1
Highly Linear GaN Class AB
Power Amplifier Design
Pedro Miguel Cabral, José Carlos Pedro and Nuno Borges Carvalho
Instituto de Telecomunicações – Universidade de Aveiro, Campus Universitário de Santiago 3810-193 Aveiro
Corresponding author:
Pedro Miguel da Silva Cabral
Instituto de Telecomunicações
Campus Universitário de Santiago
3810-193 Aveiro – Portugal
Phone: +351 234 377 900
Fax:
+351 234 377 901
e-mail: pcabral@av.it.pt
Appropriate Conference Topics:
1. Solid State Devices and Circuits
3. High-Power Devices and Techniques
26. Wide Band Gap Semiconductor Devices
2
Highly Linear GaN Class AB
Power Amplifier Design
Pedro Miguel Cabral, José Carlos Pedro and Nuno Borges Carvalho
Instituto de Telecomunicações – Universidade de Aveiro, Campus Universitário de Santiago 3810-193 Aveiro
Extended Abstract:
Modern digital telecommunication systems demand a steady improvement of the RF frontend’s performance in terms of bandwidth, power added efficiency, and signal fidelity. This is
especially true in microwave power amplifiers for which many advances have been made
public. In this respect, one of the most promising device technologies is the one based on
wide band-gap materials like GaN HEMTs. These devices already offer power transistors
capable of delivering very high output powers, which are expected to be accompanied by also
interesting linearity figures.
This paper presents a highly linear 2W Class AB GaN Power Amplifier. All design stages
are explained, from the matching networks (input and output) up to the bias circuitry. For
supporting the nonlinear design of the PA, a convenient GaN HEMT equivalent circuit model
was especially built and integrated in a harmonic-balance simulator.
To validate the design strategy, the measured power amplifier performance was compared
with the one predicted by the nonlinear PA model. This way, its linear and nonlinear
predictive capabilities could be studied.
The linear predictions are illustrated with S-parameter data, while the nonlinear ones
focused on CW output power, Gain, power added efficiency and two-tone intermodulation
distortion values.
The observed performance data and the simulated and measured results’ comparison were
considered very good, fully validating, this way, the adopted PA design methodology.
3
Highly Linear GaN Class AB
Power Amplifier Design
Pedro Miguel Cabral, José Carlos Pedro and Nuno Borges Carvalho
Instituto de Telecomunicações – Universidade de Aveiro, Campus Universitário de Santiago 3810-193 Aveiro
Abstract—This paper presents a highly linear 2W Class AB GaN Power Amplifier. All design stages are explained,
from matching networks up to bias circuitry. The obtained Power Amplifier performance is compared with the one
predicted by a nonlinear device model especially built for these transistors. This model’s linear and nonlinear
predictive capabilities are studied. The former are illustrated with the comparison between measured and simulated
S-Parameters while the latter with output power, power added efficiency and intermodulation distortion data.
I. INTRODUCTION
The deployment of modern digital telecommunication systems, with continuously increasing capacity, has
demanded a steady improvement of the RF front-end’s performance in terms of bandwidth, power added
efficiency, PAE, and signal fidelity. This is especially true in microwave power amplifiers, PAs, for which many
advances have been made public.
In this respect, one of the most promising device technologies is the one based on wide band-gap materials
like GaN HEMTs. Despite the recognized device processing infancy, it already offers power transistors of
unbeaten breakdown voltages, therefore capable of delivering very high output power, Pout, figures [1].
Also significant is the high linearity provided by these GaN HEMTs. In fact, the observed valleys of
intermodulation distortion, IMD, versus input drive level patterns, which have been frequently observed in class
AB PA designs, constitute a great help in achieving the aimed compromise between nonlinear distortion and
power added efficiency [2],[3].
Unfortunately, the critical dependence of these IMD valleys, the so-called large-signal IMD sweet-spots, on
almost unsuspected issues like: out-of-band terminations [4], device’s strong and mild nonlinearities [2],[3] and
quiescent point (not unusually in ranges of only a few tenths of Volt) have raised the quality standards of
common PA design methodologies and nonlinear device models.
This paper addresses the design of a microwave GaN PA, paying particular attention to the prediction of
small- and large-signal Pout and IMD. Section II introduces the most important design stages and Section III
presents all tests made and compares the results obtained from measurements with the ones predicted by a
nonlinear global model especially conceived for this kind of active devices. Finally, Section IV summarizes all
the work done.
II. 900 MHZ CLASS AB POWER AMPLIFIER DESIGN
The active device chosen for our PA was a 2mm GaN HEMT on Si substrate, encapsulated in a standard high
power microwave package, similar to the one used in [5].
Our PA design goals were to simultaneously optimize Pout, PAE and signal to intermodulation ratio. Gate
voltage was selected to maximize signal to IMD ratio and, as it is widely known, an active device operating
under class AB is able to provide the best compromise between linearity, efficiency and output power. Drain
voltage was set to 20 V taking full profit of this wide band-gap device output voltage and current excursion
capabilities.
After a few VGS tests around VT (i.e., close to class B and AB) it became clear that best IMD performance
could be achieved when the HEMT presented an IMD vs Pin pattern with double minima. This led to a quiescent
point VGS1 = -4.20 V or 4% of IDSS.
Maximization of Pout and PAE demanded a careful selection of the Cripps load-line and fine tuning of the
even harmonics [6]. Fig. 1a shows the schematic used to determine the output matching network requirements in
order to achieve drain constraints. Fig. 1b shows the simulated IDS vs VDS characteristics, for six different VGS
biases and superimposed to it the desired and obtained drain load line.
4
Rd
Ld_B
Ld
1.0
R31
OUT_Match
50 Ω
Drain
IDS (A)
Cds
0.5
C31
34 Ω @ 900 MHz
a
0 Ω @ 1800 MHz
0
5
10
15
20
VDS (V)
25
30
35
b
Fig. 1. (a) Schematic used to determine output matching network requirements. (b) Simulated IDS vs VDS characteristics, for six different VGS
biases (-) and desired (--) and obtained (-x-) drain load line.
A two-stub output matching network was designed to guarantee the calculated intrinsic 34 Ω load-line at 900
MHz central frequency and a short-circuit at the 1.8 GHz (2nd harmonic).
Fig. 2 shows the simulated output match response at the drain from 900 MHz to 1800 MHz.
1800 MHz
900 MHz
34 Ω
Fig. 2. Simulated output match response seen at the drain from 900 MHz to 1800 MHz.
After designing the output network, next stage was to conceive an input network capable of providing
possible source matching and optimized gain without in-band instability. As it is known, that is important to
compensate for the expected gain loss caused by the PA output mismatch. After this, a broad band stability
analysis was conducted which showed potential problems at VHF. This was solved by the design of convenient
lossy gate and drain bias networks.
However, since it is known that the bias circuitry also determines the device terminations at the envelope
frequencies and thus nonlinear distortion performance [3], they were retuned to guarantee very low impedances
at most of the envelope bandwidth (4 MHz).
Fig. 3 shows the simulated output match response at the drain from 30 kHz to 4 MHz.
4 MHz
30 KHz
Fig. 3. Simulated output match response seen at the drain from 30 kHz to 4 MHz.
In fact, it could be confirmed during the simulation, and then in the PA testing, that these low frequency
terminations can either jeopardize IMD performance or even introduce undesired sideband asymmetries (a
symptom of long term PA memory effects) [7].
The PA was then implemented in MIC technology using a RT/Duroid high frequency laminate with a εr=10,2.
Fig. 4a shows the complete PA schematic and Fig. 4b a photograph of the implemented amplifier board.
40
5
VGS
VDS
Output
Input
a
b
Fig. 4. (a) Complete PA schematic. (b) Photograph of the implemented PA MIC board.
III. 900 MHZ CLASS AB POWER AMPLIFIER TESTING
In order to validate our PA design, several experimental tests were conducted and the results thus obtained
compared with the ones predicted by the model presented in [5].
A. Small-Signal S-Parameter Measurements
The first PA test was a set of broad band small-signal S-parameter measurements.
0
-5
10
-5
|S11| (dB)
-10
0
-10
-15
-20
0
|S22| (dB)
20
|S21| (dB)
0
a
0.2
0.4
0.6 0.8 1.0
Freq (GHz)
1.2
1.4
-20
0
-10
-15
b
0.2
0.4
0.6 0.8 1.0
Freq (GHz)
1.2
1.4
-20
0
c
0.2
0.4
0.6 0.8 1.0
Freq (GHz)
1.2
1.4
Fig. 5. Measured (x) and modeled (–) PA |S11| (a), |S21| (b) and |S22| (c).
As seen in Fig. 5a, Fig. 5b and Fig. 5c, there is a reasonable good agreement between measured and modeled
results.
50
30
40
20
30
10
20
0
-10
-20
a
-15 -10
-5
0
5
Pin (dBm)
10
15
10
0
20
17
16
Gain (dB)
40
PAE (%)
Pout (dBm)
B. Large-Signal One-Tone Measurements.
The second test step consisted in several CW experiments to evaluate Gain, Pout and PAE versus input drive
level. As seen in Fig. 6a, the PA presents a 1dB compression point of 2 W with an associated Gain of 15 dB,
and a PAE of nearly 32 %. Compared to the model predictions, it is clear that the efficiency came somewhat
lower than expected, while the output power and gain deviations were within the measurement error.
Nevertheless, one remarkable result that should be pointed out is the correct prediction of the Gain versus Pin
pattern, Fig. 6b, despite the rather complex behavior of gain compression followed by gain expansion to end up
again in gain compression. This is a direct consequence of the selected bias point as it can be related to the
double minima IMD pattern aimed at the PA design phase [3].
15
14
13
-20
b
-15 -10
-5
0
5
Pin (dBm)
10
15
20
Fig. 6. (a) Measured (x) and modeled (–) Pout and PAE under CW operation. (b) Measured (x) and modeled (–) Gain vs Pin under CW
operation.
6
40
40
20
20
20
0
-20
0
-40
-40
-60
a
-10
-5
0
Pin (dBm)
5
10
15
-60
-80
0
-20
-20
-40
-80 -15
Pout & IM3 (dBm)
40
Pout & IM3 (dBm)
Pout & IM3 (dBm)
C. Large-Signal Two-Tone Nonlinear Distortion Measurements.
In nowadays communication systems, the wide variety of distinct modulation schemes and wideband signals
present a statistical amplitude distribution that is quite different from the one of a simple CW or two-tone
excitation [8]. Therefore, tailoring the IMD versus Pin pattern is crucial. Fortunately, the GaN HEMTs under
study showed a very flexible IMD control via VGS bias.
In order to evaluate our device’s model flexibility of accurately reproducing the dramatic changes observed
on the IMD vs Pin pattern with bias variations, several two-tone excitations (tones centered at 900 MHz, with a
frequency separation of 100 kHz) were applied at our PA input, for three different VGS values (VGS1 = -4,20 V,
VGS2 = -4,15 V and VGS3 = -4,10 V). All these gate voltages guaranteed class AB PA operation.
Fig. 7a, Fig. 7b and Fig. 7c, show Pout and IM3 vs Pin for the three VGS values previously mentioned.
b
-15
-10
-5
0
Pin (dBm)
5
10
15
-60
-80
c
-15
-10
-5
0
Pin (dBm)
5
10
15
Fig. 7. Measured (x) and simulated (–) PA Pout and IM3 vs Pin for VGS1 (a), VGS2 (b) and VGS3 (c).
As seen from the data depicted in Fig. 7a, Fig. 7b and Fig. 7c, there is again a good agreement between the
predicted and measured results. As VGS values are increased, it is possible to see a displacement in the minima
position. In Fig. 7a there are two minima, Fig. 7b is similar to the previous one but now the minima are closer
and in Fig. 7c they are overlapped.
Therefore, controlling the IMD pattern represents an enormous advantage when designing a PA with linearity
and efficiency constraints.
IV. CONCLUSIONS
A 2W class AB power amplifier circuit was built and all goals and design stages explained. The test results
were compared with the ones predicted by a nonlinear global model. Indeed, a remarkable good agreement
between measured and simulated Pout, PAE and two-tone IM3, was obtained in a practical circuit. This
represents an important achievement as GaN modeling studies are still in their early stages of development.
ACKNOWLEDGMENT
The authors would like to acknowledge Eng. João Paulo Martins for the development of the automatic
measurement benches extensively used throughout this work, Nitronex Corporation for providing the GaN
HEMT devices, Portuguese Science Bureau, F.C.T., for the Ph.D. grant Ref. 11323/2002, given to the first
author and financial support provided under Project POCTI/ESE/45050/2002 MEGAN. This work was also
partially supported by EC NoE TARGET.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
K. Joshin, T. Kikkawa, H. Hayashi, T. Maniwa, S. Yokokawa, M. Yokoyama, N. Adachi and M. Takikawa, “A 174 W high-efficiency
GaN HEMT power amplifier for W-CDMA base station applications”, in Proc. IEEE International Electron Devices Meeting Technical
Dig., pp. 12.6.1-12.6.3, Dec. 2003.
N. B. Carvalho and J. C. Pedro, “Large and Small Signal IMD Behavior of Microwave Power Amplifiers”, IEEE Trans. on Microwave
Theory and Tech., vol. 47, No. 12, pp. 2364-2374, Dec. 1999.
J. C. Pedro and N. B. Carvalho, Intermodulation Distortion in Microwave and Wireless Circuits, Artech House, 2003.
F. Palomba, M. Pagani, I. De Francesco, A. Meazza, A. Mornata, G. Procopio and G. Sivverini, “Process-Tolerant High Linearity
MMIC Power Amplifiers”, in Proc. Gallium Arsenide Applications Symposium Proc., Munich, pp. 73-76, Oct. 2003.
P. M. Cabral, J. C. Pedro and N. B. Carvalho, “New Nonlinear Device Model for Microwave Power GaN HEMTs”, to be published in
the IEEE Int. Microwave Symposium Dig., Fort Worth, Jun. 2004.
S. C. Cripps, RF Power Amplifiers for Wireless Communications, Artech House, 1999.
N. B. Carvalho and J. C. Pedro, "A Comprehensive Explanation of Distortion Sideband Asymmetries", IEEE Trans. on Microwave
Theory and Tech., vol. 50, No. 9, pp. 2090-2101, Sept. 2002.
J. C. Pedro and N. B. Carvalho, “Designing Band-Pass Multisine Excitations for Microwave Behavioral Model Identification”, to be
published in the IEEE Int. Microwave Symposium Dig., Fort Worth, Jun. 2004.