Beruflich Dokumente
Kultur Dokumente
Application of Non-Linear Models in a range of challenging GaN HEMT Power Amplifier Designs
Ray Pengelly, Brad Millon, Don Farrell, Bill Pribble and Simon Wood Cree Inc., Research Triangle Park, NC 27709
Outline
Attributes of GaN HEMTs Cree GaN HEMT Models Design Examples Broadband CW Amplifiers Linear WiMAX Amplifier Future Model Improvements Conclusions
Model Schematic
R R3 R=0.01 Ohm SDD6P SDD6P1 I[1,0]=(vg)/5e8 I[2,0]=id1 I[3,0]=(_v3)*gdsc I[4,0]=(_v4)*gmc I[5,0]=-vd*idt I[6,0]=(vdf)/5e4 C[1]= Cport[1]= vg1 pncap3 X12 sc=sc scf=scf cg1=cgs pF cg2=cg2 cg3=cg3 vgg0=vgg0 V R R1 R=(ri/(sc*scf)) Ohm pncap3 X15 sc=sc scf=scf cg1=cgd pF cg2=0.6 cg3=0.1 vgg0=-21 V
Drain current
SRL SRL4 R=(rd/(sc*scf)) Ohm L=(ld/sc+ld1/(sc*scf)) nH
Port P1 Num=1
C C7 C=1.0 uF C C9 C=(cds*sc*scf) pF
R R5 R=1e6 Ohm
R R6 R=1e6 Ohm
Port P2 Num=2
vd1g
Thermal resistance
R R2 R=rth Ohm C C8 C=30.0 nF Port P8 Num=8
Based on 13-element MESFET model (H. Kondoh 1986 MTT-S) ADS version shown using non-linear equation-based elements Easily changed during design process Speed comparable to C-coded version AWR version uses C code with model wizard
5 0
p1
-15
-10
-5 Voltage (V)
The model includes new capacitance functions as well as modeling of the drain-source breakdown and self heating. The model has four ports, with the extra port providing a measure of the temperature rise. The voltage between the external thermal circuit port and the source node is numerically equal to the junction temperature rise in degrees C. This occurs because the current source in the thermal circuit is numerically equal to the instantaneous power dissipated in the FET and the resistance, R_TH is numerically equal to the thermal resistance. The RC product of the thermal circuit is the thermal time constant. The model addresses the sharp turn-on knee in GaN HEMTs leading to the accurate prediction of IMD sweet spots in Class A/B operation.
0.5
permute(Is_high.i)
-4 -3 -2 -1 0 1 2
0.10
gm
0.05 0.00
Vlow
Vhigh
Transconductance curve fit to Gm from small-signal model fits over bias range Output conductance dispersion model Peak current and knee voltage fit from load-pull - includes trap effects Pinch-off fit from DC IV-characteristics gives model of drain current IV function similar to Fager-Statz formulation good model of pinch-off needed to accurately predict intermodulation distortion
36.5 output power 36 35.5 35 34.5 34 0 50 100 chuck temp 150 200
Drain current is only temperature dependent model element Drain current scales to provide -0.1 dB/10oC reduction in power for current-limited load-line Self-heating included using a thermal resistance calculated from finite element analysis of die and package. Thermal performance due to package needs to be included where appropriate
Drain Voltage
Feedback capacitance is a strong function of drain voltage Inclusion of this effect necessary to fit small-signal data Non-linearity changes harmonic generation from the model effects efficiency and linearity predictions Output Capacitance CDS is linear no voltage dependence (weak anyway)
Input capacitance is a strong function of gate voltage CGS is also a function of drain voltage, but this non-linearity is not included at present The gate-voltage non-linearity also effects models harmonic generation
model
25
bg ag
20
15
freq, Hz
On-wafer S-parameters of 0.5 mm HEMT 25OC baseplate Major challenge of modeling for high power circuits scaling from reasonable test cell to large periphery output stages successfully implemented for scaling factors >100:1 Non-linear model fits small-signal parameters over a range of bias voltages All measurements performed using 1% duty cycle, 20s pulsed bias to control thermal effects
measured model
On-wafer load-pull of 0.5 mm HEMT Measured at 3.5 GHz, VDS=48V, Id~25%IDSS, 25OC chuck temperature PAE contours not used for modeling due to sensitivity to harmonic loading PAE verified using hybrid amplifier measurements
240 Watts CW RF from a 28.8mm HEMT operating at 50 volts drain voltage Assume 60% DC to RF conversion efficiency 160 watts dissipated heat Active chip area is 2.5 sq. mm so heat density is > 40 kilowatts per square inch ! Much emphasis on new amplifier architectures to improve drain efficiencies
Broadband Amplifier Performance Trade-Off Analysis - Background Broadband (0.8 to 2.2 GHz) push-pull amplifier to provide 100 watts peak power Two GaN HEMT die in Gemini package HEMTs attached to composite material shims within Cu-Mo-Cu package Study drain efficiencies over the band and impact on thermal management Comparison of different matching approaches and termination impedances
Basic Amplifier
Different matching topologies Drain-to-Gate Feedback Lossy Match Multi-section reactive Lossy Match with Feedback Concentrate on Lossy Match case Input Match Schematic
DCVS ID=V1 V=28 V
DC RF
2
DC & RF
DC DC & 1 RF RF
2
TLIN ID=TL4 CAP Z0=18.75 Ohm ID=C4 EL=86.31 Deg C=1.006 pF F0=1.5 GHz
TLIN ID=TL5 CAP Z0=35.25 Ohm ID=C5 C=0.6011 pF EL=93.57 Deg F0=1.5 GHz
Power=10 PORT1 P=1 Z=50 Ohm Pwr=Power dBm ATTEN ID=U1 R=50 Ohm LOSS=3 dB SUBCKT ID=S1 NET="IMN Lossy Match"
1 2 1 2 1 3 2
BIASTEE ID=X1
3
Efficiency
Gain = 13.2 dB 0.8 dB POUT = 46 to 59 watts Drain Efficiency = 49 to 66% Worst Case Dissipated Heat is 54 watts (per transistor)
POUT
Thermal Performance
Assuming TJ limit of 200OC, maximum dissipated heat of 108 watts and theta-jc of 1.1 deg C/Watt leads to a maximum case temperature of 81 deg C The thermal characteristics of the die and the package are very important
The design requires a composite material shim such as silver-diamond (theta jc =550 W/mK) mounted on a Super-CMC package flange (theta jc = 370 W/mK)
Pdiss (W/mm)
Pdiss
3.5
2.5
1.5
0.5
Before full electrical design is completed broadband amplifiers require thermal design even with GaN HEMTs!
0 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Frequency (GHz)
Pdiss
Multi-section reactive
3.5
P is (W m d s /m )
2.5
1.5
Feedback
0.5 0 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Lossy Match
Frequency (GHz)
Large-Signal Modeling of Broadband Amplifiers invaluable in selecting optimum topology for both electrical and thermal performance
25 to 50 ohm coupler
Drain Efficiency at Device Sat
70.00
60.00
50.00
40.00
30.00
20.00
100.00
10.00
80.00
0.00 0.000 0.500 1.000 1.500 Frequency (GHz) 2.000 2.500 3.000
Psat (W)
60.00
40.00
20.00
0.00 0.000
0.500
1.000
2.000
2.500
3.000
Full amplifier with coupler insertion losses Average Gain = 15 dB Psat > 80 watts and Drain Efficiencies > 45%
Lossy Match
IND ID=L1 L=200 nH IND ID=L2 L=200 nH
20
p2
10 0
p3 p1
60 dBm and %
dB
40
DB(|S(1,1)|)[X,2] 500 to 2700 MHz Stage DB(|S(2,1)|)[X,2] 500 to 2700 MHz Stage DB(|S(2,2)|)[X,2] 500 to 2700 MHz Stage
p1: Pwr = -3 dBm p2: Pwr = -3 dBm p3: Pwr = -3 dBm
20
2.5
2.7
Worst Case Heat Dissipation is 9 watts Theta-jc of packaged transistor is 5 deg C/watt Max. channel temperature at 85 deg C case is 130 deg C.
Output Power
26 24 22 20 12 14 16 18 20 22 24 26 28 30 32 34 PIN (dBm)
Saturated Output Power and Drain Efficiency Saturated Output Power and Efficiency
dB
Efficiency Power
p1 p4 p3 p6
0.25
0.5
0.75
1.25
1.5
1.75
2.25
2.5
2.75
3.25
Frequency (GHz)
CG H40045F_r1
1
CAP ID=C5 C=0.03 pF
GaNg28v2_r3 ID=G aNv2sc1 T NOM =25 SC=20 SCF 1.44 = RT 2.7 H= VDD=50
M SUB Er=9.6 H= m 20 il T 1.4 m = il Rho=1 T and=0 ErNom =9.6 Nam e=Alum _pkg1
MD L X
1 2
P O RT P =1 Z =50 Oh m
M ST EP $ I D=TL 8
P OR T P =2 Z=5 0 O h m
MD L X
SU BC KT ID =S 2 N ET ="Gate B ias F eed w ith T rans Line_02 _06_06" MLE F ID= Open W = 30 m il L=9 3 m il
M OD catc 06 03001 ID =AT C_6 00S _C 1 C =0.85 pF M SU B= Sim _m ode= 0 Tol era nc e=1 PAD W =35 m il
MD LX
2 1 2
PO R T P= 1 Z =5 0 O hm
PO R T P= 1 Z=5 0 O hm
MSTE P$ ID =TL14
MSTE P$ ID =TL10
M STEP $ I D=T L9
M STE P$ ID =TL8
P O RT P =2 Z= 50 O hm
MSU B Er= 3.66 H= 20 m il T= 1.4 mil Rh o=1 Ta nd= 0.004 ErN om =3.66 Na me= Roge rs 1
3
SU BC KT ID =S8 N ET= "Input_S hunt_GN D _03_21_ 071"
ML I N I = Db i as L 2 D W = W b a s mi l i L L b i as m l = i
1
M B END 90 X I D= M S 4 W = W o u t mi l M =0 . 5
M B E N D9 0 X ID = MS 1 W = W b a s mi l i M = 0 .5
ML I N I = Db i as L 1 D W = W b a s mi l i L= 2 4 0 m i l P O RT P=1 Z = 5 0 O hm
M L EF I D= O S t u b1 W = 1 1 0 mi l L =3 0 m il
ML E F I = OS t u b 2 D W = 11 0 m l i L= 3 0 mi l
L st u b _ A = 50 W o u t = 34 L o ut = 6 1 L st u b _ M= 7 0 W d l n e = 1 00 i W b i as = 4 0 L b a s = 25 5 i
M L IN I D = T L3 W = W o ut m i l L = L o ut m i l
M LI N I D= T L 7 W=Wout m l i L = L ou t m l i
M L EF I D =O S t u b 4 W =4 0 m l i L =1 7 4 m il M LI N I D= T L 6 W = 3 4 mi l L = 60 m l i
3
M L EF I D= O S t u b6 W = 7 2 mi l L =2 6 m il
M DL X
1 2 1 2 1 2
M B E ND 90 X I D= M S 2 W = W o u t mi l M = 0. 5
M CR OS S$ I D =T L 9
1 3
M STEP$ I D= T L 1 3
1
M LI N I D= T L 1 W = 4 4 mi l L = 1 44 m l i
ML I N I = T L1 5 D W = 70 m i l L= 2 0 0 m i l
M LI N I D= T L 1 0 W = 2 0 mi l L = 1 30 m l i
M STEP$ I D =T L 4
ML I N I = T L1 1 D W = 70 m i l L 2 0 0 m il =
M L IN I D =T L 1 6 W =1 6 0 mi l L =2 6 5 m il
MS T P $ E I D= T L 19
M LI N I D= T L 2 0 W =1 1 0 m il L =2 m il
ML I N I D= Db i a sL W = W b i as m l i L =2 4 0 mi l
M OD ca t c 10 0 B0 0 1 I D= A T C _1 0 0 B_ C1 C= 8 . 2 pF M S UB = Si m _m o d e= 0 T o l era n c e= 1 PA D W = 1 10 m l i
ML I N I D= T L 2 W = W o u t mi l L =5 0 m l i
M LI N I D= T L 8 W = W o u t mi l L = 45 m i l
M L IN ID = T 1 7 L W = 48 m l i L = 1 2 mi l
M CR OS S $ 4 I D =T L 1 8
M STEP$ I D =T L 1 2
M STEP$ I D= T L 1 4
P ORT P =2 Z = 5 0 Oh m
M L EF I D =O S t u b 3 W =4 0 m l i L =1 7 4 m il
M L EF I D= O S t u b5 W = 7 2 mi l L=2 6 m il
M B E N D9 0 X ID = MS 6 W = W b a s mi l i M = 0 .5
1
M SUB E r= 3 . 6 6 H = 0 ml 2 i T =1 .4 m i l R h =1 o T a n d =0 . 0 0 4 E rN o m= 3 . 4 8 N a e = Ro g ers 4 3 50 m
M LI N ID= Db i a sL 3 W = W b i a s m il L = Lb i a s m i l
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 Frequency (GHz)
3.1 3.2
40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% 42.0
24.0
26.0
28.0
30.0
32.0
34.0
36.0
38.0
40.0
EVM (%)
EVM @ 26dBm
3.50 3.00 2.50
2.400
2.500
2.600
Frequency (GHz)
EVM (%)
Behavioral models to allow direct simulation of various digital waveforms Improved active and passive switch models Improved models for switch mode PAs Additional noise models Support for simulators other than ADS and MWO
Conclusions
Successful development of GaN HEMT large-signal models Models are scale-able over > 100:1 gate width ratio Models Are broadband Accurately predict DC, s-parameters, dynamic load lines, non-linearities Include self-heating Can be used in a range of amplifier types Demonstrated a variety of hybrid circuit applications Equally useful for MMIC amplifier designs Future extensions to include other features such as noise