High Linearity Class B Power Amplifiers in GaN HEMT Technology

1
High Linearity Class B Power Amplifiers in GaN
HEMT Technology
S. Xie, V. Paidi, R. Coffie, S. Keller, S.
Heikman, A. Chini, U. Mishra, S. Long, M.
Rodwell Department of Electrical and Computer
Engineering, University of California, Santa
Barbara
Email paidi_at_ece.ucsb.edu, sxie_at_engr.ucsb.edu
Phone [phone] Fax [phone]
2002 Topical Workshop on Power Amplifiers
September 2002, UCSD
2
Outline
UCSB
S. Xie, V. Paidi

• Introduction
• How does Class B PA work?
• Why single-ended Class B?
• Highly linear single-ended Class B PA design and
  simulation
• Measurement results
• Conclusions

3
How does push-pull Class B PA work?
UCSB
V. Paidi, S. Xie
Vout VDS1 VDS2
Vin

• Two identical devices working in 50 duty cycle
  with 180 phase shift.
• Half sinusoidal drain current on each device, but
  full sinusoidal drain voltage.
• Even harmonics are suppressed by symmetry gt wide
  bandwidth (limited by the power combiner).
• Class B Ideal PAE 78.6 feasible PAE 40-50
  (typical GaN HEMT at X-band) Class A Ideal PAE
  50, feasible PAE 20-30.

4
Why not push-pull for RF MMIC
UCSB
V. Paidi, S. Xie
To obtain high efficiency (78), a
half-sinusoidal current is needed at each drain.
This requires an even-harmonic short. This can be
achieved at HF/VHF frequencies with transformers
or bandpass filters. However,

1. Most wideband microwave baluns can not provide
   effective short for even-mode. Efficiency is
   then poor.
2. They occupy a lot of expensive die area on MMIC.

5
Single-ended Class B push-pull
UCSB
V. Paidi, S. Xie
Push-pull Class B
Even harmonics suppressed by symmetry
Single-ended Class B with Bandpass filter
Conclusion From linearity point of view,
push-pull and single-ended Class B with bandpass
filter B are equivalent same transfer function.
Even harmonics suppressed by filter
Bandwidth restriction lt 21
6
Why is biasing critical for Class B?
UCSB
V. Paidi, S. Xie
Ideal Class B
Bias too low Class C
Bias too high Class AB
7
Voltage Transfer Function as a Function of Bias
Voltage
UCSB
V. Paidi, S. Xie
Class AB, nonlinear
Class B, linear
Class C, nonlinear
Class B is linear given that the current transfer
function is linear
8
A source of IM3 distortionTransconductance
distortion
UCSB
V. Paidi, S. Xie
Highly linear characteristics of GaN HEMT on SiC
The third order term in the Taylor expansion is
very small when biased at pinch off.
Bias point
The distortion will be minimum when the amplifier
is biased at Class B by using GaN HEMT on SiC.
9
Single-ended Class B Power Amplifier
UCSB
S. Xie, V. Paidi

• Lossy input matching network to increase the
  bandwidth
• Cds is absorbed into the ?-section output lowpass
  filter

? - section lowpass filter
Lossy input matching
10
Simulation performance of Class B
UCSB
S. Xie, V. Paidi
Waveforms of drain voltage and current
Simulation of class B amplifier _at_10GHz
Saturated output power 37 dBm
Saturated PAE 48
Class B bias Saturated output power 37
dBm, Saturated PAE 48
11
Simulation of Intermodulation Suppression and
PAE _at_10GHz
UCSB
S. Xie, V. Paidi
Best IM3 suppression is achieved at Class B and
Class A
12
Device Performance of GaN HEMTs
UCSB
V. Paidi, S. Xie
Performance of 12 fingers (1.2mm) device

• Lg 0.25um
• Idss 1A/mm

• ft 55 GHz ( 50 GHz for dual gate)
• Vbr 40V ( 55V for dual gate)

RF Performance
DC I_V curve
1.2 A
Ft55 GHz
13
Chip photograph of Class B power amplifier
UCSB
V. Paidi, S. Xie
Gate 2
Gate 1
Source
Air bridges
Drain
(Approximately 6mmX1.5mm)
14
Measurement setup
UCSB
S. Xie, V. Paidi
Measurements

• Single tone from 4 GHz to 12 GHz
• Two-tone measurement at f1 8 GHz, f2 8.001
  GHz
• Bias sweep Class A (Vgs -3.1V), Class B (Vgs
  -5.1V, Class C (Vgs - 5.5 V) and AB (Vgs -4.5
  V).

15
Class B power amplifier measurement result
UCSB
S. Xie, V. Paidi
Gain vs. frequency
Class AB
Class B
3 dB bandwidth 7GHz - 10GHz
16
Class B bias _at_Vgs - 5.1V
UCSB
S. Xie, V. Paidi
Single tone performance _at_ f0 8GHz
Saturated output power 36 dBm
PAE (maximum) 34
f1,f2
Two tone performance _at_ f18GHz, f28.001GHz
2f1-f2, 2f2-f1

• Good IM3 performance
• 40dBc at Pin 15 dBm, and
• gt 35 dBc for Pin lt 17.5 dBm

17
Class A bias _at_Vgs - 3.1V
UCSB
S. Xie, V. Paidi
Single tone performance _at_ f0 8GHz
Saturated output power 36 dBm
PAE (maximum) 34
Two tone performance _at_ f18GHz, and f28.001GHz
f1,f2
Saturated output power each tone 33dBm
2f1-f2, 2f2-f1
Good IM3 performance at low power level but
becomes bad rapidly at high power levels
18
Summary of IM3 suppressions
UCSB
S. Xie, V. Paidi
Class B
Class A
Class AB
Psat
Class C

• Low output power levels (Pout lt 24 dBm), Class A
  and Class B both exhibit good linearity (Class B
  gt 36 dBc, Class A gt 45 dBc).
• Higher output power levels, Class A behaves
  almost the same as Class B.
• Class AB and C exhibit more distortion compared
  to Class A and B.

19
Class B vs. Class A
UCSB
S. Xie, V. Paidi
IM3 suppression and PAE of two-tone
PAE of single tone
Class A
Class B
Class A
Class B
While maintains the same IM3 suppression as Class
A, Class B can get more than 10 of PAE.
20
Conclusions
UCSB
S. Xie, V. Paidi

• For a less than octave bandwidth, a push-pull
  Class B amplifier can be replaced by a
  single-ended Class B amplifier with bandpass or
  lowpass filter.
• A single-ended Class B MMIC power amplifier in
  GaN HEMT is designed and 36dBm of saturated power
  and 35dBc of IM3 suppression are obtained.
• Class B is better than Class A because it can get
  good IM3 performance comparable to that of Class
  A, while providing PAE 10 higher than that of
  Class A.

21
Acknowledgements
UCSB
S. Xie, V. Paidi
This work was supported by the ONR under grant
(N00014-00-1-0653) Special thanks to Dr. Walter
Curtice, who provide us the C_FET3 model for
simulation Thanks to L.-Y. (Vicky) Chen and
Likun Shen, who helped us for the measurement.