Gan Power Amplifier Design

GaN Power Amplifier Design
Presented to IEEE Long Island MTT Chapter

June 18, 2014
Ed Niehenke, PhD, PE
Baltimore, MD 21075
(410) 796796-5866, e.niehenke@ieee.org
Copyright 2014, Edward C. Niehenke, Niehenke Consulting, Inc. All rights
reserved. No part of this document may be copied or reproduced in any manner
without the written permission of Edward C. Niehenke.
1
2014 Niehenke Consulting Inc.
GaN Power Amplifier Design -1
GaN Power Amplifier Design
This lecture introduces attendees to the GaN transistor, its

properties, various structures, including the latest GaN power
amplifier (PA) design techniques.
The properties of GaN will be presented showing the
advantage of these devices over GaAs and Si. GaN HEMT
transistors will be shown delineating the various geometries,
semiconductor processes and structures with associated
performance.
Guidelines for reliable operation will be presented considering
device junction temperature including thermal management
techniques.
The nonlinear models of GaN HEMT devices necessary for the
CAD of PAs will be presented.
Design considerations for both constant amplitude envelope
signals as well as the nonnon-constant amplitude envelope
signals will be presented.
StepStep-byby-step design procedures will be shown for various
GaN PA examples including different classes of operation as
well as the popular Doherty PA.
EDWARD C. NIEHENKE, Ph.D., PE,

Consultant, Baltimore, Maryland, USA.
Niehenke has pioneered the development of statestate-ofof-thethe-art RF,

microwave, and millimeter wave components at
Westinghouse/Northrop Grumman for 34 years. Circuits include low
noise amplifiers, low noise oscillators, mixers, power amplifiers,
amplifiers,
phase shifters, attenuators, limiters, frequency multipliers, low
low-phase
noise millimeter wave fiber optical links, and miniature integrated
integrated
assemblies and subsystems. He previously worked in cryogenic
electronics research at MartinMartin-Marietta. He now consults and lectures
on linear/nonlinear and wireless transmit/receive circuits and
systems. Since 1983 he lectured to over 3000 professionals
throughout the world for Besser Associates and the Continuing
Education of Europe. He holds nine patents, one George
Westinghouse Innovation Award, and has authored numerous
papers on RF, microwave, and millimeter wave circuits.
Niehenke is active in IEEE MTTMTT-S activities serving on three technical
committees and is their Ombudsman . He was technical program
chair of the 1998 IMS, chair 1986 IMS, 1986/87 IEEE Distinguished
Distinguished
Microwave Lecturer, and served as a member of ADCOM for 9 years
Niehenke taught electricity and magnetism for 3 years at Johns
Hopkins University. He was a recipient of the IEEE Centennial and
and
Millennium Medals, is a fellow of the IEEE, and is a registered
professional engineer in the State of Maryland.
What is Important for a Power Amplifier
High Power
High Efficiency
High Reliability
Good Frequency Range
Non Complex Matching
with wide bandwidth
Linear for Varying
Amplitude Modulation
Low Cost
Lets Examine a Transistor
To see to see what parameters are important

For Output Power
For Efficiency
Power and Efficiency Class A

Current (I)
Imax
IV Trajectory
VDC
V DC
Idc
Imin
VDC
Vmin
Voltage (V)
I
Vmax
Transistor Current
GaN HEMT Basic Transistor Circuitry

Drain to Source
DC Bias Voltage
Gate to Source
DC Bias Voltage
GaN HEMT
Tcase = 250C
Rth = 3.80C/W
Gate
Biasing
Drain
Biasing
Output Port
Load Resistance
RF Source DC
Blocking
Cap
DC
Blocking
Cap
Drain
to Source
Voltage
and
Current
Monitors
Temperature Increase
Monitor Probe
Material Properties Comparison

RF Power = (Vmax Vmin)Imax/8
Material
Band Gap
Energy
(eV)
eV)
Critical
Breakdown
Field
(MV/cm)
Thermal
ConducConductance
(W/cm(W/cm-0K)
Mobility
(cm2/V/V-s)
Saturated
Velocity
(107 cm/s)
Relative
Dielectric
Constant
Si
1.1
0.3
1.5
1300
11.9
GaAs
1.4
0.4
0.5
6000
1.3
12.9
4H SiC
3.2
3.3
3.7
610
9.7
6H SiC
3.0
3.0
4.9
310
2.0
9.7
GaN
3.4
3.0
1.5
1500
2.7
9.0
High
High Power
High Power
Temperature (High Vmax)
(High Imax)
Operation
High Power GaN on 4H SiC
(Low Thermal Heating)
GaN mobility enhanced with HEMT structure for higher gain and frequency operation
Differences Between 4H and 6H SiC Substrates
4H and 6H SiC are different crystalline structures of the

material with different properties
6H SiC is conductive and used for LED
LEDs (one contact)
4H SemiSemi-insulating, better crystalline matched to GaN,
GaN, and
easier to make
GaAs Low Noise Operation
The mobility of GaN is so much lower than GaAs.

GaAs.
For Ultra Low Noise GaAs PHEMTs are used over
GaN.
GaN.
The mobility shown is for GaN but GaN HEMT
HEMTs
have higher mobility. The noise figure is very good
for these devices not as low as the GaAs PHEMTs.
PHEMTs.
For example the Crey 0.4 mm device:
Where the GaN HEMT
will shine is to have
reasonable NF with
high power capability
so a receiver protector
could be eliminated
also they would have
a high IP3
Power vs. Frequency for Solid State Devices

FP2=K
1000
BJT (Pulsed)
MOSFET
Power (W)
100
BJT
LDMOS
FET
10
GaN HEMT
HBT
HEMT
PHEMT
0.1
0.01
0.1
10
100
Frequency (GHz)
Silicon LDMOS Versus GaN HEMT Transistor
The bandgap energy of silicon is only 1.1eV but Si

LDMOS are used for high power applications.
Silicon as used in LDMOS can be run at 1500C for
high reliability, not 180 to 2250C for GaN (Higher
bandgap).
bandgap).
LDMOS has good breakdown voltages and also can
use a large device to get the current for power.
However is typically limited to use around 3 GHz.
The GaN HEMT is a much higher frequency device.
Cost is the issue and GaN price is now coming
down.
High Efficiency ()
= k (Vmax
(Vmax Vmin)/(Vmax + Vmin)
Vmin)
k = 50% Class a, 78.5% Class b and higher for
class c, d, e, f, and f-1
We will discuss later in the lecture all the details of
the various classes of operation
With high voltage operation (high Vmax),
Vmax), and
reasonable low Vmin (knee voltage) the ratio of
(Vmax Vmin)/(Vmax + Vmin)
Vmin) will close to unity as
opposed to low voltage operation where the ration
takes a big hit on efficiency
Operation with high standing waves
With a highly mismatched load, the transistor will

see a higher voltage depending on the phase of the
mismatch, so for the worse case the breakdown
voltage must be larger than normal for this
condition.
GaN Transistor Meets Objectives
High Power
High breakdown voltage
High peak current, (Saturated velocity)
High current capability
Good efficiency
Low knee voltage (Vmin
(Vmin))
High Vmax --- High breakdown voltage
Can cutoff current at high voltage low Imin
Low semiconductor and circuit losses
High Reliability
High temperature operation (High band gap material)
High reliability process
Low thermal resistance
Lets Look at a GaN Device to See How it Performs

Compared to the Calculated Performance
1.9 GHz
28V DC operation
GaN HEMT Device on 100 m of 4H SiC
0.25 Gate Length
1.8 mm Wide
Nonlinear Model Developed by Dr. Walter Curtice
Load the file cfet9.dll into AWR model directory
Open up the file hb_cfet_1p8mmClassA.emp for
class A operation and hb_cfet_1p8mmClassB.emp
for class B operation
1.8 mm GaN HEMT
Note that the IV has a Negative Slope

In the Saturated Region
This is due to self heating of the transistor. As the

transistor heats up the current decreases
Self heating turned off
With self heating
1.8 mm GaN HEMT Class A
Vmax = 58.29, Vmin =0

Vmin = 3.2 V, I max = 1.12 A
RL = (Vmax
(Vmax Vmin)/Imax
Vmin)/Imax
RL = (58.29 3.2)/1.12 = 49.2
49.2
P = (58.29 3.2) (1.12)/8
P = 7.72 W or 38.9 dBm
Pdc = 28 * 0.62A = 17.36 W
= 7.72/17.36 = 45%
= 0.5 * (58.28(58.28-.2)/(58.28+3.2)
= 45%
PAE = (Pout Pin)/Pdc

PAE = (1 1/G)
G = Gain
For high gain PAE =
Load Line Selection
Examine the load line and adjust Pin, Vgs and

RLoad until you get simultaneously a good voltage
and current swing.
There are many answers so you should also look at
the data of Pout, PAE, Gain and Gain Compression
and choose the best.
When choosing possibilities, always compare
possibilities with the same gain compression
amount
I chose a condition is where the gain is compressed
about 1 dB in comparing the various possibilities.
You can also do an optimization procedure.
10

Calculated parameters similar to Harmonic Balance results
P = 38.9 dBm
= PAE = 45%

Gain and Phase VS Pin
Note low Gain and Phase variation versus Pin ideal for low ACPR and
EVM for digitally amplitude variant modulated signals
11
Thermal Considerations
Pdc in
Prf in
Prf out
Tj = Ths + P dissipated *
Tj = Junction temperature
Ths = heat sink temperature
= Thermal resistance
Pdiss = Pdc in + Prf in Prf out
Class a has high temperature when not obtaining output power
because dc power and inputs power are the heat inputs.
With higher input powers, significant power is extracted,
reducing the dissipated power and junction temperature rise
Thermal Port Voltage
The thermal port voltage is calibrated to read the

junction temperature rise over heat sink
temperature. The heat sink temperature is an input
to the model.
You just put a voltage monitor on it and read the
voltage value and change volts to 0C.
12
1.8 mm GaN HEMT Class A Thermal

Considerations
Models of transistors include a thermal port to check thermal

rise as part of nonlinear program and also vary RF modeling
parameters as a function of heating
Note high thermal rise for class a for low input powers
Lets work in class b and examine properties
Operation under Pulsed Conditions
Some transistor models have the capability to consider

pulsed conditions.
In the model below, the current source represents the
instantaneous power in the transistor.
The RC represents the thermal time constant of the device.
The voltage out is the temperature rise above the heat sink.
So you can run pulsed
conditions with specific
duty cycles to determine
the temperature
13
Thermal resistance vs time
Crey device
Effect of Conduction Angle on Ans

An Coefficients Normalized to Imax = 1A
Constant Gm
0.6
0.5
A1
0.4 Fundamental
A0
DC Component
0.3
A2
0.2
Second Harmonic
A3
Third
0.1
A4
Harmonic
Forth
0
Harmonic
-0.1
0
45
90
135 180 225 270 315 360
Class
B
AB
Conduction Angle (Degrees)
ECN FT11.MCD
14
Normalized Efficiency (NORM), Normalized Power (PNORM), and

Normalized Load Conductance (GNORM) Versus Amplifier Conduction
Angle - Constant Transconductance - Resistive (R) and Tuned (T) Load
1.2
1.1
PNORM (R)
GNORM (R)
1
0.9
PNORM and GNORM (T)
0.8
NORM (T)
0.7
NORM (R)
0.6
0.5
PRF = PNORM (Imax - Imin) (Vmax - Vmin)/8
0.4
0.3
GRF = GNORM (Imax - Imin)/ (Vmax - Vmin)
0.2
0.1
NORM
/F
F = (1 + Imin/(A0(Imax - Imin)) (1 + Vmin/((1 A0)(Vmax - Vmin)))
45
90
135
180
225
270
315
360
Conductance Angle (Degrees)

1.2
Relative Output Power (P/Pnorm), and Efficiency (

vs Input Power (p2)and Conduction Angel (CA)
Class A and B linear, Class AB nonlinear
improved
improved for class AB especially at low drive level (p)
P/Pnorm
1.1
1
2700
0.9
2250
1800
0.8
0.7
2250
0.6
3600
2700
0.5
0.4
0.3
1800
2250
2700
3600
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Normalized Variable Input Power (p^2)

15
Class B Operation - Tuned Load, 78.5%

Current (I)
Imax
IV Trajectory
Imin
VDC
VDC
Class B, A0 1 / , A1 0.5
R
-180
Vmax
-90
0
-90
t (Degrees)
180
t (Degrees)
Vmin
Transistor Current (I)
VDC
Transistor Voltage (V)
1.8 mm GaN HEMT Class B
Second harmonic short added to circuit for good class b efficiency

efficiency
Power same for class a and b
Efficiency better, = 78.5 (56.26 3.2)/(56.26+3.2) = 70%
16
1.8 mm GaN HEMT Class B
Gain and phase varying VS Pin for class B

May need redesign for better linearity
Gain Compression with GaN Devices
Some GaN HEMTs devices show a soft

compression, so in my example I took the
performance where I got the Pout
If the product is used for a constant envelop, then
work the PA where the PAE is maximum
If the PA is used for varying envelop device, then
design the semiconductor process and matching
techniques for low AM/AM and AM/PM
characteristics, and you have a winner
17
AM to PM of Power Amplifiers
AM/PM is caused by a nonlinear parameter or

temperature parameter changing with input power.
For example the transistor has nonlinear capacitors
(varactors)
varactors) which vary depending on the input
power to the device.
Also as the temperature changes (input power), this
can cause a phase shift
Thermal Considerations
For Class b, current low at low output power
Pdc in
Prf out
Prf in
18
1.8 mm GaN HEMT Class B Thermal

Considerations
Models of transistors include a thermal port to check thermal

rise as part of nonlinear program and also vary RF modeling
parameters as a function of heating
Temperature rise only 420C now not 1600C rise
GaN Transistor Meets Objectives
High Power
High breakdown voltage
High peak current, (Saturated velocity)
High current capability
Good efficiency
Low knee voltage (Vmin
(Vmin))
High Vmax --- High breakdown voltage
Can cutoff current at high voltage low Imin
Low semiconductor and circuit losses
High Reliability
High temperature operation
High reliability process
Low thermal resistance
Thermal management critical because of the very high power
density in a small area and heat must be removed with low
thermal resistance
19
Class A, B, AB, C, D, E, F, and F-1

Operation
Constant transconductance, tuned load
1max-Imin=1
IV Trajectory
Imax
Bias Point
D, E, F, F-1
C
Imin
Vmin
A
AB
VQ
C
Vmax
-90
-180
AB
t (Degrees)
is angle from origin where current is at Imin

Conduction angle is 2
Class
Conduction Angle Efficiency
Gain
Linearity
50 %
High
Good
A
3600
50 - 78.5% -3 - -6 dB Harmonics
AB
180 - 3600
78.5 %
-6 dB
Harmonics
B
1800
>78.5 %
Low
Harmonics
C E, F, F0-1 - 1800
C, D,
Class D, E, F, and F-1

Class F
Class D
Vcc
Idc
Iout(t)
RFC
/4
Vce(t)
f0
V0ut(t)
fop01.vsd
Class F-1
Class E
20
Class D1 (Push-Pull)
Odd mode balance places virtual short circuit in

symmetry plane providing low Z at second
harmonic
Series LL-C provides open at harmonics especially
important for 3rd and 5th harmonic
This enhances Vce to look like square wave with
value when Icd = 0 and Vce = 0 with current value
Result: very high efficiency
Vbe
Vcc
Ice1
Cser
Q1
Lser
Vce1
RG
RL
Vce2
Q2
ClassD1.VSD
Ice2
Power Amplifier Operating Class Summary

A
Operating
Class
Conduction
Angle
Gain Reduction
To Class A
DegDegrees
360
dB
Ideal Efficiency
Efficiency
Reduction
50
Poor
fo
Harmonic
Suppression
nfo
Peak Transistor
Voltage
Vmax
Vcc
Volts
Vo
Volts
F
Inverse
65
78
85
100
100
100
100
100
12Vo/Vcc
See
Notes
1Vo/Vcc
1Vo/Vcc
1Vo/Vcc
Good
Very
Good
Poor
Fair
Good
180
Poor, Class D, D1, E, and F need to be driven

hard to obtain square wave pulse required for
high
<20 MHz KW P
<2 GHz Watts P
Short
All nf0
2Vcc2Vcc-Vo
Volts
Knee Voltage
D1
Short 2f0
Supply Voltage
<180
<100
Good
180
GHz
Pout/Pin Gain
Linearity
180
to
360
1-Vo/Vcc
Efficiency at
Reduced Power
Frequency
AB
Series LC circuit
resonant at fo
VccVcc-Vo
Poor
3.5Vcc3.5Vcc2.5Vo
<20
Short even n
Open odd n
2Vcc2Vcc-Vo
Open
even n
Short
odd n
V0+(Vcc
-Vo)
Poor (Good with system solution)
21
Conclusions
We have seen that the GaN HEMT meets theoretical

performance limits
Lets now examine substrates for the GaN HEMT
Transistor Configurations
With and without field plate
Packaging for high density transistors
Cree Devices, on 4HSiC
Pout, PAE, Reliability, Thermal Improvements
Nitronex Devices on high resistivity Silicon
Packaging of High Power Density Transistors
GaN-on-SiC MEMT Transistors and MMICs Enter the Mainstream,

Ray Pengelly, Cree RF and Microwave Products
Presented to IEEE AP/MTT Chapter, November 3, 2011
22
Properties of Relevant Materials

Cree
Field plate is connected to source. This reduces the

gate to drain capacitance due to field plate insertion
23
Double Recess Process to Increase

Breakdown Voltage
Many GaN HEMTs employ a double recess process to

increase the breakdown voltage similar to GaAs
HEMTs.
HEMTs.
Also many GaN HEMTs employ a field plate similar to
many LDMOS transistors
Cree GaN Devices with and without Field Plate

24

Thermal Improvements

25
Typical Performance of Cree Field Plate HEMTs

GaN ON Si HEMT Devices Nitronex
The AlGaN/GaN HEMT structure results higher current and mobility

compared to a FET structure
The field plate improves the breakdown voltage
Metal organic chemical vapor deposition (MOCVD) is the preferred
method of producing GaN-based thin films on Si or SiC for RF
applications.
GaN-based devices and structures have been mass-produced using
MOCVD for lighting applications and the same supporting high
volume epitaxial manufacturing infrastructure can be leveraged for
RF applications.
26
GaN HEMT Devices Nitronex
Can operate at 1800C with MTBF of 106
Nitronex GaN on Si HEMT Device
27
GaN HEMT Structures Higher Frequencies
BAE Non Field Plate 0.2 m Gate MMIC Process
ft = 50 GHz, fmax = 220 GHz, 5W/mm, 46% PAE at 30 GHz
High Efficiency Ka/Q Band PHEMT Power Amplifier MMICs Dr. James J. Komiak, BAE Systems,
IMS2011
Workshop
2014
Niehenke Consulting Inc.
28
HRLs Baseline T Gate Structure
GaN T-Gate and Field Plate Technology for Applications Below 45 GHz Harris
Moyer, HRL Laboratories, IMS2011 Workshop, Introduction to GaN MMIC
Design
Moyer, HRL, IMS2011 Workshop, Introduction to GaN MMIC Design
29
Moyer, HRL, IMS2011 Workshop, Introduction to GaN MMIC Design
MMIC: B. Kim et al., A Linear, High-Efficiency GaN Power Amplifier Operating at 74 GHz, GOMAC
2011, March 2011
Presentation: IMS2011 Workshop WMA, Q/V-Band Linear Power Amplifiers using Envelope Tracking
and Digital Pre-distortion James Schellenberg, QuinStar Technology, Inc., Contributors: Bumjin Kim,
Jonmei Yan*, Donald Kimball* *University of California, San Diego, CA, USA 92093
30
Millimeter Wave Results

[1] L. A. Samoska, Towards terahertz MMIC
amplifiers: Present status
and trends, in IEEE MTT-S Int. Dig., San
Francisco, 2006, pp.
333336.
[2] W. R. Deal, Solid-state amplifiers for
terahertz electronics, in IEEE
MTT-S Int. Dig., Anaheim, CA, 2010, pp. 1122
1125.
[3] V. Radisic, K. M. Leong, X. Mei, S. Sarkozy,
W. Yoshida, P. Liu, J.
Uyeda, R. Lai, and W. R. Deal, A 50 mW 220
GHz power amplifier
module, in IEEE MTT-S Int. Dig., Anaheim, CA,
2010, pp. 4548.
[5] M. Micovic et al., GaN MMIC PAs for E-band
(71 GHz95 GHz)
radio, in Proc. IEEE CSICs08, Monterey, CA,
2008, pp. 14.
[6] M. Micovic et al., W-Band GaN MMIC with
842mWoutput power at
88 GHz, in IEEE MTT-S Int. Dig., Anaheim, CA,
2010, pp. 237239.
.
This Work
Dual-Gate GaN MMICs for MM-Wave Operation
Ruediger Quay, Senior Member, IEEE, A. Tessmann, R.
Kiefer, S. Maroldt, C. Haupt, U. Nowotny, R. Weber, H.
Massler, D. Schwantuschke, M. Seelmann-Eggebert, A.
Leuther, M. Mikulla, and O. Ambacher
IEEE MICROWAVE AND WIRELESS COMPONENTS
LETTERS, VOL. 21, NO. 2, FEBRUARY 2011, pp. 95-97
New Epitaxial Material for GaN
Less Strain using IN0.17Al0.83N on GaN compared to

existing AlGaN on GaN
Higher Reliability, Higher Power, Less short channel
effects meaning can use shorter gate lengths for
higher frequency operation
31
IN0.17Al0.83N on GaN
InAlN/GaN heterojunction structures offer several potential

advantages over AlGaN/GaN.
AlGaN/GaN. These include
1) a improved lattice matched structure with much reduced
lattice stress
2) higher reliability and robustness due to the improved lattice
match
3) higher output current and current density and thus higher
output power where the breakdown condition is preserved
4) potentially higher chemical and thermal stability due to the
higher temperature the structure can withstand
5) potentially improved control of surface instabilities
6) thinner barrier and shorter gate structures which will lead
to higher power performance at higher frequencies into the
millimeter range.
Addressed in IMS2013 Workshop WMA: Advancements in
InAlN/GaN Device and Microwave/MMW Circuit Technology
Design a 30 W (P1dB) PA using the Crey

CGH35030F GaN HEMT at 1.9 GHz
Operate at 28 Vds and design for Class a and Class b

32
Stability
For stable operation k>1

Amplifier unstable
Stability
For stable operation No Circles in Smith chart
33
Stabilization
Add Series Resistance
Unstable
Region
Unstable
Region
Add Shunt Resistance

Unstable
Region
Unstable
Region
Stabilization
Unstable
Region
Unstable
Region
Add Series or
Shunt Resistance
Unstable
Region
Unstable
Region
34
AWR Schematic Class A

Stability achieved by placing a series resistor and a
parallel resistor on the input
Unconditionally Stable (K>1)
35
Unconditionally Stable
Load Line
Vgs = -1.17 V, RL = 15 . Pin = 34.2 dBm
36
AWR Current Voltage Class A
Pout @P1dB = 31.6 W

PAE = 50.37%
37
Low AM/AM and AM/PM
Output Match
Output needs to see 15

15
50
50 28.70
2.13 pF
38
Input Match
0.13 +j0.02 to 1
50 28.6
28.60
50
80
6 pF
0.13 +j0.02
2 pF
50
Match Input and Output with Low Pass

Matching Circuits
39
Gate Current and Reliability
High gate current leads to metal migration which can open

circuit the gate. With GaAs devices, the rule of thumb is not to
exceed 1 ma per mm of gate periphery.
So under normal operation the gate current is low, but when
driven with higher power than normal, gate current is
experienced.
One way to reduce gate current with higher powers is to place
a resistor in the DC portion of the gate bias circuitry
Many times the manufacturer will suggest a gate resistor for
their device.
The metal migration process does not happen immediately,
and it takes time to develop.
Always monitor the gate current for your PA design.
As a precaution, do not overdrive the PA, an place a resistor
in the gate dc portion of the DC bias circuit.
Example
Rgate = 0 Ohms
40
Example
Rgate = 50 Ohms
Example
Rgate = 100 Ohms
41
Example
Rgate = 1K
Conclusion
For this transistor, the Pout at P1dB does not

change for gate resistors 1 to 1000 Ohms
The gate current starts at 4 dB above the P1dB
point. So do not operate the transistor here
Operating the transistor there does not make
sense, because this is past the peak PAE point
Adding 1K helps tremendously should one by
accident overdrive the PA 4 dB or more above the
P1dB point
42
At P1dB, Second Harmonic down 17.3 dB

and third down 23.2 dB
Second Harmonics follow 2dB/dB
Third Harmonic follows 3dB/dB
Input Matched
Note Zin changes with input power a cause of AM/PM
43
Pdc = 49.2 W for low RF Input Power
T = 25 + 171 = 196 0C for low RF input power
44
Investigate Working at Class B
Better PAE
Similar power
Lower temperature for low input powers
Better PAE at lower power compared to Class a
However Class B has a gain and phase variation
versus input power so design will sacrifice on Pout
with digitally modulated signals with varying input
amplitude envelop as experienced with modern
digitally modulated signals like WCDMA, LTE, and
WIMAX
Use Tuner for Optimization

Optimum values for
Good Pout@P1dB
PAE and Low
AM/AM and AM/PM
45
Translate the Impedances into a Circuit

Circuit
Representation
Model
HBTUNER
ID=TU1
Mag1=Mag1
Ang1=Ang1 Deg
Mag2=Mag2
Ang2=Ang2 Deg
Mag3=Mag3
Ang3=Ang3 Deg
Fo=1.9 GHz
PORT
Zo=50 Ohm
P=1
Z=50 Ohm
Second
Harmonic
Short
RES
ID=R2
R=50 Ohm
Matching at Second Harmonic
First we will design the second harmonic circuit

close to the transistor
It want to see a = 1@141.80
50 Line 19.10 Long

Short
46
Matching at Second Harmonic

It want to see a = 1@141.80
Add a 50 ohm line
19.10 long and
terminate it
with a short
and you have the
circuit as seen by
the transistor
This is 9.550 long
at the fundamental
Short
Matching at Second harmonic

Short at Second harmonic
Transistor
= 1@141.80
47
Matching at Fundamental
Now work at fundamental with the second harmonic

circuit installed
= 0.612@157.4
0.612@157.40
50 9.550 line
Transistor
50
= 0.612@157.40
23 900 line
Results Pout = 20 W @P1dB, PAE = 72%, 11.9 dB SS Gain
48
Match 5.5 Ohm Input

Low temp rise 350C, DC current rises with Pin
49

Unconditionally stable, Input Matched, however AM/AM & AM/PM variation

Unconditionally stable, Input Matched, however AM/AM & AM/PM variation
50

Unconditionally stable, Input Matched, Low AM/AM & AM/PM,
Low Harmonics
Summarization of AM/AM and AM/PM for

Class A and B
Class A
Class B
Pout = 20 W @P1dB
PAE = 51%,
14 dB SS Gain
Pout = 20 W @P1dB
@P1dB
PAE = 72%,
72%,
11.9 dB SS Gain
51
Summarization of AM/AM and AM/PM for

Class A and B
Class A: Phase flat then phase increases rapidly

with increasing power past around P1dB
Class B: phase not flat with increasing power
because dc average current constantly changing
with increasing power. Phase increases slighly
with increasing power then decreases slightly
around P1dB.
Modeling GaN HEMT, GaN Nuances
Trapping effects and associated currentcurrent-knee collapse:

Increases knee voltage at which electron velocity occurs
Bias dependencies:
Source resistance and drain to source resistance
important
Sub threshold valid modeling important for designers
for Class B, C, D, E, and F high efficiency operating
modes since not all models can fit behavior in this
region. Curtice FET (CFET) model is well behaved in this
region
Models should be checked for gate voltages at or below
threshold, if this is important
Electrothermal modeling important for high voltagevoltagecurrent products
Measurements must use pulsed IV characteristics
L. Dunleavy, C. Baylis, II, W. Curtice, and R. Connick, Modeling GaN: Powerful but
Challenging, IEEE Microwave Magazine, pp82- 96, October 2010.
52
Modeling GaN HEMT
As GaN technology has developed, first in research

laboratories and more recently in multiple
commercial device manufacturers, the demand for
improved nonlinear models has grown alongside
the device process improvements.
The need for improved models for GaN is twofold:
First, GaN devices have unique nuances in
behavior to be addressed
Second, there is a desire for improved accuracy
to take full advantage of the performance wins to
be gained by GaN HEMT performance in the
areas of high efficiency and highhigh-power
operation.
Modeling GaN HEMT

Pulsed IV measurements are required
Comparison of pulsed IV (solid lines without symbols) and static IV for a GaN
HEMT. Pulse conditions were 0.2 ms pulse width and 1-ms separation with
quiescent bias set at Vdsq 5 0, Vgsq 5 0. Vgs is varied from 25 to 21 V in 1 V steps.
53
Modeling GaN HEMT

Pulsed IV Measurements Important for accurate modeling
Pulsed S-parameter and static Sparameter comparison for a 10 W
GaN HEMT. Pulse conditions:
5 ms pulse width, 0.1% duty cycle
(IF BW 5 1/pulse width 5 ,200 kHz).
Cree CGH3503F_r6 Device

IV Curves with and Without Self Heating
Self Heating Turned Off in Model
Self Heating Turned ON
Electrothermal model feature essential for accurate design
54
Measurements for Modeling GaN HEMT

Main considerations for non-linear (NL) Empirical transistor
models
IMS2011 WMJ Workshop: Modeling Considerations for GaN HEMT and Higher Level IC
Devices, Dr. Larry Dunleavy, Dr. Jiang Liu, Modelithics, Inc.
Modeling GaN HEMT

Nonlinear transistor modeling process
55
Modeling GaN HEMT

Comparison of example FET models used for
GaAs, silicon, and GaN FET/HEMT devices.
Cree (Modified Fager - Statz) 18+
Yes
Yes
HEMT
Modeling GaN HEMT

Comparison of example FET models used for
GaAs, silicon, and GaN FET/HEMT devices.
Cree (Modified Fager - Statz) 18+
Yes
Yes
HEMT
Electrothermal Models
[4] Y. Tajima, Introduction of new large signal model (LS7) for MESFET family of devices, presented
at Workshop 38th European Microwave Conf.: WFR-15: Advances in Model-based HPA Design,
Amsterdam, The Netherlands, Oct. 2008.
[5] W. R. Curtice, Users Guide for the C_FET Model for Agilents Advanced Design Simulator.
Washington Crossing, PA: W. R. Curtice Consulting, June 2004.
[11] I. Angelov, K. Andersson, D. Schreurs, D. Xiao, N. Rorsman1, V. Desmaris, M. Sudow, and H.
Zirath, Large-signal modelling and comparison of AlGaN/GaN HEMTs and SiC MESFETs,
in Proc. Asia-Pacific Microwave Conf. 2006, Dec. 2006, pp. 279282.
56
GaN MMIC Design and Modeling Bill Pribble, Jim Milligan,

Jeff Barner, Jeremy Fisher, Thomas Smith Cree, Inc.
IMS2011
Workshop,
Introduction to GaN GaN
MMIC
Design
2014 Niehenke
Consulting Inc.
Power Amplifier
Design -113
Modeling GaN HEMT

Topology for the Angelov GaN HEMT model
The electrothermal model elements Rtherm and Ctherm enable

estimation of channel temperature rise due to power dissipation
And varies model parameters to account for temperature rise.
Angelov, K. Andersson, D. Schreurs, D. Xiao, N. Rorsman1, V. Desmaris, M. Sudow, and H.
Zirath, Large-signal modelling and comparison of AlGaN/GaN HEMTs and SiC MESFETs,
in Proc. Asia-Pacific Microwave Conf. 2006, Dec. 2006, pp. 279282.
57
Millimeter Wave MMIC Foundries
BAE 0.2mm Non field Plate HEMT, ft = 50 GHz, fmax = 220 GHz, 5W/mm, 46% PAE at 30 GHz
10W WIMAX PA Design

Ed Niehenke
Niehenke Consulting
eniehenke@ieee.org
+1 410410-796796-5866
Copyright 2013, Edward C. Niehenke, Niehenke Consulting, Inc. All rights
reserved. No part of this document may be copied or reproduced in any manner
without the written permission of Edward C. Niehenke.
116
58
10W WIMAX PA Design
Transistor:
Eudyna 10W GaN HEMT amplifier, EGN010MK
Vds = 50V
Ids = 100 ma (no RF power) deep class b
Frequency range: 3.4 to 3.8 GHz
Design objectives: (at P1.5 dB)
Power > 41 dBm
PAE > 70%
> 80%
G > 8.5 dB
Stable 0.1 to 10 GHz
Input return loss > 15 dB
PA Design Steps
Step 1:
Examine IV Curves and note knee voltage (one point on
load line) and no current point at Vgs = 95 V
Note input dc voltage for no gate current
Determine Vgs for 100 ma of current
AWR circuit GaN HEMT WIMAX PA Step1
Step 2:
Examine Load Pull data and determine optimum load for
max power at Pin = 30 dBm
Determine circuit using transmission line. Hint try a
length of line of length theta and impedance Zmatch
Examine Pout vs Pin and Pout vs freq at Pin = 30 dBm
Optimize circuit for operation over full frequency
59
PA Design Steps
Step 3:
Match input at Pin = 30 dBm over frequency. Need to use a 10
degree 50 length of line on input in order to solder input
transistor lead
Suggestion: try a shunt cap and series which will also serve
as an input blocking cap
Examine circuit stability (k, MU2, input stability plane
Step 4:
Now examine stability (k factor and MU2 as well as input
stability plane) and completely stabilize circuit with minimal
degradation of gain and rematch input circuit. This is the
hardest step
Suggestion: See schematic next page
Try to keep Rsh equal to or greater than 200 so that gain is
not severely reduced
Suggested Circuit Schematic for Stability
60
Results: Step 1
Results: Step 1
61
Results: Step 2
Results: Step 2
Optimize circuit for Pout > 43 dBm,

dBm, PAE > 80% with
Pin = 30 dBm and f = 3.5 GHz
62
Results: Step 2
Review parameters versus Pin
Results: Step 2
Optimize over frequency: 3.4 to 3.8 GHz
63
Results: Step 3 Input Match
Unmatched
Matched
Results: Step 3 Unstabilized Stability
64
Results: Step 3 Unstabilized Stability
Results: Step 4 Stabilized Stability
65
Results: Step 4 Stabilized Stability
Results: Step 4 : Input Match
66
Results: Step 4 : Power Sweep
Results: Step 4 : Frequency Sweep Pin = 33 dBm
67
Final Schematic
Final Schematic
68
Doherty Amplifier Design
Main Amp Class AB, Aux amp Class C

At -6dB for high power, aux amp turned off
Each amp changes load that each sees resulting in
better efficiency and linear power
69
Design Doherty PA Using 10W WIMAX PA
AWR Balanced PA for Comparison
AWR Balanced PA for Comparison
PORT_PS1
P=1
Z=50 Ohm
PStart=6 dBm
PStop=38 dBm
PStep=1 dB
1
SUBCKT
ID=S2
NET="Main PA"
QHYB
ID=U1
R=50 Ohm
COUPL=3 dB
LOSS=0 dB
0
2
-90
0
-90
RES
ID=R1
R=50 Ohm
QHYB
ID=U2
R=50 Ohm
COUPL=3 dB
LOSS=0 dB
SUBCKT
ID=S1
NET="Main PA"
4
1
RES
ID=R2
R=50 Ohm
3
-90
0
-90
PORT
P=2
Z=50 Ohm
70
length set to 430 for high output Reflection

Coefficient () Aux PA when turned off
TLIN
ID=TL5
Z0=50 Ohm
EL=43 Deg
F0=3.5 GHz
Performance the same for both circuits with equal

gate biases for each PA
71
Examine DC Current of Aux PA and adjust so that at 6 dB down the

Transistor DC current is zero so the aux PA should be turned off
6 dB
Examine DC Current of Aux PA and adjust so that at 6 dB down the

Transistor DC current is zero so the aux PA should be turned off
6 dB
72
Performance enhanced with similar PA output with significant enhancement

enhancement of PAE
Suspect there is a problem with nonlinear model operating in deep
deep class C
The exercise does show technique and better performance but not that of
theoretical performance (Same PAE at 6 dB backoff)
backoff)
Conclusions
GaN HEMT for power amplifiers

High breakdown voltages with Vds dc operation
at 28 to 50 V depending on manufacturer
How power in a small footprint requiring care to
get the heat out
High reliability operation MTBF 106 to 108
More easily matched due to high voltage
operation and small equivalent capacitances and
high bandwidth operation
Design techniques shown for class a and class b
operation
Design techniques shown for digitally modulated
signals as well as Dougherty PA operation
73

Gan Power Amplifier Design

Uploaded by

Copyright:

Available Formats

Gan Power Amplifier Design

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gan Power Amplifier Design

Uploaded by

Copyright:

Available Formats

GaN Power Amplifier Design

Presented to IEEE Long Island MTT Chapter

2014 Niehenke Consulting Inc.