Volume 43, Number 4
Volume 43, Number 4, 2009
A forum for the exchange of circuits, systems, and software for real-world signal processing
In This Issue
2 Editors’ Notes; New Product Introductions
3 Isolation in High-Voltage Battery Monitoring for Transportation Applications
6 Synchronizing Device Clocks Using IEEE 1588 and Blackfin Embedded Processors
11 Using Isolated RS-485 in DMX512 Lighting Applications
1 3 Designing High-Performance Phase-Locked Loops with High-Voltage VCOs
1 7 Adjustable-Gain Difference Amplifier Circuit Measures Hundreds of Volts, Rejects Large
Common-Mode Signals
19 Automobile Tail-Lamp and Brake-Lamp Controller
www.analog.com/analogdialogue
Editors’ Notes
PRODUCT INTRODUCTIONS: Volume 43, number 4
IN THIS ISSUE
Isolation in High-Voltage Battery Monitoring for Transportation
Applications
Battery stacks for transportation can provide hundreds of volts.
These high voltages can prove lethal to human beings—and even
lower voltages can damage electronic equipment—so safety is a
key concern. Although these stacks are inherently dangerous,
they must still communicate with the cell-monitoring electronics.
Galvanic isolation is thus required to make the communications
method safe and reliable. Page 3.
Synchronizing Device Clocks Using IEEE 1588 and Blackfin
Embedded Processors
IEEE 1588 defines a protocol to synchronize distributed clocks
on a network. The preferred clock synchronization method for
many applications, it is cost-effective, supports heterogeneous
systems, and provides nanosecond-level synchronization. The
ADSP-BF518 Blackfin® processor includes dedicated hardware
support for IEEE 1588. This article shows clock synchronization
performance obtained using this solution. Page 6.
Using Isolated RS-485 in DMX512 Lighting Applications
Theatrical lighting applications have evolved from lanterns in
open-air theaters into the more complex systems available today.
Modern lighting equipment includes dimmers, flashing lights,
moving lights, colored lights, and gobos. These lighting systems
are often controlled over long distances—up to 4000 feet—using
the DMX512 communications protocol. Page 11.
Designing High-Performance Phase-Locked Loops with
High-Voltage VCOs
Phase-locked loops are used to provide the local oscillator in
radio receivers and transmitters; for clock distribution and
noise reduction, and as the clock source for high-sampling-rate
ADCs. This article considers the basics of PLLs, examines the
current state of the art in PLL design, discusses pros and cons
of typical architectures, and introduces some alternatives to
high-voltage VCOs. Page 13.
Adjustable-Gain Difference Amplifier Circuit Measures Hundreds
of Volts, Rejects Large Common-Mode Signals
To monitor power-line voltages or other large signals, a differential
amplifier in a feedback loop with an inverting op amp is useful
for measuring differential signals up to 500 V. This circuit also
rejects large common-mode voltages and allows the differential
gain to be set by a ratio of resistors, enabling the user to select the
level of attenuation. Page 17.
Automobile Tail-Lamp and Brake-Lamp Controller
Light emitting diodes (LEDs) are recently finding uses in
automobiles, where they provide signaling, daytime running
lights, and interior lighting. As this technology hits the road,
manufacturers continue to investigate new ways to apply it, taking
advantage of the styling possibilities afforded by LED headlights
and taillights. Page 19.
Dan Sheingold [[email protected]]
Scott Wayne [[email protected]]
2
Data sheets for all ADI products can be found by entering the part
number in the search box at www.analog.com.
October
Amplifier, instrumentation, micropower ............................... AD8235
Controller, dc-to-dc, step-down, 2-output ......................... ADP1877
Sensor, inertial, six-degrees-of-freedom . ......................... ADIS16362
Supervisor, voltage, 4-channel, ±supplies ........................ ADM12914
November
ADC, SAR, 16-bit, 2.5-MSPS ............................................... AD7985
ADC, SAR, 16-bit, 10-MSPS, 1.5-LSB INL ......................... AD7626
Amplifier, operational, quad, rail-to-rail ........................ ADA4692-4
DAC, TxDAC+, dual, 16-bit, 1.2-GSPS . ............................... AD9122
Gain Blocks, RF/IF, 4-GHz .............................. ADL5601/ADL5602
Generator, clock, 10-output, Ethernet .................................. AD9571
Gyroscope, yaw-rate, digital-output ................................ ADIS16265
Power Supply, 1.2-A, 16-bit level-setting DACs .................... AD5560
Regulator, low-dropout, 150-mA ......................................... ADP150
Regulator, low-dropout, dual, 200-mA .............................. ADP5030
Transceiver, RS-485, 16-Mbps, full-duplex . .................. ADM1490E
December
Amplifier, difference, dual, gain of ½ or 2 . ........................... AD8279
Amplifier, difference, precision, high-voltage ........................ AD8209
Amplifier, operational, micropower, rail-to-rail . ............. ADA4051-1
Amplifier, operational, zero crossover distortion ................... AD8505
Amplifier, ultralow-noise, dual, selectable gain ..................... AD8432
ADC, SAR, 14-bit, 2.5-MSPS ............................................... AD7944
ADC, sigma-delta, 8-channel, 24-bit, 4.8-kHz ...................... AD7194
ADCs, pipelined, 10-/12-/14-bit,
80-MSPS . .............................................. AD9609/AD9629/AD9649
ADCs, pipelined, 14-/16-bit, 125-MSPS . ................ AD9255/AD9265
Codecs, audio, stereo, 24-bit, 96-kHz . ......... ADAU1381/ADAU1382
Controller, digital, isolated power supply ......................... ADP1043A
Controllers, synchronous buck, 20-V . ............... ADP1872/ADP1873
Driver, current/voltage, programmable . ................................ AD5751
Driver, laser-diode, differential, 11.3-Gbps ........................ ADN2531
Drivers, MOSFET, dual 2-A . ............ ADP3629/ADP3630ADP3631
Filter, video, 6-channel, SD/ED/HD .............................. ADA4424-6
Gyroscopes, yaw-rate, digital-output ........... ADIS16260/ADIS16265
Isolators, digital, 4-channel, 1-kV rms isolation . ............. ADuM744x
Mixer, balanced, 500-MHz to 1700-MHz .......................... ADL5367
Mixer, balanced, 1200-MHz to 2500-MHz ........................ ADL5365
Mixer, balanced, dual, 1200-MHz to 2500-MHz ............... ADL5356
Multiplexer, CMOS, 4-channel, differential, 4.5-Ω ........... ADG1439
Multiplexer, CMOS, 8-channel, 9.5-Ω .............................. ADG1438
Multiplexer, CMOS, 8-channel, differential, 4.5-Ω ........... ADG1607
Multiplexer, CMOS, 16-channel, 4.5-Ω ............................ ADG1606
Prescaler, divide-by-4, 4-GHz to 18-GHz ......................... ADF5001
Receivers, IF, dual/quad ........................................ AD6642/AD6657
Regulators, very-low-dropout, 300-mA ................ ADP122/ADP123
Reset Circuits, microprocessor,
ultralow-power . ............................................. ADM632x/ADM634x
Rheostats, digital, 1024-/256-position, 1% accuracy ............. AD527x
Switch, digital crosspoint, 4.25-Gbps, 16 × 16 ................... ADN4604
Switch, iCMOS, octal SPST, 9.5-Ω .................................... ADG1414
Switch, iCMOS, SPDT, 2.1-Ω . .......................................... ADG1419
Switches, iCMOS, dual SPST, 2.1-Ω ................................. ADG142x
Switches, iCMOS, SPST, 1-Ω ........................... ADG1401/ADG1402
Transceiver, multiband, 3G femtocell ................................ ADF4602
Transceivers, RS-485,
isolated signal and power .......................... ADM2582E/ADM2587E
Analog Dialogue, www.analog.com/analogdialogue, the technical magazine of
Analog Devices, discusses products, applications, technology, and techniques
for analog, digital, and mixed-signal processing. Published continuously
for 43 years—starting in 1967—it is currently available in two versions.
Monthly editions offer technical articles; timely information including recent
application notes, new-product briefs, pre-release products, webinars and
tutorials, and published articles; and potpourri, a universe of links to important
and relevant information on the Analog Devices website, www.analog.com.
Printable quarterly issues feature collections of monthly articles. For history
buffs, the Analog Dialogue archive includes all regular editions, starting with
Volume 1, Number 1 (1967), and three special anniversary issues. If you wish
to subscribe, please go to www.analog.com/analogdialogue/subscribe.html.
Your comments are always welcome; please send messages to dialogue.
[email protected] or to Dan Sheingold, Editor [[email protected]]
or Scott Wayne, Publisher and Managing Editor [[email protected]].
ISSN 0161-3626 ©Analog Devices, Inc. 2011
By John Wynne
Cars with wheels driven by battery-powered electric motors,
continuously or intermittently, have become a hot topic. These
“green” vehicles rely on batteries of series-connected cells to obtain
sufficiently high voltage to operate the motor efficiently. Such
high-voltage (HV) stacks are used in all-electric vehicles (EV)—as
well as hybrid electric vehicles (HEV), which rely on an internalcombustion engine (ICE) for charging and (in many cases) shared
propulsion. EVs must be plugged into a power source for charging;
some newer hybrids are designed as plug-in hybrid electric vehicles
(PHEV), which are considered to be essentially EVs with an ICE
for range extension.
HV stacks are already used in many industries and applications
outside of the transportation industry—typically in uninterruptible
power supplies (UPS) to store energy from the grid in dc form;
as emergency dc supplies in 48-V communications equipment; as
emergency supplies in crane and lift systems; and in wind turbines
for feathering the blades in an emergency. Although we discuss
here the use of battery stacks in vehicles, the underlying issues are
common to all types of stacks.
Battery stacks for transportation can typically involve 100 or more
cells, providing hundreds of volts. Since it is generally accepted
that more than 50 V or 60 V can prove lethal to human beings,
and even lower voltages can damage electronic equipment—
considering the stability concerns about cells using some types of
electrochemical reactions—safety is a key concern. Although these
stacks are inherently dangerous, they must still communicate with
the cell monitoring electronics, which are usually located within
the battery enclosure. Thus, the communications method must
be safe and reliable.
of window comparators. It is beyond the scope of this article to
discuss these products in any depth, but it is worth noting how such
devices communicate in a stack configuration. Each cell establishes
the common-mode level for the measurement input from the one
above it. A daisy-chain interface allows each individual AD7280 in
a stack to communicate directly with the next AD7280 above it or
below it (and thus pass digital information up or down the stack)
without needing isolation. The SPI interface of the bottommost
AD7280 is used to exchange data and control signals for the
whole stack with the system microcontroller. It is at this point that
high-voltage galvanic isolation must be employed to protect the
low-voltage electronics elsewhere in the system.
CELL
MONITOR
BIDIRECTIONAL
DAISY-CHAIN BUS
BETWEEN PACKS
CELL
MONITOR
TOTAL BATTERY
STACK VOLTAGE
Isolation in High-Voltage
Battery Monitoring for
Transportation Applications
ISOLATION
SAFETY
SWITCH
CELL
MONITOR
BIDIRECTIONAL
COMMUNICATION BUS
CELL
MONITOR
ISOLATION
MICROPROCESSOR
WITH CAN
CONTROLLER
CAN
BUS
Organizing Cells in HV Stacks
The original equipment manufacturer generally specifies the
physical packing of the cells into enclosures called packs, which
typically contain from six to 24 cells in series. Packs containing
large numbers of cells are physically larger and more awkward to fit
into typical vehicle spaces. The cell-monitoring integrated circuits
associated with the cells are physically close to the monitored
cells and are powered by the cells themselves. Whether it is
essential to monitor the voltage of each cell depends on the cell
chemistry. For instance, the behavior of HV stacks based upon
nickel-metal hydride (NiMH) chemistry is very well understood,
and generally no effort is made to measure individual cell voltages;
it is sufficient to measure the total voltage of all the cells within
a particular pack. With stacks based upon lithium-ion (Li-Ion)
chemistry, however, it will be necessary to monitor the voltage
of each cell to detect an over- or undervoltage condition on any
individual cell in the string. It is not generally necessary to measure
the temperature of each Li-Ion cell, but the facility to do so should
be available. The electronics for monitoring a NiMH stack are thus
considerably simpler than those for a Li-Ion stack. Figure 1 shows
a common approach to building and monitoring an HV stack.
Cell-monitor ICs typically handle six or 12 cells. Currently, two
application-specific special-purpose (ASSP) products are available
from Analog Devices for cell monitoring: the AD7280,1 intended
for use as a primary monitor, is based on a high-speed multiplexed
12-bit analog-to-digital converter; another device, intended
for use as a backup, or redundancy monitor, is based on a series
Analog Dialogue Volume 43 Number 4
CURRENT
SHUNT
SIGNAL
CONDITIONING
ISOLATION
SIGNAL
REPRESENTING
STACK CURRENT
Figure 1. Serial cell monitoring and isolation in a battery stack.
In Figure 1, the string of serially connected cells has a switch or
contactor placed in the middle of the string. Normally, this switch
is closed at all times, whether the vehicle is in normal operation or
parked. For vehicle maintenance or in emergency situations the
switch is physically pulled or removed from its position to disable
the stack voltage from appearing at the stack terminals. In order
not to compromise the isolation provided by the open switch,
it is important not to have any electronic components bridging
the switch terminals. Thus, the top half of the stack should be
electrically isolated from the bottom half with the switch open.
This means that cell data from the top half of the stack must be
communicated via its bottommost cell monitor across an isolation
barrier to the microprocessor or microcontroller that is managing
the flow of data into and out of the complete stack. Similarly,
the bottom half of the stack must also be isolated from this
microprocessor or microcontroller, so it has an identical isolation
barrier to that of the top half.
In addition to the cell monitors, a current monitor is positioned
somewhere in the stack to measure and report on the stack
current. This monitor is generally placed at the bottom of the
stack; it also needs to be considered for isolation. Hall-effect
3
current sensors have inherent galvanic isolation and need no
further isolation circuitry. If, however, the current sensor uses a
shunt element, the associated shunt monitor circuitry will require
an individual isolation barrier. Current sensing using shunts
is becoming very popular; it is much more stable and accurate
than, yet price competitive with, Hall-effect sensing. The
use of low-value shunt resistors with low-cost high-resolution
monitoring electronics—such as the AD820x and AD821x
families of AEC-Q100 qualified current shunt monitors, which
have shipped over 100M channels into automotive sockets to
date—minimizes self-heating, a traditional objection to this
approach. Thus, the system in Figure 1 requires three separate
isolation barriers, unless the current-sense monitor can feed into
the bottommost cell monitor, sharing its isolation barrier.
Another popular approach to organizing cells in a battery stack is to
group the battery packs into a series of electrically separate clusters
(Figure 2). The bottommost monitor of each cluster communicates
the local cell conditions across a dedicated isolation barrier back
to the microcontroller on the nonisolated side.
CELL
MONITOR
ISOLATION
ISOLATED
COMMUNICATION BUS
TOTAL BATTERY
STACK VOLTAGE
CELL
MONITOR
ISOLATION
SAFETY
SWITCH
CELL
MONITOR
ISOLATION
CELL
MONITOR
ISOLATION
MICROPROCESSOR
WITH CAN
CONTROLLER
CAN
BUS
discharges or surges to enter a piece of equipment—and for
interfering signals to escape, either by conduction of the spurious
signals on the I/O lines or by radiation from the I/O cable. Adding
more cables to a battery stack can reduce its EMC performance
significantly unless careful attention is paid to the robustness of
the signals and the communication protocol chosen. Because of
this, the EMC performance of the I/O device connected to the
port is crucial to the EMC of the entire equipment.
The popular SPI communication protocol is suitable for
communicating between devices on the same printed circuit board
(PCB); but single-ended signals can be difficult to transmit reliably
over 24 to 36 inches of wire, especially in a noisy environment.
Where digital signals are to be transmitted off board, prudent system
design might include differential transceivers, such as the ADM485.
These transceivers can be powered from the low-side power source,
so no power is drawn directly from the cells in the stack.
Isolation Technology Is Key to Stack Communications
For battery stack voltages to get higher in order to satisfy the
demands of higher power electric motors found in heavier
private vehicles, as well as light delivery trucks and vans, the
number of cells in battery stacks must increase. In addition to
increased numbers of serially connected cells, many battery
packs now contain paralleled strings of cells in order to
increase the ampere-hour (AH) capacity of the overall battery
pack. The cells of each parallel string must be monitored—
resulting in the collection of a lot of data. The cell monitor
data associated with all of these cells must be transmitted
back to the battery-measuring-system (BMS) microcontroller
reliably and within the system loop time requirements set by
the system integrators.
Accordingly, the difficulties associated with providing reliable data
communications across system-to-system boundaries have also
increased. A key element to providing reliable communications across
so many isolated boundaries inside a typical battery stack is automotivequalified isolation technology, now available from Analog Devices.
The basis of the technology is magnetic isolation, with transformers
fabricated in a planar fashion using cost-effective standard CMOS
processes (see Figure 3). This facilitates the integration of multiple
isolation channels into a single component—or the integration of
isolation channels with other semiconductor functions, such as line
drivers and analog-to-digital converters (for example, the AD7400
isolated ∑-∆ modulator).
16 VDD2
VDD1 1
CURRENT
SHUNT
SIGNAL
CONDITIONING
ISOLATION
SIGNAL
REPRESENTING
STACK CURRENT
Figure 2. Battery stack with parallel access to packs.
The increased use of digital isolators makes this approach
somewhat more expensive than the system shown in Figure 1, but
it offers the possibility of reading back all the cell data in a much
shorter time, with all cell clusters simultaneously being asked
to report on what the cell monitors are seeing within the packs.
Another important benefit is that it allows backup monitoring to
continue in the presence of problems developing with the daisy
chain, such as broken wires or poor connector contacts. The data
from “off-the-air” packs can still be determined by correlating the
remaining pack voltages with the overall stack voltage.
It does require more cabling, which can be problematic, since
up to 75% of electromagnetic-compatibility (EMC) problems are
considered to occur in relation to input/output (I/O) ports. The I/O
port is an open gateway for electrostatic discharges or fast transient
4
GND1 2
15 GND2
VIA 3
ENCODE
DECODE
14 VOA
VIB 4
ENCODE
DECODE
13 VOB
VOC 5
DECODE
ENCODE
12 VIC
VOD 6
DECODE
ENCODE
11 VID
VE1 7
GND1 8
10 VE2
9 GND2
Figure 3. Functional block diagram of ADuM1402 quad isolator.
These iCoupler® digital isolators that, unlike optocouplers, do
not degrade over the lifetime of the vehicle can accommodate
the harsh operating conditions often encountered through the
changing seasons. The recently released family of devices listed in
Table 1—AEC-Q100 qualified to 125°C—uses the same materials
as its well-established counterparts in the ADI family of iCoupler
products, with more than 300 million channels of isolation shipped
to date. The 2-channel, 3-channel, and 4-channel digital isolator
Analog Dialogue Volume 43 Number 4
Table 1. AEC Q100-Qualified iCoupler Isolators
Part Number
2.5 kV rms Isolation
ADuM120xA/WS
ADuM120xB/WT
ADuM120xC/WU
ADuM130xA/WS
ADuM130xB/WT
ADuM140xA/WS
ADuM140xA/WS
Reverse
Direction
Options
Total
Number
of
Channels
2
3
4
0
1
2
•
•
•
•
•
•
•
•
•
•
—
—
—
—
—
•
•
•
•
•
•
Output
Max
Max
Data Propagation
Rate
(Mbps) Delay (ns)
1
10
25
1
1
1
10
150
50
45
100
32
100
50
ADuM1300W
•
—
—
—
—
—
—
—
—
—
—
•
•
•
•
•
•
•
Supply
Range
(V)
Max
Temperature
(°C)
Package
Price
($U.S.)
3.0 to 5.5
3.0 to 5.5
3.0 to 5.5
3.0 to 5.5
3.0 to 5.5
3.0 to 5.5
3.0 to 5.5
125
125
125
125
125
125
125
8-lead SOIC_N
8-lead SOIC_N
8-lead SOIC_N
16-lead SOIC_W
16-lead SOIC_W
16-lead SOIC_W
16-lead SOIC_W
1.21/2.13
1.76/3.11
2.44/4.30
1.61/2.42
2.42/3.62
2.15/3.22
2.22/4.82
used for signal isolation, power can now be transferred across an
isolation barrier—allowing fully integrated isolation for remotely
powering the data isolators in the battery packs. Local power is
supplied to an oscillating circuit that switches current through a
chip-scale air core transformer. Power transferred to the isolated
side is rectified and regulated to either 3.3 V or 5 V. The isolatedside controller provides feedback regulation of the output by
creating a PWM control signal that is sent back to the local side
by a dedicated iCoupler data channel. The PWM control signal
modulates the oscillator circuit to control the power being sent to
the isolated side. The use of feedback permits significantly higher
power and efficiency.
The ADuM540xW devices are 4-channel digital isolators that
include an isoPower®, integrated, isolated dc-to-dc converter, which
provides up to 500 mW of regulated, isolated power at either 5.0 V
from a 5.0-V input supply or 3.3 V from a 3.3-V supply. As with
the standard iCoupler devices, a variety of channel configurations
and data rates is available. Because an isoPower device uses
high-frequency switching elements to transfer power through its
transformer, special care must be taken during PCB layout to
meet emissions standards. Refer to AN-0971 Application Note,
Recommendations for Control of Radiated Emissions with isoPower
Devices, for details on board-layout considerations. The ADuM540x
family is currently undergoing AEC-Q100 qualification.
References
1
I n for mat ion on all A DI components ca n be fou nd at
w w w.analog.com.
One of the most exciting developments of iCoupler technology is
the integration of both power transmission and signal transmission
within the same package. With microtransformers similar to those
ADuM1201W
Z
•
Two of the most distinguishing features of the iCoupler technology
are the ability to support high data rates and to operate with low
supply currents. The supply current drawn by an iCoupler channel
is largely a function of the data rate it is carrying. For 3-V operation,
the total power supply current—for both sides and all four channels
of the ADuM140xWS—is 1.6 mA typical (4 mA maximum) at a
data rate up to 2 Mbps. Low-power operation is important since,
on the isolated or “hot” side of the ADuM140xWS, the power
comes from the cells themselves through a voltage regulator. The
monitors are also powered from this same voltage source, so the less
power taken by all elements of the monitoring and communicating
circuitry the better. All isolation products are available in small,
low-profile, surface-mount 8-lead SOIC_W or 16-lead SOIC_W
packages and come with safety certifications from UL, CSA, and
VDE. They feature isolation ratings up to 2.5 kV rms and working
voltages up to 400 V rms.
ADuM1200W
L
•
The planar transformers are inherently bidirectional; therefore,
signals can pass in either direction. All possible combinations of
drive and receive channels within the total number of channels are
available. For instance, the 2-channel ADuM120xW, 3-channel
ADuM130xW, and 4-channel ADuM140xW, alone or together,
offer seven different channel configurations (4-0, 3-1, 2-2, 3-0,
2-1, 2-0, 1-1), ensuring an optimized solution for all situations.
Figure 4 summarizes the various configurations available.
EN
H
•
families in the table have data rates up to 25 Mbps and propagation
delays as low as 32 ns.
iCoupler Technology Begets isoPower Devices: Integrated,
Isolated Power
Default
Author
John Wynne [[email protected]] is a precision converter
marketing manager at Analog Devices.
ADuM1301W
ADuM1400W
ADuM1401W
ADuM1402W
Figure 4. Seven different configurations with the ADuM120xW/ADuM130xW/ADuM140xW.
Analog Dialogue Volume 43 Number 4
5
Synchronizing Device Clocks
Using IEEE 1588 and Blackfin
Embedded Processors
and operating conditions of the individual oscillators would limit
the ability of the clocks to operate synchronously. Some possible
simplistic solutions to address these limitations include:
By Jiang Wu and Robert Peloquin
• All the devices could utilize oscillators with nearly identical
characteristics. This approach is impractical due to the difficulty
of acquiring nearly identical oscillators and keeping them from
drifting apart over time. More importantly, each oscillator will
be subjected to different operating conditions.
Introduction
The IEEE 1588 standard, introduced in 2002, defines a protocol
to synchronize distributed clocks on a network. It is becoming
the preferred clock synchronization method for many different
applications, including test and measurement, telecommunications,
and multimedia streaming. This standardized method for
synchronizing clocks is cost-effective, supports heterogeneous
systems, and provides nanosecond-level synchronization precision.
This article provides an introduction to both the original
IEEE 1588-2002 standard and the enhancements incorporated
as part of the updated IEEE 1588-2008 version. Dedicated
hardware support for IEEE 1588 has been integrated into the
ADSP-BF5181 Blackfin® embedded processor because of the
increasing importance of IEEE 1588 in some of its targeted
applications. An overview of its capabilities is provided, followed
by an example showing clock synchronization performance results
obtained by an ADSP-BF518 processor solution.
What Time Is It?
It is common for a system to need to maintain its own sense of time
using a local oscillator. Figure 1 shows how hardware and software
combine to generate time information within a system.
FREQUENCY
HARDWARE
OSCILLATOR
COUNTER
CONVERSION
SOFTWARE
TIME
HOUR:MIN:SEC
EPOCH
DIVIDER
NUMBER OF
PULSES
H:M:S TIME
API FUNCTIONS
CLOCK OUTPUT
LOCAL TIME SYSTEM
APPLICATION
APPLICATION HARDWARE
APPLICATION SOFTWARE
Figure 1. Local timekeeping.
This time information can be used by both hardware and software
resources within the system. In hardware, one or more physical
clock signals (clock outputs) are derived from the oscillator’s clock
and can be used to drive or trigger other parts of the system. The
time maintained in software is typically referred to as system time.
The system time can be represented in the form of numbers of clock
pulses or in second/nanosecond notation. The system software
derives the time from the number of oscillator clock pulses and
its frequency information, and provides application-programminginterface (API) functions that other parts of the software use to
retrieve and set the time. If an absolute time is desired, the provided
time is associated with a predefined epoch, which identifies a
reference point in time.
Synchronize Your Watches
Many applications require two independent devices to operate in
a synchronized fashion. If each device were to rely solely on its
own oscillator, differences between the specific characteristics
6
• All the devices could use a single physical oscillator. This is only feasible
for distributed systems in close proximity; a high-frequency clock
signal cannot be reliably delivered over a long distance.
• If all the devices are interconnected via a communications network
such as Ethernet, they can dynamically adjust their individual clocks
to a single “master” clock by exchanging time messages over the
network. With network time protocol (NTP), the traditional time
synchronization protocol, every device in the system adjusts
its clock according to the time information it retrieves from
an NTP time server. However, this protocol can only achieve
synchronization accuracy on the order of milliseconds.
IEEE 1588 defines a newer protocol capable of nanosecond
synchronization accuracy. How it can achieve this level of clock
synchronization is discussed in the following sections.
What IEEE 1588 Does
The IEEE 1588 standard defines a protocol for time-synchronizing
devices that are geographically dispersed but interconnected by
some form of communications technology, for example, Ethernet.
By exchanging timing messages between devices they can maintain
the same absolute system time, which is represented in seconds
and nanoseconds.
An intuitive way to achieve this goal is for one device, which has
the “best” (most accurate) clock, and is designated as the masterclock device, to broadcast its time to the other devices. The other
devices will adjust their times to match the time sent by the master
clock. This solution has several limitations, though:
1. The master-clock device cannot broadcast the time at
infinitesimal intervals, so the “slave” clock devices have to use
their own independent and “inferior” oscillators to interpolate
the time points between two broadcasts from the master-clock
device. This results in degraded synchronization during the
time between updates from the master clock.
2. Delays inevitably exist on the broadcast path, with magnitudes
depending on the communications technology—the time that
a physical signal takes to travel along a wire from one device to
another, for example. This delay results in an additional offset
between the master clock and each slave clock.
3. Differences among the broadcast paths between the masterclock device and each slave-clock device will further degrade
the synchronization between individual slave-clock devices.
IEEE 1588 specifies a protocol that solves the second and third
problems by measuring path delay. It also allows the slave clock
to be adjusted to match the master clock’s pace so as to mitigate
the first problem. Where possible, the first problem can be
further reduced by using smaller broadcasting intervals and
higher-quality oscillators.
How IEEE 1588 Measures Communication Delay
IEEE 1588-2002 2 def ines four messages to measure the
communication delay of the forward (master to slave) and
backward (slave to master) paths: Sync, Followup, DelayReq, and
DelayResp. The newer version, IEEE 1588-2008,3 provides further
mechanisms to measure the peer-to-peer delay with three additional
messages: PdelayReq, PdelayResp, and PdelayRespFollowup.
Analog Dialogue Volume 43 Number 4
Among these messages, Sync, DelayReq, PdelayReq, and
PdelayResp, so-called event messages, must be time-stamped
(recording the local time) when they leave and arrive at a device.
There are two techniques to time-stamp packets:
1. Software time-stamp occurs when the messages are handled
by the software. Usually occurring in the message’s receive/
transmit interrupt service routine (ISR), the time-stamp is the
current value of the system time.
2. Hardware time-stamp occurs when the messages physically
arrive at or leave the device. The time-stamp operation is
executed by hardware, which maintains its own continuous
time information.
Either time-stamp method is acceptable in IEEE 1588, but a
hardware time-stamp can provide significantly better precision,
as will be shown below.
Delay from Master-Clock Device to Slave-Clock Device
software has the two times, Ts1' (Sync arrival time) and Tm1'
(Sync departure time). The master-to-slave path delay, Tmsd, is
determined by Equation 1.
Delay from Slave-Clock Device to Master-Clock Device
The DelayReq message is sent by the slave-clock devices, and the
DelayResp message is sent by the master-clock device in response.
With these messages, the slave-clock devices can calculate the
communication path delay from the slave-clock device to the
master-clock device.
At time Ts3 (Figure 3), the slave-clock device software reads
the current local system time (Ts3), inserts it into a DelayReq
message, and sends the message out. After the message is sent,
the slave-clock device software reads the time-stamp to get the
departure time of the message, Ts3', and waits for the response
from the master-clock device.
The messages Sync and Followup are sent by the master-clock
device; it is a slave-clock device’s responsibility to receive them
and calculate the communication path delay from the master-clock
device to the slave-clock device.
SLAVE
SOFTWARE
MASTER
HARDWARE
SLAVE
SOFTWARE
Ts1'
SYNC
CARRIES Tm1
SLAVE GETS Ts1'
Ts1
Ts2'
Ts2
FOLLOWUP
CARRIES Tm1'
SLAVE GETS Tm1'
HARDWARE
TIME-STAMP
POINT
HARDWARE
TIME-STAMP
POINT
SOFTWARE
TIME-STAMP
POINT
Figure 2. Measuring communication delay between
master-clock and slave-clock devices.
After the Sync message has been sent, the master-clock device
software reads the Sync message’s departure time, Tm1', through
the time-stamping unit, inserts it into a Followup message, and
sends that message out at Tm2. This message is received by slaveclock device software at Ts2. At this point, the slave-clock device
Analog Dialogue Volume 43 Number 4
MASTER
RECEIVES DELAYREQ
SAVES Tm3'
Tm3
Tm4
MASTER
SENDS DELAYRESP
WHICH CARRIES Tm3'
Tm4'
Ts4'
Ts4
SLAVE GETS Tm3'
HARDWARE
TIME-STAMP
POINT
HARDWARE
TIME-STAMP
POINT
SOFTWARE
TIME-STAMP
POINT
The DelayReq message arrives at the master-clock device at a later
time, Tm3', and is processed by the master software at Tm3. The
software then reads the time-stamp to get the arrival time, Tm3',
puts it into the DelayResp message, and sends to a slave-clock
device at Tm4. When the slave-clock device software receives
the DelayResp message at Ts4, it can extract the time, Tm3', and
calculate the slave-to-master delay, Tsmd, by Equation 2.
(2)
In both Equation 1 and Equation 2, there is an unknown
variable, the master-slave time difference, Tms. So it is not
possible to get either Tmsd or Tsmd individually. However,
if one makes the usually acceptable assumption that the
communication path is symmetric
SOFTWARE
TIME-STAMP
POINT
MASTER
SOFTWARE
Figure 3. Measuring slave-master communication delay.
SLAVE
HARDWARE
Tm2'
MASTER
HARDWARE
Tm3'
SOFTWARE
TIME-STAMP
POINT
Tm1'
Tm2
MASTER
SENDS
FOLLOWUP
COMMUNICATION
PATH
Ts3'
COMMUNICATION
PATH
MASTER
SENDS SYNC
Tm1
SLAVE
HARDWARE
SLAVE SENDS
DELAYREQ
Ts3
In Figure 2, at time Tm1, the master-clock device software
reads the current local system time (Tm1, the software timestamp), inserts it into a Sync message, and sends the message
out. The message leaves the master-clock device at a later
time, Tm1', which is the hardware time-stamp. It arrives at
slave-clock hardware at Ts1' (slave-clock device local time),
and is received by the slave-clock device software at a later
time, Ts1. The software will read the hardware time-stamp
to get Ts1'. If there is no communication delay, Ts1' should
be equal to (Tm1' + Tms), where Tms is the time difference
between master clock and slave clock. The protocol’s ultimate
goal is to compensate for this difference.
MASTER
SOFTWARE
(1)
(3)
—a key assumption for IEEE 1588 to work correctly—then, adding
Equation 1 and Equation 2 gives
(4)
All these calculations are performed by the slave-clock devices,
since it is they who seek to synchronize themselves to the masterclock device. They get Tm1' from master-clock device’s Followup
message, Ts1' from their Rx (reception) time-stamping, Ts3'
from their Tx (transmission) time-stamping, and Tm3' from the
master-clock device’s DelayResp message.
7
How to Calculate the Time Difference Between a Slave Clock and
Master Clock
Once the communication path delay, Td, is obtained, the slavemaster time difference is easy to calculate, using either Equation 1
or Equation 2, as shown in Equation 5 and Equation 6.
(5)
(6)
How to Adjust the Time of a Slave-Clock Device
With the time difference from the master clock known, each slaveclock device needs to adjust its own local time to match the master
clock. This task has two aspects. First, slave-clock devices need to
adjust their absolute time by adding the time difference to make
their time perfectly match the master-clock time at this moment.
Then, each slave-clock device needs to adjust its clock frequency
to match the frequency of the master clock. We cannot rely on the
absolute time alone, since the time difference is applied only at a
certain period and could be either positive or negative; as a result,
the adjustment will make the slave-clock time jumpy or even run
backward. So, in practice, the adjustment takes two steps.
1. If the time difference is too big, for example, larger than
one second, absolute time adjustment is applied.
2. If the time difference is small, a percentage change of frequency
is applied to slave clocks.
Generally speaking, the system becomes a control loop, where
master-clock time is the reference command, slave-clock time is
the output tracking the master-clock time, and their difference
drives the adjustable clock. PID control, which is commonly
used by many IEEE 1588 implementations, could be used to
achieve specific tracking performance. Figure 4 illustrates this
control loop.
MASTER
TIME
IEEE 1588 CONTROL LOOP
+
CONTROL LAW
–
ADJUSTABLE
CLOCK
SLAVE
TIME
Figure 4. IEEE 1588 control loop.
Peer-to-Peer Delay
The revised version, IEEE 1588-2008, introduces a new
mechanism for measuring path delay, called peer-to-peer (P2P)
delay. By contrast, the master-slave mechanism discussed in the
previous sections is end-to-end (E2E) delay. In an IEEE 1588-2008capable network, a master-clock device can be linked to slave-clock
devices either directly or through multiple hops (stages). The E2E
delay is actually the total delay from a master-clock device to a
slave-clock device, including all the hops in between. However,
the P2P delay is limited to two directly connected devices. The
overall delay along the path is the sum of the P2P delay of all the
hops. From the perspective of preserving path symmetry, the P2P
mechanism provides better accuracy.
As noted earlier, IEEE 1588-2008 includes three additional
messages, PdelayReq, PdelayResp, and PdelayRespFollowup,
to measure P2P delay. They work in a manner similar to that
explained above. Reference 3 provides more details.
Factors Affecting Synchronization Performance
Well-designed IEEE 1588 devices are capable of achieving
highly accurate clock synchronization, but it is important to
recognize the key factors that directly affect performance.
Some of these include:
8
1. Path delay: As noted earlier, the path delay measurement of
IEEE 1588 assumes that the communication-path delays are
symmetrical, that is, the transmission delay of the forward path
is equal to the reverse transmission delay. In addition, the delay
should not vary during the delay measurement. Variation in
delay during measurement will produce asymmetry and delay
jitter, which will have a direct impact on the synchronization
precision. While delay symmetry and jitter cannot be controlled
outside the boundaries of an IEEE 1588 device, both path
symmetry and jitter can be improved within the device if
measurements are based on hardware time-stamping. Hardware
time-stamping eliminates the significant jitter resulting from
software time-stamping—due to interrupt latency, context
switch, and thread scheduling.
2. Drift and jitter characteristics of clocks: The frequency
and phase of the master clock represent the inputs of the
tracking control system, and the slave clock is the control
object. Any time-varying behavior of the master clock will
act as a disturbance to the control system and result in both
steady-state- and transient errors. Clocks with less drift and
jitter will, therefore, improve synchronization accuracy.
3. Control law: The control method determines how the errors in
the slave-clock-device time are corrected in the adjustment of
the slave clock. The control-law parameters, including settling
time, overshoot, and steady-state error, will directly affect clock
synchronization performance.
4. Resolution of the clocks: As shown in Figure 1, the resolution
of the local time is determined by the frequency of the clock; the
minimum increment of time is one period of the clock signal.
The IEEE 1588 protocol runs on a time with a resolution of
1 ns for IEEE 1588-2002 and 2 –16 ns for IEEE 1588-2008. It
is not practical to have a clock of 216(!) GHz (or even 1 GHz).
The quantization of the local clocks is expected to affect the
precision of local time measurement and control.
5. How often Sync messages are issued: The frequency
with which the slave clocks are updated ultimately affects
the precision of synchronization. A longer period usually
leads to larger time errors observed at the next Sync, since
the time error is the integral accumulation of the slave-clock
frequency error.
6. How often delay measurement is conducted: Delay
measurement is performed periodically, at intervals based on
the expectation that the delay does not change significantly
between adjacent samples. If the IEEE 1588 network experiences
large delay variations, then increasing the delay-measurement
frequency will improve clock-synchronization performance.
Which Is the Master-Clock?
Having considered how to accurately determine the time difference
between master-clock devices and slave-clock devices, a relevant
question is how to determine which device, among possibly hundreds
of interconnected devices, will serve as the master clock.
IEEE 1588 defines a method called the best master clock (BMC)
algorithm to choose the master clock device. For this approach,
every device of an IEEE 1588 network maintains a data set
describing the nature, quality, stability, unique identifier, and
preference of its local clock. When a device joins an IEEE 1588
network, it will broadcast the dataset of its own clock and receive
the datasets from all other devices. Using the datasets of all the
participating devices, every device runs the same BMC algorithm
to decide on the master clock and its own future status (master
clock or slave clock). Because the same algorithm is executed
independently by all the devices on the same data, all will come
Analog Dialogue Volume 43 Number 4
to the same conclusion without requiring any negotiation among
them. More information about the details of the BMC algorithm
can be found in References 2 and 3.
ADSP-BF518 Processor’s Support for IEEE 1588
The Analog Devices ADSP-BF518 processor recently joined ADI’s
Blackfin DSP family. Like its predecessor, the ADSP-BF537,4 it
has a built-in Ethernet media-access controller (EMAC) module. Its
capability to support EMAC functionality within the IEEE 1588
standard is extended by an additional TSYNC module, as well as
extra features to support a wide range of IEEE 1588 applications
on Ethernet. Figure 5 shows the block diagram of the TSYNC
module. The ADSP-BF51x Blackfin Processor Hardware Reference
provides additional information.5
CLOCK
OUT
CLOCK OUTPUT
DRIVER
ADSP-BF518
TSYNC MODULE
PPS
GENERATOR
PPS
OUTPUT
EXTERNAL
CLOCK
MII/RMII
CLOCK
SYSTEM
CLOCK
EVENT
FLAG
CLOCK SOURCE
MUX
ADDEND
ADJUSTABLE
CLOCK
LOCAL TIME
COUNTER
ALARM
GENERATOR
EVENT DETECTION
PTP Tx EVENT
PACKET DETECTION
INTERRUPT
MII Rx
PTP Rx EVENT
PACKET DETECTION
INTERRUPT
REQUEST
Rx
TIMESTAMPING
Figure 5. Block diagram of the ADSP-BF518 processor's
TSYNC module.
Packet Detection
The ADSP-BF518 processor can detect and provide hardware
time-stamps for all IEEE 1588 event messages, including both
incoming and outgoing packets. The precision of an IEEE 1588
system depends significantly on both the accuracy of the eventmessage time-stamps and on where they are taken, as these affect
the requirement for symmetry and constancy of path delay. The
ADSP-BF518's TSYNC module keeps monitoring the hardware
interface between the MAC controller and the Ethernet physical
interface transceiver (PHY), that is, the media independent interface
(MII), and produces a hardware time-stamp whenever it detects
an event message—a capability promoting higher synchronization
precision with the ADSP-BF518.
The detection of event messages, designed to be programmable,
can basically be configured to support either IEEE 1588-2002
(default) or IEEE 1588-2008. Furthermore, this programmability
allows for the support of future versions of IEEE 1588, as well as
other general protocols that require time-stamping—including
being configured to time-stamp every Ethernet packet coming
into and out of the processor.
Flexible Clock Sources
The properties of local clocks are important for the performance
of an IEEE 1588 system. To satisfy requirements of a variety of
applications, the ADSP-BF518 processor allows three options for
the local clock source: system clock, external clock, or Ethernet
clock. If the application has a specific clock requirement, it can
choose external clock and provide a customized clock source. The
Ethernet clock option can offer good precision if the master-clock
devices and the slave-clock devices are connected back-to-back,
since the clock is inferred from the Ethernet lines, and the two
Analog Dialogue Volume 43 Number 4
The selected source clock is also driven by the TSYNC module as
an output of the processor, via the specific pin Clockout, to be used
by other parts of the system for local time information.
PPS Output
The pulse-per-second (PPS) signal is a physical representation
of time information. It is nominally a 1-Hz signal with a pulse
at each one-second transition of time. It can be used to control
local devices or to provide an auxiliary time channel in case of
network failure. It can also be used in testing. The phase difference
between PPS signals at two devices is a physical measurement of
their time offset.
The ADSP-BF518 processor provides a flexible PPS output. It
uses a programmable start time (PPS_ST) and period (PPS_P)
to generate a sig nal wit h pulses occur r ing at t he times
( PPS _ ST + n × PPS_P), where n = 1, 2, 3… In the basic
usage, t he PPS sig nal ca n be created si mply by set t i ng
PPS _ P to 1 second, and PPS_ST to any future instant as a
multiple of seconds. This PPS output capability allows for its use
as the reference for generation of a periodic signal with a fully
programmable frequency and start time.
Auxiliary Snapshot
Tx
TIMESTAMPING
MII Tx
devices are running on the same clock. A general application can
take the processor’s system clock as its clock source.
Some applications may need to time-stamp a certain event
indicated by the toggle of a flag signal. The ADSP-BF518's
TSYNC module facilitates this request by providing an auxiliary
snapshot function, using a dedicated pin to accept an external flag.
Toggling the flag will trigger the module to capture the current
local time in a time-stamp register for software to access.
Alarm
If an application needs to execute a task at a specific time, it can
make use of the alarm feature of the TSYNC module. This feature
allows an absolute local time to be set so as to trigger a processor
interrupt when the time arrives. The software can then service
the interrupt and run the task.
Adjustable Clock
The adjustable clock of the TSYNC module is an addend-based
clock. As shown in Figure 6, it takes a fixed input clock signal and
outputs a “pulse-steal” version of the input: the value of addend is
added to the accumulator at each input clock, and each time the
accumulator overflows the carry bit drives the local-time counter,
which gives the local time in terms of the number of pulses counted.
The frequency of the local clock can be adjusted by changing the
addend, since the addend decides how often the accumulator
overflows, and so how often the local-time counter increments. If
the frequency of the input clock is Fin, and the value of addend is
A, then the local clock frequency will be
(7)
ADDEND-BASED ADJUSTABLE CLOCK
32-BIT
ADDEND
CARRY
INPUT CLOCK
32-BIT
ACCUMULATOR
SUM
64-BIT
LOCAL TIME
COUNTER
SUM
32-BIT
OFFSET
Figure 6. Addend-based adjustable clock.
9
Implementation of IEEE 1588 on the ADSP-BF518 Processor
A complete IEEE 1588-2008-compliant system was built on an
ADSP-BF518 processor as shown in Figure 7.
ADSP-BF518 IEEE 1588 systems. 6938 measurements were taken
over a period of approximately 1700 seconds. The resulting mean
error is 0.015 ns, and the standard deviation is 12.96 ns. A Sync
message interval of 0.25 seconds was used for this test.
Conclusion
APPLICATION
ADSP-BF518
PROCESSOR
IEEE 1588-2008 STACK SOFTWARE
TSYNC MODULE
DRIVER SOFTWARE
MAC CONTROLLER
DRIVER SOFTWARE
TSYNC MODULE
HARDWARE
MAC CONTROLLER
HARDWARE
PROCESSOR
SYSTEM CLOCK
PPS
OUTPUT
The IEEE 1588 standard provides a highly accurate, low-cost
method for synchronizing distributed clocks. While hardware
support is not explicitly required for IEEE 1588, hardware-assisted
message detection and time stamping is critical to achieve the
highest level of synchronization precision. The ADSP-BF518
processor provides hardware support for both IEEE 1588-2002
and IEEE 1588-2008, including features that can support a wide
range of applications. High-precision clock synchronization has
been demonstrated by implementing IEEE 1588 technology using
the ADSP-BF518 processor and the IXXAT IEEE 1588-2008
protocol software.
References
1
ADSP-BF518 data sheet. http://www.analog.com/en/embeddedprocessing-dsp/blackfin/adsp-bf518/processors/product.html.
2
IEEE Std. 1588-2002. IEEE Standard for a Precision Clock
Synchronization Protocol for Networked Measurement and Control
Systems. http://ieee1588.nist.gov/PTTI_draft_final.pdf.
ETHERNET PHY
IEEE 1588 PTP
SYSTEM
3
ETHERNET
IEEE Std. 1588-2008. IEEE Standard for a Precision Clock
Synchronization Protocol for Networked Measurement and Control
Systems. http://ieee1588.nist.gov.
Figure 7. An implementation of IEEE 1588 on the ADSP-BF518.
The TSYNC module of the processor detects incoming and
outgoing IEEE 1588 messages and uses hardware to time-stamp
event messages. The IEEE 1588 stack software, provided
by IXXAT (IXXAT Automation GmbH), implements the
message-exchange protocol required by the standard. It makes
use of the TSYNC driver to read, write, and adjust the TSYNC
clock, and uses the MAC controller driver to send and receive
messages on the Ethernet MAC layer (Layer 2 of the OpenSystems Interconnection Model). It also implements the control
law and filtering of P2P delay measurements. The Ethernet PHY
is National Semiconductor DP83848,6 chosen because of its low
jitter delay characteristics. For simplicity, the processor’s system
clock (80 MHz) was chosen to be the TSYNC module clock source.
1400
NUMBER OF SAMPLES
1200
1000
800
600
400
200
0
–60
–40
–20
0
20
40
60
ERROR (ns)
Figure 8. Histogram of slave-clock error of an IEEE 1588
system on ADSP-BF518.
Figure 8 shows the clock synchronization performance of the
device as a histogram of the measured error between two identical
10
4
ADSP-BF537 data sheet. http://www.analog.com/en/embeddedprocessing-dsp/blackfin/adsp-bf537/processors/product.html.
5
ADSP-BF51x Blackfin Processor Hardware Reference Preliminary,
Revision 0.1 (Preliminary). January 2009. Analog Devices, Inc.
http://www.analog.com/static/imported-files/processor_manuals/
bf51x_hwr_rev_0-1.pdf.
6
AN-1507: DP83848 and DP83849 100Mb Data Latency. 2006.
National Semiconductor Corporation. http://www.national.com/
an/AN/AN-1507.pdf.
Authors
Dr. Jiang Wu [[email protected]] joined
A nalog Dev ices i n 2006, where he works
on real-time embedded systems and digital
processing. In 2004, he received a PhD in
electrical engineering from the University of
Rhode Island. In 1996 and 1993, respectively, he
received MS and BS degrees in automation, both
from the Beijing Institute of Technology. His
graduate research work was on modeling and instrumentation
of biomedical systems. Prior to ADI, Jiang worked as a system
engineer at Vivoda Communications.
Robert Peloquin [[email protected]]
joined Analog Devices in 2000. As a senior
applications engineer, he contributes to many
different aspects of ADI's digital signal processor
solutions. Some recent areas include Ethernet
AVB and IEEE 1588. Prior to joining ADI, Bob
worked at the Naval Undersea Warfare Center
performing applied research for unmanned
undersea vehicle technologies. He holds a BSEE degree from
Syracuse University and a MSEE degree from the University
of Massachusetts, Dartmouth.
Analog Dialogue Volume 43 Number 4
Using Isolated RS-485 in DMX512
Lighting Applications
By Hein Marais
Introduction
The western theatrical age started in Greece circa 500 B.C.
Open-air theaters used the sun as their main lighting source, but
often added lanterns to indicate nighttime scenes. These early
lighting applications have evolved into the more complex systems
available today.
Modern lighting equipment includes dimmers, flashing lights,
moving lights, colored lights, and gobos1 (GOes Before Optics).
These lighting systems are often controlled over long distances—up
to 4000 feet—using the DMX512 communications protocol.2
What is DMX512?
DMX512 is a standard developed by the Entertainment Services
and Technology Association (ESTA).3 This standard describes
a method of digital data transmission between controllers and
controlled lighting equipment and accessories, including dimmers
and related equipment. The physical electrical interface is specified
by EIA/TIA-485,4 also known as RS-485.
The DMX512 standard, which specifies an 8-bit asynchronous
serial protocol and 250-kbps data rate, is designed to carry
repetitive control data from a single controller to one or multiple
receivers. The control data on the primary link consists of up to 513
slots that are sent in packets over a balanced transmission line.
Information, in an 8-bit format, is sent sequentially to the various
nodes. Values range from 0 to 255, where 0 indicates the off
condition and 255 indicates the on condition. A break condition
lasting two frames indicates the start of a sequence of 512 values.
A high level for at least 8 µs indicates the start of the first byte.
Discrete DMX512 System Configuration
A DMX512 port consists of four signals (Data1+, Data1–, Data2+,
and Data2–) plus a common reference. The primary data link is
formed by Data1+ and Data1–. An optional secondary data link is
formed by Data2+ a nd Data2–. The paired links are configured
as a bidirectional half-duplex RS-485 communications network.
The DMX512 standard specifies a system using ground-referenced
transmitting devices and isolated receiving devices. The receiving
devices are isolated to protect expensive lighting equipment from
harmful current surges.
Figure 1 shows a discrete implementation of an isolated DMX512
receiver. An isolated power supply is generated by a transformer
driver driving the primary side of a transformer. The output of
the transformer is rectified and regulated to create an isolated 5-V
supply. The data and control signals for the RS-485 transceiver
are isolated using optical isolation.
The ADM485 RS-485 transceiver converts a control signal
received on the A and B pins into a serial output on the RXD
pin. This signal is optically isolated and connected to the UART
input of the ADuC7020 precision analog microcontroller. The
ADuC7020 software decodes the message and outputs logic-level
signals to a digital-to-analog converter (DAC). The ADTL084
JFET-input op amps buffer the DAC outputs to provide 0 V to
10 V signals.
The ADuC7020 software sends a response to indicate that
the message was received correctly. This signal is optically
isolated from the ADM485, which outputs a signal on the A
and B pins.
DMX512 System Configuration Using the ADM2487E
Figure 2 shows a simpler, more integrated, isolated DMX512
receiver using the ADM2487E isolated RS-485 transceiver. An
on-chip oscillator outputs a pair of square waves that drive an
external transformer to provide the 3.3-V isolated power required
by the bus side. The logic side of the device is powered with a
3.3-V or 5-V supply. The ADM2487E employs iCoupler ®5 digital
isolation technology to combine a 3-channel isolator, a 3-state
differential line driver, and a differential input receiver into a
single package. TXD, RXD, DE, and RE pins connect directly
to the ADuC7020 UART.
The ADM2487E transceiver, suitable for half-duplex or fullduplex communication on multipoint transmission lines,
operates with data rates up to 500 kbps. It provides 2.5-kV rms
isolation—certified to Underwriters Laboratory (UL) and VDE
standards—and ±15-kV ESD protection.
VIN
ISOLATION
BARRIER
ISO +5V
OUT
10𝛍𝛍F
IN
ADP3330
+
GND SD
22𝛍𝛍F
0.1𝛍𝛍F
10𝛍𝛍F
MLC
5V
VCC
D1
TRANSFORMER
DRIVER
D2
EN
ON/OFF
GND
DAC0
LIGHTING
CONTROL
SIGNALS
VOUT A 0V TO 10V
DAC1
VOUT B
DAC2
VOUT C
DAC3
VOUT D
HIGH SPEED OPTO
ADTL084
TxD
VCC
TxD
LOW SPEED OPTO
A
RS-485
TRANSCEIVER
ADP485
B
GND
RTS
DE
RE
RxD
VCC
ADuC7020
HIGH SPEED OPTO
RxD
Figure 1. Discrete DMX512 receiver block diagram.
Analog Dialogue Volume 43 Number 4
11
+5V ISO
OUT
IN
ADP3330
GND2
CENTER-TAPPED
TRANSFORMER
RECTIFIER
LDO
VDD1
22𝛍𝛍F
ERR GND SD
10𝛍𝛍F
TANTALUM
100nF
GND2
GND2
D2
VDD2
DATA1+
Y
DATA1–
Z
100nF
DECODE
VDD
RTS
ENCODE
DECODE
RS-485 TRANSCEIVER
ISOLATION
BARRIER
VOUT B
ADuC7020
DAC2
VOUT C
DAC3
VOUT D
RxD
ENCODE
DIGITAL ISOLATION
GND2
DAC1
TxD
A
R
DAC0
LIGHTING
CONTROL
SIGNALS
VOUT A 0V TO 10V
ADTL084
ENCODE
DECODE
B
GND1
D1
ADM2487E
D
5V/3.3V
LOCAL
POWER
SUPPLY
iCOUPLER
RE
GND1
GND1
Figure 2. ADM2487E DMX512 receiver block diagram.
For half-duplex operation, the transmitter outputs and
receiver inputs share the same transmission line by externally
linking transmitter output Pin Y to receiver input Pin A, and
transmitter output Pin Z to receiver input Pin B. Designed for
balanced transmission lines, the ADM2487E complies with
TIA/EIA 485 A-98 and ISO 84826:1993 standards.
Current-limiting and thermal shutdown features protect against
output short circuits and situations where bus contention might
cause excessive power dissipation. Fully specified over the –40°C to
+85°C industrial temperature range, the ADM2487E is available
in a 16-lead, wide-body SOIC package.
T he A DM2487E integ rates galvanic isolation wit h t he
transceiver to create robust protection against harmful current
surges. An ideal solution to implement an isolated DMX512
receiver, it will reduce the overall form factor, improve reliability,
and increase robustness.
References
(Information on all ADI components can be found at www.analog.com)
1
http://en.wikipedia.org/wiki/Gobo_(lighting).
2
http://en.wikipedia.org/wiki/DMX512.
3
http://en.wikipedia.org/wiki/Entertainment_Services_and_
Technology_Association.
The ADM2487E features an open- and short-circuit fail-safe
receiver input design, eliminating the need for external biasing
resistors. The low-input-current receiver design (125 µA) enables
up to 256 nodes to be connected on the same bus.
4
Replacing the discrete DMX512 receiver implementation with
the ADM2487E implementation provides a space-saving, more
robust, more reliable system.
6
Sudden voltage surges between two or more interconnected circuits
can cause damage to expensive equipment. Electrical isolation is
designed to protect expensive equipment from these voltage surges.
Because of the distance between nodes, DMX512 transmitters and
receivers require different power supplies. This will increase the
impedance of the earth ground, making it more likely that ground
currents from other sources will find their way into the link’s ground
wire. Isolating the link reduces or even eliminates these problems.
Galvanic isolation, shown in Figure 3, is required if there is no
guarantee that the potentials at the earth grounds of different nodes
in the system are within the common-mode range of the receiver.
POINT B
ISOLATOR
INFORMATION FLOW
NO CURRENT FLOW
ISOLATION BARRIER
PROTECT HUMANS/
EQUIPMENT
ELIMINATE GROUNDING
PROBLEMS
IMPROVE SYSTEM
PERFORMANCE
Figure 3. Galvanic isolation allows information flow, prevents
ground-current flow.
12
www.analog.com/en/interface/digital-isolators/products/CU_
over_iCoupler_Digital_Isolation/fca.html.
www.iso.org/iso/catalogue_detail.htm?csnumber=20954.
Further Reading
Why Is Isolation Important?
POINT A
http://en.wikipedia.org/wiki/EIA-485.
5
ANSI E1.11 – 2008: Entertainment Technology—USITT
DMX512-A, Asynchronous Serial Digital Data Transmission
Standard for Controlling Lighting Equipment and Accessories.
USITT DMX512, DMX512/1990, DMX512-A.
A N - 9 6 0 A p p l i c a t i o n N o t e , R S - 4 8 5 / R S - 4 2 2 C i r c u i t
Implementation Guide.
Author
Hein Marais [[email protected]]
g raduated f rom t he Un iver sit y of
Stellenbosch, South Africa, in 2001 with a
bachelors degree in electronics. He started
his career in the South African Air Force
working on electronic warfare systems. In
2003, he joined Grintek Communication
Systems, where he was involved in the
de si g n of sof t w a r e a nd h a rdw a r e for
military radios. In 2007, Hein joined ADI,
where he is currently a product applications engineer for the
High Speed Interconnect Group, working on RS-485 and
LVDS products.
Analog Dialogue Volume 43 Number 4
Designing High-Performance
Phase-Locked Loops with
High-Voltage VCOs
2. Phase frequency detector (PFD): Derives the phase-error signal
from the reference and feedback signals.
By Austin Harney
5. VCO: Outputs a frequency that depends on the voltage
presented to its tuning port (Vtune). The VCO has gain,
K V, expressed in MHz/V. The basic VCO expression
relating output frequency to the input control voltage is
f o = f c + Kv (Vtune), where fc is the VCO offset frequency.
Introduction
The phase-locked loop (PLL) is a fundamental building block
of modern communication systems. PLLs are typically used to
provide the local-oscillator (LO) function in a radio receiver or
transmitter; they are also used for clock-signal distribution and
noise reduction—and, increasingly, as the clock source for highsampling-rate analog-to-digital (A/D) conversion.
As feature size has shrunk in integrated-circuit processing,
device supply voltages, including supplies for PLLs and other
mixed-signal functions, have followed downward. However,
the practical technology for voltage-controlled oscillators (VCOs),
a critical element of PLLs, has not decreased as rapidly. Many
high-performance VCO designs are still implemented with discrete
circuitry that may require supply voltages as great as 30 V. This
imposes a challenge for today’s PLL or RF system designer: to
interface the low-voltage PLL IC with a higher voltage VCO. The
level-shifting interface is typically implemented using active filter
circuitry—to be discussed below.
This article will consider the basics of PLLs, examine the current
state of the art in PLL design with high voltage VCOs, discuss
the pros and cons of typical architectures, and introduce some
alternatives to high-voltage VCOs.
PLL Basics
A phase-locked loop (Figure 1) is a feedback system in which a
phase comparator or detector drives a VCO in a feedback loop to
make the oscillator frequency (or phase) accurately track that of
an applied reference frequency. A filter circuit is typically required
to integrate and smooth the positive or negative error signal—and
promote loop stability. A frequency divider is often included in the
feedback path to establish the output frequency (within the range
of the VCO) as a multiple of the reference frequency. The divider
can be implemented so that the frequency multiple, N, will be
either an integer or a fractional number, characterizing the PLL
as an integer-N PLL or a fractional-N PLL.
FREF
PHASE
DETECTOR
LOWPASS
FILTER
VCO
N × FREF
÷N
COUNTER
Figure 1. A basic phase-locked loop.
Because a PLL is a negative-feedback control loop, the frequency error
signal will be forced to zero at equilibrium to produce an accurate and
stable frequency of N × FREF at the output of the VCO.
PLLs are implemented in various ways, using all-digital, all-analog,
or combined circuitry, depending on the required frequency range,
noise and spurious performance, and physical size. At present, the
architecture of choice for high-frequency, or RF, PLLs combines
all-digital blocks, such as feedback dividers and phase detectors,
with high-precision analog circuits, such as charge pumps and
VCOs. The main features of a mixed-signal PLL are:
1. Reference frequency: The stable, accurate frequency reference to
which the RF output will be phase locked. It is typically derived
from a crystal or temperature-controlled crystal oscillator (TCXO).
Analog Dialogue Volume 43 Number 4
3. Charge pump: Converts the error signal into a train of positive
or negative current pulses in proportion to the phase error.
4. Loop filter: Integrates the current pulses from the charge
pump, providing a clean voltage to the VCO tuning port.
6. N divider: Divides the output frequency down to equal the
PFD or reference frequency. It can straightforwardly divide
by an integer—or, increasingly, be implemented as a fractional
divider. The fractional divider can be simply implemented
by toggling the divide values in an integer divider to get a
fractional average (for example, to get an average of 4.25,
count to 4 three times and count to 5 once. Seventeen pulses
have been counted and 4 pulses have been created; so the
frequency ratio is 17/4 = 4.25). In practice, better results can
be achieved by borrowing from techniques used in highresolution noise-shaped converters. Thus, the fractional
engine is usually implemented using a ∑-∆ architecture, which
has the advantage of reducing spurious frequencies.
As an example of the highly integrated circuitry used in available
devices, Figure 2 shows a block diagram of a fractional-N PLL
IC, the ADF4350 wideband synthesizer with integrated VCO;
it has an output frequency range of 137.5 MHz to 4400 MHz.
(A brief summary of its capabilities can be found in the WideBandwidth PLLs with Integrated VCOs section.)
The key performance-limiting characteristics of PLLs are phase
noise, spurious frequencies, and lock time.
Phase noise: Equivalent to jitter in the time domain, phase noise is
oscillator or PLL noise as evaluated in the frequency domain. It is
the rms sum of the noise contributed by the various components in
the PLL. The charge-pump-based PLL will suppress VCO noise
inside the loop filter bandwidth. Outside of the loop bandwidth,
the VCO noise dominates.
Spurs: Spurious frequency components are caused by the charge
pump’s periodic updating of the VCO tuning voltage. They will
appear at a frequency offset from the carrier by the PFD frequency.
In a fractional-N PLL, spurs will also occur due to the action of
the fractional divider.
Lock Time: The time taken for the PLL’s phase or frequency
to return to lock range when changing from one frequency to
another or responding to a transient offset. It can be specified in
terms of frequency- or phase settling. Its degree of importance as
a specification depends on the application.
Why Do VCOs Still Use High Voltages?
High-performance VCOs are among the last electronic components
to resist the tide of silicon integration. Only in the past few years
have VCOs for cellular handsets been fully integrated into their
radio chipsets. However, cellular base stations, microwave
point-to-point systems, military and aerospace, and other
higher-performance applications are still beyond the capability
of silicon-based VCOs and are still implemented using a discrete
approach. Here’s why:
Most commercially available discrete VCOs use a variablecapacitance varactor diode as the tunable element in an LC-based
tank circuit. Varying the diode’s voltage changes its capacitance
and, thus, the resonant frequency of the tank circuit.
13
SDVDD
REFIN
10-BIT R
COUNTER
×2
DOUBLER
DVDD
AVDD
VP
RSET
VVCO
MULTIPLEXER
÷2
DIVIDER
MUXOUT
LOCK
DETECT
FLO SWITCH
SW
LD
CLK
DATA
LE
DATA REGISTER
FUNCTION
LATCH
CHARGE
PUMP
CPOUT
PHASE
COMPARATOR
INTEGER
REG
FRACTION
REG
VTUNE
VREF
VCOM
VCO
CORE
MODULUS
REG
THIRD-ORDER
FRACTIONAL
INTERPOLATOR
TEMP
OUTPUT
STAGE
MULTIPLEXER
÷1/2/4/8/16
N COUNTER
MULTIPLEXER
CE
AGND
DGND
CPGND
RFOUTA+
RFOUTA–
PDBRF
OUTPUT
STAGE
RFOUTB+
RFOUTB–
ADF4350
SDGND
AGNDVCO
Figure 2. ADF4350 PLL synthesizer block diagram.
Any voltage noise on the varactor will be amplified by the VCO
gain, K V, (expressed in MHz/V) and will translate into phase
noise. To keep VCO phase noise to a minimum, K V must be kept
as low as possible, but achieving a reasonably wide tuning range
requires a large K V. Thus, for applications that require low phase
noise and a wide tuning range, VCO manufacturers typically
design oscillators with low gain and a large input voltage range to
satisfy these conflicting requirements.
Figure 3 shows two examples of recommended active filter
topologies with prefiltering using inverting and noninverting
gains. Please note that these amplifier circuits are true timeintegrators, which force the PLL’s loop to maintain zero error at
their inputs. Outside of the loop, the topologies shown could drift
to the supply rails.
Typical voltage tuning ranges for narrow-band VCOs are 0.5 V to
4.5 V, while wideband VCOs typically tune over 1 V to 14 V, and
in some cases have ranges as wide as 1 V to 28 V.
Coaxial resonator oscillators (CROs) are another special type
of VCO that uses a very low gain and wide input tuning
voltage to achieve ultralow phase-noise performance. They
are typically used in narrow-band private mobile radio and
land mobile radio applications.
a. Inverting topology.
Interfacing to a High-Voltage VCO
Most commercial PLL synthesizer ICs have charge pump outputs
that are limited to a maximum of about 5.5 V, insufficient to
directly drive a VCO that requires higher tuning voltages if the
loop filter uses passive components alone. An active loop filter
topology using op amp circuitry must be employed to reach the
higher tuning voltages.
Figure 3. Active filter using prefiltering.
The simplest approach to achieve this would be to add a gain
stage after the passive loop filter. Although simple to design, this
approach has some pitfalls: an inverting op amp configuration
presents a low input impedance that will load the passive loop filter,
altering the loop dynamics; a noninverting configuration has input
impedance high enough not to load the filter but will amplify any
op amp noise by the active filter gain without the benefit of filtering
by the preceding passive loop filter. A much better topology is to
integrate the gain stage and filter into a single active filter block.
Prefiltering is advisable so as not to overdrive the amplifier with
the very short current pulses from the charge pump—which could
rate-limit the input voltage.
The inverting topology has the advantage of biasing the charge
pump output at a fixed voltage, typically one-half the charge-pump
voltage (V P/2)—optimal for spur performance. Care needs to be
taken to provide a clean bias voltage, ideally from a dedicated lownoise linear regulator like the ADP150, with adequate decoupling
as close as possible to the op amp input pin. The resistance values
used in the divider network should be minimized to reduce their
noise contribution. When using the inverting topology, it is
important to make sure that the PLL IC allows the PFD polarity
to be inverted, if necessary, canceling out the op amp’s inversion
and driving the VCO with the correct polarity. The ADF4xxx
family has this property.
14
b. Noninverting topology.
Analog Dialogue Volume 43 Number 4
The noninverting loop filter configuration does not require a
dedicated bias, so it can be a more compact solution. The charge
pump voltage, instead of being biased at a fixed level, will now
vary across its operating range. For this reason, it is much more
critical to use an op amp with rail-to-rail inputs when using this
filter type. (Input voltage range requirements are described in
the next section.)
Choice of Op Amp
The choice of op amp is the key to maximizing the potential
of an active filter. Besides bandwidth, the main performance
specifications to consider are:
• Noise voltage density—expressed in nV/√Hz
• Current noise—expressed in pA/√Hz
• Input bias current
• Common-mode voltage range
The output of the filter directly affects the generated frequency
and phase; therefore, the op amp’s noise-voltage density gives
an indication of how much phase noise will be added by the
active filter. Amplifier noise is added both within the PLL loop
bandwidth and out of band—and is most pronounced at the
loop filter’s corner frequency, especially for amplifiers with high
noise voltage density. Thus, it is important to keep the amplifier’s
noise low to fulfill the mission of the amplifier and high-voltage
VCO: to provide lower phase noise. A good design target would
be <10 nV/√Hz. Current noise is usually quite small compared to
the error current pulses, so its effects tend to be much smaller
than those of voltage noise.
Op amps that have significant input bias currents in relation to the
PFD output current can result in large spurs on the PLL output
spectrum. To keep the VCO tuning voltage constant and the PLL
in lock, the charge pump must replace the bias current drawn by
the op amp inputs on each PFD cycle. This modulates the V TUNE
voltage at the PFD frequency and causes spurs around the carrier
at an offset equal to the PFD frequency. The higher the input bias
current, the greater the modulation of the V TUNE voltage and the
higher the spur amplitudes.
Common-mode voltage range, or input voltage range (IVR),
another important op amp specification, is often overlooked,
causing serious problems in the end design. IVR determines
the clearance needed at the input terminals between the
ma x i mu m /m i n i mu m sig na ls a nd t he posit ive /neg at ive
power rails.
Early op amps that ran on ±15 V had typical IVRs of ±12 V.
Slow lateral PNP input stages, added later, allowed the IVR
to include the negative rail, thus providing single-supply
capability. Although any op amp will run on ground and
a positive supply, it is necessary to observe the distance
from the rails.
For example, the popular OP27 has an IVR of ±12.3 V with
a ±15-V supply. This means that the input voltages need
to be at least ±2.7 V from the positive and negative rails.
This limitation at the lower end of the range would make it
unattractive for use with wide input swings in single-supply
operation. A dual-supply design option, if available, allows a
much greater choice of op amps (and simplifies the input bias
question). If single-supply design is necessary, use op amps that
allow the input voltage to swing from rail to rail (but many of
them may have higher noise-voltage specs). Thus, for best
results, one needs an op amp with low noise voltage density
for low phase noise, low input bias currents for low spurs,
and rail-to-rail inputs for single-supply operation. Table 1
lists some Analog Devices op amps and their specifications
in relation to the above design criteria.
Analog Dialogue Volume 43 Number 4
Table 1. Recommended Op Amps to Use in PLL Active Loop Filters
Op Amp
Voltage
Noise,
f = 1 kHz
(nV/√Hz)
Current
Noise,
f = 1 kHz
(pA/√Hz)
Input
Bias
Current
(Typ)
Input
Voltage
Range,
Clearance
from Lower
Rail (V)
VSUPPLY
Max,
SingleSupply (V)
AD820
16
0.8
2 pA
–0.2
36
OP184
3.9
0.4
60 nA
0
36
AD8661
12
0.1
0.3 pA
–0.1
16
OP27
3
0.4
10 nA
+2.7
36
AD8099
2
8
100 nA
+1.3
12
The choice of op amp depends on the application. If the PFD spurs
are far outside the loop bandwidth (for example, in fractional-N
synthesizers), then a bipolar-junction-transistor-input (BJT) op
amp—like the OP184 or OP27—will be suitable. The PFD spurs
caused by the high input bias current of the BJT will be well
attenuated by the loop filter, and the PLL can take full advantage
of the BJT op amp’s low noise voltage density.
If the application requires the PFD-to-loop-bandwidth ratio to
be small (for example, in an integer-N synthesizer), a compromise
should be reached between noise and spur levels; the AD820 and
AD8661 could be good choices here.
It is worth noting that although active filters tend to contribute
noise to the PLL, their ability to act as buffers provides a
performance advantage over passive filters in some application
niches. For example, if the VCO has high leakage current on its
tuning port leading to high PFD spurs, an op amp can be used to
reduce the spur levels. The op amp’s low impedance output can
easily supply the tuning port leakage current.
Design Example
Consider an example where the LO has the following specifications:
• Octave tuning range from 1000 MHz to 2000 MHz
• Phase noise requirement of –142 dBc/Hz at 1 MHz offset
• Spurs < –70 dBc
• 250-kHz channel spacing
• Lock time < 2 ms
• Single supply of 15 V or 30 V available
To cover the 1-GHz band, while meeting the phase-noise target,
it is necessary to use a high-voltage VCO and an active loop
filter. The phase-noise and spur specifications, and single-supply
restriction, will drive the choice of op amp. To meet the spur
specifications the op amp input bias current must be low, while
best phase noise will be achieved with an op amp with low voltage
noise. A compromise between the two can be reached by choosing
a JFET-input op amp, such as the AD8661, which has an input bias
current of 0.3 pA and voltage noise of 12 nV/√Hz. This device
can also handle the single-supply requirement. The RFMD
UMS-2000-A16 VCO was selected to cover the octave range.
The best place to start is with a simulation involving the activeTM
filter topologies supported in the ADIsimPLL tool. Two of the
recommended filter types are shown in Figure 3, but ADIsimPLL
supports other configurations as well.
For the PLL, the ADF4150, which can be operated in either integer
or fractional mode, was chosen; it also has output divider options
of 2/4/8/16/32—which allow continuous coverage from 2 GHz
down to 31.25 MHz. The ADF4150 is similar to the ADF4350
shown in Figure 2, but it allows the choice of an external VCO
for applications that need to meet more stringent phase-noise
requirements. In the simulation, the PLL loop filter was set at
20 kHz to attempt to minimize the op amp noise contribution,
while keeping the PLL lock time less than 2 ms.
15
Figure 4 shows a plot of noise (dBc) as a function of frequency
offset in simulated and measured systems, using an ADF4150
PLL, a UMS VCO, and an AD8661-based filter. A –90-dBc peak
at about 20 kHz, due to the noise added by the active loop filter,
can be seen in both noise profiles, but the –142-dBc/Hz target at
1 MHz offset is still being met. To reduce the in-band noise, a
lower-noise op amp, such as the OP184 or OP27, could be used
at the expense of higher spurs; or the PLL loop bandwidth could
be decreased to below 20 kHz.
–80
MEASURED AD8661
ADIsimPLL AD8661
–90
PHASE NOISE (dBc)
–100
–110
–120
–130
This shows the real benefit of moving VCO designs from a discrete
to a silicon-based solution: significant levels of integration can be
achieved in minimal area, allowing greater design flexibility. For
example, the ADF4350 also integrates a programmable outputdivider stage that allows frequency coverage from 137.5 MHz all
the way up to 4.4 GHz—a very attractive feature for radio designers
who want to reuse the same design across multiple frequencies
and standards.
The ADF4350 comes in a 5-mm square LFCSP package—as
compared with the standard 12.7-mm square VCO package.
Performance levels approach those of discrete designs, with phase
noise of –114 dBc/Hz at 100-kHz offset and –134 dBc/Hz at 1-MHz
offset (see Figure 2).
–140
–150
–160
–170
–180
1k
shown in Figure 2. In this case, the VCO is integrated on chip.
The inherent trade-off of wide tuning range and low phase noise
discussed above is avoided by using a multiband VCO approach. In
the ADF4350, three separate VCOs are integrated on chip, each of
which has 16 overlapping subbands, giving a total of 48 subbands.
Each time the frequency is updated, an automatic calibration is
initiated to select the appropriate VCO subband.
10k
100k
1M
2.8
10M
FREQUENCY (Hz)
Figure 4. ADIsimPLL simulation vs. measured performance
using AD8661 as an op amp in PLL active filter.
PHASE NOISE (dBc)
0.8
4600
4400
4200
4000
3800
3600
3400
3200
3000
FREQUENCY (MHz)
–120
Figure 6. Plot showing the 48 distinct bands in the ADF4350
VCO voltage vs. frequency relationship.
–130
–140
–150
–160
–170
–180
1k
2800
–110
2600
0
2400
0.4
1800
–100
1.2
2200
MEASURED AD8661
MEASURED OP27
ADIsimPLL OP27
–90
1.6
2000
–80
2.0
VTUNE (V)
Figure 5 shows the approximately 6-dB improvement when the
OP27 is used. In this case, spurs do not increase significantly, as
the loop bandwidth is relatively narrow. Lowering the bandwidth
further will improve phase noise for offsets below 100 kHz at the
expense of increased PLL lock-time. All of these trade-offs can be
tested with ADIsimPLL simulations prior to going into the lab.
2.4
10k
100k
1M
10M
For more on one of the broadest PLL portfolios in the industry,
including integer-N, fractional-N, integrated VCO, and highvoltage PLL ICs—pushing performance boundaries and easing
design challenges for PLL- and radio designers worldwide—
consult the PLL Synthesizers/VCOs website.
FREQUENCY (Hz)
Figure 5. PLL measured performance using AD8661 vs.
OP27 in active loop filter.
Breaking News: High Voltage PLLs
Up to this point the discussion has been driven by the need to
use active filters to interface a low-voltage PLL device to a highvoltage VCO. High-voltage PLLs are becoming available, however,
greatly reducing the necessity for an active filter. An example is the
ADF4113HV PLL, which integrates a high-voltage charge pump
and has –212-dBc/Hz normalized phase-noise floor. In this case,
the PLL charge pump output can be as high as 15 V, allowing a
simpler passive filter before the VCO.
This family of high-voltage PLLs will soon be further augmented
by devices that increase the maximum voltage to 30 V and
fractional-N PLLs that have high-voltage charge pumps. Consult
the PLL website for updates and new-product information.
Wide-Bandwidth PLLs with Integrated VCOs
Another alternative to using an active filter with high-voltage VCO
is to use a fully integrated high-performance PLL like the ADF4350,
16
References
1. Applied Radio Labs Forums
http://www.radiolab.com.au/Forums/default.asp.
2.Best, Roland E. Phase-Locked Loops. Design, Simulation, and
Applications. McGraw Hill.
3. Curtin, Mike and Paul O’Brien. 1999. “Phase-Locked Loops
for High-Frequency Receivers and Transmitters—Part 2.”
Analog Dialogue, Volume 33.
4. Information on all A DI components can be found at
www.analog.com.
Author
Austin Harney [[email protected]]
graduated in 1999 with a BEng from University
College, Dublin, Ireland, and joined Analog
Devices following graduation. He is currently
an applications engineer for the ISM-band
wireless product line, based in Limerick. In his
spare time, Austin enjoys football, music, and
spending time with his daughters.
Analog Dialogue Volume 43 Number 4
Adjustable-Gain Difference Amplifier
Circuit Measures Hundreds of Volts,
Rejects Large Common-Mode Signals
Laser wafer trimming provides resistor matching for commonmode rejection (CMR) of up to 86 dB minimum.
By Moshe Gerstenhaber and Michael O’Sullivan
A high common-mode difference amplifier, in a feedback loop with
an inverting op amp, is a useful aid for performing high-voltage
differential measurements up to 500 V.
Two common solutions to monitor power-line voltages or other
large signals—using low-voltage electronics—involve a highresistance voltage divider to attenuate the input, followed by an
op amp buffer (Figure 1a); or an inverting op amp with high input
resistance (Figure 1b). These methods have several shortcomings:
they are useful mainly for single-ended measurements, making
them prone to ground noise; they cannot reject common-mode
voltages; and the resistors dissipate different amounts of power,
leading to gain errors.
(2)
Capacitor C1 (100 nF), with resistor R2 (20 kΩ), provides loop
stability and sets the bandwidth at approximately 1 kHz.
VOUT
R
A highly useful application, shown in Figure 3, uses the AD629
and an op amp, such as the AD708,2 to provide adjustable gain. A
differential input signal is applied to difference amplifier A1, which
rejects any common-mode voltage on the input. The differential
input signal appears at the output with a gain of 1. The output
voltage is fed into the input of operational amplifier A2, which
is configured as a voltage inverter with a gain of –R2/R1. The
inverter’s output voltage is applied to the difference amplifier’s
reference pin (REF+). The voltage applied to this pin is multiplied
by a gain of 19, and is added to the output of A1. Solving for the
output of this negative feedback loop,
VIN
R/20
R/20
21.111k𝛀𝛀
R
VOUT
VIN
–VDIFF/2
Figure 1. Single-ended measurements using op amps.
1
A better way involves the use of a difference amplifier. The AD629
unity-gain difference amp, shown in Figure 2, can reject extremely
high common-mode signals (up to 270 V with 15-V supplies).
To achieve this high common-mode voltage range, a precision
internal resistor divider attenuates the noninverting signal by a
factor of 20. Other internal resistors restore the gain to provide
a differential gain of unity. The complete transfer function, with
reference inputs grounded, is
(1)
REF(–) 1
–IN 2
+IN 3
–VS 4
21.1k𝛀𝛀
380k𝛀𝛀
380k𝛀𝛀
380k𝛀𝛀
20k𝛀𝛀
AD629
8
NC
7
+VS
6
OUTPUT
5
REF(+)
NC = NO CONNECT
Figure 2. Functional block diagram of the AD629
difference amplifier.
Analog Dialogue Volume 43 Number 4
380k𝛀𝛀
380k𝛀𝛀
VCOMMON
+VDIFF/2
380k𝛀𝛀
A1
AD629
OUTPUT
C1
C1
20k𝛀𝛀
R2
R1
AD708
A2
Figure 3. Difference amplifier in adjustable-gain configuration.
The maximum differential signal that the circuit can handle is
limited by the output range of amplifiers A1 and A2. When R2/R1
is less than or equal to 1, amplifier A1’s output will saturate first.
With ±15-V supplies, the circuit can handle differential voltages
up to 520 V p-p. Systems with ±10-V or ±5-V supply voltages
can handle differential voltages up to 320 V p-p and 120 V p-p,
respectively. With ratio R2/R1 greater than 1 and an amplifier with
rail-to-rail output for A2, these ranges can be extended closer to
600 V p-p, 400 V p-p, and 200 V p-p.
Negative feedback around the loop helps to reduce the voltage
seen by the inputs of op amp A1. This means that the circuit has
the input range to reject large common-mode voltages, even in
the presence of large differential voltages.
17
a 400-V p-p differential signal is measured in the presence of a
50-V common-mode voltage. As can be seen, this circuit allows
very large differential inputs; rejects common-mode voltages; and
allows the differential gain to be set by the ratio of resistors R2 and
R1, enabling the user to select the appropriate level of attenuation.
The resistors inside the difference amplifier are precision lasertrimmed and manufactured with low-drift thin film material, so
the system does not suffer from resistor self-heating errors.
OUTPUT 20V p-p 120Hz
1
References
1
www.analog.com/en/amplifiers-and-comparators/current-senseamplifiers/ad629/products/product.html.
2
www.analog.com/en/amplifiers-and-comparators/operationalamplifiers-op-amps/ad708/products/product.html.
2
INPUT 400V p-p 120Hz
0V COMMON MODE
CH1 10V
CH2 100V
M4.00ms
250kS/s
A CH1
0.00V
4.0𝛍𝛍s/pt
Moshe Gerstenhaber [[email protected]
analog.com] is a Division Fellow at Analog
Devices. He began his career at ADI in 1978
and has held various senior positions over the
years in manufacturing, product engineering,
and product design. Moshe is currently the
design manager of the Integrated Amplifier
Products Group. He has made sig nif icant
contributions in the field of amplifier design, especially veryhigh-precision specialty amplifiers, such as instrumentation
and difference amplifiers.
OUTPUT 20V p-p 120Hz
1
2
INPUT 400V p-p 120Hz
50V COMMON MODE
CH1 10V
CH2 100V
M4.00ms
250kS/s
A CH1
0.00V
4.0𝛍𝛍s/pt
Figure 4. 400-V p-p differential signal measured with 0-V
and 50-V common-mode signal.
Figure 4 shows two plots: in the first plot, a differential signal of
400 V p-p is measured using ±15 V supplies; in the second plot,
18
Authors
Michael O’Sullivan [[email protected]
analog.com] has worked at Analog Devices since
2004. Currently the product and test engineering
manager of the Integrated Amplifier Products
Group, he supports product characterization
and release of very-high-precision specialty
amplifiers, such as instrumentation and difference
amplifiers. Mike worked as a product engineer in
the semiconductor field for over 14 years.
Analog Dialogue Volume 43 Number 4
Automobile Tail-Lamp and
Brake-Lamp Controller
By Chau Tran
Light emitting diodes (LEDs), long used in consumer electronic
products, are recently finding uses in automotive lighting, where
they now provide signaling functions, daytime running lights, and
interior lighting in production vehicles. As this lighting technology
hits the road, manufacturers continue to investigate new ways to
apply it, taking advantage of the styling possibilities afforded by
LED headlights and taillights.
Red LEDs are now widespread for rear lighting. Cost remains
an issue, but factors such as safety, environmental friendliness,
and styling flexibility weigh heavily in favor of LEDs. One of the
most popular applications is center brake lights. This design idea
shows a way to use the same LED array for both taillights and
brake lights.
FZT951
BATTERY
+
VPLUS
SHUNT
5
VSENSE
5V
7
350k𝛀𝛀
Timer A2 converts the signal into a pulse-width-modulated signal
with a duty cycle determined by R3, R4, R5, and C2. The pulse
width is determined by:
LED
LOAD
REF
2
+
The LEDs are turned on and off depending on the digital voltage
on a CMOS-compatible PWM pin (AD8240 Pin 3). This voltage
can be continuous for a simple on/off control, or PWM for dimming
control. The PWM frequency should be less than 500 Hz, with a
duty cycle from 5% to 100%. Typical values are 5% for running and
95% for braking. In Figure 2, the PWM frequency is determined
by R1, R2, and C1 of timer A1. The pulse period is:
With R1 = 49.9 kΩ, R2 = 10 kΩ, and C1 = 0.1 μF, the period is
4.84 ms, or about 206 Hz.
8
1
Costs are further reduced by eliminating the inductor required
for a switching design; and a switching driver is not required
because the LED lamps operate at much lower power than
incandescent lamps.
T = 0.693 (R1 + 2 × R2) C1
VO
6
measures the voltage across an external current shunt, detecting
an open LED when the measured current drops below a preset
threshold. Output current limiting is provided by latching off the
output voltage when the current reaches a level set by the value
of the external current shunt. When the sense amplifier output
exceeds 5 V, an internal comparator causes the driver to latch off
the output voltage. The latch is reset during the next PWM cycle.
An overcurrent condition can also be detected by measuring the
sense amplifier output.
Pulse width = 1.1 × R × C2
LATCH OFF
PWM
GENERATOR
DRIVER
3
PWM
250k𝛀𝛀
AD8240
4
GND
Figure 1. Running braking lamps controller.
The brightness of the LED is controlled by a simple switch,
allowing dim lighting while running and bright lighting while
braking. The block diagram, shown in Figure 1, includes an
AD8240 LED driver/monitor, PNP pass transistor, and PWM
generator. The AD8240 supplies a constant voltage to drive LED
lamps. It also provides cost-effective LED lamp monitoring and
short-circuit protection. The output is regulated at 12 V when the
battery is between 12.5 V and 27 V.
Figure 2 shows the PWM generator, which consists of two
555 timers. The PWM signal controls the LED brightness. VO is
turned on when the PWM input is high and turned off when the
input is low. The AD8240 is designed to work with a frequency up to
500 Hz and a typical duty cycle of 5% to 95%.
5% 50% 95%
5V
+
LM555
R3
LM555
10k𝛀𝛀
R4
R5
R1
1
1
GND
VCC 8
2
TRIG
DISCH 7
3
OUT
4
RESET
A1
1
GND
VCC 8
2
TRIG
DISCH 7
THRES 6
3
OUT
THRES 6
CONT 5
4
RESET
R2
C1
OFF
SWITCH
A2
0.01𝛍𝛍F
CL
PWM SIGNAL
Figure 2. PWM signal generator.
With its low-power operation, the AD8240 provides a low-cost
solution in a small package. An internal current-sense amplifier
Analog Dialogue Volume 43 Number 4
Note that the brightness of the LEDs increases as the duty cycle
increases. When the brake is applied, the duty cycle is 95%, and the
LED array is at maximum brightness. During normal operation,
the duty cycle is at 5%, and the LED array is dimmed. Using a
single LED array for both operations reduces cost.
If a short circuit or an overload condition occurs, the voltage at
VSENSE (Pin 1) falls to zero, and the output shuts down. This resets
during the next PWM cycle. If the condition persists, the AD8240
attempts to drive the output to 12 V, shutting down and restarting
after each PWM cycle.
This circuit presents a way to use a constant voltage, driving and
monitoring the LEDs with only two wires (power and ground). In
many cases, this can be reduced to one wire when the chassis or
shared ground return is used. Currently, these lamps are controlled
and driven by the body control ECU (electronic control unit).
With this constant voltage architecture, the control and drive
function for the LEDs can remain in the ECU with minimal
design modifications.
Author
CONT 5
1nF
0.01𝛍𝛍F
2
where R is equal to R5, the parallel resistance of R3 and R5, or the
parallel resistance of R4 and R5, depending on the switch position.
With R3 = 2.37 kΩ, R4 = 45.7 kΩ, R5 = 42.4 kΩ, and 
C2 = 0.1 μF, the duty cycle is 5% when the switch is in
Position 1, 50% when the switch is in Position 2, and 95% when
the switch is in the OFF position.
Chau Tran [[email protected]] joined
Analog Devices in 1984, where he works in the
Instrumentation Amplifier Products (IAP) Group
in Wilmington, MA. In 1990, he graduated with an
MSEE degree from Tufts University. Chau holds
more than 10 patents and has authored more than
10 technical articles.
19
Analog Devices, Inc.
Worldwide Headquarters
Analog Devices, Inc.
Three Technology Way
P.O. Box 9106
Norwood, MA 02062-9106
U.S.A.
Tel: 781.329.4700
(800.262.5643,
U.S.A. only)
Fax: 781.461.3113
Analog Devices, Inc.
Europe Headquarters
Analog Devices, Inc.
Wilhelm-Wagenfeld-Str. 6
80807 Munich
Germany
Tel: 49.89.76903.0
Fax: 49.89.76903.157
Analog Devices, Inc.
Japan Headquarters
Analog Devices, KK
New Pier Takeshiba
South Tower Building
1-16-1 Kaigan, Minato-ku,
Tokyo, 105-6891
Japan
Tel: 813.5402.8200
Fax: 813.5402.1064
Analog Devices, Inc.
Southeast Asia
Headquarters
Analog Devices
22/F One Corporate Avenue
222 Hu Bin Road
Shanghai, 200021
China
Tel: 86.21.2320.8000
Fax: 86.21.2320.8222
©2011 Analog Devices, Inc. All rights reserved.
Trademarks and registered trademarks are the property
of their respective owners.
M02000434-0-2/11
www.analog.com/analogdialogue
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