This document is owned by Agilent Technologies, but is no longer kept current and may contain obsolete or
inaccurate references. We regret any inconvenience this may cause. For the latest information on Agilent’s
line of EEsof electronic design automation (EDA) products and services, please go to:
www.agilent.com/fi nd/eesof
Agilent EEsof EDA
February 26, 2008 1
Nickolas Kingsley, PhDYusuke Tajima, PhD
Auriga Measurement Systems650 Suffolk Street, Suite 410
Lowell MA, 01854 USA
Large signal model extraction of GaAs MESFETs and GaN
HEMT Devices
2
Introduction
1. Modeling steps will be described for small signal, noise, and large signal models
2. Model extraction examples will be presented for GaAs pHEMTs and GaN HEMT devices
3
Topics
1. Model extraction programs2. Small signal and noise model
extraction3. Large signal model extraction steps
4
Model extraction programs
1. ICCAP(Agilent)Versatile extraction platform.Customer can add extraction process for different models.
2. Model Station (Auriga)Limited to FET family of devices.Good for small signal, noise and large signal models.Generates models compatible for ADS and MWO.
5
Model Station Dialog Box
6
Data Inputs to Model Station
Small signal data inputs
Large signal data inputs
7
Topics
1. Model extraction programs2. Small signal and noise model
extraction3. Large signal model extraction steps
CAPID=CdsC=0.3498 pF
CAPID=CgdC=0.1065 pF
CAPID=CgsC=1.906 pF
INDID=LdL=0.118 nH
INDID=LgL=0.1378 nH
INDID=LsL=0.001114 nH
RESID=RiR=0.5284 Ohm
1
2
3
4
VCCSID=U1M=0.1935 SA=0 DegR1=1817 OhmR2=88.48 OhmF=0 GHzT=2.5 ps
INDID=LgdL=0 nH
RESID=RgdR=0 Ohm
T1 2
RESTID=Rd1R=1.029 OhmT=16.85 DegC
T1 2
RESTID=Rg1R=2.536 OhmT=16.85 DegC
T
1
2
RESTID=Rs1R=0.4443 OhmT=16.85 DegC
CAPID=CggC=0 pF
RESID=RggR=1e5 Ohm
PORTP=1Z=50 Ohm
PORTP=2Z=50 OhmID=EN1 ID=EN2
ID=EN3
ID=EN4
ID=EN5
ID=EN6 ID=EN7ID=EN8 ID=EN9
ID=EN10
ID=EN11
ID=EN12
ID=EN13Cval=.5
8
Small signal model
9
150x80 1.
01
.0-
1.0
10.
0
10.0
-10.0
5.0
5.0
-5.0
2.0
2.0
-2.0
3.0
3.0
-3.0
4.0
4.0
-4.0
0.2
0.2
-0.2
0.4
0.4
-0.4
0.6
0.6
-0.
6
0.8
0.8
-0.
8
S11 and S22Swp Max26.4GHz
Swp Min0.2GHz
S(1,1)FET Schematic
S(1,1)SParm Data
S(2,2)FET Schematic
S(2,2)SParm Data
0
15
30
45
60
75
90
105
120
135
150
165
-180
-165
-150
-135
-120
-105 -9
0
-75
-60
-45
-30
-15
S12Swp Max
26.4 GHz
Swp Min0.2 GHz
Mag Max0.05
0.01Per Div
S(1,2)FET Schematic
S(1,2)SParm Data
0
15
30
45
60
75
90
105
120
135
150
165
-180
-165
-150
-135
-120
-105 -9
0
-75
-60
-45
-30
-15
S21Swp Max
26.4 GHz
Swp Min0.2 GHz
Mag Max20
5Per Div
S(2,1)FET Schematic
S(2,1)SParm Data
10
Extraction of Models at multiple bias points
11
Small signal model with noise sources
CAPID=CdsC=0.03942 9 pF
CAPID=CgdC=0.02953 pF
CAPID=CgsC=0.125579 pF
INDID=LdL=0.007666 nH
INDID=LgL=0.000293 nH
INDID=LsL=0.001929 nH
RESID=RiR=0.5 Ohm
T1 2
RESTID=RgR=2.07231 OhmT=16.85 DegC
T
1
2
RESTID=RsR=1.98319 OhmT=16.85 DegC
1
2
3
4
VCCSID=U1M=0.09605 SA=0 DegR1=979421 OhmR2=190.957 OhmF=1e18 GHzT=0.391458 ps
1
2
3
4
VINCORID=VIN1V1SQ=0.020672I2SQ=1043.23CorrR=0.123144CorrI=0.303625
T1 2
RESTID=RdR=2.60038 OhmT=16.85 DegC
INDID=LgdL=0 nH
RESID=RgdR=0 Ohm
CAPID=Cgg1C=0 pF
RESID=Rgg1R=100000 Ohm
PORTP=1Z=50 Ohm
PORTP=2Z=50 Ohm
Noise is generated from parasitic resistive elements as well as intrinsic noise sources as shown here. Noise sources are defined by voltage and current noise sources and correlation (complex) between them.
12
Noise parameters simulated from extracted model are compared with the measured results (Vd=0.9V, 8mA)
Freq 2-26GHz
13
Topics
1. Model extraction programs2. Small signal and noise model
extraction3. Large signal model extraction steps
14
Large signal model extraction steps
1. Extraction of Auriga model- Charge conservation satisfied
2. Our modeling steps- Define de-embedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature test- Loadpull verification
15
Many models draw non-physical gate current because they were violating Charge Conservation Law
Cgs(Vgs, Vds)
Vgs Vds
0
0.1
0.2
0.3
0.4
0.5
0.6
-1.5 -1 -0.5 0 0.5 1
Cgs(pF) vs Vgs with Vds as a parameter
Vd=0V
Vd=0.5V
Vd=1VVd=2V
Vd=3VVd=4V
Vd=5V
Vd=6V
Typical Cgs dependency on gate and drain voltages
16
Charge conservationAs is well known, if the charge conservation is not
satisfied at terminals, the device model will draw non-physical current in order to maintain charge neutrality under RF drive.
VgsVgdVgsCgd
VgdVdsVgsCgs
∂∂
=∂
∂ ),(),(
Capacitors are constrained by the equation above in new Auriga model (LS6)
17
0 10 20 30 32Power (dBm)
Gate Current 50ohms A
0
0.01
0.02
0.03
0.04A
0
0.01
0.02
0.03
0.04
A
p2
|Icomp(I_METER.AMP2,0)|[*,X] (L)50ohm power simulation LS6
|Icomp(I_METER.AMP1,0)|[*,X] (R, A)50ohm power simulation LS6
|Icomp(I_METER.AMP1,0)|[1,X] (R, A)50ohm power simulation LS5
|Icomp(I_METER.AMP2,0)|[*,X] (L)50ohm power simulation LS5
p1: Freq = 2.6 GHz
p2: Freq = 2.6 GHz
Traditional(LS5)
Data
Gate current under RF driveMeasured data vs traditional model (LS5)
Without charge conservation, a large non-physical current flows through gate terminal. This is not real.
18
Charge Conserved Model (LS6)vs Traditional model (LS5)
(Example: Gate current behavior under RF drive.)0 10 20 30 32
Power (dBm)
Gate Current 50ohms A
0
0.01
0.02
0.03
0.04A
0
0.01
0.02
0.03
0.04
A
p2
|Icomp(I_METER.AMP2,0)|[*,X] (L)50ohm power simulation LS6
|Icomp(I_METER.AMP1,0)|[*,X] (R, A)50ohm power simulation LS6
|Icomp(I_METER.AMP1,0)|[1,X] (R, A)50ohm power simulation LS5
|Icomp(I_METER.AMP2,0)|[*,X] (L)50ohm power simulation LS5
p1: Freq = 2.6 GHz
p2: Freq = 2.6 GHz
Traditional(LS5)
Charge conserved (LS6)
Data
With charge conserved model, gate current under rf drive is well simulated.
19
Large signal model extraction steps1. Extraction of Auriga model
- Charge conservation satisfied
2. Our modeling steps- Define de-embedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature test- Small signal and Loadpull verification
20
De-embedding
235um 235um
A A’B B’
Reference planesMeasurement A-A’Modeling B-B’
1
2
3
Auriga_LS6_FETID=MF1DESC=8x100AFAC=1AL=1Temp=23.85 DegC
MLINID=TL1W=0.076 mmL=0.3 mm
MLINID=TL2W=0.076 mmL=0.3 mm
PORTP=1Z=50 Ohm
PORTP=2Z=50 Ohm
AA’B
B’
Input de-embedding box
Output de-embedding box
21
Our modeling steps
- Define deembedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature test- Small signal and Loadpull verification
22
DC IV measurement
1. This is an important step to screen chips and select the chips for modeling.
2. DC is used to measure gate diode characteristic for both forward and reverse.
23
Measurement of Diodes between G-S and G-D
VdIg
VgG
D
S
VdIg
VgG
D
S
D1
D2
0.0V
0.5V
1.0, 1.5, 2.0V
VdCurrent limit
D2D1+D2
24
Our modeling steps
- Define deembedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature test- Small signal and Loadpull verification
25
Pulsed IV and S parameter measurements
Characterization using pulse technique becomes only tool to measure large devices within thermal limits.
- Pulsed IV- Pulsed S parameters
26
Principle of Pulsed IV
All the IV data are measured in short pulses while the bias point is kept at the quiescent point .
0
1000
2000
3000
4000
5000
6000
0 10 20 30 40 50 60 70 80 90 100
Id (m
A)
pe
ak 5
.3 a
mps
Vd (volts)
QuiescentVd = 50 vId = 150 mA
Vg 2.0
Vg 1.0
Vg 0
Vg -0.5
Vg -1.0
27
Pulsed IV system AU4550
AU4550
Pulser
Pulser
DUT
28
Why Pulsed IV?1. Thermal 2 Field induced traps
Quiescent condition 0Vd, 0Vg Quiescent condition 6Vd,-4Vg
Pulsed IV data of a pHEMT at different quiescent conditions
29
Pulsed IV and IV Model
(Blue) Pulsed IV Data Quiescent point (50V, 200mA)(Red) Modeled Pulsed IV
0 20 40 60 80 100Vds (Volts)
IV Curves
0
1
2
3
4
5
6
Id(A
)
30
Different results for Cgd and Rds between CW(10Vds) and Pulsed bias (10Vds peak, 50V quiescent)
Strong quiescent bias dependency on Cdg and Rds!
CW and Pulsed Cgd
0
1
2
3
4
5
-2 -1.5 -1 -0.5 0
Vgs (V)
Cgd
(pF)
Pulsed
CW
CW and Pulsed Rds
1.00E+00
1.00E+01
1.00E+02
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
-2 -1.5 -1 -0.5 0Vgs (V)
Rds
(Ohm
)
Pulsed
CW
31
Similar Results for Cgs and Gm Between CW(10V) and Pulsed Bias (10Vds Peak, 50V Quiescent)
CW and Pulsed gm 10Vds
0
0.5
1
1.5
2
2.5
-2 -1.5 -1 -0.5 0
Vgs(V)G
m (S
)
Pulsed
CW
CW and Pulsed Cgs
0
10
20
30
40
50
-2 -1.5 -1 -0.5 0
Vgs (V)
Cgs
(pF)
CW
Pulsed
Cgs, Gm showed little dependency on quiescent bias condition
32
Pulsed S Parameter Set up
Bias Tee
Bias Tee
PNA
Bias Tee
Bias Tee
VNA
DUT
VNA was used to take S parameters during the pulsed bias applied through bias tees.
33
S parameter data
1. S parameters are taken over 50-150 bias points to derive small signal equivalent circuit model as a function of terminal voltages.
2. Voltage dependent Cgs, Cds and Cgd are modeled and parameterized into the Auriga large signal model.
34
Our modeling steps
- Define deembedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature effect- Small signal and Loadpull verification
35
How short the pulses need be to be isothermal?
50 nSec sliding data window
Vds
Ids
Vq
Iq
t =0 t=tmaxT1 T2
dwell time2Vg
1Vg
0Vg
-1Vg
Id mA
IV Curves were measured between 0.2 and 50µsec on a GaN HEMT. No noticeable difference was observed. It can be concluded that channel temperature stays the same in this time frame for this device.
36
Temperature dependent IV curve model
-0.0344 50 100Vds (Volts)
IV Curves
-0.5
0
0.5
1
1.5
Id(A
)
25C
-0.0362 50 100Vds (Volts)
IV Curves
-0.5
0
0.5
1
1.5
Id(A
)
Model was derived from 25C pulsed IV data. Then scaled to other temperatures (0, 80, 150C) and compared to the pulsed IV data.
Blue: Pulsed IV DataRed: ModelVd=0-100VVdq=50V
0C
37
Temperature dependent IV curve model (2)
-0.0254 50 100Vds (Volts)
IV Curves
-0.5
0
0.5
1
1.5
Id(A
)
-0.0348 50 100Vds (Volts)
IV Curves
-0.2
0
0.2
0.4
0.6
0.8
1
Id(A
)
80C
Blue: Pulsed IV DataRed: ModelVd=0-100V
150C
38
Temperature model
))(*1(*)())(*1(*)())(*1(*)(
TnomTempTcoefRRdTRdTnomTempTcoefRRsTRsTnomTempTcoefIdsTIds
−+=−+=−+=
Temperature and Size scaling factors are available on the schematic.
39
Model Scaling
Scalable parameters-Intrinsic device parameters including currents, capacitors and resistive components (Ids, Cgs, Cgd, Rg, Rs, Rd)
Non Scalable parameters- Inductive components (Ld, Lg, Ls) - External circuit
Scalable models save time and
money
40
Our modeling steps
- Define deembedding parameters.- DC IV- Pulsed IV- (Pulsed) S parameters- Temperature effect- Small signal and Loadpull verification
41
Large signal model (Auriga LS6) Small signal verification (0.4-12GHz)
0 1.0
1.0
-1.
0
10.
0
10.0
-10.0
5.0
5.0
-5.0
2.0
2.0
-2.0
3.0
3.0
-3.0
4.0
4.0
-4.0
0.2
0.2
-0.2
0.4
0.4
-0.4
0.6
0.6
-0.
6
0.8
0.8
-0.
8
S11 and S22Swp Max
11.99GHz
Swp Min0.38GHz
S(1,1)FET Schematic
S(1,1)SParm Data
S(2,2)FET Schematic
S(2,2)SParm Data
0
15
30
45
60
75
90
105
120
135
150
165
-180
-165
-150
-135
-120
-105 -9
0
-75
-60
-45
-30
-15
S21Swp Max
11.99 GHz
Swp Min0.38 GHz
Mag Max20
5Per Div
S(2,1)FET Schematic
S(2,1)SParm Data
0
15
30
45
60
75
90
105
120
135
150
165
-180
-165
-150
-135
-120
-105 -9
0
-75
-60
-45
-30
-15
S12Swp Max
11.99 GHz
Swp Min0.38 GHz
Mag Max0.06
0.02Per Div
S(1,2)FET Schematic
S(1,2)SParm Data
S11, S22 S21 S12
40Vd –1.0Vg
S parameter simulation from the large signal model was verified over a wide bias range.
42
Load Pull Testing and Simulation Results
DUTFixtured device
Focus iTuner708
ElectronicTuner
Load pull system
PA
Source
ReceiverBias Tee
Bias Tee
Power Supply
Power Supply
43
Loadpull simulation circuit
Prematch Fixture
Prematch Fixture3:Bias
12
HBTUNER2ID=TU1Mag1=0.58Ang1=-168 DegMag2=0.56Ang2=69.4 DegMag3=0Ang3=180 DegFo=2 GHzZo=50 Ohm
0 mA
SU
0 mA
3:Bias
1 2
HBTUNER2ID=TU2Mag1=0.56Ang1=84.9 DegMag2=0.4Ang2=-63.1 DegMag3=0Ang3=180 DegFo=2 GHzZo=50 Ohm
149 mA149 mAPOP=Z=
Device Model
Pin Pout
44
Device Simulation vs Single Tone Load Pull Data @ 2GHz
0 10 20 30 34Power (dBm)
Load pull adjusted source and load termination
10
20
30
40
50
0
20
40
60
80
p2
p1AMtoAM(PORT_2)[1,X,*] (L, dBm)Adjusted Load PullPAE(PORT_1,PORT_2)[1,X,*] (R)Adjusted Load Pull|Icomp(DCVS.V3,0)|[1,X,*] (R)Adjusted Load Pull|Icomp(DCVS.V4,0)|[1,X,*] (L)Adjusted Load Pull
PlotCol(1,2) (L)30LF run 1 w AMPMPlotCol(1,4) (R)30LF run 1 w AMPMPlotCol(1,6) (R)30LF run 1 w AMPM
p1: Freq = 2 GHzVg = -1.08
p2: Freq = 2 GHzVg = -1.08
%dBm Max Power47dBm
Max Eff67%
45
Device Simulation vs Two Tone Load Pull Data @ 2GHz
IM3
IM5
SimulationdBm
46
Summary
Modeling extraction steps for small signal and large signal model are described in this presentation using GaAs pHEMTs and GaNHEMT device data.Difficulties of modeling large devices, such as GaN HEMTs with 100W output power, were discussed.These models were used successfully to simulate 2.1GHz 100W GaN HEMT amplifier.
Thank you
www.agilent.com/fi nd/emailupdatesGet the latest information on the products and applications you select.
www.agilent.com/fi nd/agilentdirectQuickly choose and use your test equipment solutions with confi dence.
Agilent Email Updates
Agilent Direct
www.agilent.comFor more information on Agilent Technologies’ products, applications or services, please contact your local Agilent office. The complete list is available at:www.agilent.com/fi nd/contactus
AmericasCanada (877) 894-4414 Latin America 305 269 7500United States (800) 829-4444
Asia Pacifi cAustralia 1 800 629 485China 800 810 0189Hong Kong 800 938 693India 1 800 112 929Japan 0120 (421) 345Korea 080 769 0800Malaysia 1 800 888 848Singapore 1 800 375 8100Taiwan 0800 047 866Thailand 1 800 226 008
Europe & Middle EastAustria 0820 87 44 11Belgium 32 (0) 2 404 93 40 Denmark 45 70 13 15 15Finland 358 (0) 10 855 2100France 0825 010 700* *0.125 €/minuteGermany 01805 24 6333** **0.14 €/minuteIreland 1890 924 204Israel 972-3-9288-504/544Italy 39 02 92 60 8484Netherlands 31 (0) 20 547 2111Spain 34 (91) 631 3300Sweden 0200-88 22 55Switzerland 0800 80 53 53United Kingdom 44 (0) 118 9276201Other European Countries: www.agilent.com/fi nd/contactusRevised: March 27, 2008
Product specifi cations and descriptions in this document subject to change without notice.
© Agilent Technologies, Inc. 2008
For more information about Agilent EEsof EDA, visit:
www.agilent.com/fi nd/eesof