time and frequency domain transistor modeling based on ... · issues with compact device models •...
TRANSCRIPT
Time and frequency domain transistor modeling based on Nonlinear Vector Network Analyzer databased on Nonlinear Vector Network Analyzer data
D. E. Root, J. Horn, J. Xu, M. Iwamoto, F. Sischka, and Y. Yanagimoto
Agilent Technologies IncAgilent Technologies Inc.
MOS-AK/GSA WorkshopSan Francisco, CADecember 8, 2010
© Copyright Agilent Technologies 2010
Page 1Page 1 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Three Modeling Flows based on Nonlinear Data
IC-CAP
Measurement, Modeling, & Design Capabilities
parameterextraction fromNVNA data Adv. Dynamic Model
ADSNVNA
DI
Adv. Dynamic Model
IC design21j dsgs φ φ TV V
Measurement-based compact model
X-parameters
© Copyright Agilent Technologies 2010
Page 2Page 2 D. E. RootD. E. RootDecember 8, 2010
NVNA-based devicecharacterization Measurement-based
behavioral modelMOS-AK/GSA
Outline
• Introduction: NVNA Measurements – complex spectra & waveforms
• Compact ModelsCompact model validation and parameter extractionGaAs transistor model identification and validationGaAs transistor model identification and validation
• Load-dependent X-parameter model of packaged GaN transistor & validation with harmonic load-pullp
• Discussion and Comparison
• ConclusionsConclusions
© Copyright Agilent Technologies 2010
Page 3Page 3 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Introduction: NVNA measurements complex spectra and waveforms [1]complex spectra and waveforms [1]
2 kA1kA
B 2kB
pkBpkA1kB 2k
Port IndexHarmonic Index
I I Includes:S1I 2I S-paramsWaveformsLoad-linesX-params
© Copyright Agilent Technologies 2010
Page 4Page 4 D. E. RootD. E. RootDecember 8, 2010
time timep
MOS-AK/GSA
Outline
• Introduction: NVNA Measurements – complex spectra & waveforms
• Compact ModelsCompact model validation and parameter extractionGaAs transistor model identification and validationGaAs transistor model identification and validation
• Load-dependent X-parameter model of packaged GaN transistor & validation with harmonic load-pullp
• Discussion and Comparison
• ConclusionsConclusions
© Copyright Agilent Technologies 2010
Page 5Page 5 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Compact Transistor Models: time domain
Thermal Subcircuit (Two-Poles)( )
Standard GaAs FET Model
Model: Coupled nonlinear ordinary differential equations
© Copyright Agilent Technologies 2010
Page 6Page 6 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Conventional Empirical Device Modeling Flow
IC CAPIC-CAP
VNA0.09
ADS0.03
0.04
0.05
0.06
0.07
0.08
Cgd
(pF)
CGD vs bias
P t
80
100
120
140
160
180-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
0.01
0.02
Vds
Parameter Analyzer
0 1 2 3 4 5 6 7 8 90
20
40
60
80
Your Compact
Model HereDoes it actually work in NL?
Circuit design
Parameter extraction for std. compact model
NVNA-based devicecharacterization
DC + S-parameter(or C-V data)
© Copyright Agilent Technologies 2010
Page 7Page 7 D. E. RootD. E. RootDecember 8, 2010
p( )
MOS-AK/GSA
Issues with compact device models
• Parameters typically extracted from DC and linear S-pars Ironic for a nonlinear model– Some devices may not be able to be characterized under DC and static
operating conditions (power, temperature)Ad d t b d d l t b id tifi bl f– Advanced measurement-based models may not be identifiable from only DC and S-parameter data.
– No direct evidence that these nonlinear models will reproduce large-i l b h i d t l diti f !signal behavior under actual conditions of use!
• Large-signal data from Nonlinear Vector Network AnalyzerLarge signal data from Nonlinear Vector Network Analyzer (with additional advanced modeling techniques) can solve all these problems
© Copyright Agilent Technologies 2010
Page 8Page 8 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Nonlinear Modeling Flow with NVNA data (1)
IC CAPIC-CAPNVNA
ADS
S BiSync. BiasSupply
120
140
160
180
Your Compact
Model Here0 1 2 3 4 5 6 7 8 9
0
20
40
60
80
100
Circuit designParameter extraction for
std. compact model NVNA-based devicecharacterization
Dynamic Load-lineWaveform meas.
© Copyright Agilent Technologies 2010
Page 9Page 9 D. E. RootD. E. RootDecember 8, 2010
pusing NVNA data as targetDC-IV, S-pars
MOS-AK/GSA
Model parameter extraction from Nonlinear Data (1)
NVNA data vs HB simulationusing initial parameter values extracted from DC + CV
Modify parameter values (optimize) to better fit large-signal NVNA data
• Get optimal parameter• Get optimal parameterset for given model
• trade-off DC, SP, fornonlinear performance
• App-dependent tuning
© Copyright Agilent Technologies 2010
Page 10Page 10 D. E. RootD. E. RootDecember 8, 2010
Ref. [10]
• App-dependent tuning• Explore model limits
MOS-AK/GSA
ICCAP comparison of simulated & NVNA measured results
Extract or Tune parameters with this capabilityBased on NVNA data
Large-signal Validation for free!Evaluate model under realistic operationSee where it needs improvement
© Copyright Agilent Technologies 2010
Page 11Page 11 D. E. RootD. E. RootDecember 8, 2010
Based on NVNA dataDetails are model-dependent
p
MOS-AK/GSA
Outline
• Introduction: NVNA Measurements – complex spectra & waveforms
• Compact Models Compact model validation and parameter extractionGaAs transistor model construction and validationGaAs transistor model construction and validation
• Load-dependent X-parameter model of packaged GaN transistor & validation with harmonic load-pullp
• Discussion and Comparison
• ConclusionsConclusions
© Copyright Agilent Technologies 2010
Page 12Page 12 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Nonlinear Modeling Flow with NVNA data (2)[4] J. Xu, M. Iwamoto, J. Horn, D. E. Root, IEEE MTT-S International Microwave Symposium Digest, May, 2010.
NVNA Advanced Non-parametricDynamic Model
DI
ANN
)()( tVtI
jT
IG QG IDQD
ADS2_emitC
dsV 2
2_emitR
2_captR1_emitC
gsV 1
1_emitR
1_captR
)()( tVtIthC
thR
0T21j dsgs φ φ TV V
80
100
120
140
160
180
0 1 2 3 4 5 6 7 8 90
20
40
60
80
Circuit designNVNA-based device
characterization
Dynamic Load-lineWaveform meas.DC + Spars
Direct construction of complicated constitutive relations
© Copyright Agilent Technologies 2010
Page 13Page 13 D. E. RootD. E. RootDecember 8, 2010
DC + SparsNot really parameter extraction!
MOS-AK/GSA
Motivations
Modeling modern III-V FETs requires the identification and self-consistent coupling of
(1) slowly varying dynamical variables
e.g., temperature, slow emission of traps
(2) rapidly varying dynamical variables(2) rapidly varying dynamical variables
e.g., the instantaneous intrinsic terminal
lt d f t tvoltages and fast capture processes.
© Copyright Agilent Technologies 2010
Page 14Page 14 D. E. RootD. E. RootDecember 8, 2010 MOS-AK/GSA
Model Formulation [4] J. Xu, M. Iwamoto, J. Horn, D. E. Root, IEEE MTT-S International Microwave Symposium Digest, May, 2010.
IG QG IDQD
© Copyright Agilent Technologies 2010
Page 15Page 15 D. E. RootD. E. RootDecember 8, 2010
Time (ns) Time (ns)
MOS-AK/GSA
NVNA data for identification of constitutive relations
))()()()()(())()()()()(()( t ,t ,tT ,tV ,tVQdtd t ,t ,tT ,tV ,tVItI 21jdsgsD21jdsgsDdrain
Auxiliary Terminal Auxiliary Terminal u a yVariables
e aVoltages
u a yVariables
e aVoltages
NVNA Constitutive Relation Compiled Waveforms Identification / Fit
F(Vgs,Vds, Aux. variables)Terminal Voltages(Vgs,Vds)
pModel in ADS
Auxiliary Variable Generation using Auxiliary variables fixed at their
steady-state large-signal valuesth0j RtV tITT )()(
))(( tVMin gs1
))(( tVMax ds2
AuxiliaryVariables
jT 1 2
steady-state large-signal values established by nonlinear measurements
© Copyright Agilent Technologies 2010
Page 16Page 16 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Artificial Neural Networks (ANN)An ANN is a parallel processor made up of simple, interconnectedprocessing units, called neurons, with weighted connections.
sigmoidweights biases
x1
...
baxwsvxxFI
i
K
kikkiiK
1 11 ),...,( xk
• Universal Approximation Theorem: Fit “any” nonlinear function of any # of variables• Infinitely differentiable: good for distortion.• Easy to train (fit) with appropriate SW
© Copyright Agilent Technologies 2010
Page 17Page 17 D. E. RootD. E. RootDecember 8, 2010
• Works well on non-gridded data and data with irregular boundaries (intrinsic FET)
MOS-AK/GSA
NVNA Measurements with Trap State ]1003[][ ,., 21 p ][][ ,, 21
Dynamic Load-lines
200
250
100
150
Ids
50
0
50
-2 0 2 4 6 8 10 12-50
Vds
Red line is one example of the 191 waveforms for this trap state
© Copyright Agilent Technologies 2010
Page 18Page 18 D. E. RootD. E. RootDecember 8, 2010
191 waveforms for this trap state
MOS-AK/GSA
Artificial Neural Network Training Procedure
(t)DI
kEAdjust wID and wQD
waveformthk for (t),measDI
jW
IFFT
j
FFT
QD_ANNID_ANN
th0j RtV tITT )()(
))(())(( tVMax,tVMin ds2gs1 wQDwID
))(())(( , ds2gs1
waveformthk for , (t) (t) 21jdsgs TVV
))()()()()(())()()()()(()( tttTtVtVQdtttTtVtVItI
© Copyright Agilent Technologies 2010
Page 19Page 19 D. E. RootD. E. RootDecember 8, 2010
))()()()()(())()()()()(()( t,t,tT,tV,tVQdt
t,t,tT,tV ,tVItI 21jdsgsD21jdsgsDdrain
MOS-AK/GSA
Intrinsic Model CharacteristicsID @ two different trap state valuesD @
Ids (mA) Ids (mA)
))()()()()(())()()()()(()( t ,t ,tT ,tV ,tVQdtd t ,t ,tT ,tV ,tVItI 21jdsgsD21jdsgsDdrain
120
140
160
180
120
140
160
180
“knee walkout”
60
80
100
120
60
80
100
120knee walkout
0
20
40
60
0
20
40
60
0 1 2 3 4 5 60
0 1 2 3 4 5 6Vds (V) Vds (V)
fixed Tj and fixed trap state = [-2.0, 8.0] Isothermal DC )080255( TVVI )55( VVTVVI
© Copyright Agilent Technologies 2010
Page 20Page 20 D. E. RootD. E. RootDecember 8, 2010
)080255( .. 21jdsgsD , ,T ,V ,VI )55( ds2gs1jdsgsD V,V,T,V ,VI
MOS-AK/GSA
Model Validation- DC
Model (___ ) Measured data ( o )( )
Vgs=1.2V
Ids (mA)Vgs=0.0V
Vgs=-3.0V
© Copyright Agilent Technologies 2010
Page 21Page 21 D. E. RootD. E. RootDecember 8, 2010
Vds (V)
MOS-AK/GSA
Model Validation- Frequency dispersion of small-signal characteristics
Frequency : 0.5 GHz to 50 GHz
-12 12 -12 1212 12 12 12
-12 12 -12 12
-0.6 0.6 -0.2 0.2
))()()()()(())()()()()(()( tttTtVtVQdtttTtVtVItI
Excellent agreement over the entire bias range
© Copyright Agilent Technologies 2010
Page 22Page 22 D. E. RootD. E. RootDecember 8, 2010
))()()()()(())()()()()(()( t,t,tT,tV,tVQdt
t,t,tT,tV ,tVItI 21jdsgsD21jdsgsDdrain
MOS-AK/GSA
Model Validation- Gain, Bias Current, Harmonic Distortion vs. Power
Simulated (___ ) Measured data (symbols)
Simulated (___ ) Measured data (symbols)Measured data (symbols)
© Copyright Agilent Technologies 2010
Page 23Page 23 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Model Validation- Load Line (Freq=2GHz, Bias=[-0.2, 6])
180
200Id (mA) (.): DC
120
140
160
180 (o): MeasuredLine: Simulated
60
80
100
Load
-20
0
20
40
This Load was not usedin the model extraction
0 2 4 6 8 10 1220
Vds (V)
Demonstrates model can go far beyond the range over which ti l DC d li S t b t k
© Copyright Agilent Technologies 2010
Page 24Page 24 D. E. RootD. E. RootDecember 8, 2010
conventional DC and linear S-parameters can be taken
MOS-AK/GSA
Model Validation- Gate Current Contours(4 t f diff t d l d F 4GH )
Ig (A) Symbol: Data
(4 contours for different powers and loads, Freq=4GHz)
Ig (A) Symbol: Data
0.01
0.015
yLine: Simulated
0.01
0.015
g ( ) yLine: Simulated
0
0.005
0
0.005
-0.01
-0.005
-0.01
-0.005
-2 -1.5 -1 -0.5 0 0.5 1-0.015
Vg (V)0 2 4 6 8 10 12
-0.015
Vd (V)
© Copyright Agilent Technologies 2010
Page 25Page 25 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Model Validation- Gain vs. Pout @26GHz
Model was extracted using 2GHz and 4GHz dataai
n (d
B)
Ga
Pout (dBm)
© Copyright Agilent Technologies 2010
Page 26Page 26 D. E. RootD. E. RootDecember 8, 2010
( )
MOS-AK/GSA
Outline
• Introduction: NVNA Measurements – complex spectra & waveforms
• Compact ModelsCompact model validationParameter extractionParameter extractionGaAs transistor model identification and validation
• Load-dependent X-parameter model of packaged GaN p p p gtransistor & validation with harmonic load-pull
• Discussion and Comparison
• Conclusions
© Copyright Agilent Technologies 2010
Page 27Page 27 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
S-parameters Solve All Small-Signal ProblemsBut devices must operate linearlyp y
Reflected Transmitted
Incident Freq. Domain ModelB1 S11A1 + S12A2
Measure
Agilent Vector Network AnalyzerS
B1 = S11A1 + S12A2
B2 = S21A1 + S22A2
g y
S-Parameters
ReflectedDesign
What about large-signal
nonlinear problems?
Reflected
Incident
1 2S_ParamSP1
Step=0.1 GHzStop=10.0 GHzStart=1.0 GHz
S-PARAMETERS
freq (1 000GHz to 26 00GHz)
$TC
700.
.S(2
,1)
© Copyright Agilent Technologies 2010
Page 28Page 28 D. E. RootD. E. RootDecember 8, 2010
nonlinear problems?freq (1.000GHz to 26.00GHz)
MOS-AK/GSA
X-parameters Solve Nonlinear Problems [1,8]Same use model as S-parameters, but much more powerfulp , p
Scattered Scattered
Model
Measure
Incident
11
, 11
*
( )
( )
( )
F mpm pm
S m npm qn qn
T m n
B X A P
X A P A
X A P A
Agilent Nonlinear Vector
Network Analyzer X
, 11( )pm qn qnX A P A
Reflected
Design X-Parameters
Reflected
Incident
1 2Agilent ADSEDA Software
© Copyright Agilent Technologies 2010
Page 29Page 29 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
X-Parameters: Nonlinear Spectral MapsFrom S-Parameters to “Harmonic Time-Domain Load-Pull”
2 kA1kA 1 1 1 1 1 2 2 1 2 2( , , , . . . , , , . . . )k kB F D C A A A A
2 2 1 1 1 2 2 1 2 2( , , , . . . , , , . . . )k kB F D C A A A A
Port IndexHarmonic (or carrier) Index
1kB 2kB magnitude & cross-frequency phase
Port Index
X-parameters allow us to simplify the general B(A) relations:Trade efficiency, practicality, for generality & accuracyPowerful correct and practical; Native Freq Domain ModelPowerful, correct, and practical; Native Freq. Domain Model
1 1 2 2( )) (i i iB S DC A S DC A The simplest X-parameters are just linear S-parameters
,,
( ) *1
( )11
,1
( ), 1
,1( ,| |) ), ( ,( )
ef gef gh hef
S f hgh
g h
T f hg
F fe
g hf hB X DC A X DP C A DC A P AP A X
, ,
( )11 21
( ), 11
( ) *11 2112 ( , ,| |,( ,| ( , ,| |,|,| ) )| ),
ef ghghf efe
F fe f
S f h T f hggh hB X DC A A X DC A X DC AA A AP AP P
© Copyright Agilent Technologies 2010
Page 30Page 30 D. E. RootD. E. RootDecember 8, 2010
, ,hg h g
MOS-AK/GSA
NVNA+Load-Pull = Instant Large-Signal Model
• Drag and drop measured X-parameters into ADS for immediate nonlinear circuit design based on measured X-pars
NVNA +Load-Pull
© Copyright Agilent Technologies 2010
Page 31Page 31 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Load-Dependent X-Parameters of a FETWJ FP2189 1W HFET
G. Simpson et al IEEE ARFTG Conference, December, 2008
15
20
0.25
0.30
Me
ag
eg
e Sim
Measured and Simulated Voltage and Current Waveforms
Measurements X-par Simulation
Pout Contour (dBm)
WJ FP2189 1W HFET
0
5
10
0.10
0.15
0.20
ea
sure
dC
urre
ntM
ea
sure
dV
olta
Sim
ula
ted
Vo
lta
mu
late
dC
urre
nt
0.2 0.4 0.6 0.80.0 1.0
-5 0.05
time, nsec
Measured and Simulated Voltage and Current Waveforms0 30
Measured and Simulated Dynamic Load Line
8
10
12
14
16
0.2
0.3
0.4
0.5
Me
asu
red
Cu
sure
dV
olta
ge
late
dV
olta
ge S
imu
late
dC
u0.15
0.20
0.25
0.30
asu
red
Cur
ren
tm
ula
ted
Cu
rre
nt
0.2 0.4 0.6 0.80.0 1.0
4
6
2
0.1
0.0
time, nsec
urre
ntM
ea
sS
imu
urre
nt
0 2 4 6 8 10 12 14 16-2 18
0.10
0.05
MeasuredVoltage
Me
SimulatedVoltage
Sim
E i t l H i B l X t if S t d l d ll
© Copyright Agilent Technologies 2010
Page 32Page 32 D. E. RootD. E. RootDecember 8, 2010
Experimental Harmonic Balance X-parameters unify S-parameters and load-pull
MOS-AK/GSA
Load-Dependent X-Parameter Model for GaN HEMT:C[12] J Horn G Simpson D E Root
GPIB
Bias
Cree CGH40010 GaN HEMT
[12] J. Horn, G. Simpson, D. E. Root IEEE CSIC Symposium Oct. 4, 2010
NVNA
MDC S l
Bias Tees
NVNA
10 W packaged transistor
900 MHzMaury Software
U
DC Supply NVNA Firmware
• 900 MHz• Measure Load-dependent X-parameters vs power at 9 impedances
DUT Maury Tuner
USB
Maury Tuner
vs power at 9 impedances• 4 harmonics measured• probe tones at 2nd and 3rd
harmonicsTunerTuner
9 load states at f1
harmonics• harmonic impedancesuncontrolled
X parameters > ADS for independent validation© Copyright Agilent Technologies 2010
Page 33Page 33 D. E. RootD. E. RootDecember 8, 2010
X-parameters -> ADS for independent validation
MOS-AK/GSA
Harmonic Load-pull Setup: For Validation Only[12] J. Horn, G. Simpson, D. E. Root IEEE CSIC Symposium Oct. 4, 2010
•Waveforms measured
NVNA
GPIB
Bias Tees
•Waveforms measured versus power at each set of 729 harmonic loads as controlled independently
Maury Software
DC Supply NVNA Firmware
by the tuners.•Fundamental, second, and third complex impedances setSoftware
USB
Firmware impedances set independently
DUT Maury Tuner Z1
Maury Tuner
Maury Tuner Z2
Maury Tuner Z3
© Copyright Agilent Technologies 2010
Page 34Page 34 D. E. RootD. E. RootDecember 8, 2010
9 states 9 states 9 states
MOS-AK/GSA
Load-Dependent X-Parameters versus Harmonic Load-Pull
Load-dependent X-pars• One output tuner to vary load at
Harmonic load-pull validation• Three output tuners to vary• One output tuner to vary load at
fundamental frequency. At each load inject small tones at 2nd and 3rd
harmonic freqs
• Three output tuners to vary loads at fundamental, second, and third harmonics independentlyharmonic freqs
(9x(1+2x2) = 45 measurements,actually ~125 measurements)
• Measured DC 4th harmonic
independently (9x9x9 = 729 measurements)
• Measured DC – 4th harmonic
• Take into ADS. Present 729 independent loads to model
• Measured DC - 4th harmonic
Compare waveforms PAE dynamic load lines etcADS
© Copyright Agilent Technologies 2010
Page 35Page 35 D. E. RootD. E. RootDecember 8, 2010
Compare waveforms, PAE, dynamic load-lines, etc.
MOS-AK/GSA
Prediction of GaN HEMT Harmonic-Load Dependencefrom Fundamental-Only Load-Dependent X-Parameters
60
70
80
PAE
30
40
50Z1Z2 Z3 Cree
Harmonic loads
6 8 10 12 14 16 184 20
20
10
CGH40010 GaN HEMT
Pin (available)2.0
y
Id [A] 70 2.0
Vd VdId
0.5
1.0
1.5
Id [A]
30
40
50
60
0 5
1.0
1.5
Vd IdVdId
10 20 30 40 50 600 70
0.0
-0.5
Vd [V]X parameter model
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.20.0 2.4
10
20
30
0
0.0
0.5
-0.5
Time (nanoseconds)
© Copyright Agilent Technologies 2010
Page 36Page 36 D. E. RootD. E. RootDecember 8, 2010
X-parameter model Harmonic time-domain load-pull measurements
Time (nanoseconds)
MOS-AK/GSA
Prediction of GaN HEMT Harmonic-Load Dependencefrom Fundamental-Only Load-Dependent X-Parameters
50
60
70
ZCree
CGH40010
PAE
20
30
40Z1Z2 Z3
CGH40010 GaN HEMT Harmonic loads
6 8 10 12 14 16 184 20
10
2.0
y a c oad e
Pin (available)70 2.0
VdId
VdId
1.0
1.5Id [A]
30
40
50
60
1.0
1.5
Vd
0.0
0.5
0.5
0
10
20
-10
0.0
0.5
-0.5
© Copyright Agilent Technologies 2010
Page 37Page 37 D. E. RootD. E. Root
0 10 20 30 40 50 60-10 70
December 8, 2010
Vd [V] 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.20.0 2.4
Time (nanoseconds)
MOS-AK/GSA
70
PAE vs. Available Input Power
Prediction of GaN HEMT harmonic-load dependencefrom fundamental-only load-dependent X-pars
50
60
70
Cree CGH40010
PAE
20
30
40Z1Z2 Z3
CGH40010 GaN HEMT Harmonic loads
6 8 10 12 14 16 184 20
10
y 60
70
1 5
2.0
Vd IdVdId
Pin (available)
0 5
1.0
1.5
2.0
30
40
50
60
0.5
1.0
1.5
Id [A]
-0.5
0.0
0.5
-1.0
10
20
0
-0.5
0.0
-1.0
© Copyright Agilent Technologies 2010
Page 38Page 38 D. E. RootD. E. Root
10 20 30 40 50 600 70 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.20.0 2.4
December 8, 2010
Vd [V] Time (nanoseconds)
MOS-AK/GSA
10W GaN Device Model Comparisonp
Harmonic Loads 75
50(%)Z1
0
25PAE
Z2
0 100 200 300Phase of Γ2 (degrees)
45
Z3
40
el(d
Bm
)
35
Pde
0 100 200 300Ph f Γ (d )
© Copyright Agilent Technologies 2010
Page 39Page 39 D. E. RootD. E. RootDecember 8, 2010
Phase of Γ2 (degrees) MOS-AK/GSA
10W GaN Device Model Comparisonp
Harmonic Loads 75
50
E(%
)
Z1
0
25PAE
Z2
0 100 200 300Phase of Γ2 (degrees)
45)Z3
40
el(d
Bm
)
35
Pde
0 100 200 300Phase of Γ (degrees)
© Copyright Agilent Technologies 2010
Page 40Page 40 D. E. RootD. E. RootDecember 8, 2010
Phase of Γ2 (degrees) MOS-AK/GSA
Fundamental-only load-dependent X-parameters
• Full two-port nonlinear functional block model for simulationAccounts for load tuning dependence of device performance without the– Accounts for load-tuning dependence of device performance without the requirement of independently controlling harmonic loads
– Use to design matching networks, multi-stage amps, Dougherty amps.– Accurate model of GaN device demonstrated
• Large reduction of data / time compared to harmonic load-pull
• Harmonic load-pull may be unnecessary– Source-pull unnecessary (see [12]) except for power transfer
© Copyright Agilent Technologies 2010
Page 41Page 41 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
Comparison of approaches presented hereApproach Advantages DisadvantagesApproach Advantages Disadvantages
Compact Works in all simulation modes (TA, D l i lCompact
Device model
( ,HB, CE)
More scalable Simple statistics Noise models;
Development time long Extraction difficult Technology-dependent
Noise models; dynamic self-heating common
What-if scenarios possible
X-parameter Device model
Technology independent Extremely accurate within
characterization range Complete IP protection
Limited by NVNA BW Large file size, especially
for multi-tone models Complete IP protection Works for packaged parts Automated extraction Often converge better than compact
Limited memory effects in products (but demonstrated in [15])
© Copyright Agilent Technologies 2010
Page 42Page 42 D. E. RootD. E. RootDecember 8, 2010
models in circuits
MOS-AK/GSA
Conclusions
• NVNA is a powerful, valuable tool for device modeling1 NVNA data (“HB data”) can be used to validate tune extract optimal1. NVNA data ( HB data ) can be used to validate, tune, extract optimal
parameters, explore compact model limitations & suggest improvements2. NVNA data can be used to identify complicated constitutive relations for
advanced compact models not possible using conventional dataadvanced compact models - not possible using conventional data.3. Load-dependent X-parameters, measured on an NVNA, is a powerful,
efficient, and accurate way to model important, new devices for PA and i it d i C l t h t t d licircuit design. Complementary approach to compact modeling.
• Compact models & X-parameters compared for device modeling applicationsg pp
• Every nonlinear modeling station should use an NVNA instead of a conventional linear VNA.
© Copyright Agilent Technologies 2010
Page 43Page 43 D. E. RootD. E. RootDecember 8, 2010
MOS-AK/GSA
NVNA, ANNs, and X-parameters fill in nonlinear puzzleElectronic design
automation softwareAgilent Nonlinear Vector
Network Analyzer
Nonlinear Nonlinear
CustomerNonlinear
Simulation & DesignMeasurements
Customer Applications
Nonlinear Modeling
© Copyright Agilent Technologies 2010
Page 44Page 44 D. E. RootD. E. RootDecember 8, 2010
*11 , 11 , 11( ) ( ) ( )F S m n T m n
pm pm pm qn qn pm qn qnB X A X A P A X A P A
MOS-AK/GSA
References[1] http://www.agilent.com/find/nvna[1] http://www.agilent.com/find/nvna
[2] Chiu et al “Characterization of annular-structure RF LDMOS transistors using Polyharmonic Distortion Model,” IEEE MTT-S International Microwave Symposium Digest, June, 2009, pp 977-980
[3] D. Gunyan et al, “Nonlinear validation of arbitrary load X-parameter and measurement-based device models,” IEEE MTT-S ARFTG Conference, Boston, MA, June 2009
[4] J. Xu, M. M. Iwamoto, J. Horn, D. E. Root, “Large-signal FET model with multiple time scale dynamics from nonlinear vector network analyzer data,” IEEE MTT-S International Microwave Symposium Digest, May, 2010.
[5] J. Horn, “Harmonic load-tuning predictions from X-parameters,” IEEE PA Symposium, San Diego, Sept. 2009
[6] G. Simpson et al, “Load-pull + NVNA = Enhanced X-parameters for PA designs with high mismatch and technology-independent large-signal device models,” IEEE ARFTG Conference, Portland, OR December 2008.g
[7] J. Verspecht and D. E. Root, “Poly-Harmonic Distortion modeling,” in IEEE Microwave Theory and Techniques Microwave Magazine, June, 2006
[8] D. E. Root et al “X-parameters: The new paradigm for measurement, modeling, and design of nonlinear RF and microwave components,” Microwave Engineering Europe, Dec. 2008 pp 16-21.
[9] A il t Ad d D i S t d t ti N li D i[9] Agilent Advanced Design System documentation: Nonlinear Devices
[10] E. Vandamme et al, “Large-signal network analyzer measurements and their use in device modeling,” MIXDES 2002, Wroclaw, Poland.
[11] http://www.cree.com/products/pdf/CGH40010.pdf
[12] J. Horn, G. Simpson, and D. E. Root IEEE MTT-S CSIC Symposium , Monterey, CA 2010.
[13] O. Jardel et al, “An electrothermal model for AlGaN/GaN power HEMTs including trapping effects to improve large-signal simulation results on high VSWR,” IEEE Transactions on MTT., vol. 55, Dec., 2007.
[14] J Verspecht, "Black box modelling of power transistors in the frequency domain," INMMiC'96 Conference Record, Duisburg, 1996
[15] J. Verspecht et al, “Extension of X-parameters to include long-term dynamic memory effects,” IEEE MTT-S International Microwave Symposium Digest, 2009. pp 741-744, June, 2009.
© Copyright Agilent Technologies 2010
Page 45Page 45 D. E. RootD. E. Root
y g
December 8, 2010
MOS-AK/GSA