Energy Efficient RF Transmitter Design using Enhanced Breakdown

Energy Efficient RF Transmitter Design using Enhanced Breakdown
Energy Efficient RF Transmitter Design using Enhanced Breakdown Voltage
SOI-CMOS Compatible MESFETs
by
Mohammad Reza Ghajar
A Dissertation Presented in Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy
Approved May 2012 by the
Graduate Supervisory Committee:
Trevor Thornton, Chair
Bertan Bakkaloglu
Michael Goryll
James Aberle
ARIZONA STATE UNIVERSITY
August 2012
ABSTRACT
The high cut-off frequency of deep sub-micron CMOS technologies has
enabled the integration of radio frequency (RF) transceivers with digital circuits.
However, the challenging point is the integration of RF power amplifiers, mainly
due to the low breakdown voltage of CMOS transistors. Silicon-on-insulator
(SOI) metal semiconductor field effect transistors (MESFETs) have been
introduced to remedy the limited headroom concern in CMOS technologies. The
MESFETs presented in this thesis have been fabricated on different SOI-CMOS
processes without making any change to the standard fabrication steps and offer
2-30 times higher breakdown voltage than the MOSFETs on the same process.
This thesis explains the design steps of high efficiency and wideband RF
transmitters using the proposed SOI-CMOS compatible MESFETs. This task
involves DC and RF characterization of MESFET devices, along with providing a
compact Spice model for simulation purposes. This thesis presents the design of
several SOI-MESFET RF power amplifiers operating at 433, 900 and 1800 MHz
with ~40% bandwidth. Measurement results show a peak power added efficiency
(PAE) of 55% and a peak output power of 22.5 dBm. The RF-PAs were designed
to operate in Class-AB mode to minimize the linearity degradation.
Class-AB power amplifiers lead to poor power added efficiency,
especially when fed with signals with high peak to average power ratio (PAPR)
such as wideband code division multiple access (W-CDMA). Polar transmitters
have been introduced to improve the efficiency of RF-PAs at backed-off powers.
A MESFET based envelope tracking (ET) polar transmitter was designed and
i
measured. A low drop-out voltage regulator (LDO) was used as the supply
modulator of this polar transmitter. MESFETs are depletion mode devices;
therefore, they can be configured in a source follower configuration to have better
stability and higher bandwidth that MOSFET based LDOs. Measurement results
show 350 MHz bandwidth while driving a 10 pF capacitive load.
A novel polar transmitter is introduced in this thesis to alleviate some of
the limitations associated with polar transmitters. The proposed architecture uses
the backgate terminal of a partially depleted transistor on SOI process, which
relaxes the bandwidth and efficiency requirements of the envelope amplifier in a
polar transmitter. The measurement results of the proposed transmitter
demonstrate more than three times PAE improvement at 6-dB backed-off output
power, compared to the traditional RF transmitters.
ii
TABLE OF CONTENTS
Page
LIST OF TABLES....................................................................................................... v
LIST OF FIGURES .................................................................................................... vi
CHAPTER
1
INTRODUCTION .......................................................................................... 1
1.1. SOI-MESFET Device Operation ............................................. 6
1.2. RF Power Amplifier Theory and Figure of Merit.................... 9
1.2.1. Figures of Merit ...................................................... 10
1.2.2. RF Power Amplifier Classes of Operation ............. 14
1.2.2.1. Linear Power Amplifiers .................................... 14
1.2.2.2. High Efficiency Power Amplifiers..................... 22
1.3. Summary and Discussion ....................................................... 27
2
CHARACTERIZATION AND MODELING ............................................. 29
2.1. DC Characterization ............................................................... 30
2.2. RF Characterization ................................................................ 34
2.3. TOM3 Modeling ..................................................................... 42
2.3.1. DC Model Characterization .................................... 44
2.3.2. RF Modeling ........................................................... 47
2.3.3. Test Power Amplifier ............................................. 49
2.4. Summary ................................................................................. 51
3
SOI-MESFET POWER AMPLIFIERS ....................................................... 53
3.1. Wideband Class AB SOI-MESFET RF PA........................... 56
iii
CHAPTER
Page
3.2. Cascoded MOSFET-MESFET RF Power Amplifier ............ 63
3.3. On-Die Measurements of MESFET RF-PAs......................... 69
3.4. Summary of SOI-MESFET RF Power Amplifiers ................ 77
4
WIDEBAND SUPPLY MODULATOR FOR POLAR TRANSMITTER.78
4.1. RF Power Amplifiers Supply Modulators ............................. 83
4.2. SOI-MESFET for PA Supply Regulation .............................. 85
4.3. LDO Measurement Results .................................................... 88
4.4. Summary ................................................................................. 96
5
BACKGATE MODULATED POLAR TRANSMITTER .......................... 97
5.1. Substrate Voltage Effect on Device Characteristics .............. 99
5.2. Transient Behavior of Substrate Contact ............................. 101
5.3. Efficiency Enhancement Using Backgate Modulation ........ 103
5.4. Backgate Modulation versus Supply Modulation ................ 109
5.5. Summary ............................................................................... 111
6
CONCLUSION ........................................................................................... 113
REFERENCES ........................................................................................................ 116
APPENDIX
A
ENVELOPE TRACKING TEST SETUP.................................................. 122
iv
LIST OF TABLES
Table
Page
1. Measurement results of transformer-less MESFET RF-PAs (MESFET has
LG=200 and LaS=LaD= 200nm and 500nm with Width=4mm). ................ 77
2. Advanced modulation schemes specifications [46]. ................................. 78
v
LIST OF FIGURES
Figure
Page
1. CMOS scaling trend. ................................................................................... 2
2. Class A RF power amplifier ....................................................................... 3
3. Maximum output power of a Class-A amplifier versus gate length under 5
Ohm loadline condition............................................................................... 4
4. Cross section view of a partially depleted MESFET structure. .................. 7
5. Cross section view of a standard N-channel MOSFET. ............................. 8
6. Power Amplifier block diagram................................................................ 11
7. Power Gain vs Output Power .................................................................... 12
8. An example of power amplifier output spectrum ..................................... 13
9. Third Order Inter-modulation products .................................................... 14
10. Class A amplifier ...................................................................................... 15
11. The quiescent bias point on the transistor transfer characteristic [28]...... 17
12. PA efficiency vs class of operation [28]. .................................................. 17
13. Transistor Family of Curves ..................................................................... 18
14. Transistor power gain versus frequency. .................................................. 19
15. Different impedance transformers using lumped components. ................ 20
16. Matching network efficiency versus impedance transformation ratio for
different inductor quality factors. ............................................................. 22
17. Inverse Class F Amplifier ......................................................................... 23
18. Class F drain waveforms ........................................................................... 24
19. Class E amplifier with lumped elements matching network .................... 25
vi
Figure
Page
20. Simplified class E amplifier ...................................................................... 26
21. Class E efficiency vs load resistance [28]................................................. 27
22. Gummel plots for LG=150, 200, 300 and 400 nm on a 150 nm SOI CMOS
technology. ................................................................................................ 30
23. Threshold voltage extracted for each of the devices fabricated on 150nm
and 350nm technologies. .......................................................................... 31
24. The breakdown voltage of MESFETs (a) on a 150nm SOI process. (b) on
a 45nm SOI process. Dashed line shows the trend and the red cross shows
to the peak measured breakdown voltage. ................................................ 32
25. The family of curves for an enhanced VBD MESFETs with LG= 200nm,
LaS = LaD = 1 µm. The gate voltage is stepped from +0.5V (uppermost
curve) to -0.5V in 0.25V steps. ................................................................. 33
26. Shows the exponential scaling of the peak cut-off frequencies for
MESFETs manufactured using different SOI CMOS processes. ............. 34
27. The peak cut-off frequency and RS as a function of LaS for MESFETs
fabricated on a 45nm SOI CMOS technology. ......................................... 35
28. Measured extrinsic resistances versus LaD for MESFETs with LG=400 nm
and LaS=300nm on a 150 nm process. ...................................................... 37
29. MESFETs intrinsic capacitances and gm versus LaD for MESFETs with
LG=400 nm and LaS=300 nm on a 150nm process. ................................... 38
30. Measured MAG versus frequency for different MESFETs on a 45nm SOI
CMOS process. ......................................................................................... 39
vii
Figure
Page
31. shows the cut-off frequencies and fMAX versus drain current for selected
devices (s) on the 150 nm SOI CMOS process (b) on the 45nm SOI
CMOS. ...................................................................................................... 41
32. TOM3 model circuit schematic [38]. ........................................................ 43
33. Measured extrinsic RD, RS and RG versus frequency. ............................... 43
34. Measured (solid line) and simulated (open circles) results for the silicon
MESFETs fabricated on the 45nm SOI process. The line at 1V represents
the approximate maximum operating voltage of MOSFETs on the same
process [40]. .............................................................................................. 45
35. Measured (solid) and modeled (open circles) transconductance. ............. 45
36. Model fit of the saturation current scaling fk (circles) compared to
extracted fk from measured drain current (solid line) [40]. ...................... 46
37. Gummel plots depicting drain and gate current of the MESFET across a
wide drain bias range of 4V and 0.1V. Measured results are solid lines
and modeled open circles. Additional drain voltages have been omitted
for clarity but show a good fit to measured data from Vd = 0 to 4V [40]. 47
38. Measured (solid line) and modeled (Open red symbols line) S Parameters
at Vd=2V and Vg=0.25V. ......................................................................... 49
39. Measured and simulated H21 and MAG at the same bias conditions. ..... 49
40. Class AB, 433MHz, RF Power amplifier schematic ................................ 50
41. Measured (symbols) and simulated (solid line) of gain, output power,
power added efficiency (PAE) and drain efficiency (DE). ....................... 51
viii
Figure
Page
42. A general Power Amplifier block diagram. .............................................. 55
43. Source/Load Pull Simulation Setup [28]. ................................................. 56
44. MESFET RF performance before and after packaging. (Red shows before
packaging) (Black shows after packaging) ............................................... 57
45. Family of Curves for MESFET with LG=LaS=LaD=300nm and
Width=4.2mm. .......................................................................................... 58
46. PAE and Power contours at drain and gate. Red shows PAE and Blue is
Power contours.......................................................................................... 59
47. Schematic of the simulated class AB SOI-MESFET PA.......................... 60
48. Simulation results of the 433MHz class AB MESFET PA. ..................... 60
49. Photograph of fabricated class AB 433MHz SOI-MESFET PA. ............. 61
50. Measurement results of 433MHz class AB MESFET PA. ....................... 62
51. Output spectrum of the MESFET PA driven by a GSM signal. ............... 62
52. Class AB RF PA measurement results; PAE, DE, Pout and gain versus
frequency................................................................................................... 63
53. The proposed cascoded MOSFET-MESFET pair. ................................... 64
54. (a) conventional NMOS-NMOS cascode structure (b) proposed MOSFETMESFET power amplifier......................................................................... 66
55. Cascoded MOSFET-MESFET’s Family Of Curves (FOC). .................... 67
56. Time domain drain’s current and voltage ................................................. 68
57. Pout, Gain and PAE versus Input power. ................................................. 69
ix
Figure
Page
58. MESFETs breakdown voltage versus LaD on a 45nm SOI-CMOS process.
................................................................................................................... 70
59. Theoretical peak output power of class-A amplifier versus LaD (a) device
loadline of 5 Ω (b) device loadline slope of 50 Ω. ................................... 71
60. On-Die MESFET RF-PA measurement setup. ......................................... 72
61. The presented load to the MESFET’s drain at fundamental and harmonic
frequencies. ............................................................................................... 73
62. PAE, Pout vs Pin. Class-A 400 MHz MESFET-PA with LG=200nm,
LaS=LaD=500nm. VG=0.0V ....................................................................... 74
63. 400 MHz MESFET-PA performance at different classes of operation.
VDD=4 V, VSUB=4V, Pin 6 dBm. .............................................................. 75
64. MESFTE RF-PA performance versus supply voltage. VG=-0.4 V,
VSUB=4V, Pin=6dBm. ............................................................................... 76
65. MESFET Class-AB amplifier efficiency versus normalized output power.
................................................................................................................... 79
66. Class-A transient drain voltage. ................................................................ 80
67. Efficiency improvement by reducing the DC bias point. ......................... 80
68. Kahn Envelope Elimination and Restoration (EER) Transmitter [52] ..... 81
69. (a) Series SMA LDO supply modulator in EER (b) Parallel Amplifier and
SMA supply modulator in EER. ............................................................... 84
70. (a) N-MOSFET based LDO (b) P-MOSFET based LDO ........................ 85
71. LDO envelope tracking regulator using the SOI-MESFET. ..................... 86
x
Figure
Page
72. Frequency response of the proposed LDO................................................ 88
73. Measured LDO voltage gain for different capacitive loads. ..................... 89
74. Measured LDO efficiency......................................................................... 90
75. LDO output spectrum when it is fed with a two tone signal at 100MHz and
separated by 200kHz. ................................................................................ 90
76. Transient response of the LDO to a rectified 1.2MHz sine wave applied as
the envelope signal. ................................................................................... 91
77. Measured PSR of the proposed LDO........................................................ 92
78. SOI-MESFET envelope tracking test setup. ............................................. 93
79. Effect of supply voltage swing on linearity. ............................................. 93
80. Transient input and output voltage of LDO for EDGE envelope signal... 94
81. Normalized output spectrum of MESFET ET for EDGE signal. ............. 95
82. Photograph of the 350 MHz SOI-MESFET LDO. ................................... 96
83. Simplified block diagram of a traditional polar transmitter...................... 97
84. Cross sectional view of a partially depleted MESFET (a) VG=0, VBG<0 (b)
VG<0, VBG<0 (c) VG<0, VBG=-2. ............................................................ 100
85. Effect of substrate voltage on SOI-MESFET I/V curve. ........................ 101
86. Measured input capacitance of the backgate terminal. ........................... 102
87. Transient behavior of a SOI-MESFET PA when the backgate is driven by
a 100 KHz square wave. ......................................................................... 103
88. Block diagram of the proposed backgate modulated polar transmitter. . 104
xi
Figure
Page
89. Effect of the backgate voltage on the MESFET RF-PA output power.
VDD=3V, VG=-0.2V. ............................................................................ 105
90. SOI-MESFET Power Amplifier performance versus the substrate voltage.
................................................................................................................. 105
91. PAE versus output power comparison between the traditional class-AB
PA and a backgate modulated polar transmitter. .................................... 106
92. Measured AM/PM characteristics of SOI-MESFET 1.8GHz PA at fixed
input power of 6-dBm. ............................................................................ 107
93. Measured slope of AM/PM curves for VDD and VBG sweeps. ............ 108
94. Output spectrum of EDGE signal for backgate modulated SOI-MESFET
ET transmitter. ........................................................................................ 108
95. PAE of backgate modulated PT at different input power levels. ............ 110
96. Efficiency comparison between the hybrid supply modulator [63],
backgate PT and pre-distorted backgate PT............................................ 111
xii
CHAPTER 1
INTRODUCTION
Modern digital communication systems have rapidly expanded in
capability which has required significant advancements of both digital CMOS and
RF front-end performance.
The expectations of more features and extended
battery lifetime has made bandwidth, power and system efficiency even more
critical. The standard platform for system integration at relatively low cost has
traditionally been silicon CMOS. The high cut-off frequency (fT) of deep submicron CMOS technologies now makes it possible to integrate RF and digital
CMOS circuits. This has led to a trend of building single chip CMOS transceivers
[1], [2], [3]. Yet, the advancement and scaling of CMOS has also led to reduced
operating voltages which makes analog and RF design more difficult due to
limited voltage headroom. The nominal supply voltage and cut-off frequency of
CMOS transistors for various technology nodes are shown in Figure 1 [4], [5].
The high cut-off frequency of deep sub-micron CMOS transistors benefits
both digital and analog/RF circuits. This trend has enabled the RF front ends to
operate at several GHz. The reduction in supply voltage has also helped to
minimize the power dissipation in digital circuitries. However, due to aggressive
signal-to-noise ratio, output power, dynamic range and linearity requirements of
modern communication standards, analog and RF circuits suffer intensely from
this supply reduction. The sticking point is the integration of front end stages,
specifically power amplifiers, since inductive elements used in their design lead to
high voltage swings around the drain terminal.
1
Figure 1.. CMOS scaling trend.
RF PAs are power hungry blocks that need to handle large voltage and
current swings on the output node to achieve a desired transmitted power. Based
upon the application, the output stage should be capable of delivering a fixed
amount of maximum power. As we will see in the following sections, the supply
voltage of a technology limits either the output power, efficiency, or adds design
complexities.
A very simplified form of a Class
Class-A
A RF amplifier is shown in Figure 2.
The operational behavior of the Class
Class-A
A amplifier will be discussed more in detail
in the subsequent sections of this thesis. However the basic output power
calculation is shown here.
2
Figure 2. Class A RF power amplifier
The output power is shown in Equation 1 for a Class-A amplifier with an
ideal transistor. For a fixed supply voltage, VDD, reducing RL increases the output
power, POUT. As will be shown later, in practice there are some limitations on
lowering the RL value [6]. In this example it is assumed to have RL=5 Ω.
2 Equation 1
Substituting the collected data shown in Figure 1 into Equation 1, the
maximum output power for different single device PA on CMOS technologies,
for the case of a Class A output stage, was calculated and shown in Figure 3. As it
can be seen, the maximum POUT for a standard 45nm CMOS technology is less
than 60-mW. This amount of power is much lower than what is required for most
handset device applications. For example, the GSM 900/1800 [7] standards
require a POUT of 1-4 Watt. This is less of an issue for other higher voltage
3
technologies such as GaN, GaAs or LDMOS which can encompass power ranges
of 1-10 Watt [8], [9], [10].
Output Power (dBm)
40
35
30
25
20
15
10
0.01
0.1
1
Gate Length (um)
Figure 3. Maximum output power of a Class-A amplifier versus gate length under
5 Ohm loadline condition.
As a result of the reduced output power capability for deep-submicron
CMOS devices, high power and expensive discrete units like SiGe and GaAs are
often employed. Nevertheless, many efforts have been made to employ CMOS
transistors in PA modules [11], [12]. The main motivation is the low fabrication
cost of CMOS compared to all the other PA technologies. Moreover, almost all
the other transceiver circuit blocks can currently be designed and integrated on the
same CMOS chip. Thus, the CMOS power amplifier is the last challenge towards
a single chip transceiver solution.
4
Many techniques have been studied to overcome the headroom limitation
of CMOS transistors. One such solution is higher impedance transformation [13]
which is often used to reduce the voltage swing on the transistor nodes.
Unfortunately this adds additional losses to the matching network [6]. Another
option is the cascode architecture [14], [15]. The cascode allows for a higher
supply voltage since the voltage is divided between two or more transistors. This
method can slightly increase output power. Some more specialized processes also
include bipolar transistors; however, these processes are typically much more
expensive than standard bulk CMOS processes [16]. A low cost solution which is
possible on many of today’s processes is using thick gate oxide MOSFETs [17].
These transistors can typically handle voltages that are 1.5-2 times higher than
standard MOSFETs. Parallel amplification is another approach that uses a power
combiner that adds up output power from parallel PAs. This method requires
extra passive components, which results in larger die area and worse efficiency.
In summary, there are tricks and architectures that can enhance the output
power for a class-A amplifier, but each of them has their own drawbacks.
Designing high efficiency amplifiers, such as class–E, are even more challenging
since the voltage swing is 2-3.5 times larger than the supply voltage [18]. All
these issues together, have led to a vital need for high breakdown voltage, low
cost, RF transistors on CMOS processes.
The idea of a high voltage, silicon on insulator (SOI)-CMOS compatible
MESFET was proposed and patented by T.J. Thornton, et al. [19]. The patented
MESFET has some outstanding characteristics that make it one of the best device
5
solutions towards fabrication of low cost, fully integrated transceivers. The main
advantage of MESFETs over MOSFETs is their enhanced breakdown voltage.
These MESFETs have supply voltage capabilities that are several times higher
than the MOSFETs on the same processes. On certain technologies, breakdown
voltages that are 20-30 times higher than the MOSFETs have been achieved. This
greatly relaxes the design constraints on CMOS PAs and can significantly
increase the peak output power.
The remaining portion of this thesis will focus mainly on demonstrating
the feasibility of high power, high speed and linear RF output stage design using
the proposed SOI-MESFET devices. This task would not be possible without
having a good understanding of both RF power amplifier theory and SOIMESFET device operation. Hence, the first chapter is focused on understanding
these two concepts.
1.1.
SOI-MESFET Device Operation
Metal-semiconductor field-effect-transistors (MESFETs) can be integrated
on any SOI or SOS (Silicon on Sapphire) CMOS processes without altering the
process or adding any additional fabrication steps [20], [21], [22], [23]. Other
reports [24], [25], [26] have considered silicon MESFETs on SOI, SOS, and bulk
CMOS processes but in each case, none of them used a standard CMOS process
flow.
A cross section view of the proposed SOI-MESFET is shown in Figure 4.
Unlike most MOSFETs, MESFETs are depletion mode devices. They are also a
majority carrier device. This means that a channel exists between drain and
6
source, even when the gate is floating. In order to reduce the drain current and
eventually “pinch-it off” a negative gate bias voltage is needed to invert the
channel under the gate. Spacers regions are added between the drain and gate,
LaD, and the source and gate, LaS, to avoid shorting the device terminals.
Depending on the channel thickness, the MESFET can operate as a fully or
partially depleted transistor. MESFETs used in this work are partially depleted
(PD), which means that the silicon channel below the gate is sufficiently thick to
control the conduction of the device.
Like a CMOS transistor, the RF performance and current drive capability
of a MESFET is highly dependent on the minimum gate length that can be
achieved. The PD-MESFET drain current (ID) equations are also fairly similar to
standard MOSFETs. Current and parasitic capacitance calculations are discussed
in the modeling chapter. The rest of this section talks about structural differences
between the MOSFET and MESFET and the benefits of MESFETs in certain
applications such as PAs.
Figure 4. Cross section view of a partially depleted MESFET structure.
7
A cross sectional view of an N-channel MOSFET on a bulk CMOS
technology is shown in Figure 5. One of the dissimilarities between MESFETs
and MOSFETs is the gate contact. The MOSFET’s high impedance gate oxide
layer does not exist in MESFETs. Instead a MESFET gate is formed by a metalsilicide contact on a lightly doped silicon region. This introduces much higher
quiescent gate current (IG) in MESFETs, which increases off-mode power
dissipation. However, it benefits the MESFET in terms of breakdown voltage
behavior. Oxide breakdown is one of the key CMOS breakdown mechanisms and
it continues to get worse as the oxide thickness, tox, scales with technology nodes.
, Equation 2
Figure 5. Cross section view of a standard N-channel MOSFET.
The drain to source electric field is another source of breakdown in
MOSFETs. The LaS and LaD regions added in the MESFET structure, increase the
distance between drain to source and thus lower the electric field at this critical
junction. Therefore, the spacers can also be used to improve the breakdown
voltage. This is a similar concept to that used in LDMOS.
8
The MESFET breakdown is thought to be caused mostly by avalanche
ionization and tunneling mechanisms. As the MESFET approaches soft
breakdown, which is reversible, the surface electric field can become large
enough to lower the barrier height at the gate and allow electrons to tunnel into
the channel from the gate metal. Consequently this leads to an exponential
increase in drain-to-gate current. If the drain voltage is further increased and/or
the gate becomes more negatively biased, the electric field will become even
larger and avalanche ionization will begin to occur. Eventually this will lead to a
non-reversible hard breakdown for the MESFET [27].
The cost of enlarging the spacer regions includes added parasitic
resistance for the MESFET and additional die area needed to lay the devices out.
Source resistance degrades both DC gain and RF performance. Thus there is a
tradeoff between enhancing the breakdown voltage and RF performance.
Although the RF performance degrades as VBD increases, the overall MESFET
performance has evident advantages over MOSFETs for up to 5GHz PA design
applications. Measured characteristics of fabricated MESFETs are discussed in
Chapter II of this thesis. The following subsection is an overview of how device
parameters can affect the PA performance.
1.2.
RF Power Amplifier Theory and Figure of Merit
RF power amplifier design methodology is somewhat different and more
challenging than low power or low noise amplifiers. Specific figures of merit
(FoMs) have been defined for characterization and comparison purpose of power
amplifiers. Among them power added efficiency (PAE), drain efficiency (DE), 19
dB compression point, harmonic suppression, bandwidth, gain, AM/AM and
AM/PM characteristics, are the most widely used. In the next section, the FoMs
and different classes of operations will be presented.
1.2.1. Figures of Merit
A large portion of the total energy consumed in any radio transmitter is
devoted to the output stages. Therefore, the efficiency is a critical factor when
evaluating different PAs. It characterizes the effectiveness of the DC-to-RF power
conversion of an amplifier. It is also a measure of dissipated power as heat in
percent. Low efficiency can be seen in high electricity cost, low battery life and
heating effects. Heating can significantly compromise the reliability and lifetime
of the whole transmitter; hence, in some cases, special cooling systems are
required [28].
Figure 6 shows a general block diagram of a power amplifier. The ratio of
the fundamental RF power delivered to the load, RL, (
to the DC power
handled from the VDD, supply voltage, ! is called drain efficiency (DE) and is
calculated in Equation 3.
10
Figure 6: Power Amplifier block diagram
" !
#! Equation 3
Transistor power gain (GT or G) is finite and drops as frequency increases.
Therefore a more inclusive definition for power efficiency should also include the
effect of input power ( $% ). Power added efficiency (PAE) is defined in Equation
4 [28].
& ' $%
!
' $%
1
(1 ' +
#! #! *
1
" (1 ' +
*
Equation 4
Transistor gain is highly dependent on bias conditions, and it drops
quickly at higher output powers. This behavior contributes to the main part of
nonlinearities. Power amplifiers directly affect the signal integrity. Therefore, the
RF PA nonlinearity should be low enough in order to satisfy the modulation
scheme specifications. Figure 7 shows the generic nonlinear behavior of a power
11
amplifier. The 1-dB compression point is noted on the plot. It is defined as the
point where the gain drops by 1 dB compared to the small signal gain [28].
Figure 7. Power Gain vs Output Power
Communication standards put some requirements on both the in-band
signal quality and out of band harmonic emission of the transmitted RF signal.
Second and third harmonic suppression characterizes nonlinearities and out of
band emission behavior of a power amplifier. Harmonic suppression is defined as
the ratio of power delivered to load at the second or third harmonic to the amount
of power located at the fundamental frequency, while applying a single tone
sinewave to the PA’s input. Figure 8 shows an example of a PA output spectrum.
Second and third harmonic suppression levels are also marked on the plot [28].
12
Figure 8. An example of power amplifier output spectrum
Inter-modulation distortion is another consequence of the PA nonlinearity.
In many applications, the modulated signal has a wide bandwidth and PA
nonlinearity introduces new distortion products which fall inside the signal
bandwidth. A conventional approach for in-band distortion characterization is a
two-tone test. If two sinusoidal signals, with frequencies equal to f1 and f2, are
injected at the input of a nonlinear amplifier, the output signal contains additional
frequency components called inter-modulation products.
Figure 9 shows the generation of third-order inter-modulation products
[29], [30]. Besides the input signal harmonics, the amplifier generates new
frequencies located at nf2-mf1 where m and n are integers. Among these products,
2f2-f1 and 2f1-f2, third order inter-modulation products, are the most critical as they
may fall within the channel bandwidth and interfere with the desired signal [29],
[30]. The inter-modulation distortion to carrier signal power ratio is usually
measured and used to characterize the PA nonlinearity.
13
Figure 9. Third Order Inter-modulation products
A great deal of effort has been employed to design high efficiency power
amplifiers satisfying the required linearity on the transmitter side. Different
classes of operation and more advanced transmitter architecture have been
developed [28].
1.2.2. RF Power Amplifier Classes of Operation
1.2.2.1.
Linear Power Amplifiers
Depending on biasing and matching network conditions, the RF PA can
operate at different classes. Each class of operation has its own tradeoff between
linearity and efficiency. Among all the different classes, class-A is known as the
simplest and most linear PA, but the least efficient class of operation.
Figure 10 shows the schematic of a class-A amplifier. IDC is the transistor
quiescent current.
14
Figure 10. Class A amplifier
The main assumption in a class-A design is that the bias levels are chosen
so that the transistor operates in the linear region and has 360 conduction angles
[32]. This condition implies that the transistor passes the entire waveform at the
input. In theory, class A does not add any distortion to the signal, but in practice
the amplifier gain is a function of input power. So large signal swing at the gate
causes gain variation which is a source of distortion.
The matching network between the FET drain and load has an inherent
filtering effect which improves the out of band harmonics suppression. It also
transforms the load impedance; 50 Ohm is used in most of the cases, to the
desired impedance seen from the drain at the fundamental frequency. Maximum
output power (Pout) of a class-A amplifier is derived in Equation 5.
15
,
2 -
Equation 5
This shows that the maximum output power depends on supply voltage
and the impedance presented to the transistor by the matching network.
Theoretically, the peak class-A efficiency is 50%, when the DC current (IDC) is
equal to the RF signal amplitude (IRF). Equation 6 shows class-A drain efficiency
under the above conditions [28].
" .30 /2 #12 .30 /2 ,56 $7819:;
./0 #12
<========> " 50%
!
#12 Equation 6
Based on the above formula, to improve the efficiency, the DC current has
to be decreased. One method is to decrease the conduction angle of the waveform
which ultimately distorts the output signal. Deviating from a 360 degree
conduction angle changes the class of operation from class A to class AB, B, C,
with better efficiency. Figure 11 shows different classes of operation depending
on quiescent current (ID). Vgs is DC voltage across the transistor gate. Also their
efficiency versus conduction angle is plotted in Figure 12. Class C gives the
highest efficiency at the cost of significant nonlinearity [28], [30], [31], [32].
16
Figure 11. The quiescent bias point on the transistor transfer characteristic [28].
Figure 12. PA efficiency vs class of operation [28].
17
All the above calculations were done based on the fact that the RF
transistor is ideal. However, in practice rds=∞, VKNEE=0, G=∞ and maximum POUT
do not happen. An example of a class-A PA design using a non-ideal transistor
explained here shows how efficiency and linearity degrades based on our device
choice.
Figure 13. Transistor Family of Curves
An example of a more realistic transistor family of curves (FOC) is shown
in Figure 13. The drain voltage swing is limited to VBD from the top and VKNEE
from the bottom. For example, if VKNEE=0.5V the peak to peak voltage swing is
reduced to 2.5V instead of 3V for the ideal case. Drain efficiency decreases as the
knee voltage increases. In this example it drops by 9%:
" BCC DBEFGG HIJ K65L3MK
BCC
<========> " 41%
18
Equation 7
Another consideration is transistor gain at GHz frequencies. Figure 14
shows an example of the maximum available power gain for a transistor versus
frequency. For this example, the maximum power gain at 1GHz is assumed to be
10 dB. Putting this number in the PAE equation (4), the peak PAE of class-A in
this example drops to 37%, comparing to the 50% of an ideal transistor.
1
& " (1 ' + 41 1 ' 0.1 37%
*
Equation 8
Figure 14. Transistor power gain versus frequency.
Maximum available frequency, fMAX, is the point where the gain plot
crosses the x-axis, corresponding to MAG=0 dB. Typically, to achieve 10dB of
power gain, a device with 5-10X higher fMAX than the desired operational
frequency is needed. Measurement results in Chapter II shows that most of the
fabricated MESFETs have a fMAX greater than 10GHz, which is enough for a 2.5
GHz PA application.
19
Another practical design consideration is the peak output power. Based on
Equation 5, for a fixed VDD, the peak POUT would increase by reducing ROPT. The
ratio of ROPT over RL is defined as the impedance transformation ratio, r. If the
supply voltage reduces by a factor of two, then r should increases by a factor of
four to achieve the same power. High r values though increase matching network
sensitivity, loss and reduce the PA bandwidth [30], [31], [32].
Impedance transformation can be done using different topologies. A few
of the most commonly used transformers on CMOS technology, L-matched, Tmatched and π-matched networks, are shown in Figure 15. The following
calculations are based on an L-network, which has the minimum number of
passive components.
Figure 15. Different impedance transformers using lumped components.
A detailed analysis of the transformer is discussed in [31], pages 119-130.
There, the quality factor, Q, of the matching network is defined as:
Q √r ' 1
Equation 9
20
Knowing Q and the quality factor of inductors, QIND, the efficiency and
bandwidth of the transformer can be calculated, where the f0 is the center
frequency:
BW WX
Y
WWZ[\5]J^.\LK\ Equation 10
1
Y
1_Y
Equation 11
$%
Figure 16 shows the efficiency of the matching network versus different
inductor quality factors and impedance transformation ratios. A typical Q value
number for a high quality, on-chip inductor on a CMOS process is around 10-15.
Therefore, the matching network of a 1W single transistor class-A PA on a 45nm
CMOS technology has less than 60% efficiency, which is due to the high
impedance transformation ratio and low quality inductors on CMOS processes.
The previous examples in this Chapter show that peak class-A PAE, considering
gain and kneevoltage non-idealities, is around 37%. Based on Figure 16, this
value is much lower, 19-22 % after the matching network.
21
Matching Network Efficiency (%)
100
90
80
70
60
50
40
30
Q=5
Q = 10
Q = 15
Q = 20
20
1
10
100
T ran sfo rm atio n R atio (r)
Figure 16. Matching network efficiency versus impedance transformation ratio for
different inductor quality factors.
1.2.2.2.
High Efficiency Power Amplifiers
All the above calculations show that a practical class A efficiency can be
much lower than the 50% calculated in Equation 6. There is a tradeoff between
efficiency and linearity. When the linearity constraints are somewhat relaxed, it is
definitely better to aim for higher efficiency. Figure 12 shows how efficiency
improves based on bias conditions. Another approach to achieve high efficiency is
through a proper matching network design. Class F, F-1 are the examples of very
energy efficient amplifiers which are designed based on waveform engineering
and harmonic tuning at the load. An example of inverse class-F circuit is shown in
Figure 17.
22
Figure 17. Inverse Class F Amplifier
The series LC resonators at the drain act as short circuit at odd harmonics
of the center frequency (`X ). For infinite number of resonators, the drain voltage
waveform consequently appears ideally as a (half) sinusoid and the current
waveform as a square wave [30], [31], [32]. Example of a voltage and current
waveform at the drain/collector of an ideal inverse class-F amplifier is shown in
Figure 18. 100% efficiency is ideally achievable. In practice, due to transistor
parasitics, the finite quality factor of matching components and the finite number
of resonators this value degrades more than 15%. The factor also depends on the
device technology and fabrication process and materials used.
23
Figure 18. Class F drain waveforms
Switching amplifiers are another example of high efficiency amplifiers.
Linear amplifiers like class A, AB, B or C are designed based on the transistor
transconductance which is acting as voltage dependent current source. Since a
pseudo class-E amplifier is designed in this work, the theory behind this class of
operation is described in more detail in ensuing sections.
Unlike these classes, in switched mode PAs, the transistor is operating as a
switch having two states; ON or OFF. In theory they can achieve almost 100%
efficiency by satisfying the following condition; when the transistor is turned ON,
the voltage across the switch is almost zero, and high current is flowing through
the device [30], [31], [32]. In other words, the transistor acts as a low resistance
(closed switch) during this first part of the period. When the transistor is turned
OFF, current through the switch is zero, and a high voltage is applied across the
device, i.e. the transistor acts as a high resistance (open switch) during the other
24
part of period. For this reason, SMPAs such as classes D, E are inherently
nonlinear but highly efficient. All these classes ideally achieve 100% power
efficiency with nonzero output power [30], [31], [32].
Figure 19. Class E amplifier with lumped elements matching network
The simplest form of class E topology is sketched in Figure 19 [28]. The
RF choke provides a DC path to the supply voltage and has very high impedance
at high frequency. High efficiency is achievable under zero voltage switching
(ZVS) condition. Figure 20 shows a simplified class E. The transistor is replaced
with a switch and the inductor is split into two inductors La and Lb. (La-C)
resonates at desired center frequency [28].
25
Figure 20. Simplified class E amplifier
Referring to the “RF power amplifier” book [31], section 5.6.4, satisfying
the ZVS condition results in the following formulas for the inductor and
capacitors.
ab fg dd 8
0.1836
_ 4`
`
d ' 4
2d _ 4` ab
8
. d _ 4
Equation 12
Equation 13
Equation 14
Class E operation is not guaranteed for the whole frequency range. There
is a maximum frequency (fmax) of operation which is:
WL56 0.05066
.
a.
Equation 15
As Equation 8 shows, the class E output power (Po) depends on load
resistance R. Lower resistance gives higher output power. On the other hand,
efficiency goes down as the resistance decreases. Besides the analysis which is
26
only valid for class E operation, higher impedance transformation results in higher
loss in the matching network.
Figure 21 plots the relation between class E efficiency and load resistance.
Figure 21. Class E efficiency vs load resistance [28].
As it was discussed class E, F and inverse F improves the peak efficiency,
however in many applications the average power is much less than the peak
output power. There are more advanced transmitter architectures, such as polar
transmitter and Doherty amplifiers that improve the average efficiency. Few of
these methods will be overviewed in the last chapter of this thesis. A novel high
efficiency transmitter based on MESFET characteristics will also be proposed
[28].
1.3.
Summary and Discussion
Problems associated with RF power amplifier design on deep submicron
CMOS technologies were discussed and SOI-MESFET was presented as an
27
alternative cheap solution. The SOI-MESFET device structure, behavior and its
advantages and disadvantages comparing MOSFET devices on the same process
were compared. The effect of device non-idealities on the PA performance, like
PAE, gain and POUT were then overviewed.
The following chapters will talk about measured characteristics of SOIMESFET devices and how they impact power amplifier design using this
technology. To demonstrate the idea of MESFET PA, a few RF output stages
have been fabricated and measured. Their results will be presented in subsequent
chapters. A fast and low cost approach to design these circuits is to start with
simulations. Therefore, a compact large signal model was developed for measured
devices.
28
CHAPTER 2
CHARACTERIZATION AND MODELING
Circuit design, and more specifically RF circuit design, would not be
possible without knowing accurate characteristics of active and passive devices
over a wide range of frequencies. Moreover, having a compact computer aided
design (CAD) model reduces the design cost and time to market. The accuracy of
the model is dependent on the precision of the measurements and its capability to
capture non-idealities. Thus to have a solid design based on simulations, the
MESFETs need to be well characterized in advance. DC and RF characterization
of SOI-MESFETs are discussed in this chapter. A scalable compact large signal
model is also developed for our fabricated devices.
The MESFETs have been fabricated at different SOI-CMOS technology
nodes and foundries. Two of the earlier MESFET demonstrations made use of
partially depleted (PD) SOI technologies at the 600 nm [22] and 350 nm CMOS
technology nodes [20], [27]. In this chapter recently taken data from PDMESFETs fabricated on 150 nm and 45nm SOI CMOS processes will be
presented. These devices represent the most aggressively scaled and highest
performing Si-MESFETs to date with gate lengths as short as 150 nm. Where it is
applicable, the results are compared to those from devices fabricated using the
earlier 350 nm process. This comparison between different foundries provides
insight into the MESFETs performance improvement from one technology node
to the next.
29
2.1.
DC Characterization
The SOI MESFETs are depletion mode devices, thus they have negative
threshold voltage. An example of a MESFET drain and gate current versus gate
voltage at a fixed drain voltage for four different devices is shown in the Gummel
plot in Figure 22.
Lg = 150 nm
Current (mA/mm)
102
0
10
Lg = 400 nm
Lg = 200 nm
Lg = 300 nm
-2
10
-4
10
Vd = 2 V
LaS = LaD = 300 nm
-6
10
-2
-1.5
-1
-0.5
0
Gate Voltage (V)
0.5
1
Figure 22. Gummel plots for LG=150, 200, 300 and 400 nm on a 150 nm SOI
CMOS technology.
From Figure 22 it is evident that the aggressively sized gate lengths on the
150 nm process can be affected by short channel effects (SCE). The gate length at
which SCE ceases to be an issue becomes clearer in Figure 23 which extracts the
threshold voltage, Vt, for various MESFETs. From the figure we conclude that
gate lengths Lg ≥ 400 nm are required to avoid SCE altogether in the 150 nm
technology, while Lg ≥ 600 nm is required for the 350 nm process. Presumably,
30
the SOI channel in the 150 nm technology is thinner and more heavily doped than
in the older 350 nm technology which allows the MESFETs to be scaled to
shorter gate lengths before SCE become significant. This observation is consistent
with the threshold voltage model developed by Chiang et al. for short-channel
SOI MESFETs [33], [27].
Threshold Voltage, Vt (V)
0.0
150 nm SOI CMOS Process
-0.5
-1.0
350 nm SOI CMOS Process
-1.5
-2.0
-2.5
0
0.2
0.4
0.6
Gate Length (µm)
0.8
1
Figure 23. Threshold voltage extracted for each of the devices fabricated on
150nm and 350nm technologies.
MESFETs have a much higher breakdown voltage compared to the
MOSFETs on a same process. The breakdown measurement approach used in this
work is based on the drain-current injection technique [34]. The biasing metric of
1 mA/mm was used as the constant current source forced into the drain.
Breakdown voltage measurements for the MESFETs on 45nm and 150nm SOI
CMOS technologies are shown in Figure 24.
31
Voltage Breakdown, VBD (V)
14
L =L
g
12
aS
= 600 nm
10
L = 400 nm, L
g
8
aS
L = 600 nm, L
g
6
aS
= 600 nm
= 300 nm
4
2
L = 150 nm, L
g
aS
= 300 nm
L =L
g
aS
= 300 nm
0
0
1
2
3
Drain Access Length, L
aD
4
(µm)
5
(a)
Soft Breakdown Voltage (V)
30
25
20
L
15
aS
(nm)
200
300
10
500
2000
5
0
500
1000
1500
Drain Access Length, L
2000
aD
2500
(nm)
(b)
Figure 24. The breakdown voltage of MESFETs (a) on a 150nm SOI process. (b)
on a 45nm SOI process. Dashed line shows the trend and the red cross shows to
the peak measured breakdown voltage.
32
To date the peak measured breakdown on the 150 nm process was ~12 V
and ~28 V on a 45nm process. These are considerably lower than ~55 V [22]
achieved on the 350 nm process. The reason for the significant drop is likely the
results of several different factors. Among them are channel doping, channel
thickness, and various implants introduced on deep submicron processes that may
not have been blocked during layout, but are not seen on the 350 nm process.
Overall, MESFETs have shown 2-30X higher VBD than standard MOSFETs. As a
reference, the maximum steady-state operating voltages of the standard CMOS
devices is 0.9 V, 1.95 V, and 3.5 V for the 45 nm, 150 nm and 350 nm processes
respectively. VBD improvement is not free of charge, and we will see in the
subsequent chapters how it affects the RF performance of the proposed
MESFETs.
Lg = 200nm LaS = 1000nm LaD = 1000nm
0.08
Drain Current (A/mm)
0.07
0.06
Vg = -0.5 to 0.5V in 0.25V Steps
0.05
0.04
0.03
0.02
0.01
0
0
5
10
15
20
Drain Voltage (V)
Figure 25. The family of curves for an enhanced VBD MESFETs with LG= 200nm,
LaS = LaD = 1 µm. The gate voltage is stepped from +0.5V (uppermost curve) to 0.5V in 0.25V steps.
33
To visualize the explained breakdown voltage improvement, FOCs, for a
high VBD MESFET on a 45nm technology, are plotted in Figure 25.
2.2.
RF Characterization
RF characterization was performed by on-wafer probing using select
devices with ground-signal-ground (GSG) pad configurations. Also included on
the die was an accompanying set of open- and short-circuit test structures to deembed the devices and remove the parasitics of the GSG pads. Measurements
were taken using Agilent 8510C vector network analyzers. From the deembedded S-parameters, WinCal [35] was used to extract the cut-off frequency,
fT, of the MESFETs which was defined as the point where current gain, |h21|2,
Peak Cut-off Frequency, fT (GHz)
equaled 0 dB [27].
50
150 nm SOI CMOS Process
LaS =LaD = 300 nm
30
20
10
8
6
4
350 nm SOI CMOS Process
LaS = LaD = 1 µm
2
0.1
0.2
0.3 0.4
0.6
1.0
2.0
Gate Length (µm)
Figure 26. Shows the exponential scaling of the peak cut-off frequencies for
MESFETs manufactured using different SOI CMOS processes.
34
Similar to a MOSFET, the cut-off frequency depends on gate length. The
fT trend is shown in Figure 26. This graph shows the measured peak fT for various
gate lengths on different SOI CMOS technologies. Cut-off frequency is a function
of device transconductance and gate capacitance:
W h
1 iL
2d ajk
Equation 16
CGS decreases as gate length shrinks and at the same time gm improves as
well. However, LaS and LaD can also significantly affect the MESFET
transconductance. From a circuit perspective, LaS and LaD appear as parasitic
resistances in series with the channel of the MESFET [36], [27]. Source resistance
degenerates the device transconductance and results in lower gm. This directly
affects the fT performance of a MESFET. Figure 27 shows the measured RS and
the peak fT versus access length for MESFETs with the same gate length. As
expected, fT reduces by increasing LaS.
15
fT
RS
25
T
12
20
9
15
S
6
10
Extrinsic R (Ω
Ω)
Cut-off Frequency, f , (GHz)
30
3
5
0
0
0
500
1000
L
aS
=L
aD
1500
2000
(n m )
Figure 27. The peak cut-off frequency and RS as a function of LaS for MESFETs
fabricated on a 45nm SOI CMOS technology.
35
Cut-off frequency is a widely used term in the literature for evaluating RF
performance of a technology. However, maximum available gain (MAG) and fMAX
are more critical parameters for RF power amplifier designers. The breakdown
voltage also plays an important role. As shown in Figure 24, VBD depends mainly
on LaD for a fixed LG, and it improves by enlarging the drain spacer, LaD. The
effect of LaD and other parasitics on MESFET RF and breakdown voltage
performance is presented below:
fMAX is a function of cut-off frequency but it also depends on the gate to
drain capacitance and gate resistance as well [31]:
W,
lm
W
1
iL
l
m
8daj j 4d aj ajk j
Equation 17
Based on Equation 17, fMAX depends on gm, RG, CGD and CGS. RS also
appears in the MESFET source and reduces the overall transconductance by
degenerating the intrinsic gm. Hence, it is important to calculate these parasitics
for different MESFET geometries. RG, RD and RS can be considered extrinsic
components; therefore they do not depend on the biasing conditions. A wellknown method called cold-FET extraction [37] is used in this work to measure
these extrinsic resistances. Measured resistances versus LaD for various fabricated
MESFETs are shown in Figure 28. RS and RG do not change with LaD for a fixed
LaS and LG, therefore the MESFET’s fT and fMAX remain constant.
36
Resitance (mΩ.
Ω.
Ω.mm)
6
5
RG
RD
RS
4
3
2
1
0
300
400
500
600
L
aD
700
800
900
1000
(n m )
Figure 28. Measured extrinsic resistances versus LaD for MESFETs with LG=400
nm and LaS=300nm on a 150 nm process.
Intrinsic capacitances also affect the MESFET fMAX. Figure 29 shows that
LaD does not change CGD, while CGS and gm scale proportionally with LaD.
Overall, increasing LaD does not degrade the MESFET RF performance and
benefits it in the form of higher breakdown voltage. It should be mentioned that
there is a limit on the VBD improvement by just changing LaD as it is shown in
Figure 24.a. Furthermore, increasing LaD increases the overall layout area which
can be significant in a high current device.
37
90
800
C gs
C gd
85
C ds
600
500
80
400
75
300
200
gm (mS/mm
Capacitance (fF/mm)
700
gm
70
100
65
0
300 400 500 600 700 800 900 1000
L
aD
(n m )
Figure 29. MESFETs intrinsic capacitances and gm versus LaD for MESFETs with
LG=400 nm and LaS=300 nm on a 150nm process.
The above discussion underlines the importance of appropriately sizing
LaS and LaD for specific applications. Measured MAG for different MESFETs is
plotted in Figure 30. All the measured devices have more than 15dB available
gain for frequencies less than 2.5 GHz, which is an adequate amount of gain for
most RF PA applications. The peak measured fMAX on the 45nm process is
~35GHz for MESFET with LG=LaS=LaD=200nm.
38
Maximum Available Gain (MAG) (dB)
40
200nm
500nm
2000nm
1000nm
35
30
25
20
15
10
5
0
8
10
9
10
10
10
11
10
Frequency (Hz)
Figure 30. Measured MAG versus frequency for different MESFETs on a 45nm
SOI CMOS process.
The measured cut-off frequency and fmax for a few selected devices on
45nm and 150nm processes is plotted in Figure 31 as a function of drain current.
The results from the 150nm MESFETs show the impact of RG on fmax as the gate
lengths is scaled. It can also be explained using the fmax approximation in Equation
17. Values for RG have been extracted for each device using the cold-FET
technique and are shown in the inset to Figure 31. A power law fit shows that RG
varies proportionally with 1/LG as indicated by the solid line in the inset. For the
300 nm gate length device the peak fmax is ~30% higher than the peak fT.
However, as RG increases with decreasing LG this trend is reversed, with the peak
fmax of the 150 nm device being 30% lower than the peak fT.
39
40
L = 150nm
30
g
10
40
L = 200nm
30
g
20
R G (Ω)
Frequency (GHz)
20
10
40
30
60
50
40
30
100 200 300
LG (nm)
Lg = 300nm
- - - fmax
20
___
10
-3
-2
10
10
Drain Current (A)
(a)
40
f
T
40
35
30
LaD=LaS=1000nm
LG=200nm
25
20
15
10
fMAX
fT
5
0
0 .1
40
35
30
1
10
100
10
100
10
100
LaD=LaS=500nm
LG=200nm
25
20
15
10
5
0
0.1
1
40
35
30
LaD=LaS=200nm
LG=200nm
25
20
15
10
5
0
0.1
1
Drain Current (mA)
(b)
Figure 31. shows the cut-off frequencies and fMAX versus drain current for selected
devices (s) on the 150 nm SOI CMOS process (b) on the 45nm SOI CMOS.
41
2.3.
TOM3 Modeling
Simulation is an essential part of circuit design as it greatly reduces the
design time and considers non-idealities of a device or technology. Device
behavior and non-idealities should be accurately captured and reflected in device
models for circuit simulation. Various compact models have been developed for
different fabricated MESFETs. The modeling process and approach is the same
for all our devices. The following describes the compact model developed for one
of the SOI MESFETs fabricated on a 45nm SOI CMOS process.
Silicon MESFETs were fabricated at a commercial 45nm semiconductor
foundry with LaS = LaD = 100nm, LG = 184nm and gate width LW = 300µm.
Initially, twenty different devices were measured using a DC probe station and an
average device was chosen to fit a model to. This device showed the most average
gate, drain and source currents compared to all measured results. The TriQuint
Own Model, TOM3 [38], was chosen to model the results because it is an
extension of the Curtice-Statz [39] model and most importantly has a continuous
expression for drain current across sub threshold, linear and saturation regions of
device operation. Additionally, the model is widely available for different circuit
simulators including Agilent ADS and Cadence. Figure 32 shows the circuit
schematic of the complete TOM3 model [40].
42
Figure 32. TOM3 model circuit schematic [38].
As the first step, extrinsic components, such as RG, RD and RS, are
extracted using cold-FET techniques explained in [37]. Measured extrinsic
resistances versus frequency are shown in Figure 33.
Resitance (Ω)
10
8
6
4
RD
2
RG
RS
0
100
1000
Frequency (MHz)
Figure 33. Measured extrinsic RD, RS and RG versus frequency.
43
2.3.1. DC Model Characterization
First, the model was developed to describe the device operating at DC and
then charge parameters were determined for RF as described in the subsection
below. Initially, family of curves measurements were used to fit the model to the
drain current of the MESFET, which is described in the TOM3 model equations
as
I ds = β × (VG ) × f k × (1 + λVds )
Q
Equation 18
where β is the transconductance, VG is a scaled nonlinear voltage as a function of
region of operation, Q is the power generalizing the square-law, fK relates the
saturation parameter α approximately as
f k ≈ tanh(αVds)
Equation 19
where Vds is the drain to source voltage and λ is the channel length modulation
parameter. VG can initially be thought of as the gate voltage in a simple squarelaw, but is in fact determined nonlinearly in the model as a function of internal
gate diode voltage, threshold voltage and drain voltage. Figure 34 shows the
model fit (open circles) to measured data (solid line) for the fabricated MESFET
with Vds to 4.5V and Vgs from -0.5 to 0.5V. It is important to note that the
nominal MOSFET device operating voltage is less than 1V for this process.
The transconductance in units of mS/mm as a function of gate voltage is plotted
in Figure 35 from low (0.1V) to high (4V) drain biases [40].
44
0.030
Vg = -0.5 to 0.5 in 0.25V Steps
Drain Current (A)
0.025
0.020
0.015
0.010
0.005
0.000
-0.005
0.00
1.00
2.00
3.00
4.00
Drain Voltage (V)
Figure 34.
Measured (solid line) and simulated (open circles) results for the
silicon MESFETs fabricated on the 45nm SOI process. The line at 1V represents
the approximate maximum operating voltage of MOSFETs on the same process
[40].
g m (m S /m m )
150
Vd=4V
100
Vd=2V
50
Vd=0.5V
0
-50
-1
Vd=0.1V
-0.5
0
Gate Voltage (V)
0.5
Figure 35. Measured (solid) and modeled (open circles) transconductance.
45
In order to get a good agreement to the measured data across drain
voltages, the solution to fk must be precise. Using measured data for Ids, we can
solve Equation 18 for fk (Figure 36) to get an empirical fit for α. This enables a
smooth and correct transition from linear to saturation region in the model [40].
1.5
Vg=0.5V
fK
1
0.5
0
0
1
2
3
Drain Voltage (V)
4
Figure 36. Model fit of the saturation current scaling fk (circles) compared to
extracted fk from measured drain current (solid line) [40].
After matching the drain current, the next step is to model the gate current,
leakage and breakdown of the MESFET which is represented by internal diodes in
the TOM3 model (Figure 32 D1-D4). The MESFET gate is a Schottky diode and
under forward bias can be determined by the saturation current Is and ideality
factor N.
Additional diodes depict the leakage current under reverse bias.
However, other effects such as edge effects, device doping and DIBL are not
completely captured by the model and there is a slight difference in gate current
between measured and modeled results. Our current research includes additional
methods to capture these effects. Figure 37 shows the measured and modeled
46
Gummel plots of the MESFET which depict both absolute gate and drain current.
The model fits across a wide drain bias range which is critical for reliable
simulations [40].
Vd=4V
-2
10
Drain Current (A)
Vd=0.1V
-4
10
10-6
10-8
-10
10
-1
-0.5
0
0.5
Drain Voltage (V)
Figure 37. Gummel plots depicting drain and gate current of the MESFET across
a wide drain bias range of 4V and 0.1V. Measured results are solid lines and
modeled open circles. Additional drain voltages have been omitted for clarity but
show a good fit to measured data from Vd = 0 to 4V [40].
2.3.2. RF Modeling
In order to use the model for microwave simulations, charge parameters
must be determined to describe the parasitic capacitances of the MESFET Cgs,
Cgd and Cds (Figure 32). S Parameter measurements from 100MHz to 40GHz
were performed on an RF probe station on GSG MESFET structures also used for
the DC model extraction. The total gate charge in the TOM3 model can be
described as
47
QGG = QGL × T + QGH × (1 − T )
Equation 20
where QGG is the total charge on the gate, QGL is the low power region function,
QGH is the high power region function and T describes the transition between
regions as
T = exp(−QGGB × I ds × Vds )
Equation 21
with QGGB as a transition coefficient [40]. In order to get the model to agree
across different power levels, Equation 20 is simultaneously fit to the measured S
parameters across different gate and drain bias levels. Figure 38 shows measured
and simulated S Parameters for a drain bias of 2V and gate bias of 0.25V. The
maximum available gain (MAG) and H21 were calculated for both measured and
simulated data and are shown in Figure 39.
0 .0 0
S11 and S12 (dB)
S11
- 1 0 .0
- 2 0 .0
S 12
- 3 0 .0
- 4 0 .0
- 5 0 .0
0 .0
5 .0
10
15
20
25
30
F re q u e n c y (G H z )
48
35
40
S22 and S21 (dB)
15
10
S21
5
0
-5
S22
-10
0.0
10
20
30
40
Frequency (G H z)
Figure 38. Measured (solid line) and modeled (Open red symbols line) S
Parameters at Vd=2V and Vg=0.25V.
H21 and MAG (dB)
5 0 .0
4 0 .0
H21
3 0 .0
MAG
2 0 .0
1 0 .0
0 .0 0 8
10
10
9
10
10
10
11
F re q u e n c y ( G H z )
Figure 39. Measured and simulated H21 and MAG at the same bias conditions.
2.3.3. Test Power Amplifier
In order to validate the large signal behavior of our model, a test Class AB
power amplifier operating at the ISM band of 433MHz was designed, simulated
and then fabricated on an FR4 PCB board. A MESFET device with LW=2.7mm
49
was packaged and bonded to a Kyocera A309 package utilizing multiple
bondwires to reduce parasitic inductances. The MESFET was chosen so that it
would be optimally matched to 50Ω and simplify matching networks for the
demonstration. Figure 40 shows the schematic of the PCB board. The power
amplifier was tested with a network analyzer that was used to supply varying
input power at 433MHz and a power meter to measure the output power. Figure
41 shows the simulated model compared to measured amplifier results for key
parameters such as output power, gain, drain efficiency (DE) and power added
efficiency (PAE) [40].
Figure 40. Class AB, 433MHz, RF Power amplifier schematic
50
60
20
PAE
50
15
DE
40
POUT
30
10
20
Efficiency (%)
Pout (dBm), Gain (dB)
GAIN
5
10
0
-10
-5
0
5
0
10
Pin (dBm)
Figure 41. Measured (symbols) and simulated (solid line) of gain, output power,
power added efficiency (PAE) and drain efficiency (DE).
2.4.
Summary
Various MESFETs fabricated on different SOI-CMOS processes were
measured. DC and RF characteristics of these devices were shown in this chapter.
The first chapter discussed the need for high breakdown voltage devices on
CMOS technologies for power amplifier applications. 2-30X breakdown voltage
enhancement were achieved using MESFETs comparing to MOSFETs on the
same SOI-CMOS process.
RF measurements proved that MESFETs can be used in power amplifiers
operating in the frequency range of DC-2.5GHz. It will be shown in the next
chapter that RF performance can be improved at higher frequencies. Also a
TOM3 model was developed and validated for silicon MESFETs fabricated on a
51
45nm CMOS process through simulation of measured devices and a 433MHz
Class AB power amplifier.
52
CHAPTER 3
SOI-MESFET POWER AMPLIFIERS
The design of RF power amplifiers are certainly a complex task.
Nevertheless, this complexity, when coupled with extra limitations related to the
fabrication technology, can lead to even more challenging design factors. This is
particularly true in the case of CMOS power amplifiers in sub-micron
technologies. Indeed, one of the main obstacles in new CMOS technologies is the
low supply voltage that results in very low output power.
In addition, this
technology is characterized with a very low drain breakdown voltage that
prevents adequate implementation of advanced power amplifier classes of
operation, such as class E, due its excessive peak drain’s voltage that can reach as
much as 3.5×VDD. Furthermore, passive RF components, inductors and
capacitors, are another source of challenges in a fully integrated CMOS design.
This manifests in low quality factor of inductors that leads to excessive losses and
consequently to power efficiency degradation. In addition the large size of passive
components, when compared to the RF transistors, limits the practical output
matching network topologies for fully integrated implementation [28]. Chapter I
discussed how the lower breakdown voltage of CMOS transistors lead to a need
for more passive components and therefore lower efficiency and higher
fabrication cost due to the large die area of inductors. SOI-MESFET devices were
introduced as an alternative solution for the limited headroom of CMOS PAs
which allows digital and RF circuit’s integration on a same silicon die. The first
53
reported fabricated RF power amplifiers using the proposed SOI-MESFET
devices are introduced in this chapter.
Agilent Advanced Design Software (ADS) was used to simulate MESFET
RFPAs before fabrication. An advanced simulation tool called Load/Source Pull
was used to find the optimum device size, geometry and desired matching
network characteristics to achieve the best performance. Load/source pull
approach is explained in the following.
A general power amplifier block diagram is shown in Figure 42. The
maximum power transfer happens when ZOUT is the conjugate of ZDRAIN.
However, the conjugate matching condition does not necessarily result in the
maximum efficiency. The first chapter discussed how the efficiency can be
improved by either tuning the harmonics of the load or waveform engineering
approach. Few classical methodologies like class C, F, F-1, … were introduced
that ideally can give 100% efficiency. These RF-PAs are designed based on
harmonic tuning in the matching network. The equations and load conditions
given for these classes are based on an assumption that the RF transistor is an
ideal unilateral device. In practice though, optimum source/load conditions for a
packaged transistor, operating at giga hertz range, might be very different than
what the class E and F equations suggests. Load/source pull technique helps
taking into account all the non-idealities such as finite reverse isolation and gain
compression of the RF transistor [28].
54
Figure 42. A general Power Amplifier block diagram.
Figure 43 shows a basic block diagram of a load/source pull setup. This
technique has been widely used for many years [41] and it is based on impedance
tuning at the drain and gate of the transistor to find the optimum source/load
conditions. This system can be used to design for the best linearity, efficiency,
bandwidth or maximum power delivery to the load. Design steps are the same for
all these cases, but the goal function might be different for different applications.
The optimization steps are explained below:
Step I: Set initial values for the variable load (ZOUT) and source (ZIN) at the
fundamental frequency and its harmonics (3, 4 harmonics).
Step II: Sweep the fundamental load or source impedance all over the smith chart
and find a point on the smith chart which gives the maximum PAE, linearity, etc
(goal function or solving the existing tradeoffs).
Step III: Change the load/source impedances to the new point calculated in the
step II.
55
Step IV: Sweep the other harmonics at the load and source and perform step II
and III (this time for harmonic frequencies).
Step V: Repeat the step III and IV until results converge to the best possible
performance.
Step VI: Repeat the step II to V until achieving best possible PAE or goal
function [28].
It has to be mentioned that this is not a blind load manipulation approach,
and the initial values should be calculated by knowing the RF transistor family of
curves, device parasitics and design goals. Biasing conditions and initial values of
load at the 2nd and 3rd harmonics can highly affect the final result.
Figure 43. Source/Load Pull Simulation Setup [28].
Load/source pull simulation is used to demonstrate the feasibility of the
proposed MESFET power amplifiers. The TOM3 model developed for the
fabricated MESFETs has been used for these simulations in ADS. The first
fabricated power amplifier using SOI-MESFETs is an overdriven class AB PA
operating at 433MHz.
3.1.
Wideband Class AB SOI-MESFET RF PA
A MESFET transistor fabricated on a 150nm SOI-CMOS process was
used to design an RF-PA operating at the center frequency of 433MHz with more
than 40% bandwidth. This device has ~5V breakdown voltage and 22GHz of cutoff frequency. Due to some limitations this device had only one pad for each
drain, source and gate terminals. As a result, the RF performance was
significantly degraded after packaging. Figure 44 shows the MESFET RF current
56
and power gains versus frequency before and after packaging. Although the
device has more than 15 dB power gain up to 2GHz before packaging, bond wires
have extremely degraded the RF gain for frequencies above 500MHz. This
problem can be eliminated by adding more bond pads, which has been included in
more recently fabricated MESFETs [42].
Figure 44. MESFET RF performance before and after packaging. (Red shows
before packaging) (Black shows after packaging)
Chapter I discussed how the impedance transformation reduces the overall
PA efficiency and limits the bandwidth. High transformation ratio is required
when the transistor breakdown voltage is much less than the required voltage
swings on the 50 Ω load. Therefore the main advantage of using MESFETs
instead of MOSFETs in power amplifiers is relaxing the impedance
transformation requisites or even eliminating it whiling meeting the standard
requirements. MESFET used in this work has LG=LaS=LaD=300nm and device
57
width of W=4.2mm. It has a peak drain current of 200mA and can handle up to
~5-6 V. The MESFET’s family of curves is shown in Figure 45. A class AB
power amplifier operating at 433MHz with peak output power of 19dBm was
designed and built using this packaged SOI-MESFET.
Figure 45. Family of Curves for MESFET with LG=LaS=LaD=300nm and
Width=4.2mm.
A load/source pull simulation has been completed in Agilent ADS using
the TOM3 model to find the optimum load/source conditions that give the peak
efficiency. Simulated power and PAE contours are plotted in Figure 46. Drain
(load) PAE contours (Red) at the fundamental frequency show that a close to ~50
Ω load termination results in the maximum efficiency. The optimum gate
impedance that gives the peak PAE and output power is around ~25 Ω. However,
58
PAE degrades by less than 3% if the source termination of 50 Ω is used instead of
25 Ω. Figure 46 shows the existing tradeoff between power and efficiency [42].
Figure 46. PAE and Power contours at drain and gate. Red shows PAE and Blue
is Power contours
Load/source pull simulation results are consistent with load line
calculations. A class AB PA based on the above calculations was designed and
simulated in ADS. A band-pass series LC filter centered at 433MHz is used for
harmonic suppression of the class AB PA. Packaging and bondwire effects are
included in simulations.
59
Figure 47. Schematic of the simulated class AB SOI-MESFET PA
The RF PA shown in Figure 47 was simulated in ADS. Power added
efficiency (PAE), drain efficiency (DE), power gain and output power are plotted
as a function of the input power in Figure 48. The maximum PAE of 58% was
achieved [42].
Figure 48. Simulation results of the 433MHz class AB MESFET PA.
60
The discussed class AB PA was fabricated using FR4 material. A
photograph of the PA module is shown in Figure 49. Measurement results are
shown in Figure 50. A maximum PAE of 48% was measured while delivering
17dBm to a 50Ω load. Few percent variations between measurement and
simulations are due to several reasons. The TOM3 model is initially calibrated
using small signal measurements at different biasing conditions, thus
nonlinearities might not be captured fully in the model. Additionally, die to die
variation, capacitor and inductor models along with inaccuracy in the PCB’s
model are the other possible sources [42].
Figure 49. Photograph of fabricated class AB 433MHz SOI-MESFET PA.
The PA linearity was tested by driving it with a GSM signal at 433MHz.
The input GSM signal power was increased until it reached the 1-dB compression
point of the PA. As specified in the standard, adjacent channel power should be
61
less than -30dBc with respect to the carrier at -/+200KHz and -60dBc at /+400KHz. The PA output spectrum was measured from 432MHz to 434MHz and
the measurement resolution bandwidth is set to 10kHz. The measured output
spectrum is shown in Figure 51 [42].
Figure 50. Measurement results of 433MHz class AB MESFET PA.
Figure 51. Output spectrum of the MESFET PA driven by a GSM signal.
62
The matching network bandwidth directly limits the PA frequency
response. As it is mentioned before, the load/source-pull simulations show that
there is no need for impedance transformation to match this SOI-MESFET device
to a 50 Ohm load. As a result, this PA has a high bandwidth of 200 MHz centered
on 400 MHz. In this work, the band-pass (BP) filter was designed to suppress
harmonics of the RF signal at 400MHz. This limits the overall PA bandwidth, but
it can be either eliminated or be replaced with another BP filter and can be used at
other frequencies as well. Figure 52 shows measured PA performance versus
frequency where the drain efficiency stays above 50% from 280 to 480 MHz [42].
Figure 52. Class AB RF PA measurement results; PAE, DE, Pout and gain versus
frequency.
3.2.
Cascoded MOSFET-MESFET RF Power Amplifier
The previous sub section measurement results demonstrate that SOI-
MESFETs can be used in power amplifier modules to relax matching network
63
requirements of a high power PA. RF measurements of MESFETs with VBD of
28/20 V showed a good performance up to 2.5GHz. The frequency response of
MESFET PAs can be further improved by using a cascode MOSFET-MESFET
pair shown in Figure 53. The following explains how this technique improves the
frequency response of MESFET PAs.
Figure 53. The proposed cascoded MOSFET-MESFET pair.
The breakdown voltage of the proposed structure is the summation of two
devices breakdown voltages. The RF signal is fed to the MOSFET gate, which
has few times higher cut-off frequency than MESFETs on a same process. Since
the MOSFET converts the RF signal into current, the whole cascode pair has
higher fT than the MESFET’s cut-off frequency. The small signal analysis of
cascode pair shows that the overall fT is approximately the geometric mean of the
two devices fT. Following are the small signal fT analysis of the cascode structure
shown in Figure 53:
64
I] jb noI] `ajkb p oI] 'n s!qr and j 'nob `ajk
I
tuv
xv qr
b iLb oI] p b 'n s!
and 'iL ob
p j n
w
s!tuy Iz
I
tuv
Equation 22
Equation 23
wxy
KCL at drain1 node:
b j _ 1 _ n
s!tuy
wxy
and
iLb I]
`ajk
'n
(1 _ n
+
`ajkb
iL
Equation 24
iLb
1
` b
1
'n
'n
I]
`ajkb (1 _ n `ajk +
` {1 _ n ` |
` iL
Equation 25
|~b | | I z ` ` | 1, so ` ? ,
1 _ sy€
Equation 26
` b ' 1 l ‚` ` b ` Equation 27
To find fT
I
qr
y
s€v
y
s€
sy
€y
Solving the above equation:
` ` √2
m1 _ 4
Equation 27 shows that cascode architecture improves the frequency
performance of MESFET PAs.
Comparing cascoded MOSFET-MESFET and MOSFET-MOSFET, the
first structure has much higher breakdown voltage while the RF performance is
comparable with the later design for frequencies less than 10GHz. Figure 54
compares the conventional NMOS-NMOS cascode with the MOSFET-MESFET
65
cascode topology. The transconductance (gm) of a cascode block is mostly
determined by the input device; therefore in Figure 54.b, the MOSFET is used as
the input transistor since it has higher gain and fT.
Figure 54 (a) conventional NMOS-NMOS cascode structure (b) proposed
MOSFET-MESFET power amplifier
In Figure 54.a, a steady-state drain-to-source voltage of each transistor
should not exceed the nominal 3V supply voltage given by the 150nm SOI CMOS
technology. Therefore as a result of the cascode, the supply voltage can be
theoretically increased by a factor of two. In practice however, the VDD scaling
will be less than two because the RF signal division ratio between the top and
bottom transistors is not 1:1 with the top transistor receiving a larger share of the
voltage swing. This implies that higher VDD and consequently a higher output
power level for the PA can be achieved if the top transistor has higher breakdown
voltage. While the MESFET does have a noticeably lower fT and gm compared to
the MOSFET, the overall performance of the PA’s frequency operation and gain
66
should only marginally degrade since the performance of those metrics is
dominated by the bottom transistor. On the 150nm SOI CMOS process the
maximum voltage swing for the circuit in Figure 54.a is ~5V. Replacing the
MOSFET with the MESFET in Figure 54.b, it can be increased to 8 to 12 V
depending on MESFET’s geometry. A cascoded MOS-MES structure was
implemented in Agilent ADS as shown in Figure 54. b. The MESFET’s gate
length was chosen to be 300nm which allowed a 6V signal swing on the common
gate transistor. In our simulations, a MOSFET with LG=180nm transistor is used
as the bottom MOSFET, it has slightly higher breakdown voltage comparing to
SOI-MOSFET with gate length of 150nm [43].
Figure 55. Cascoded MOSFET-MESFET’s Family Of Curves (FOC).
Load/Source Pull simulations were performed and device widths in this
simulation were chosen to deliver 24dBm power to a 50 ohms load. As the first
step the family of curves for the cascoded device is plotted in Figure 55 to find
67
optimum biasing point. As it is shown in this plot, the drain voltage can go up to
10V without even seeing the soft breakdown effect. A high efficiency RF
amplifier at 1GHz was targeted and simulation results showed a maximum PAE
of 75% with around 28dB power gain. Time domain voltage and current at the top
transistor drain is shown in Figure 56. Maximum voltage swing on the common
gate transistor is less than 6V, which guarantees safe operation of the MESFET
device. Gain, output power and PAE are plotted as a function of input power in
Figure 57 [43].
Figure 56. Time domain drain’s current and voltage
68
Figure 57. Pout, Gain and PAE versus Input power.
3.3.
On-Die Measurements of MESFET RF-PAs
Various MESFETs breakdown voltages were measured and are plotted versus
LaD in Figure 58. These devices were fabricated on a 45nm SOI-CMOS process
that has a nominal supply voltage of 0.9V. The peak output power of a class-A
amplifier can be calculated using Equation 5 in the first chapter. The theoretical
peak output powers of different MESFETs are calculated from the breakdown
voltage plot of Figure 58. The theoretical maximum POUT versus drain access
length (LaD) for two different drain terminations of 5 Ω and 50 Ω are shown in
Figure 59. Figure 59.a shows that peak output power of ~20 Watt can be achieved
if a MESFET with LaD = 2µm and loadline of 5 Ω is used. This is much higher
than what most cellphone applications require.
69
Figure 58. MESFETs breakdown voltage versus LaD on a 45nm SOI-CMOS
process.
Figure 59.b shows peak output powers for MESFETs with a 50 Ω
loadline. The output power range is between 22dBm to 33dBm. This range of
power covers many of the applications such as GSM [44] and EDGE [45]. As
mentioned in the first chapter, using a 50 Ω loadline transistor can possibly result
in a higher efficiency and wider bandwidth of the RF power amplifier. The term “
transformer-less “ will be used, in the rest of this thesis, for single ended class
A/AB RF-PAs designed with 50 Ω loadline SOI-MESFETs.
70
Figure 59. Theoretical peak output power of class-A amplifier versus LaD (a)
device loadline of 5 Ω (b) device loadline slope of 50 Ω.
MESFETs breakdown voltage can be controlled by appropriate sizing of
spacer regions between drain, source and gate. Therefore, a transformer-less RFPA can be designed for any application with the maximum output power in the
range of 0.3-2 Watt. The design steps toward transformer-less MESFET PAs are
explained here.
The minimum required breakdown voltage for a given peak output power
is calculated in Equation 28.
ƒ" { 2 | 1-†‡X 5]ˆ wI‰K] -
l
<==================>
2 „…
Š h 2‚2 50 Equation 28
The optimum LaS and LaD can be extracted from Figure 58 to achieve the
above breakdown voltage. Also, the device width should be correctly set to ensure
71
a 50 Ω loadline. Using Equation 29 and knowing the MESFET’s current drive,
device width can be calculated.
#,
Š ' ‹%ŒŒ
<=========> Ž… ‘
/\\K]0 ˆ\I‰K
50
Equation 29
The explained methodology was used to design RF power amplifiers using
the SOI-MESFET. The idea of transformer-less MESFET PA was initially
explored by using the setup shown in Figure 60. The shown test system is similar
to the load/source pull system described at the beginning of this chapter. The main
difference is that here the load and source impedances are constant at all
frequencies, while in the load-pull approach; a load tuner is used to search for the
optimal performance. As a result, the reported efficiencies in this work are not the
highest achievable performances that can be obtained by SOI-MESFETs.
On-die open, short and thru structures in a Ground-Signal-Ground (GSG)
format were used to calibrate out all the losses associated with filters, bias-T,
cables and pad loss.
Figure 60. On-Die MESFET RF-PA measurement setup.
72
A low pass filter used in this work pushes the location of harmonic
impedances to the edge of the smith chart. Therefore the impedance seen from the
drain of MESFET is close to open at the second and third harmonics. The
locations of load impedances on the smith chart are shown in Figure 61. A
parallel L-C resonator is required to filter out harmonics in the traditional class
A/AB amplifiers. In this work we have used a low-pass filter that represents a
series L-C resonator. This technique is known as low-pass class E amplifier [MS].
Although it is not a switching amplifier, it still improves the efficiency, by
influencing the time domain drain voltage and current waveforms. The drawback
of this approach is the higher nonlinearity comparing to the traditional class-A
RF-PAs.
Figure 61. The presented load to the MESFET’s drain at fundamental and
harmonic frequencies.
73
MESFETs fabricated on a 45nm process were tested in this setup.
MESFETs with LG=200nm and LaS=LaD=500nm have breakdown voltage of ~1213. The device used in this work has a loadline of ~80 Ω, thus it can deliver 22
dBm of power to a 50 Ω load. Measurement results at 400 MHz show peak output
power of 22.5 dBm and 50 % power added efficiency. Efficiency and output
power versus input power for a class-A bias condition at two different supply
voltages are shown in Figure 62.
25
40
35
20
30
25
15
20
15
10
PAE (%)
Output Power (dBm)
VDD=6
VDD=4
10
5
5
-12 -10
-8
-6
-4
-2
0
2
0
4
Input Power (dBm)
Figure 62. PAE, Pout vs Pin. Class-A 400 MHz MESFET-PA with LG=200nm,
LaS=LaD=500nm. VG=0.0V
PA class of operation was changed by sweeping gate bias voltage. Figure
63 shows MESFET PA performance versus gate voltage. As the plot shows the
74
output power goes down for lower gate voltages (class-C), but efficiency
increases to 52 %.
60
50
20
40
15
30
10
PAE (%)
Pout (dBm), Gain (dB)
25
Pout
20
Gain
PAE
5
-0.6
-0.4
-0.2
0
0.2
10
0.4
Gate Voltage (V)
Figure 63. 400 MHz MESFET-PA performance at different classes of operation.
VDD=4 V, VSUB=4V, Pin 6 dBm.
The next chapter discusses polar transmitters as a solution to efficiency
degradation at backed-off powers. In detail operation of polar transmitters will be
discussed in Chapter IV. One of the key parameters in designing these type of
transmitters is RF-PA performance versus supply voltage. MESFET RF-PA PAE,
gain and output power versus supply voltage were measured and plotted in Figure
64. As it is shown the peak output power increases with increasing the supply
voltage. Accordingly, the 1-dB compression point increases and results in a better
linearity.
75
Pout (dBm), Gain (dB)
60
20
50
15
40
10
30
PAE (%)
25
Pout
Gain
20
5
PAE
10
0
1
2
3
4
5
6
7
VDD (V)
Figure 64. MESFTE RF-PA performance versus supply voltage. VG=-0.4 V,
VSUB=4V, Pin=6dBm.
Similar
measurements
were
taken
for
a
LG=LaS=LaD=200nm fabricated on a 45nm Technology.
MESFET
with
Three RF power
amplifiers operating at 433, 900 and 1800 MHz were measured. The peak output
power of 21 dBm was measured at 433 MHz. The peak measured power added
efficiency is close to 55% and MESFET PA has 15 dB of gain at 433MHz. The
measured PA performances at different frequencies are shown in Table 1.
76
Table 1. Measurement results of transformer-less MESFET RF-PAs (MESFET
has LG=200 and LaS=LaD= 200nm and 500nm with Width=4mm).
Frequency
(MHz)
LG, LaS, LaD
(µm)
VDD
(V)
VG
(V)
PAE
(%)
POUT
(dBm)
Gain(dB)
433
0.2-0.2-0.2
2
0.1
55
18.4
15
433
0.2-0.5-0.5
6
0.0
48
21.2
20
900
0.2-0.2-0.2
3
0.1
42
19
9.5
1800
0.2-0.5-0.5
3
-0.2
48
21
11
1800
0.2-0.5-0.5
7
0.0
35
24
14
3.4.
Summary of SOI-MESFET RF Power Amplifiers
The first demonstration of SOI-MESFET RF power amplifiers were
presented in this chapter. SOI-MESFETs on 45nm process provides enough RF
gain for power amplifier applications below 2.5 GHz. The RF gain can be further
increased by using a MOSFET device as the input transistor and MESFET as the
cascode device. A 1 GHz RF-PA using the proposed cascode structure was
designed and simulated in ADS. Simulation results show ~75 % PAE and 24dBm
output power.
MESFET only RF-PAs were designed and measured at 433, 900 and 1800
MHz. Peak PAE of 55% and output power of 22.5 dBm were measured.
The following chapter discusses RF polar transmitter design using the
fabricated MESFET PA at 433 MHz.
77
CHAPTER 4
WIDEBAND SUPPLY MODULATOR FOR POLAR TRANSMITTERS
The rapid evolution of wireless communication technologies has increased
the need for handheld devices that can support high data rate communication
standards. Radio Frequency (RF) front ends, especially the RF power amplifier,
are the main challenges in implementation of these systems due to aggressive
linearity and dynamic range requirements. Power Amplifiers (PAs) dominate the
sources of distortions and power consumption in the RF-front end. They are
typically operated in linear classes in order to minimize the linearity degradation.
However, linear operation leads to poor average power efficiency, especially
when fed by signals with high peak to average power ratio (PAPR) such as Long
Term Evolution (LTE) [47] and Wideband Code Division Multiple Access (WCDMA) signals [48]. Table 2 shows PAPR, bandwidth and power specifications
of few high data rate standards [46].
Table 2. Advanced modulation schemes specifications [46].
Channel
Bandwidth
Standards
Modulation Scheme
PAPR (dB)
IEEE 802.11b
(WLAN)
DBPSK, DQPSK, CCK
~10
GSM EDGE
8-PSK
~3.2
WCDMA/UMTS
QPSK
~3.5-7
20 MHz
200 KHz
5 MHz
1.25 MHz
QPSK, BPSK, 8-PSK,
~4-9
16QAM
An example of MESFET class-AB RF power amplifier efficiency versus
CDMA 2000
output power is shown in Figure 65. W-CDMA signal average power is 5-6 dB
78
below the peak power. Therefore if the PA is fed by this signal, the average
efficiency drops to 10% comparing to the maximum power added efficiency of
55%. Reduced efficiency in the power amplifier is the main contributor to the
significant the battery life reduction of handset devices. Low efficiency also
causes high power dissipation in the device and results in severe heating effects.
70
Efficiency (%)
60
PAE
DE
50
40
30
20
10
-6
-5
-4
-3
-2
-1
0
N orm alized O utput Pow er (dB )
Figure 65. MESFET Class-AB amplifier efficiency versus normalized output
power.
To remedy this issue, some advanced methods like Doherty amplifier and
drain modulation based architectures were introduced [49], [50], [51]. Although
the Doherty technique improves efficiency at backed off power, it has bandwidth
limitations. Polar modulation on the other hand, can potentially have larger
bandwidth and achieve close to peak PA efficiency operation over a range of
input powers by tracking the envelope variation of the modulated signal at the PA
supply node [28].
79
The main idea behind the polar architecture is to improve the drain
efficiency by reducing the DC power with respect to the output power. Transient
drain voltage of a class-A amplifier at different power levels are plotted in Figure
66. (a) and (b) have the same DC power consumption, while (a) has higher output
power than (b), therefore it has higher drain efficiency than (b). An envelope
amplifier is added to the class-A circuit of Figure 67. The envelope amplifier
controls the supply voltage, VDD, based on the peak-to-peak voltage swing at the
drain of the PA’s transistor. As a result, the DC power consumption is adaptively
being changed to achieve higher efficiency at lower output powers.
Figure 66. Class-A transient drain voltage.
Figure 67. Efficiency improvement by reducing the DC bias point.
80
The idea of polar transmitter was first introduced in [52]. Figure 68 shows
the proposed circuit by Kahn in [52] known as Envelope Elimination and
Restoration (EER).
Low Frequency
Amplitude
CF
Envelope
Detector
PA
Modulated RF
Signal
Limiter Constant Amplitude
Phase
Figure 68. Kahn Envelope Elimination and Restoration (EER) Transmitter [52]
Contrary to traditional Cartesian transmitters, polar transmitters transform
the in-phase and quadrature components of the input signal into amplitude (r) and
phase (θ) representation.
‚# _ Y , ’ “ZDb ”0
$0
Equation 30
In an EER transmitter, the amplitude information is used to control the
drain supply voltage and consequently the gain. The modulated phase signal is a
constant amplitude signal which ensures the peak efficiency operation of the RFPA. Since the input signal has a constant amplitude, one of the main attributes of
the EER technique is to make possible the deployment of high efficiency PAs,
81
which are strongly nonlinear, while meeting the linearity requirements. The
overall efficiency of the whole output stage is the PA efficiency multiplied by the
Envelope path efficiency. Therefore, the peak efficiency reduces, but the average
efficiency is improved at back off powers.
As mentioned above, a nonlinear PA can be used in EER architecture
which improves the transmitter’s efficiency. But this technique has two main
drawbacks: (i) phase information has much more bandwidth than the original I/Q
signals (ii) high input power leaks to the output and limits the dynamic range of
the transmitter. To eliminate the bandwidth and dynamic range problem of EER
techniques, an alternative form of polar transmitter known as Envelope Tracking
(ET) has been introduced [53], [54]. In ET architecture, the PA input terminal is
fed by the original modulated signal which has less BW while the supply is
controlled by amplitude modulated signal. This increases the PA dynamic range
and improves the backed-off power efficiency at the same time. Since the input
signal has amplitude variation, a linear RF power amplifier has to be used in ET
architecture. The RF PA of these architectures has to have good performance
versus supply voltage variation, while the ET amplifier is linear, EER can be a
nonlinear PA. Later sections discuss a new ET architecture using the proposed
four terminal SOI-MESFETs.
As mentioned above, the polar transmitter needs phase (θ) and magnitude
(r) information instead of Inphase (I) and Quadrature (Q) signals. I/Q to r/θ
transformation are nonlinear functions, which affect signal BW. It increases the
phase BW approximately by ~10-20 times and amplitude BW by ~5X comparing
82
to the original I and Q signals. Therefore, the power amplifier and supply
modulator must have wider bandwidth comparing to the traditional Cartesian
transmitters. Previous chapters discussed how SOI-MESFETs can be used to
design very wideband RF amplifiers. This chapter talks about wideband supply
modulator design using the proposed SOI-MESFET device.
4.1.
RF Power Amplifiers Supply Modulators
Analog amplifiers, buck convertors or mixed analog/digital circuitries are
the candidates for the envelope path of polar transmitters. Buck convertors and
switched mode amplifiers (SMAs) have traditionally being used to modulate the
PA’s supply node. Although switched mode circuits have high efficiency, they
suffer from bandwidth and linearity limitations. They also generate high levels of
switching noise ripple at the output terminal. SMAs are more applicable for low
bandwidth (BW) signals like EDGE/GSM [55]. For large BW signals such as WCDMA and LTE, a very fast analog amplifier is needed. Analog amplifiers can
provide much more bandwidth than buck convertors, but result in low efficiency.
More advanced architectures use the combination of a buck convertor for low
frequencies and an analog amplifier for high frequencies [49], [55].
Various hybrid linear/SMA regulators were proposed to achieve high
bandwidth and efficiency at the same time [49], [56]. In general, the linear
amplifier can be in series or in parallel with the SMA. Figure 69 shows both
groupings. The series combination of the switching and linear amplifiers helps
remove switching noise. The drawback of this tactic is that an LDO is required to
function as the linear amplifier [56]. CMOS LDOs suffer from bandwidth
83
limitations, stability issues and limited voltage swings [57]. The following section
discusses how SOI- MESFETs can improve LDO stability and bandwidth.
(a)
(b)
Figure 69. (a) Series SMA LDO supply modulator in EER (b) Parallel Amplifier
and SMA supply modulator in EER.
84
4.2.
SOI-MESFET for PA Supply Regulation
Different LDO topologies based on P-channel or N-channel pass
transistors have been introduced [57]. Both P- and N-channel LDOs are shown in
Figure 70, where they each have their own advantage and disadvantages. Output
impedance (ZOUT) of the P-channel transistor in Figure 70.b is proportional to rds
of the device. However Figure 70.a uses an N-MOSFET which has lower output
impedance in an LDO circuit which gives higher bandwidth compared to Pchannel LDOs or better stability. The drawback using N-channel LDOs is that a
charge pump circuit is required to overcome the threshold voltage drop for
enhancement mode N-channel devices [58].
Figure 70. (a) N-MOSFET based LDO (b) P-MOSFET based LDO
In this section, we present a design methodology for a SOI-MESFET
based LDO that features a depletion-mode N-type device as the pass transistor.
This combines the advantages of both P and N devices allowing for low cost,
design simplicity, stability, high current drive and bandwidth.
85
As discussed in the introduction, LDO regulators can be used to improve
the overall efficiency and linearity of a RF transmitter. The proposed SOIMESFET based LDO is shown Figure 71. An N-channel MESFET is used as the
pass transistor of the LDO in a common drain configuration. The simplified
equivalent small signal circuits of MESFET and PA are also shown in Figure 71.
At low frequencies, the resistance seen by the LDO output node is proportional to
1⁄iL ; this pushes the pole contributed by the pass transistor to higher frequencies
compared to a P-FET device which has output resistance of Rds. This stabilizes
the LDO for larger capacitive loads.
Figure 71. LDO envelope tracking regulator using the SOI-MESFET.
86
The loop gain transfer function of the LDO, LG(S), depends on the opamp transfer function, A(S), and the common drain transistor frequency response,
M(S).
f– *“Z f* & Ž
Equation 31
It can be shown that:
Ž iL _ awJ 
1
1
— _ ˆJ _ iL _ awJ 
l
iL _ awJ 
1
_ iL _ awJ _ a 
Equation 32
gm is much larger than 1/RL, therefore the pole associated with M(S) is
located at:
`-, l
iL !˜†X
<==> `-, ` awJ _ a
Equation 33
For CL=0, the transistor’s dominant pole is located at the MESFET cut-off
frequency, fT. MESFETs used in this work have more than 10GHz cut-off
frequency. The op-amp used in this work has a unity gain bandwidth of 350MHz.
Its first pole is located at 35 kHz and the second pole at 500 MHz [59]. Since the
pole related to the MESFET is at a much higher frequency than the op-amp’s
second pole, it has minimal effect on the LDO phase margin. Figure 72 shows the
open loop frequency response of the LDO in the presence of low capacitive loads,
less than 5pF. As it is shown in the measurements section, higher values of CL
push the MESFET pole to lower frequencies and cause instability. This issue can
be resolved by using a slower op-amp with the cost of a reduced bandwidth.
87
Figure 72. Frequency response of the proposed LDO.
4.3.
LDO Measurement Results
An N-channel MESFET LDO for polar and envelope tracking transmitter
applications is designed and built using a packaged MESFET fabricated on a
150nm SOI-CMOS process. A 350MHz bandwidth AD8061 op-amp from Analog
Devices [59] was used to drive the MESFET gate. As explained in the previous
section, the overall LDO bandwidth is limited mostly by the op-amp unity gain
frequency. The LDO transfer function, under different load conditions has been
measured and plotted in Figure 73. This plot demonstrates that a bandwidth of up
to 350MHz can be achieved.
88
Figure 73. Measured LDO voltage gain for different capacitive loads.
The efficiency of a supply regulator directly affects the overall polar
transmitter performance. Assuming the LDO voltage supply remains constant, the
efficiency will drop as the output power decreases as explained in Equation 34 .
f"„Œ^^ % #.5ˆ
100 h 100 #!
Equation 34
To prevent this issue, a buck regulator can be used in series with the LDO
to control the supply voltage, VDD, and amplify the low frequency components of
a modulated signal. The LDO efficiency with (PAE) and without (DE) the
presence of an error amplifier is shown in Figure 74. Measured peak output power
and efficiency are 26.2 dBm and 77% respectively when the LDO is connected to
a 20Ω resistor and a 10pF capacitor.
89
Figure 74. Measured LDO efficiency
The LDO linearity was tested by driving it with a two tone sinusoidal
signal at 100MHz separated by 200 kHz. The output spectrum is shown in Figure
75. Measured IMD3 is approximately -55dBc when the input power is at -20dBc
with respect to the full voltage swing.
Figure 75. LDO output spectrum when it is fed with a two tone signal at 100MHz
and separated by 200kHz.
90
Figure 76 shows the transient response of the LDO with input rectified
sinusoidal. Fast slope changes at the signal edges show the broadband behavior of
the LDO.
Figure 76. Transient response of the LDO to a rectified 1.2MHz sine wave
applied as the envelope signal.
One of the advantages of using an LDO as the linear amplifier is to
suppress the switching noise coming from the switch mode power supply. Figure
77 shows power supply noise rejection behavior of the proposed LDO. More than
50dB noise rejection has been achieved for frequencies below 300MHz.
91
Figure 77. Measured PSR of the proposed LDO.
The linearity and transient performance of the proposed LDO was tested
in an envelope tracking configuration. In this work, LDO drives the 433 MHz
SOI-MESFET power amplifier discussed in the previous chapter. Figure 78
shows the test setup block diagram used in this work. Appendix-A talks about the
test setup in more details. As the first step a two tone signal was applied to the PA
input and the envelope was used to drive the LDO. Effect of supply voltage
variation on linearity of this ET transmitter is shown in Figure 79. For a fixed PA
VDD of 2 V, IM3 components are 35dB lower than the fundamental tones.
Measurements show that IM3 starts growing when envelope signal swing on the
supply voltage is more than 0.8 V.
92
Normalized Output Power (dBc)
Figure 78. SOI-MESFET envelope tracking test setup.
0
-5
Fundumental
-10
-15
-20
IM3
-25
-30
-35
-40
0
0.2
0.4
0.6
0.8
1
VDD p-p Swing (V)
Figure 79. Effect of supply voltage swing on linearity.
93
1.2
1.4
An RF modulated EDGE signal was used to drive the PA input and the
envelope of the EDGE signal was used to drive LDO in an ET architecture. The
transient input and output voltage of LDO was captured using an Infiniium
Agilent oscilloscope and is plotted in Figure 80. RMS value of weighted error
signal is approximately -57 dB. Effect of envelope voltage swing on linearity of
EDGE signal is shown in Figure 81. This plot shows the tradeoff between supply
voltage swing (efficiency improvement) and linearity. Larger envelope swing
results in better efficiency with the cost of more nonlinearity.
LDO Input/Output Voltage (V)
2.6
2.4
2.2
2
1.8
1.6
1.4 Input (Red)
Output (Black)
1.2
50
100
Time (µ−
c
e
s
1
0
150
200
)
Figure 80. Transient input and output voltage of LDO for EDGE envelope signal.
94
Normalized Output Spectrum (dB)
0
-10
VP-P=0
VP-P=2.2
VP-P=1.7
VP-P=1.2
-20
-30
-40
-50
-60
-70
-80
398.5
399 399.5 400
400.5 401 401.5
Frequency (MHz)
Figure 81. Normalized output spectrum of MESFET ET for EDGE signal.
LDO was fabricated on a FR4 board using a packaged SOI-MESFET and
AD8061 Analog Device op-amp. The photograph of the LDO board is shown in
Figure 82.
95
Figure 82. Photograph of the 350 MHz SOI-MESFET LDO.
4.4.
Summary
This chapter discussed the need for RF transmitters which have less
efficiency degradation at backed-off powers. Polar transmitter was introduced as a
solution, and different polar architectures were outlined. Polar transmitter
introduces switching noises on the PA supply node. To remedy this problem, a
high speed and stable LDO is required. Depletion mode behavior of SOIMESFETs makes it possible to achieve several hundred megahertz of bandwidth
and decent stability at the same time. This idea was validated by designing and
measuring an SOI-MESFTE LDO with 350 MHz bandwidth. LDO was tested in
an ET architecture driving a MESFET PA. Measurement results show that the
proposed ET transmitter satisfies EDGE requirements.
96
CHAPTER 5
BACKGATE MODULATED P
POLAR TRANSMITTER
Chapter IV explained
ained the idea behind building polar ttransmitters
ransmitters (PTs) and
described different polar architectures. All the efforts in PTs are toward
improving the overall transmitter efficiency at backed
backed-off
off powers, while
satisfying the linearity
arity requirements of modern communication standards. A very
simplified polar transmitter block diagram is shown in Figure 83.
Figure 83.. Simplified block diagram of a traditional polar transmitter.
At any output power level, the overall efficiency of the polar transmitter
is:
Equation 35
Therefore the envelope amplifier’s efficiency is as critical as the PA
efficiency. Designing an efficient supply modulator that meets all the linearity
specifications, can be as challenging as RF
RF-PA design itself.. Buck converters and
97
switching mode amplifiers are very efficient, but they suffer from bandwidth
limitations and switching noises. Combinations of a linear and switched mode
amplifier are introduced to regulate the PA supply voltage for wideband envelope
signals. As explained in the previous chapter, the efficiency of the hybrid
envelope amplifier is not constant for all signal levels and in fact decreases when
the envelope signal drops. For a well-designed envelope amplifier and RF-PA, the
efficiency reduction is not as significant as standalone PA efficiency degradation
in backed-off powers. Thus, it is still possible to improve the average efficiency
by using polar transmitters.
Another problem associated with polar transmitters is their low dynamic
range due to limited supply voltage swing. Large voltage swing on the PA supply
node can change the RF transistor operation mode into the linear (triode) region
and causes severe AM-AM and AM-PM distortions.
Taking everything into account, although supply modulated polar
transmitters can improve the average efficiency; they introduce many design
challenges for meeting noise, dynamic range and linearity specifications of
modern standards.
This chapter presents a novel polar architecture that can potentially
mitigate some of the problems described above. The main idea behind this work is
to control the RF gain and DC power by changing the substrate voltage of a
partially depleted transistor on an SOI process. In principle it is the same idea as
supply voltage modulation, but the advantages of the proposed technique are
higher dynamic range, less design complexity and most importantly higher
98
efficiency. Higher efficiency can be achieved because of the low power
consumption of the envelope amplifier in the proposed architecture. The buried
oxide in an SOI processes is an insulator and isolates the backgate terminal from a
MOSFET/MESFET device channel. Therefore, there is no DC current going
through the backgate terminal.
Designing a substrate modulated based transmitter requires a four terminal
device model and extraction of the associated device model. Although SOIMESFET devices are used in this work, the same technique can be applied to
MOSFET devices as well. A four terminal Verilog model have been developed
for SOI-MESFETs and reported in [62]. The following section talks about DC
and RF characterization of SOI-MESFETs fabricated on a 45nm technology.
5.1.
Substrate Voltage Effect on Device Characteristics
A cross section of a MESFET on a SOI process is shown in Figure 84.
Figure 84 (a) shows the depletion region for VG=0 and backgate voltage, VBG=0.
The depletion region grows as the gate voltage becomes more negative in Figure
84 (b) Applying a negative voltage to the substrate also depletes the channel as it
is shown in Figure 84 (c). Positive VBG has the opposite effect and reduces the
depletion region area.
99
(a)
(b)
(c)
Figure 84. Cross sectional view of a partially depleted MESFET (a) VG=0, VBG<0
(b) VG<0, VBG<0 (c) VG<0, VBG=-2.
100
The effect of substrate voltage on the channel thickness of fully depleted
transistors appears as a change in the threshold voltage, and as a result changes in
DC current. This effect has been studied for SOI-MESFET devices in [62].
Measured drain current versus gate voltage at different backgate biases are plotted
in Figure 85.
Figure 85. Effect of substrate voltage on SOI-MESFET I/V curve.
DC measurements demonstrate that the SOI-MESFET drain current can be
controlled over a wide range by changing the backgate voltage. This property can
be used in a polar transmitter to control the DC and RF power. The main barrier
toward this task is the AC response of the substrate terminal. The following
section discusses the transient and RF response of the backgate contact.
5.2.
Transient Behavior of Substrate Contact
The buried oxide layer of SOI wafers is reaching ~100-200 nm in very
deep submicron technologies such as the IBM 45nm SOI process used in this
101
work.. Therefore, a smaller voltage is required to modulate the channel thickness.
The input impedance off the backgate terminal was measured and is plotted in
Figure 86. At low frequencies,
ncies, the input impedance is a capacitive load of ~4-5pF.
~4
Figure 86.. Measured input capacitance of the backgate terminal.
The low capacitance of the backgate terminal allows us to drive the
backgate contact with a time varying signal having bandwidth of more than 100
MHz. To validate the idea, a 100 KHz
Hz square wave signal is applied to the
substrate of a class-B
B MESFET power amplifier and the output voltage is
captured with an oscilloscope. A
At the same time an
n RF signal at 433 MHz is
applied to the Schottky gate. The transient output signal is sshown in Figure 87.
This plot demonstrates that the RF power can be modulated with a baseband
signal fed to the backgate terminal.
102
Figure 87. Transient behavior of a SOI-MESFET PA when the backgate is driven
by a 100 KHz square wave.
5.3.
Efficiency Enhancement Using Backgate Modulation
The previous sub-sections described how DC and RF power of SOI-
MESFET/MOSFET PAs can be controlled using the backgate terminal.
Measurement results of a backgate modulated SOI-MESFET polar transmitter are
presented in this section. In a supply modulated EER transmitter, the output
power is controlled by changing the supply voltage of the power amplifier, while
the input power is constant. In the proposed polar transmitter, the supply voltage
and RF input power are constant, and the output power is controlled by the
substrate voltage. Figure 88 shows the block diagram of such a polar transmitter
architecture. The envelope amplifier of this transmitter drives a capacitive load of
~2-pF. Therefore, a very low power (less than 5mW) baseband amplifier can be
103
used for this purpose. Power consumption of this envelope amplifier has a
negligible effect on the overall transmitter efficiency.
Figure 88.. Block diagram of the proposed backgate modulated polar transmitter.
The static effect of the backgate voltage on the PA perform
performance
ance is shown
in Figure 89.. It shows that as the VBG goes up, higher output power can be
achieved. Output power and PAE are measured versus the substrate voltage and
plotted in Figure 90. The peak PAE of the proposed polar transmitter is the same
as the RF-PA
PA peak efficiency. The main goal though is to improve the efficiency
at backed-off
off output powers. PAE improvement using the pr
proposed
oposed method in
comparison to the traditional class
class-AB RF-PA is shown Figure 91.
91 In the
traditional approach VDD, VG and VBG are constant and the RF PA is fed by a
modulated RF input signal. To make a fair comparison, the same RF
RF-PA
PA was used
in the proposed polar transmit
transmitter.
ter. Measurement results show 15%
15
PAE
improvement at 5-dB
dB backed
backed-off power using backgate modulation.
104
25
V BG =7V
Output Power (dBm)
20
V
BG
V
15
V
10
=5V
BG
BG
=3V
=1V
5
0
-10
-5
0
5
10
15
Input Power (dBm)
Figure 89. Effect of the backgate voltage on the MESFET RF-PA output power.
VDD=3V, VG=-0.2V.
Output power and PAE are measured versus the substrate voltage and
plotted in Figure 90.
Gain (dB), P
60
50
20
40
15
30
10
20
5
10
0
0
PAE (%)
OUT
(dBm)
25
0
1
2
3
4
5
6
7
Backgate Voltage, V , (V)
8
BG
Figure 90. SOI-MESFET Power Amplifier performance versus the substrate
voltage.
105
60
Proposed Backgate Modulated
50
PAE (%)
40
30
20
10
Constant V
0
-10
-8
-6
DD
-4
and Backgate
-2
0
Normalized Output Power (dBm)
Figure 91. PAE versus output power comparison between the traditional class-AB
PA and a backgate modulated polar transmitter.
One of the challenges in implementation of supply regulated polar
transmitters is their low dynamic range. At low supply voltages, the RF transistor
goes to the triode (linear) region and causes severe AM/AM and AM/PM
distortions. A comparison between AM/PM behavior of the MESFET PA with
respect to backgate terminal and supply voltage is shown in Figure 92 and Figure
93. The voltage sweep range in this measurement translates to ~13dB dynamic
range for both backgate and supply modulation architectures. However, sweeping
VDD changes the phase by 25 degree, where the backgate modulation only causes
6 degrees of phase change.
106
AM/PM, Phase (deg)
5
-150
4
-160
3
-170
2
-180
1
-190
0
1
2
3
4
5
6
7
8
(V)
-140
rms
6
AM/AM, V
-130
0
V , V , (V)
DD
BG
Figure 92. Measured AM/PM characteristics of SOI-MESFET 1.8GHz PA at
fixed input power of 6-dBm.
The linearity performance of the proposed PT was measured in an
envelope tracking structure fed by an EDGE signal. The measured normalized
output spectrum is shown Figure 94. As it is specified on the graph, the backgate
modulated ET output spectrum satisfies the mask requirement for EDGE signals.
Efficiency improves by a factor of 1.3 for EDGE signal when the envelope signal
is applied to the backgate comparing to the constant bias condition.
107
25
2
1 .5
rms
V
15
10
1
5
0 .5
Slope (V/V)
Phase Slope (deg/V)
20
0
-5
0
1
2
3
V
DD
4
, V
BG
5
0
7
6
, (V )
Normalized Output Spectrum (dB)
Figure 93. Measured slope of AM/PM curves for VDD and VBG sweeps.
0
-10
-20
-30
EDGE mask
-40
Backgate
Modulated
-50
-60
-70
-80
f -1.5MHz
c
Fixed Backgate
fc
f +1.5MHz
c
Figure 94. Output spectrum of EDGE signal for backgate modulated SOIMESFET ET transmitter.
108
5.4.
Backgate Modulation versus Supply Modulation
Although the peak PAE of a backgate modulated PT is the same as the
RF-PA peak efficiency, PAE still drops as POUT is reduced. A similar problem is
also associated with supply modulated polar transmitters. In that case the
efficiency of the envelope amplifier reduces as the output power decreases.
Equation 35 reflects this effect in the overall efficiency calculation of the supply
modulated polar transmitter.
Efficiency drop at backed-off powers in the backgate modulated PT can be
also modeled by a PAE scaling factor, ζ. ζ is not physical power dissipation in the
envelope amplifier. It is simply a parameter which is defined for comparison
purposes. ζ is very similar to the supply modulator efficiency and plays the same
role in PAE calculation of Equation 35. Overall PAE of a backgate modulated PT
can be calculated as:
™.05M ζ ™-
_, Equation 36
ζ for a measured SOI-MESFET PA is extracted and plotted versus output
power in Figure 96. To perform a fair evaluation, the supply modulator’s
efficiency results, presented in [63], are used as a reference in this section.
Efficiencies reported in [63] are among the highest numbers reported to the date.
Figure 96 shows the comparison between the proposed technique and the
traditional supply modulated PT reported in [63]. Although the proposed method
has higher efficiency at high output powers, the supply modulated shows better
performance for normalized POUT less than -4dBm. The quicker efficiency drop of
109
the backgate modulated PT is mainly due to the RF gain reduction at low output
powers. RF input power is constant for backg
backgate
ate modulated data collected in
Figure 91.. This problem can be resolved by applying a lower input power at
backed-off
off output power. Figure 95 shows efficiency versus POUT at different
input powers. This graph shows that higher PAE is achievable for a fixed POUT if
lower PIN is used.
Figure 95. PAE of backgate modulated PT at different input power levels.
Efficiency
fficiency of the pre
pre-distorted
distorted backgate modulated PT is compared to the
hybrid supply modulator in Figure 96. The proposed method has higher efficiency
for a 6-dB range of output power. Peak efficiency is ~20% higher than the hybrid
buck and LDO supply voltage modulator.
It has to be mentioned that the power amplifier efficiency drops very fast
as VDD approaches the knee voltage of the RF transistor. Thus, even though the
110
supply voltage modulator efficiency is ~40% at -12dB backed-off,
off, the overall
efficiency can be much poorer.
Figure 96.. Efficiency comparison between tthe hybrid supply modulator [63],
[
backgate PT and pre-distorted
distorted backgate PT.
5.5.
Summary
A novel
ovel polar transmitter architecture was introduced in this chapter. The
proposed polar transmitter uses the backgate voltage to ccontrol
ontrol the output power.
Efficiency performance versus output power were measured and compared to a
traditional class-AB
AB SOI
SOI-MESFET RF-PA.
PA. The proposed technique shows ~3X
PAE improvement at 8-dB
dB backed
backed-off power.
A low power, high speed envelope amplifier can be used in this
architecture. In traditional supply modulated polar transmitter, a very high power
and low noise amplifier is needed. This adds design complexity and result in
lower efficiency comparing to bac
backgate
kgate modulated PT. More than 10% efficiency
111
improvement was achieved comparing to supply voltage modulation by using a
look-up table and pre-distorting the input signal.
112
CHAPTER 6
CONCLUSION
The low breakdown voltage of MOSFET devices on deep sub-micron
technologies limits the output power of RF transmitters and adds lots of reliability
concerns to RF PAs design. The SOI-CMOS compatible MESFET was
introduced as a potential solution for some of the issues associated with RF power
amplifier design on silicon CMOS technologies. SOI-MESFETs can be fabricated
on SOI and SOS processes with no additional cost. MESFET offers higher
breakdown voltage comparing to MOSFETs fabricated on a same process.
Therefore more robust high power RF-PA can be designed using these devices.
SOI-MESFETs fabricated on different SOI and SOS processes were
characterized from DC to 40 GHz. DC measurements demonstrates 2-30X
breakdown voltage enhancement using MESFETs compared to MOSFETs on the
same SOI-CMOS process. The cost for higher breakdown voltage is lower cut-off
frequency. This trade-off was studied in this thesis and the MESFET geometries
were optimized to achieve enough RF gain as well as high breakdown voltage. RF
measurements showed that MESFETs can be used in power amplifiers operating
in the frequency range of DC-2.5GHz.
As the first step, the idea was validated using ADS simulation tool. For
that purpose, a TOM3 model was developed and validated for silicon MESFETs
fabricated on a 45nm and 150nm CMOS processes through simulation of
measured devices. The nonlinear behavior of the model was tested by comparing
113
the simulation and measurement results of a 433MHz Class AB SOI-MESFET
power amplifier.
The first demonstration of SOI-MESFET RF power amplifiers are
presented in this thesis. SOI-MESFETs on a 45nm process provide enough RF
gain for power amplifier applications below 2.5 GHz. The RF gain can be further
increased by using a MOSFET device as the input transistor and MESFET as the
cascode device. A 1 GHz RF-PA using the proposed cascode structure was
designed and simulated in ADS. Simulation results show ~75 % PAE and 24dBm
output power. Several MESFET only RF-PAs were designed and measured at
433, 900 and 1800 MHz. Peak PAE of 55% and output power of 22.5 dBm were
measured. To achieve higher efficiency and better linearity, MESFETs need to be
characterized using load pull tuners.
SOI-MESFETs also have some advantages over MOSFETs in low dropout regulators. LDOs are used in polar transmitters to improve the efficiency at
backed-off powers. Polar transmitter introduces switching noises on the PA
supply node. To remedy this problem, a high speed and stable LDO is required.
Depletion mode behavior of SOI-MESFETs makes it possible to achieve several
hundred megahertz of bandwidth and decent stability at the same time. This idea
was validated by designing and measuring an SOI-MESFET LDO with 350 MHz
bandwidth. The LDO was tested in an ET architecture driving a MESFET PA.
Measurement results show that the proposed ET transmitter satisfies EDGE
requirements. As a future work, the error amplifier can be integrated on the same
114
die as the pass transistor. Integrating the LDO’s error amplifier helps achieving
better efficiency and wider bandwidth.
A novel polar transmitter architecture was also introduced in this thesis.
The proposed polar transmitter uses the backgate voltage to control the output
power. Efficiency performance versus output power were measured and compared
to a traditional class-AB SOI-MESFET RF-PA. The proposed technique shows
~3X PAE improvement at 8-dB backed-off power. A low power, high speed
envelope amplifier can be used in this architecture. In traditional supply
modulated polar transmitter, a very high power and low noise amplifier is needed.
This adds design complexity and result in lower efficiency comparing to backgate
modulated PT. More than 10% efficiency improvement was achieved comparing
to supply voltage modulation by using a look-up table and pre-distorting the input
signal. This idea can be more developed by implementing an EER polar
transmitter. This task requires a high voltage low frequency op-amp to be
designed.
115
REFERENCES
[1]
Masoud Zargari, David K. Su, C. Patrick Yue, Shahriar Rabii, David
Weber, Brian J. Kaczynski, Srenik S. Mehta, Kalwant Singh, Sunetra
Mendis, and Bruce A. Wooley, "A 5-GHz CMOS Transceiver for IEEE
802.11a Wireless LAN Systems", IEEE Journal of Solid State Circuits,
vol. 37, No. 12, Dec 2002.
[2]
Hooman Darabi, Shahla Khorram, Hung-Ming (Ed) Chien, Meng-An Pan,
Stephen Wu, Shervin Moloudi, John C. Leete, Jacob J. Rael, Masood
Syed, Robert Lee, Brima Ibrahim, Maryam Rofougaran, and Ahmadreza
Rofougaran, “A 2.4-GHz CMOS Transceiver for Bluetooth”, IEEE
Journal of Solid-State Circuits, vol. 36, December 2001.
[3]
Jri Lee, Yi An Li, Meng Hung, Shih-Jou Huang, “A Fully-Integrated 77GHz FMCW Radar Transceiver in 65-nm CMOS Technology”, IEEE
Journal of Solid-State Circuits, vol. 45, December 2010.
[4]
C.-H. Jan, M. Agostinelli, H. Deshpande, M. A. El-Tanani, W. Hafez, U.
Jalan, L. Janbay, M. Kang, H. Lakdawala, J. Lin, Y-L Lu, S.Mudanai, J.
Park, A. Rahman, J. Rizk, W.-K. Shin, K. Soumyanath, H. Tashiro, C.
Tsai, P. VanDerVoorn, J.-Y. Yeh, P. Bai, “RF CMOS Technology Scaling
in High-k/Metal Gate Era for RF SoC (System-on-Chip)
Applications”, IEEE International Electron Devices Meeting (IEDM),
2010.
[5]
Carlos H. Diaz, Denny D. Tang, and Jack (Yuan-Chen) Sun, “CMOS
Technology for MS/RF SoC”, IEEE Transaction on Electron Devices, vol.
50, NO. 3, March 2003.
[6]
P. Reynaert and M. Steyaert, “A 2.45-GHz 0.13-µm CMOS PA with
Parallel Amplification,” IEEE Journal of Solid-State Circuits, vol. 42, pp.
551-562, March 2007.
[7]
Qizheng Gu, “RF System Design of Transceivers for Wireless
Communications”, Springer, 2005.
[8]
Hanington, G. Pin-Fan Chen, Asbeck, P.M, Larson, L.E, “High-efficiency
power amplifier using dynamic power-supply voltage for CDMA
applications” IEEE Microwave Theory and Techniques , Volume 47, issue
8, August 2002.
[9]
Miwa, S, Kamo, Y, Kittaka, Y, Yamasaki, T, Tsukahara, Y, Tanii,
T, Kohno, M, Goto, S, Shima, A, “A 67% PAE, 100 W GaN power
amplifier with on-chip harmonic tuning circuits for C-band space
116
applications”, IEEE MTT-S International Microwave Symposium Digest,
2011.
[10]
Hossein Mashad Nemati, Christian Fager, Mattias Thorsell, and Herbert
Zirath,“High-Efficiency LDMOS Power-Amplifier Design at 1 GHz
Using an Optimized Transistor Model”, IEEE Transaction on Microwave
Theory and Technique, vol. 57, no. 7, July 2009
[11]
Ravi Gupta, Brian M. Ballweber and David J. Allstot, “Design and
Optimization of CMOS RF Power Amplifiers”, IEEE Journal Of SolidState Circuits, vol. 36, no. 2, Feb 2001
[12]
Jaemin Jang, Changkun Park, Haksun Kim, and Songcheol Hong, “A
CMOS RF Power Amplifier Using an Off-Chip Transmision Line
Transformer With 62% PAE”, IEEE Microwave and Wireless
Components Letters, vol. 17, no. 5, May 2007.
[13]
Changsik You and Qiuting Huang, “A Common-Gate Switched 0.9-W
Class-E Power Amplifier with 41% PAE in 0.25-_m CMOS, IEEE
Journal Of Solid-State Circuits, vol. 36, no. 5, May 2001.
[14]
Torsten Lehmann and Marco Cassia 1-V Power Supply CMOS Cascode
Amplifier, IEEE Journal Of Solid-State Circuits, VOL. 36, NO. 7, JULY
2001.
[15]
Tirdad Sowlati , and Domine M. W. Leenaerts, “A 2.4-GHz 0.18-um
CMOS Self-Biased Cascode Power Amplifier”, IEEE Journal Of SolidState Circuits, vol. 38, no. 8, Aug 2003.
[16]
Tak Shun Dickson Cheung, and John R. Long, “A 21–26-GHz SiGe
Bipolar Power Amplifier MMIC”, IEEE Journal Of Solid-State Circuits,
vol. 40, no. 12, DECEMBER 2005.
[17]
Mohammad Reza Ghajar, Slim Boumaiza “Concurrent Dual Band
2.4/3.5GHz Fully Integrated Power Amplifier in 0.13µm CMOS
Technology”, Proceedings of the 39th European Microwave Conference.
[18]
Riccardo Brama, et al, "A 30.5 dBm 48% PAE CMOS Class-E PA With
Integrated Balun for RF Applications", IEEE Journal Of Solid-State
Circuits, vol. 43, no. 8, Aug 2008.
[19]
Trevor J. Thornton and Michael E. Wood, “MESFETs integrated with
MOSFETs on common substrate and methods of forming the same”,
Patent number: US7589007B2, Sep 15, 2009.
117
[20]
J. Ervin, A. Balijepalli, P. Joshi, V. Kushner, J. Yang, and T. J. Thornton,
"CMOS-Compatible SOI MESFETs With High Breakdown Voltage,"
IEEE Trans. Electron Devices, vol. 53, pp. 3129-3135, 2006.
[21]
W. Lepkowski, J. Ervin, S. J. Wilk, and T. J. Thornton, "SOI MESFETs
Fabricated Using Fully Depleted CMOS Technologies," IEEE Electron
Device Letts, vol. 30, pp. 678-680, 2009.
[22]
J. Yang, A. Balijepalli, T. J. Thornton, J. Vandersand, B. J. Blalock, M.
Mojarradi, and M. E. Wood, "Silicon-Based Integrated MOSFETs and
MESFETs: A New Paradigm for Low Power, Mixed Signal, Monolithic
Systems using Commercially Available SOI," Int. J. of High Speed
Electronics and Systems, vol. 16, pp. 723 - 732, 2006.
[23]
W. Lepkowski, S. J. Wilk, S. Kim, B. Bakkaloglu and T. J. Thornton "A
Capacitor-Free LDO Using a FD Si-MESFET Pass Transistor," presented
at the 53rd IEEE MWSCAS Conference, Cancun, Mexico, 2009.
[24]
K. P. MacWilliams and J. D. Plummer, "Device physics and technology of
complementary silicon MESFET's for VLSI applications," IEEE Trans.
Electron Devices, vol. 38, pp. 2619-2631, 1991.
[25]
H. Vogt, G. Burbach, J. Belz, and G. Zimmer, "MESFETs in thin silicon
on SIMOX," Electronics Letters, vol. 25, pp. 1580-1581, 1989.
[26]
D. P. Vu and A. Sono, "Self-aligned Si MESFETs fabricated in thin
silicon-on-insulator films," Electronics Letters, vol. 23, pp. 354-355, 1987.
[27]
W. Lepkowski, M. Reza Ghajar, et al, "Scaling SOI MESFETs to 150 nm
CMOS Technologies," IEEE Transactions on Electron Devices, vol. 58,
no. 6, pp. 1628-1634, June 2011.
[28]
Mohammad Reza Ghajar, “High Efficiency CMOS Power Amplifiers for
Drain Modulation Based RF Transmitters”, Master of Science Thesis,
University of Waterloo, 2009.
[29]
Behzad Razavi, “RF Microelectronics”, Prentice Hall, Second edition,
2011.
[30]
Thomas H. Lee, “The design of CMOS radio Frequency Integrated
Circuits” Cambridge, Second edition 2004.
[31]
Andrei Grebennikov, “RF and Microwave Power Amplifier Design”,
McGraw Hill, 2005.
118
[32]
Steve C Cripps, “RF Power Amplifiers for Wireless Communications,”
2nd edition, Artech House 2006.
[33]
T. K. Chiang, Y. H. Wang, and M. P. Houng, "Modeling of threshold
voltage and subthreshold swing of short-channel SOI MESFET's," IEEE
Solid-State Electronics, pp. 123-129, 1999.
[34]
S. R. Bahl and J. A. del Alamo, "A new drain-current injection technique
for the measurement of off-state breakdown voltage in FETs," IEEE
Transaction of Electron Devices, vol. 40, pp. 1558-1560, 1993.
[35]
WinCal 3.1, Cascade Microtech, Inc., Beaverton, OR, 2002.
[36]
A. Balijepalli, R. Vijayaraghavan, J. Ervin, J. Yang, S. K. Islam, and T. J.
Thornton, "Large-signal modeling of SOI MESFETs," Solid-State
Electronics, vol. 50, pp. 943-950, 2006.
[37]
G. Dambrine, et al, “A new method for determining FET small signal
equivalent circuit”, IEEE Transaction on Microwave Theory and
Techniques, 1988.
[38]
Agilent ADS TOM3 documentation, online 2012.
[39]
R.B. Hallgren and P. H. Litzenberg, “TOM3 Capacitance Model Linking
Large and Small Signal MESFET Models in Spice,” IEEE Transaction
MTT, vol 47,1999, pp.556-561.
[40]
Seth J. Wilk, M. Reza Ghajar, William Lepkowski, Bertan Bakkaloglu,
Trevor J. Thornton, "Characterization and Modeling of Enhanced Voltage
RF MESFETs on 45nm CMOS for RF Applications", to be published at
IEEE Radio Frequency Integrated Circuit conference, June, 2012,
Montreal.
[41]
Y. Takayama,” A New Load-Pull Characterization Method for Microwave
Power Transistors”, IEEE MTT-S, International Microwave Symposium,
June, 1976.
[42]
M. Reza Ghajar, Seth J. Wilk, William Lepkowski, Bertan Bakkaloglu,
Trevor Thornton, “Wideband Class AB RF Power Amplifier using CMOS
compatible SOI-MESFET device on 150nm Technology”, IEEE Topical
Conference on Power Amplifiers for Wireless and Radio Applications
(PAWR), Jan 2012, Santa Clara, USA.
[43]
M. Reza Ghajar, et al, “Cascode MOSFET-MESFET RF Power Amplifier
on 150nm SOI CMOS Technology”, European Microwave Week
conference, Oct 2011.
119
[44]
Ichiro Aoki, Scott Kee, Rahul Magoon, Roberto Aparicio, Florian Bohn,
Jeff Zachan, Geoff Hatcher, Donald McClymont, and Ali Hajimiri,"A
Fully-Integrated Quad-Band GSM/GPRS CMOS Power Amplifier", IEEE
Journal Of Solid-State Circuits, vol. 43, no. 12, Dec 2008.
[45]
Calogero D. Presti, Francesco Carrara, Antonino Scuderi, Peter M. Asbeck
and Giuseppe Palmisano, "A 25 dBm Digitally Modulated CMOS Power
Amplifier for WCDMA/EDGE/OFDM With Adaptive Digital
Predistortion and Efficient Power Control", IEEE Journal Of Solid-State
Circuits, vol. 44, no. 7, July 2009.
[46]
Rameswor Shrestha, Ronan van der Zee, Anton de Graauw, and Bram
Nauta, "A Wideband Supply Modulator for 20MHz RF Bandwidth Polar
PAs in 65 nm CMOS", IEEE Journal of Solid State Circuits, vol. 44, April
2009.
[47]
Daehyun Kang, Dongsu Kim, Jinsung Choi, and Bumman Kim, “LTE
Power Amplifier for Envelope Tracking Polar Transmitters”, Proceedings
of the 40th European Microwave Conference.
[48]
Junxiong Deng, Prasad S. Gudem, Lawrence E. Larson and Peter M.
Asbeck, “A High Average-Efficiency SiGe HBT Power Amplifier for
WCDMA Handset Applications”, IEEE Transactions on Microwave
Theory and Techniques, VOL. 53, NO. 2, FEB 2005.
[49]
Wing-Yee Chu, Bertan Bakkaloglu and Sayfe Kiaei “A 10 MHz
Bandwidth, 2 mV Ripple PA Regulator for CDMA Transmitters” IEEE
Journal of Solid-State Circuits, vol. 40, December 2008.
[50]
J. Kim, B. Fehri, S. Boumaiza, and J. Wood, "Power Efficiency and
Linearity Enhancement using Optimized Asymmetrical Doherty Power
Amplifiers", IEEE Transaction on Microwave Theory and Techniques,
Sep. 2010.
[51]
G. Hanington, C. Pin-Fan, P. M. Asbeck, and L. E. Larson, “High
efficiency power amplifier using dynamic power-supply voltage for
CDMA applications,” IEEE Trans. Microw. Theory Tech., vol. 47, pp.
1471–1476, 1999.
[52]
L. R. Kahn, “Single sideband transmission by envelope elimination and
restoration”, Proc IRE, vol. 40, no 7, pp 803-806, July 1952.
[53]
Feipeng Wang, et al, “A Monolithic High Efficiency 2.4-GHz 20-dBm
SiGe BiCMOS Envelope-Tracking OFDM Power Amplifier”, IEEE
Journal Of Solid-State Circuits, vol. 42, No. 6, June 2007.
120
[54]
Chin Hsia, Donald F. Kimball, Sandro Lanfranco and Peter M. Asbeck,
“Wideband High Efficiency Digitally-Assisted Envelope Amplifier with
Dual Switching Stages for Radio Base-Station Envelope Tracking Power
Amplifiers”, IEEE MTT-S International Microwave Symposium Digest
(MTT), 2010.
[55]
Jennifer N. Kitchen, et al, “Polar SiGe Class E and F Amplifiers Using
Switch-Mode Supply Modulation”, IEEE Transaction on Microwave
Theory and Techniques, vol. 55, May 2007.
[56]
J. Kitchen, C. Chu, S. Kiaei and B. Bakkaloglu, “Supply Modulators for
RF Polar Transmitters”, IEEE Radio Frequency Integrated Circuits
Symposium, 2008.
[57]
G. A. Rincon-Mora and P. E. Allen, “A low-voltage, low quiescent
current, low drop-out regulator,” IEEE J. Solid State Circuits, vol. 33,no.
1, pp. 36–44, Jan. 1998.
[58]
W. Lepkowski, S.J. Wilk, S. Kim, B.Bakkaloglu, T.J.Thornton “A
Capacitor-Free LDO Using a FD Si-MESFET Pass Transistor”, IEEE
International Midwest Symposium on Circuits and Systems, 2009.
[59]
http://www.analog.com/
[60]
J. Ervin, A. Balijepalli, P. Joshi, V. Kushner, J. Yang, and T. J. Thornton,
"CMOS-Compatible SOI MESFETs With High Breakdown Voltage,"
IEEE Trans. Electron Devices, vol. 53, pp. 3129-3135, 2006.
[61]
M. Reza Ghajar, et al, “Cascode MOSFET-MESFET RF Power Amplifier
on 150nm SOI CMOS Technology”, to be published in IEEE European
Microwave Week, Oct 2011.
[62]
Asha Balijepalli, “Compact Modeling and Applications of PD SOI
MESFET”, PhD dissertation, Arizona State University, Dec 2007.
[63]
J. Kitchen, et al, “Combined Linear and –Modulated Switch-Mode PA
Supply Modulator for Polar Transmitter”, IEEE Journal of Solid State
Circuits, vol.44, Feb 2009.
121
APPENDIX A
ENVELOPE TRACKING TEST SETUP
122
EDGE signals are generated in ADS using EDGE signal source. Agilent
N8241A arbitrary waveform generator was used to generate the envelope signal
and the Agilent E4432B was used for up converting the baseband EDGE I and Q
signals. Two differential, high speeds, high resolutions digital to analog
converters (DACs) are placed in this signal generator. The sampling frequency of
N8241A is fixed to ~625 MSPs. The EDGE signal generated in ADS has only 8X
oversampling ratio. Therefore, we up-sampled the envelope signal by a factor of
300X and passed it through a low pass filter to achieve a correct analog signal for
the envelope path. This task was done using interpolation command in Matlab.
Data generated in ADS are taken into Matlab for scaling and formatting. An mfile was used to scale EDGE signal to signal generators DAC range. The m-file
script is shown at the end of this section.
Another consideration in the test setup was synchronizing the two signal
paths. N8241A have markers output and we used them to trigger the E4432B.
Basically, N8241A outputs a pulse on the marker port, which shows that the
frame is finished and re-triggers E4432B. The delay associated with cables should
be manually tuned. N8241A Control Utility software, available on Agilent
website, was used to load the data into DACs memories. E4432B was controlled
by ads in software from Agilent that works on Microsoft Excel.
All the signals should be scaled to the DACs output levels. This task was
done by the m-file script shown below.
a=EDGE_Envelope(1:16384); % import data from ADS
a=a*2-1;
123
NP=16384*300; % increase the number of points by 300X for upsampling
a=interp(a,300); % upsample and low pass filter the edge envelope
b=a*0+1;
%b=(abs(sin((2^6)*(samples)/(NP)*2*pi)));%*0+1;
mkr1=0;
mkr2=0;
%a=min(a,16383);
%a=max(a,0);
i=a;
q=b;
scale=1;
mx = max([max(abs(i)) max(abs(q))]); % scale data to the DACs range
scaleint = round(8192*scale)-1
i = i/mx*scaleint + 8191; % Make 14 bit unsigned integers
q = q/mx*scaleint + 8191;
i=round(i);
q=round(q);
i = min(i,16383); % Just to be safe
i = max(i,0);
q = min(q,16383);
q = max(q,0);
i=double(i);
q=double(q);
124
size(i)
i(1)=i(1)+mkr1*16384+mkr2*32768; % Set markers to begin segment
fk=fopen('c:\ESG\sine_AWG_q2.bin','wb'); % N8241A takes .bin format
cout=fwrite(fk,q,'int16');
fclose(fk)
fk2=fopen('c:\ESG\sine_AWG_i2.bin','wb')
cout=fwrite(fk2,i,'int16');
fclose(fk2)
plot(b)
125
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

advertising