modeling of devices for power amplifier...
TRANSCRIPT
©Agilent Technologies 2004
Modeling of Devices for Power Amplifier Applications
David E. RootMasaya IwamotoJohn Wood
Agilent Worldwide Process and Technology CentersSanta Rosa, CA USA
Contributions from
Jonathan ScottAlex Cognata
Root et al 2004 UCSD Power Amplifier Workshop Page 2
Presentation Outline
• Introduction• Nonlinear Charge Modeling• Electro-Thermal Modeling• Advanced Measurements• Mathematical CAD Techniques • Summary & Conclusions
Root et al 2004 UCSD Power Amplifier Workshop Page 3
Power Amplifier Requirements and Modeling Implications
Linearity: Harmonic and Intermodulation Distortion; ACPR; AM-AM; AM-PM
Efficiency: PAE; Fundamental Output Power; Self-biasingModeling Challenges from device physics (III-V transport), complex signals, multiple time-scale dynamics, and the wide variety of device designs in many material systemsAccuracy over bias, frequency, and temperature; powerPerspective for this talk:
Modeling of III-V HBTs & HEMTs for circuit simulationPrimarily from CAD perspective
Root et al 2004 UCSD Power Amplifier Workshop Page 4
Presentation Outline
• Introduction• Nonlinear Charge Modeling• Electro-Thermal Modeling• Advanced Measurements• Mathematical CAD Techniques • Summary & Conclusions
Root et al 2004 UCSD Power Amplifier Workshop Page 5
Nonlinear Charge Modeling: MOTIVATIONFET models with same I-V eq.s but different charge eq.s [2]
IM3
[ dBm
]
0Model A = 1 /
jCV φ−
Model B = Modified Statz Model (1987)
Charge model is critical for distortion simulationContour
Model C = "Agilent (HP) FET Model" (1991) = C dV⋅∫
Root et al 2004 UCSD Power Amplifier Workshop Page 6
Charge Modeling Problem for Circuit SimulationSpecify two (e.g. Gate and Drain) charge functions, Qi, ,i=G,D such that:
( ( ), ( ))( ) ( ( ), ( )) i GS DSi i GS DS
dQ V t V tI t I V t V tdt
= +Can calculate charge function from physics (in theory) [6]. “Inverse modeling” alternative (not just table-based):Relate model functions to small-signal data and equivalent ckt.
intIm( )G G
GS DS
D D
GS DS
GS GD GD
m GD DS GD
Q QV VYQ QV V
C C CC C C C
ω
∂ ∂⎡ ⎤⎢ ⎥∂ ∂⎢ ⎥= =∂ ∂⎢ ⎥
⎢ ⎥∂ ∂⎣ ⎦+ −⎡ ⎤
⎢ ⎥− +⎣ ⎦
DS
GCGS CGD
Cm
CDS
data Derivatives of model function
Root et al 2004 UCSD Power Amplifier Workshop Page 7
Measured Bias-Dependent FET Gate Capacitances
VDS 5.00.0
400fF
0.0VDS 5.00.0
160fF
20fF
CGDCGSVgs=+0.5
Vgs=-1.5Vgs=-1.5
Vgs=+0.5
0.25um PHEMT
DS
GCGS CGD
Cm
CDS
The measured data cannot be modeled by two, two-terminal nonlinear capacitorsA single QG function must model both CGS and CGD
Root et al 2004 UCSD Power Amplifier Workshop Page 8
A model QG function consistent w. SS data exists if and only if( )meas meas meas
GS GD GD
DS GS
C C CV V
∂ + −∂=
∂ ∂This is a constraint on
pairs of independently measured Y– parameters (or S-parameters)pairs of “measured” device capacitances (attached to gate node)
This is the modeling principle of Terminal Charge Conservation
(D. Root 2001 ISCAS Short Course)(1)
The prescription for model charge calculation is, then:model [( ) ]
GS GD GD
meas meas measG GS DS
contour
Q C C dV C dV= + −∫(2)
Similar conditions to (1) and (2) can be written at the drain
Measured FET data are very consistent with (1) at gateSomewhat less consistent at drain
Root et al 2004 UCSD Power Amplifier Workshop Page 9
Why not use independently measured capacitances?
( ) ( ( ), ( ))( ) ( )( ( ), ( )) ( ( ), ( ))
GS GD
G G GS DS
meas measGS GDGS DS GS DS
I t I V t V tdV t dV tC V t V t C V t V t
dt dt
= +
+
freqω 2ω
IG
freqω 2ω
IG
This model will fit bias-dependent capacitances perfectlyBUT: It can be shown the capacitance terms yielda spectrum with a DC component!
Terminal Charge Conserving capacitances do not produce a DC component
Enforcing Terminal Charge Conservation is a modeling trade-off. It is not physical charge cons. (KCL)
Root RAWCON98
Root et al 2004 UCSD Power Amplifier Workshop Page 10
m1indep(m1)=2PAE_contours_p=0.318 / 89.811level=77.159945, number=1impedance = Z0 * (0.818 + j0.579)
m2indep(m2)=3Pdel_contours_p=0.603 / 156.285level=26.261844, number=1impedance = Z0 * (0.258 + j0.197)
indep(PAE_contours_p) (0.000 to 40.000)
PA
E_c
onto
urs_
pm1
indep(Pdel_contours_p) (0.000 to 50.000)
Pde
l_co
ntou
rs_p
m2
freq (1.000GHz to 1.000GHz)
untit
led4
..S
(1,1
)un
title
d4..
S(3
,3)
Maximum Power-AddedEfficiency, %
77.26
Maximum PowerDelivered, dBm
26.27
PAE (thick) and Delivered Power (thin) Contours
Measured Maximum Pout
Simulated PAE Contours
Simulated Pout Contours
Measured Maximum PAE
Example: EPHEMT Large-Signal Model [24]Analytic I-V, table-based charge and gate current, + self-heating
Simulated EPHEMT Load-Pull contours for Pmax & PAE.
Max PAE: 77.26% meas. 77.16% simulated.Max Power: 26.26 dBm measured and simulated
Root et al 2004 UCSD Power Amplifier Workshop Page 11
Terminal Charge Conservation and III-V HBT Modeling [4]
Intrinsic common-emitter HBT Y-parameters are of the form:
intIm( )BB BB BB BB BB BB
m CEBE CE BC C BC C B B m B B CE
CC CC CC CC CC CC C C m C C CEm CE
BE CE BC C BC C
Q Q Q Q Q Qg gV V V I V I C g C gYQ Q Q Q Q Q C g C gg gV V V I V I
τ τω τ τ
∂ ∂ ∂ ∂ ∂ ∂⎡ ⎤ ⎡ ⎤+ − +⎢ ⎥ ⎢ ⎥∂ ∂ ∂ ∂ ∂ ∂ ⎡ ⎤+ ⋅ − + ⋅⎢ ⎥ ⎢ ⎥= = = ⎢ ⎥∂ ∂ ∂ ∂ ∂ ∂⎢ ⎥ ⎢ ⎥ + ⋅ − + ⋅⎣ ⎦+ − +⎢ ⎥ ⎢ ⎥∂ ∂ ∂ ∂ ∂ ∂⎣ ⎦ ⎣ ⎦
Where:( )
21 22Im( )CCC
C m CE
Q Y YI g g
τω
∂ += =
∂ +21Im( )CC
C C mBC
Q YC gV
τω
∂= = − ⋅∂
C C
BC C
CV Iτ∂ ∂
=∂ ∂
Necessary and Sufficient ConditionsFor model QCC consistent with ss data:
QCC can then be constructed directly from data by:
[ ]CC C BC C Ccontour
Q C dV dIτ= +∫
III-V HBT Base-Collector Delay / Capacitance model [4]Measured base-collector transit time
CC C BC C CQ C dV dIτ= +∫
Shape of delay (before Kirk effect) related to velocity-field curves
( ( , ) )difQ V I Iτ≠ ⋅
C C
BC C
CV Iτ∂ ∂
=∂ ∂
“Capacitance cancellation”effect in III-V HBTs follows
-
Root et al 2004 UCSD Power Amplifier Workshop Page 12
Device data, model equations, and physics consistent
Root et al 2004 UCSD Power Amplifier Workshop Page 13
GaAs HBT power cell (“2x20 quad-emitter” device):(4x2x20) fT & CBC vs. IC
10 20 30 40 500 60
5
10
15
20
25
30
35
0
40
Ic (mA)
ft (
GH
z)
10 20 30 40 500 60
1E-13
2E-13
3E-13
4E-13
0
5E-13
Ic (mA)C
bc
(Fa
rad
s)
VCE=0.5 to 3V w/0.5V per stepMeasurementsSimulations [25]
Accurate charge model necessary to fit both fT and CBC, simultaneously over wide range of bias.
Si-based models do not fit this behavior well, generally
Root et al 2004 UCSD Power Amplifier Workshop Page 14
GaAs HBT power cell (“2x20 quad-emitter” device):S-parameters: 0.5-25.05 GHz
freq (500.0MHz to 25.05GHz)
S21
& S
12
freq (500.0MHz to 25.05GHz)
S11
& S
22
S21/25
S12•10S11
S22
MeasurementsSimulations
VCE=1,2,3VJC=0.05mA/µm2
Root et al 2004 UCSD Power Amplifier Workshop Page 15
Nonlinear Charge Modeling: Bias-Dependent IM3 of III-V HBTs [4]
VCE VCE
Measured fTSimulated fT
fT related to ττ modeled by Q
Charge model is critical for distortion simulationIc (A)0.0010.0020.0030.0040.0050.0060.0070.0080.000 0.009
Pout
IM3 VCE=2 to 4.5V
Measured Pout, IM3Simulated Pout, IM3
-80
-70
-60
-50
-40
-30
-20
-10
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009
Ic (A)
Pow
er (d
Bm
)
IM3
Pout
Root et al 2004 UCSD Power Amplifier Workshop Page 16
6x(2x14)µm2 InGaP/GaAs HBT cell:AM-AM and AM-PM distortions
0
2
4
6
8
10
12
14
16
-35 -30 -25 -20 -15 -10 -5 0 5 10
Input Power (dBm)
Gai
n (d
Bm
)
88
89
90
91
92
93
94
95
96
Phas
e (d
egre
es)
Good collector charge model critical foraccurate AM-AM and AM-PM
Symbols: MeasurementSolid lines: Simulation
VCE=3.5VJC=0.12mA/µm2
Frequency=5GHzRL=50Ω
Root et al 2004 UCSD Power Amplifier Workshop Page 17
Presentation Outline
• Introduction• Nonlinear Charge Modeling• Electro-Thermal Modeling• Advanced Measurements• Mathematical CAD Techniques • Summary & Conclusions
Root et al 2004 UCSD Power Amplifier Workshop Page 18
Dynamic electro-thermal (self-heating) modelAlgebraic perspective
T(t) is a dynamical variable
( ) ( ( ) ( ), ( ), )c c be ceI t I V t V T tt=
( ) ( ( ), ( , , ),( ) ( ))tc tc bc c be ceQ t Q V t I V t TV T t=Charge due to transit delay is a function of temperatureTemperature evolution equation based on dissipated power
TH C CEdT T R I Vdt
τ + ∆ = simplified
T(t) calculated by the simulator self-consistently
Root et al 2004 UCSD Power Amplifier Workshop Page 19
Dynamic electro-thermal (self-heating) model [25]
Currents, Voltages, and T(t) calculated by the simulator self-consistentlyusing coupled electrical and thermal equivalent circuits
2-pole thermal circuit used. Collapsible to single node.
deltaTTT ambdev +=
Tdev=device junction temperatureTamb=device ambient (backside) temperature
Root et al 2004 UCSD Power Amplifier Workshop Page 20
Self-heating model (static case)
1.1
1.2
1.3
1.4
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0VCE (Volts)
V BE
(Vol
ts)
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0VCE (Volts)
J C (m
A/ µ
m2 )
MeasurementSelf-heating model [25]
Isothermal model
Root et al 2004 UCSD Power Amplifier Workshop Page 21
1 2 3 4 5 6 7 8 90 10
100
200
300
400
0
500
Vce (Volts)
de
ltaT
(K
elv
in)
Thermal Coupling: Current Collapse Instability Simulated with III-V HBT Model [25]
•Thermal node allows interactions between devices. •Temp limited for convergence (but see [11])•Realistic thermal constitutive relations required [10]
HBT1HBT2
delT2delT1
RRth_shnt2
RRth_shnt1
RRthx
AgilentHBT_NPN_ThHBT2SelfTmod=1
th
I_ProbeIcct
I_ProbeIcc2
I_ProbeIcc1
AgilentHBT_NPN_ThHBT1SelfTmod=1
th
V_DCSRC1Vdc=VCE
I_DCSRC2Idc=IBB
thermal network
HBT2HBT1
IcctHBT1HBT2
1 2 3 4 5 6 7 8 90 10
100
200
300
0
400
Vce (Volts)
Icct
, Ic
c1,
Icc2
(m
A)
Root et al 2004 UCSD Power Amplifier Workshop Page 22
Dynamic electro-thermal III-V HBT Model Low Freq. S-parameters and Pulse Response Measured/Simulated
Comparison of different thermal equivalent circuits in Linear (s-parameter) and Transient (time-domain) cases
1E4 1E5 1E6 1E7 1E81E3 1E9
24
25
26
27
23
28
Frequency (Hz)
S21
(dB
)
Extrapolationof measured
Mag(S21) vs. Frequency1-poleMeasured2-poles
No thermal 0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2time (msec)
I C (m
A)
Pulsed IC vs. time
No thermal1-pole
2-polesMeasured
Two poles are more accurate than one [8].Compromise between speed and accuracy (SPICE).Distributed models can be used for HB analysis [7].
Root et al 2004 UCSD Power Amplifier Workshop Page 23
6x(2x14)µm2 InGaP/GaAs HBT cell:Electro-thermal Model Simulates PAE & IDC Accurately
0.0
10.0
20.0
30.0
40.0
50.0
-15.0 -10.0 -5.0 0.0 5.0 10.0 15.0 20.0
Output Power (dBm)
PAE
(%)
10.0
20.0
30.0
40.0
50.0
60.0
I DC (m
A)
VCE=3.5VJC=0.12mA/µm2
Frequency=5GHzRL=50ΩSimulation: 1-tone HB
Symbols: MeasurementSolid lines: Simulation
Root et al 2004 UCSD Power Amplifier Workshop Page 24
6x(2x14)µm2 InGaP/GaAs HBT cell:ACPR & PAE vs. Output Power
-80.0
-70.0
-60.0
-50.0
-40.0
-30.0
2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0
Output Channel Power (dBm)
AC
PR (d
Bc)
0.0
20.0
40.0
60.0
80.0
100.0
PAE
(%)
VCE=3.5VJC=0.12mA/µm2
Frequency=1.9GHzRL=50ΩModulation: IS-95 Rev. LinkSimulation: Envelope
Symbols: MeasurementSolid lines: Simulation
Root et al 2004 UCSD Power Amplifier Workshop Page 25
Presentation Outline
• Introduction• Nonlinear Charge Modeling• Electro-Thermal Modeling• Advanced Measurements• Mathematical CAD Techniques • Summary & Conclusions
Root et al 2004 UCSD Power Amplifier Workshop Page 26
Advanced Measurements
Essential forModel DevelopmentModel ValidationDevice characterization for technology development
Can not always infer dynamic large-signal behavior from only static (DC) or linear (S-parameter) data
Preferred for characterizing limiting behavior (e.g. breakdown)
Root et al 2004 UCSD Power Amplifier Workshop Page 27
GaN Devices
1 mm 10 fingers
GaN on Si
fT ~ 30GHz
Pulse width 2us
Pulsed measurements provide much more datathan can be measured under static (DC) conditions
Root et al 2004 UCSD Power Amplifier Workshop Page 28
Pulsed data on III-V HBTCurrent dependence of breakdown [15]
5x2x14 CE GaAs HBT
fT ~ 60GHz
Pulse width 1us
Can observe and extract breakdown parameters
Root et al 2004 UCSD Power Amplifier Workshop Page 29
Pulsed Bias Measurements
Same 0.25um PHEMT device
Soak time ~20 ms Pulse width ~200 ns “Iso-dynamic”
Can use to separate thermal and trapping effects
Root et al 2004 UCSD Power Amplifier Workshop Page 30
Vector Nonlinear Network Analyzer (VNNA) Measurements:Breakdown Current
Time (ns)
Biased wellbelow pinch-off
Measure device under conditions of actual use.Examine reliability under these conditions.
Root et al 2004 UCSD Power Amplifier Workshop Page 31
Vector Nonlinear Network Analyzer (VNNA) Measurements: Forward Gate Current
Time (ns)
Biased less negativelythan before
Measure device under conditions of actual use.Forward gate current in pHEMTs is complicated.
Root et al 2004 UCSD Power Amplifier Workshop Page 32
Vector Nonlinear Network Analyzer (VNNA) Measurements: Phase of device intermodulation versus bias
measuredSimulatedAgilentHBT
Root et al 2004 UCSD Power Amplifier Workshop Page 33
Presentation Outline
• Introduction• Nonlinear Charge Modeling• Electro-Thermal Modeling• Advanced Measurements• Mathematical CAD Techniques• Summary & Conclusions
Root et al 2004 UCSD Power Amplifier Workshop Page 34
Table-Based Model LimitationNaive simulator interpolation -> poor distortion
Vd
IP3(dBM)
-1.4
-1.0
P=-20dBm
Voltage Swing
Vgs
Vds0.06V
-0.6
Vd
IP3(dBM)
(a) Vg=-1V Power=-10dBm0.2V
P=-10dBmVgs
Vds
-0.6
-1.4
-1.0
(b) Vg=-1V Power=-20dBm
Si NJFET2-tones @ 100MHz (+1MHz)
Table Model
Null in third derivative of spline w.r.t. Vgs (2nd axis) with symmetrical data points
Interpolation model is better when signal size ~ data spacing
Root et al 2004 UCSD Power Amplifier Workshop Page 35
Artificial Neural Networks
A NN is a parallel processor made up of simple, interconnected processing units, called neurons, with weighted connections.
baxwsvxxFI
i
K
kikkiiK +⎟⎠
⎞⎜⎝
⎛+=∑ ∑
= =1 11 ),...,(
sigmoidweights biases
x1
...
xk
• Universal Approximation Theorem: Fit “any” nonlinear function of any # of variables• Infinitely differentiable: good for distortion.• Easy to train (fit) using standard third-party tools (MATLAB)
Root et al 2004 UCSD Power Amplifier Workshop Page 36
Artificial Neural Networks: Excellent, smooth fit.Better than simple table-based models
+ Measured data Blue Lines ANN fit
Root et al 2004 UCSD Power Amplifier Workshop Page 37
Adjoint Neural Network [21]Constructing FET gate charge, Qg, given ( )11 12Im , Immeas measY Y
Experimental validation of terminal chargeconservation at the gate for GaAs pHEMT
g
ds
QV
∂
∂
VdsVds
g
gs
QV∂
∂
11ImYω 12ImY
ω
( , )gg gs dsC V V( , )gd gs dsC V V−
X measured___ modeled
Root et al 2004 UCSD Power Amplifier Workshop Page 38
Summary and Conclusions
• Charge Modeling shown to be critical for simulating PA FoMs• Terminal Charge Conservation modeling principle discussed in detail. • Two-dimensional FET capacitances• Voltage & Current dependent transit time and capacitance in III-V HBTs
•Electro-Thermal Effects shown to be important for PA simulation• Static and dynamic self-heating necessary to fit device characteristics• Technical issues with thermal equivalent circuit, thermal coupling,
and thermal constitutive relations summarized
•Advanced Measurements shown to reveal rich device behavior•Advanced CAD shown promising for better PA simulation models