71E - 7 ⋅ f
3.3
1
0.61
f
f
Steel
0.1
3.68E - 5 ⋅ f
1,000
0.26
f
µ Metal
0.03
3E - 4 ⋅ f
20,000
0.11
f
where so = 5.82 × 107 S/m
µo = 4p × 10-7 H/m
eo = 8.85 × 10-12 F/m
65
a
7.73
A properly shielded enclosure is very effective at preventing external interference
from disrupting its contents as well as confining any internally-generated
interference. However, in the real world, openings in the shield are often required to
accommodate adjustment knobs, switches, connectors, or to provide ventilation (see
Figure 7.74). Unfortunately, these openings may compromise shielding effectiveness
by providing paths for high-frequency interference to enter the instrument.
ANY OPENING IN AN ENCLOSURE CAN ACT AS
AN EMI WAVEGUIDE BY COMPROMISING
SHIELDING EFFECTIVENESS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
VENTILATORS
SEAMS
SWITCHES
DATA
CABLES
DISPLAY
PANEL
POWER
CABLES
a
7.74
The longest dimension (not the total area) of an opening is used to evaluate the
ability of external fields to enter the enclosure, because the openings behave as slot
antennas. Equation 7.10 can be used to calculate the shielding effectiveness, or the
susceptibility to EMI leakage or penetration, of an opening in an enclosure:
 λ 
Shielding Effectiveness (dB) = 20 log10 

 2 ⋅ L
Eq. 7.10
where λ = wavelength of the interference and
L = maximum dimension of the opening
Maximum radiation of EMI through an opening occurs when the longest dimension
of the opening is equal to one half-wavelength of the interference frequency (0dB
shielding effectiveness). A rule-of-thumb is to keep the longest dimension less than
1/20 wavelength of the interference signal, as this provides 20dB shielding
effectiveness. Furthermore, a few small openings on each side of an enclosure is
66
preferred over many openings on one side. This is because the openings on different
sides radiate energy in different directions, and as a result, shielding effectiveness is
not compromised. If openings and seams cannot be avoided, then conductive
gaskets, screens, and paints alone or in combination should be used judiciously to
limit the longest dimension of any opening to less than 1/20 wavelength. Any cables,
wires, connectors, indicators, or control shafts penetrating the enclosure should have
circumferential metallic shields physically bonded to the enclosure at the point of
entry. In those applications where unshielded cables/wires are used, then filters are
recommended at the point of shield entry.
Sensors and Cable Shielding
The improper use of cables and their shields is a significant contributor to both
radiated and conducted interference. As illustrated in Figure 7.75, effective cable
and enclosure shielding confines sensitive circuitry and signals within the entire
shield without compromising shielding effectiveness.
LENGTH OF SHIELDED CABLES DETERMINES AN
"ELECTRICALLY LONG" OR "ELECTRICALLY SHORT"
APPLICATION
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
SHIELDED ENCLOSURE B
SHIELDED ENCLOSURE A
LENGTH
SHIELDED
CABLE
FULLY SHIELDED ENCLOSURES CONNECTED BY FULLY
SHIELDED CABLE KEEP ALL INTERNAL CIRCUITS AND
SIGNAL LINES INSIDE THE SHIELD.
TRANSITION REGION: 1/20 WAVELENGTH
a
7.75
Depending on the type of interference (pickup/radiated, low/high frequency), proper
cable shielding is implemented differently and is very dependent on the length of the
cable. The first step is to determine whether the length of the cable is electrically
short or electrically long at the frequency of concern. A cable is considered
electrically short if the length of the cable is less than 1/20 wavelength of the highest
frequency of the interference, otherwise it is electrically long. For example, at
50/60Hz, an electrically short cable is any cable length less than 150 miles, where
the primary coupling mechanism for these low frequency electric fields is capacitive.
As such, for any cable length less than 150 miles, the amplitude of the interference
will be the same over the entire length of the cable. To protect circuits against lowfrequency electric-field pickup, only one end of the shield should be returned to a
67
low-impedance point. A generalized example of this mechanism is illustrated in
Figure 7.76.
CONNECT THE SHIELD AT ONE POINT AT THE LOAD
TO PROTECT AGAINST LOW FREQUENCY (50/60Hz) THREATS
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
CAPACITIVE COUPLING
TO CABLE
CABLE SHIELD
GROUNDED AT LOAD
RECEIVER
RECEIVER
en
en
en
EQUIVALENT
CIRCUITS
a
en
7.76
In this example, the shield is grounded at the receiver. An exception to this
approach (which will be highlighted again later) is the case where line-level
(>1Vrms) audio signals are transmitted over long distances using twisted pair,
shielded cables. In these applications, the shield again offers protection against lowfrequency interference, and an accepted approach is to ground the shield at the
driver end (LF and HF ground) and ground it at the receiver with a capacitor (HF
ground only).
In those applications where the length of the cable is electrically long, or protection
against high-frequency interference is required, then the preferred method is to
connect the cable shield to low-impedance points at both ends (direct connection at
the driving end, and capacitive connection at the receiver). Otherwise, unterminated
transmission lines effects can cause reflections and standing waves along the cable.
At frequencies of 10MHz and above, circumferential (360°) shield bonds and metal
connectors are required to main low-impedance connections to ground.
In summary, for protection against low-frequency (<1MHz), electric-field
interference, grounding the shield at one end is acceptable. For high-frequency
interference (>1MHz), the preferred method is grounding the shield at both ends,
using 360° circumferential bonds between the shield and the connector, and
maintaining metal-to-metal continuity between the connectors and the enclosure.
Low-frequency ground loops can be eliminated by replacing one of the DC shield
connections to ground with a low inductance 0.01µF capacitor. This capacitor
prevents low frequency ground loops and shunts high frequency interference to
ground.
68
Shielded Twisted Pair Cable Grounding Examples
The environments in which analog systems operate are often rich in sources of EMI.
Common EMI noise sources include power lines, logic signals, switching power
supplies, radio stations, electric lighting, and motors. Noise from these sources can
easily couple into long analog signal paths, such as cables, which act as efficient
antennas. Shielded cables protect signal conductors from electric field (E-field)
interference by providing low impedance paths to ground at the offending
frequencies. Aluminum foil, copper, and braided stainless steel are materials very
commonly used for cable shields due to their low impedance properties.
Simply increasing the separation between the noise source and the cable will yield
significant additional attenuation due to reduced coupling, but shielding is still
required in most applications involving remote sensors.
There are two paths from an EMI source to a susceptible cable: capacitive (or Efield) and magnetic (or H-field) coupling. Capacitive coupling occurs when parasitic
capacitance exists between a noise source and the cable. The amount of parasitic
capacitance is determined by the separation, shape, orientation, and the medium
between the source and the cable.
Magnetic coupling occurs through parasitic mutual inductance when a magnetic
field is coupled from one conductor to another. Parasitic mutual inductance depends
on the shape and relative orientation of the circuits in question, the magnetic
properties of the medium, and is directly proportional to conductor loop area.
Minimizing conductor loop area reduces magnetic coupling proportionally.
Shielded twisted pair cables offer further noise immunity to magnetic fields.
Twisting the conductors together reduces the net loop area, which has the effect of
canceling any magnetic field pickup, because the sum of positive and negative
incremental loop areas is ideally equal to zero.
To study the shielding problem, a precision RTD (Resistance Temperature Detector)
amplifier circuit was used as the basis for a series of experiments. A remote 100Ω
RTD was connected to the bridge, bridge driver, and the bridge amplifier circuit
(Figure 7.77) using 10 feet of a shielded twisted pair cable. The RTD is one element
of a 4-element bridge (the three other resistor elements are located in the bridge and
bridge driver circuit). The gain of the instrumentation amplifier was adjusted so
that the sensitivity at the output was 10mV/°C, with a 5V full scale. Measurements
were made at the output of the instrumentation amplifier with the shield grounded
in various ways. The experiments were conducted in lab standard environment
where a considerable amount of electronic equipment was in operation.
69
UNGROUNDED SHIELDED CABLES ACT AS ANTENNAS
RG
10 FEET
SHIELDED
TWISTED
PAIR
RTD
100Ω
BRIDGE
AND
BRIDGE
DRIVER
IN-AMP
OUTPUT
IN
AMP
5V FS
10mV/ºC
VERTICAL SCALE: 2mV/div
HORIZONTAL SCALE: 10ms/div
a
7.77
The first experiment was conducted with the shield ungrounded. As shown in Figure
7.77, shields left floating are not useful and offer no attenuation to EMI-induced
noise, in fact, they act as antennas. Capacitive coupling is unaffected, because the
floating shield provides a coupling path to the signal conductors. Most cables exhibit
parasitic capacitances between 10-30pF/ft. Likewise, HF magnetically coupled noise
is not attenuated because the floating cable shield does not alter either the geometry
or the magnetic properties of the cable conductors. LF magnetic noise is not
attenuated significantly, because most shield materials absorb very little magnetic
energy.
To implement effective EMI/RFI shielding, the shield must be grounded. A grounded
shield reduces the value of the impedance of the shield to ground to small values.
Implementing this change will reduce the amplitude of the E-Field noise
substantially.
Designers often ground both ends of a shield in an attempt to reduce shield
impedance and gain further E-Field attenuation. Unfortunately, this approach can
create a new set of potential problems. The AC and DC ground potentials are
generally different at each end of the shield. Low-frequency ground loop current is
created when both ends of a shield are grounded. This low frequency current flows
through the large loop area of the shield and couples into the center conductors
through the parasitic mutual inductance. If the twisted pairs are precisely balanced,
the induced voltage will appear as a common-mode rather than a differential
voltage. Unfortunately, the conductors may not be perfectly balanced, the sensor
and excitation circuit may not be fully balanced, and the common mode rejection at
the receiver may not be sufficient. There will therefore be some differential noise
voltage developed between the conductors at the output end, which is amplified and
70
appears at the final output of the instrumentation amplifier. With the shields of the
experimental circuit grounded at both ends, the results are shown in Figure 7.78.
GROUNDING BOTH ENDS OF A SHIELD PRODUCES
LOW FREQUENCY GROUND LOOPS
RG
10 FEET
SHIELDED
TWISTED
PAIR
RTD
100Ω
BRIDGE
AND
BRIDGE
DRIVER
IN
AMP
5V FS
10mV/ºC
G2
IN-AMP
OUTPUT
VERTICAL SCALE: 2mV/div
HORIZONTAL SCALE: 10ms/div
a
7.78
Figure 7.79 illustrates a properly grounded system with good electric field shielding.
Notice that the ground loop has been eliminated. The shield has a single point
ground, located at the signal conditioning circuitry, and noise coupled into the shield
is effectively shunted into the receiver ground and does not appear at the output of
the instrumentation amplifier.
71
GROUNDING SHIELD AT RECEIVER END SHUNTS LOW- AND
HIGH-FREQUENCY NOISE INTO RECEIVER GROUND
RG
10 FEET
SHIELDED
TWISTED
PAIR
RTD
100Ω
BRIDGE
AND
BRIDGE
DRIVER
IN-AMP
OUTPUT
IN
AMP
5V FS
10mV/ºC
VERTICAL SCALE: 2mV/div
HORIZONTAL SCALE: 10ms/div
a
7.79
Figure 7.80 shows an example of a remotely located, ungrounded, passive sensor
(ECG electrodes) which is connected to a high-gain, low power AD620
instrumentation amplifier through a shielded twisted pair cable. Note that the
shield is properly grounded at the signal conditioning circuitry. The AD620 gain is
1000× , and the amplifier is operated on ±3V supplies. Notice the absence of 60Hz
interference in the amplifier output.
72
FOR UNGROUNDED PASSIVE SENSORS,
GROUND SHIELD AT THE RECEIVING END
RG
+
+
SHIELDED
TWISTED
PAIR
-
G = 1000
AD620
REF
OUTPUT
EEG
ELECTRODES
G2
G1
VERTICAL SCALE: 10mV/div
HORIZONTAL SCALE: 0.2sec/div
IN-AMP
OUTPUT
a
7.80
Most high impedance sensors generate low-level current or voltage outputs, such as
a photodiode responding to incident light. These low-level signals are especially
susceptible to EMI, and often are of the same order of magnitude as the parasitic
parameters of the cable and input amplifier.
Even properly shielded cables can degrade the signals by introducing parasitic
capacitance that limits bandwidth, and leakage currents that limit sensitivity. An
example is shown in Figure 7.81, where a high-impedance photodiode is connected to
a preamp through a long shielded twisted pair cable. Not only will the cable
capacitance limit bandwidth, but cable leakage current limits sensitivity. A preamplifier, located close to the high-impedance sensor, is recommended to amplify the
signal and to minimize the effect of cable parasitics.
73
SHIELDS ARE NOT EFFECTIVE WITH
HIGH IMPEDANCE REMOTE SENSORS
CCOMP
PHOTODIODE
DETECTOR
RFB
SHIELDED
TWISTED
PAIR
HIGH
IMPEDANCE
>100MΩ
+
≈ 20pF/ft
CABLE CAPACITANCE LIMITS BANDWIDTH
CABLE LEAKAGE CURRENT LIMITS SENSITIVITY
a
7.81
Figure 7.82 is an example of a high-impedance photodiode detector and preamplifier, driving a shielded twisted pair cable. Both the amplifier and the shield are
grounded at a remote location. The shield is connected to the cable driver common,
G1, ensuring that the signal and the shield at the driving end are both referenced to
the same point. The capacitor on the receiving side of the cable shunts high
frequency noise on the shield into ground G2 without introducing a low-frequency
ground loop. This popular grounding scheme is known as hybrid grounding.
74
REMOTELY LOCATED HIGH IMPEDANCE
SENSOR WITH PREAMP
PHOTODIODE
PREAMP
RG
+
SHIELDED
TWISTED
PAIR
+
IN-AMP
-
REF
LF AND HF
GROUND
HF GROUND
G1
G2
a
7.82
Figure 7.83 illustrates a balanced active line driver with a hybrid shield ground
implementation. When a system’s operation calls for a wide frequency range, the
hybrid grounding technique often provides the best choice (Reference 8). The
capacitor at the receiving end shunts high-frequency noise on the shield into G2
without introducing a low-frequency ground loop. At the receiver, a common-mode
choke can be used to help prevent RF pickup entering the receiver, and subsequent
RFI rectification (see References 9 and 10). Care should be taken that the shields
are grounded to the chassis entry points to prevent contamination of the signal
ground (Reference 11).
75
HYBRID (LF AND HF) GROUNDING WITH ACTIVE DRIVER
BALANCED
LINE
DRIVER
BALANCED
LINE
RECEIVER
SHIELDED
TWISTED
PAIR
CM
CHOKE
-
REF
G1
G2
LF AND HF
GROUND
HF GROUND
a
7.83
To summarize this discussion, shield grounding techniques must take into account
the type and the configuration of the sensor as well as the nature of the interference.
When a low-impedance passive sensor is used, grounding the shield to the receiving
end is the best choice. Active sensor shields should generally be grounded at the
source (direct connection to source ground) and at the receiver (connect to receiver
ground using a capacitor). This hybrid approach minimizes high-frequency
interference and prevents low-frequency ground loops. Shielded twisted conductors
offer additional protection against shield noise because the coupled noise occurs as a
common-mode, and not a differential signal.
The best shield can be compromised by poor connection techniques. Shields often use
“pig-tail” connections to make the connection to ground. A “pig-tail” connection is a
single wire connection from shield to either chassis or circuit ground. This type of
connection is inexpensive, but at high frequency, it does not provide low impedance.
Quality shields do not leave large gaps in the cable/instrument shielding system.
Shield gaps provide paths for high frequency EMI to enter the system. The cable
shielding system should include the cable end connectors. Ideally, cable shield
connectors should make 360° contact with the chassis ground.
As shown in Figure 7.84, pigtail terminations on cables very often cause systems to
fail radiated emissions tests because high-frequency noise has coupled into the cable
shield, generally through stray capacitance. If the length of the cable is considered
electrically long at the interference frequency, then it can behave as a very efficient
quarter-wave antenna. The cable pigtail forms a matching network, as shown in the
figure, to radiate the noise which coupled into the shield. In general, pigtails are
only recommended for applications below 10kHz, such as 50/60Hz interference
protection. For applications where the interference is greater than 10kHz, shielded
connectors, electrically and physically connected to the chassis, should be used.
76
"SHIELDED" CABLE CAN CARRY HIGH FREQUENCY
CURRENT AND BEHAVES AS AN ANTENNA
Reprinted from EDN Magazine (January 20, 1994) © CAHNERS PUBLISHING COMPANY 1995, A Division of Reed Publishing USA
EQUIVALENT
CIRCUIT
SHIELD
ICM
ICM
ICM = COMMON-MODE CURRENT
a
7.84
77
REFERENCES: CABLE SHIELDING
1.
H.W. Ott, Noise Reduction Techniques in Electronic Systems,
Second Edition, John Wiley & Sons, Inc., New York, 1988.
2.
Ralph Morrison, Grounding and Shielding Techniques in
Instrumentation, Third Edition, John Wiley & Sons, Inc.,
New York, 1988.
3.
Systems Application Guide, Section 1, Analog Devices, Inc.,
Norwood, MA, 1993.
4.
AD620 Instrumentation Amplifier, Data Sheet, Analog Devices, Inc.
5.
A. Rich, Understanding Interference-Type Noise, Analog Dialogue,
16-3, 1982, pp. 16-19.
6.
A. Rich, Shielding and Guarding, Analog Dialogue, 17-1, 1983,
pp. 8-13.
7.
EDN’s Designer’s Guide to Electromagnetic Compatibility, EDN,
January, 20, 1994, material reprinted by permission of Cahners
Publishing Company, 1995.
8.
Designing for EMC (Workshop Notes), Kimmel Gerke Associates,
Ltd., 1994.
9.
James Bryant and Herman Gelbach, High Frequency Signal
Contamination, Analog Dialogue, Vol. 27-2, 1993.
10.
Walt Jung, System RF Interference Prevention, Analog Dialogue,
Vol. 28-2, 1994.
11.
Neil Muncy, Noise Susceptibility in Analog and Digital Signal
Processing Systems, presented at 97th Audio Engineering Society
Convention, Nov. 1994.
78
GENERAL REFERENCES: HARDWARE
DESIGN TECHNIQUES
1.
Linear Design Seminar, Section 11, Analog Devices, Inc., 1995.
2.
E.S.D. Prevention Manual
Available free from Analog Devices, Inc.
3.
B.I. & B. Bleaney, Electricity & Magnetism, OUP 1957, pp 23,24, & 52.
4.
Paul Brokaw, An I.C. Amplifier User's Guide to Decoupling, Grounding
and Making Things Go Right for a Change, Analog Devices Application
Note, Available free of charge from Analog Devices, Inc.
5.
Jeff Barrow, Avoiding Ground Problems in High Speed Circuits,
R.F. Design, July 1989.
AND
Paul Brokaw & Jeff Barrow, Grounding for Low- and High-Frequency
Circuits, Analog Dialogue, 23-3 1989.
Free from Analog Devices.
6.
International EMI Emission Regulations
Canada
CSA C108.8-M1983
FDR
Japan
CISPR (VCCI)/PUB 22
USA
VDE 0871/VDE 0875
FCC-15 Part J
7.
Bill Slattery & John Wynne, Design & Layout of a Video Graphics System
for Reduced EMI, Analog Devices Application Note (E1309-15-10/89)
Free from Analog Devices.
8.
William R. Blood, Jr., MECL System Design Handbook
(HB205, Rev. 1), Motorola Semiconductor Products, Inc., 1988.
9.
Wainwright Instruments Inc., 69 Madison Ave., Telford, PA,
18969-1829, Tel. 215-723-4333, Fax. 215-723-4620.
Wainwright Instruments GmbH, Widdersberger Strasse 14,
DW-8138 Andechs-Frieding, Germany. Tel: +49-8152-3162,
Fax: +49-8152-40525.
10.
Ralph Morrison, Grounding and Shielding Techniques in
Instrumentation, Third Edition, John Wiley, Inc., 1986.
11.
Henry W. Ott, Noise Reduction Techniques in Electronic Systems,
Second Edition, John Wiley, Inc., 1988.
12.
Robert A. Pease, Troubleshooting Analog Circuits, ButterworthHeinemann, 1991.
79
13.
Jim Williams, Editor, Analog Circuit Design: Art, Science, and
Personalities, Butterworth-Heinemann, 1991.
14.
Doug Grant and Scott Wurcer, Avoiding Passive Component Pitfalls,
The Best of Analog Dialogue, pp. 143-148, Analog Devices, Inc., 1991.
15.
Walt Jung and Richard Marsh, Picking Capacitors, Part I., Audio,
February, 1980.
16.
Walt Jung and Richard Marsh, Picking Capacitors, Part II., Audio,
March, 1980.
17.
Daryl Gerke and Bill Kimmel, The Designer's Guide to Electromagnetic
Compatibility, EDN Supplement, January 20, 1994.
18.
Walt Kester, Basic Characteristics Distinguish Sampling A/D Converters,
EDN, September 3, 1992, pp.135-144.
19.
Walt Kester, Peripheral Circuits Can Make or Break Sampling ADC
System, EDN, October 1, 1992, pp. 97-105.
20.
Walt Kester, Layout, Grounding, and Filtering Complete Sampling
ADC System, EDN, October 15, 1992, pp. 127-134.
21.
Howard W. Johnson and Martin Graham, High-Speed Digital Design,
PTR Prentice Hall, 1993.
80
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement