Post on 14-Apr-2018
transcript
7/29/2019 p11 Sadi Mos-Ak Paris 2
1/18
Modeling Intermodulation Distortion in
HEMT and LDMOS Devices Using a NewEmpirical Non-Linear Compact Model
Toufik Sadi and Frank SchwierzDepartment of Solid-State Electronics,
Technische Universitt Ilmenau,
D-98684 Ilmenau, Germany
Toufik.Sadi@tu-ilmenau.de
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
2/18
Objectives
Motivation
Non-linearities in semiconductor devices
Non-linear FET models
Compact modeling of III-V HEMTs and LDMOSFETs
Motivation
New in-house model
Validation
Summary
Outline
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
3/18
Framework: Within the COMON (COmpact MOdellingNetwork) project funded by the European Union
Aim: Development of improved universal HEMT models
Objectives: Efficient current-voltage, charge and noise models
GaAs, GaN HEMTs and other high-power devices
Focus: Non-Linearities in HEMTs
Intermodulation distortion (IMD)
Included Effects:
Self-heating; frequency dispersion; etc..
Compact Modeling of III-V HEMTs
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
4/18
Current-Voltage (I-V) Model
Accurate modeling of I-V characteristics and derivatives Inclusion of electrothermal & frequency dispersion effects
Applicable to GaAs and GaN HEMTs, and to Si LDMOS FETs
Effective parameter extraction and fitting routines
Modeling of IMD figures of merit using Volterra series analysis
Charge (C-V) Model Correct modeling of C-V characteristics is sufficient
Using simple/existing models
Non-linear HEMT Models
Design of modern microwave circuits and systemsMinimization of Intermodulation Distortion
Motivation
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
5/18
Non-Linearities in Electron Devices
Non -l inear I-V characteris t ics Distortion of the output signal shape
New frequency components appear
2nd order: 2xf
3rd order: 2xf, 3xf
nthorder: 2xf, 3xf,,nxf
0.0 0.5 1.0 1.5 2.0-15
-10
-5
0
5
10
15
Draincurren
t(a.u.)
Time
0.0 0.5 1.0 1.5 2.0-20
-10
0
10
20
30
40
Output(a.u.)
Time
Output Signal
Linear output Non-linear output
Almost everything in semiconductor electronics is nonlinear !!!
cos( )GS P
V V t
1( )d GSI t K V
2
1 2
3 4 5
3 4 5
( )d GS GS
GS GS GS
I t K V K V
K V K V K V
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
6/18
Intermodulation in HEMTs
Two-tone Input
Input with two frequency components f1and f2
Signal (Intermodu lat ion ) compon ents at new
frequencies are generated
1 2 1 1 2 2
cos cosin
V t V t V t A t A t
Example: 3rd order transfer characteristics
1 2 1 2
1 2 1 2
2
1 2
1 2
1 2
th
st
nd
1
rd
0 :
1 :
2 : ( ), ( )
(2 ), (2 ),3 :
,
2 , 2 ,
(2 ), (
3 , 3 ,
2
out
f f
DC
f f
f f
f f
V
f f
f f f f
f
t
f
2 1)
f f
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
7/18
Compact Models for III-V FETs
Physics-based
Analysis of effect of physical parameters (gate length, mobility, etc)No parameter optimizationRigorous mathematical formulaTechnology-dependentDiscontinuous (using of conditional functions)
Table-basedStoring parameters at several biases in a tableNo parameter optimizationTechnology-dependentDiscontinuities in the model elements or their derivatives
EmpiricalSimple
FlexibleContinuousTechnology-independentGood model formulation
Parameter optimization
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
8/18
Non-Linear Empirical III-V FET Models
Curtice Model (1980)Quadratic/cubic dependence of ID on VGS First empirical time-domain simulation model
Tajima Model (1981)Exponential dependence of ID on VDS and VGS First empirical frequency-domain simulation model
Materka Model (1985)Quadratic/hyperbolic dependence of ID on VGS Including drain-bias dependent pinch-off potential
Statz Model (1987)Hyperbolic/cubic dependence of ID on VGS/VDSTemperature scalability
TOM Model(s) (1990)Exponential/cubic dependence of ID on VGS/VDS Spatial/temperature scalability
ADS EEFET/EEHEMT Model(s) (1993)Rigorous formula Charge-based C-V model
Chalmers Model (1992)Hyperbolic dependence of ID on VGS/VDS First to provide a good fit for transconductance and derivatives
Auriga Model (2004)Enhanced version of the Chalmers modelMOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
9/18
Chalmers Model for HEMTs Advantages
Infinitely differentiable hyperbolic functions
Inherent reconstruction of the bell-shape ofGm(VGS) for GaAs HEMTs
Reliable modeling of the higher order
derivatives ofGm(VGS) curves
Continuity no conditional functions
Possibility of readily including several
effects, such as temperature effects,
frequency dispersion, and soft-breakdown
Simple procedure for parameter extraction
Suitability for intermodulation distortion studies Angelov et al, IEEE Trans. MTT,
vol. 40, p. 2258, 1992MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
10/18
Chalmers Model for HEMTs Limitations
max 1 max
1
Drain current at (at ) /
(
[1 tanh{ ( )}] tanh( )(
1
) ( )
)
PKPK GS PK
n i
GS n GS PK
D PK GS DS DS
i
gm gm II V V P
V P V
I I V V V
V
Limited suitability to model high-power devices and new structures such as
GaN HEMTs and LDMOSFETs(Fager et al., IEEE MTT, p. 2834, 2002; Cabral et al.,MTTS 2004)
Saturation current (ISAT) is limited to 2IPK
Improved model to provide much moreindependent control of the shape of the
current and transconductance curves while
maintaining the principal advantages of the
Chalmers model
Angelov et al,
IEEE Trans. MTT,
vol. 40, p. 2258,1992
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
11/18
New Current-Voltage Model (1)
( )( )
( ) (1 tanh
ln(1 exp{ ( ) / })ln(1 exp{
{ ( ) }) 0( ) ( tanh{ ( ) }
( / )
0
) }
)
[ ( ) ( ) ] tanh( )(1 )
GS PK
GS P
GS GS
GS G
K
S
GS PK GS
GS PK GS SAT
GS GS DS DS
f V Vf V V
V VV V
F V I f V
EC g ECEC g EC
F V I I f V
I F V F V V V
f(VGS) f(VDS)
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
12/18
New Current-Voltage Model (2)
1 2 1 2
1 2 1 2
2 2 2 2
2 2 2 2
1
1
( )
( )
( ) (
( ) (
{
{ .
( )
( )
) }
) }PK
SAT PK
TN TN TN TN
TN TN TN TN
GS n GS
GS n GS
GSN GS P
GS
GS
K
ni
i
ni
GSP
GSN GSN
G S
i
SP G P
I
I I
h V V V
h V V V
V
g V P h V
g V P h
V
V
V V
V V V V
V V V V
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
13/18
New Current-Voltage Model (3)
EC:more flexibility forI-Vcurves & derivatives
ISAT:IMAX=2IPKVTN:fine-tuning
parameters
Fager et al., IEEE MTT, p. 2834, 2002MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
14/18
I-V Model Advantages
Continuous closed-form expression
Accurate modeling of I-V characteristics and derivatives
Simple parameter extraction & fitting procedure
Applicable to GaAs, GaN HEMTs; LDMOS FETs;
LDMOS FET (Fager et al., IEEE MTT, p. 2834, 2002)GaN HEMT (Cabral et al.,MTTS 2004)
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
15/18
I-V Curves
0.25m gate-length GaAs pHEMT[1]
[1] K. Koh et al, in Proc. IEEE IMS, p. 467, 2003 [3] C. Fager et al, IEEE Trans. MTT, vol. 50, p. 2834, 2002
[2] J.-W. Lee et al, IEEE Trans. MTT, vol. 52, p. 2, 2004
VGS : -1.2V to -0.4V Step = 0.1V
0.35m gate length GaN HEMT[2]
VGS : -4V to 0V Step = 1V
LDMOS FET from[3]
VGS : 3 and 5V
Pulsed (300K)
Static DC
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
16/18
Volterra Series Analysis
Two-tone excitation input
1 2cos( ) cos( )Vin Vs t t
Results are from the GaAs pHEMT*
*K. Koh et al, in Proc. IEEE IMS, p. 467, 2003
Pin = -20dBm, RL = RS = 50 Ohm
Plin, PIM2, PIM3: linear, 2ndand 3rdorder power
IP2, IP3: 2ndand 3rdorder interception points
Modeling the contribution of the current source to non-linearities
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
17/18
Accomplished Work (5)
IMD analysis in high-power GaN HEMTs and LDMOSFETs
GaN HEMT (Cabral et al.,MTTS 2004)
LDMOS FET (Fager et al., IEEE MTT, p. 2834, 2002)
MOS-AK/GSA Workshop Paris - 7th & 8th April 2011
7/29/2019 p11 Sadi Mos-Ak Paris 2
18/18
Conclusions
New flexible empirical non-linear modelMinimized parameter fitting
Accurate calculation of higher-order derivatives
Suitable for intermodulation distortion modeling
Applicable to a wide range of devices
AcknowledgmentsThis work is funded by the European Union, in the
framework of the COMON project.
MOS AK/GSA Workshop Paris 7th & 8th April 2011