RF IV Waveform Measurement and Engineering: - Role in Supporting Non-Linear CAD Design

RF IV Waveform Measurement and Engineering
- Role in Supporting Non-Linear CAD Design -
Centre for High Frequency

Engineering
School of Engineering
Cardiff University
Contact information
Prof. Paul J Tasker tasker@cf.ac.uk
website: www.engin.cf.ac.uk/chfe
Non-Linear CAD Models:
- state function based formulation

Time domain formulations
Physics Based State function

formulation: I & Q
Four quasi-static I and Q surface

functions
Drain Current [mA/mm]
1000
800
600
400
200
0
Q1RF(V1(t),V2(t))
"Q g ( v gs (t ),v ds (t ))
"Q d ( v gs (t ),v ds (t ))
Q2RF(V1(t),V2(t))
I2(t)
I1 (V1 (t),V2(t))
"t
V2(t)
6
4
2
0
-2
-4
0
RF
"t
Non-Linear State
Function Model
V1(t)
1
2
3
4
Drain Voltage [V]
ids ( t ) = I d ( v gs (t ),v ds (t )) +
Advanced formulations include time

delays
I1(t)
igs ( t ) = I g ( v gs (t ),v ds (t )) +
Gate Charge [pC/mm]
RF
I2 (V1 (t),V2(t))
1
2
3
4
Drain Voltage [V]
This fundamental formulation is followed by all analytical models and the Root lookup table model
- behavioral black box: based formulation
Frequency or Time domain

formulations
Behavioral based formulation
b2 (% ) = f ( a1 (% ),a1( 2.% ),.....,a1( n.% ),,a 2 (% ),a 2 ( 2.% ),.....a 2 ( n.% ))
Many different formulations
i ds ( t ) = " 0 + "1v gs ( t # $ 1) + " 2 v gs ( t # $ 2 ) + " 3 v gs ( t # $ 3 ) + ......
Analytical or experimental data based
30
20
20
10
0
I1(t)
I2(t)
0
f0
2f0
3f0
2 4 6 8 10 12 14 16 18 20
Input Power [dBm]
-20
-40
-60
FFT
V1(t)
! IFFT
a1 (n" )
b1 (n" )
Black-Box
Non-Linear
Model !
!
IFFT
b 2 (n" )
V2(t)
a 2 (n" )
FFT
Phase [deg]
Output power [dBm]
180
135
90
45
0
-45
-90
174
172
170
f0
2f0
3f0
168
166
2 4 6 8 10 12 14 16 18 20
Input Power [dBm]
Generally focus on describing a specific behavior
RF I-V Waveform Measurement & Engineering

- role in CAD modelling
State Function I(V) - Q(V) Non-Linear Models
Directly Measures Model related parameters I & V
I-Q function Extraction
Data Lookup Model Generation
Analytical Model validation and Optimization
Behavioural Black Box Non-Linear Models
Directly Measures Non-Linear Behaviour
Directly Import into CAD Tool
Data Lookup behavioural model
Indirectly Import into CAD Tool
Formulated behavioural models (Volterra)

Emerging non-linear parameter equivalent to linear s-parameters (Xparameters)
- state function based formulation
Requires measurement of the state

functions: I & Q
DC I-V provides Current State Function
igs ( t ) = I g ( v gs (t ),v ds (t )) +
Static measurements: trapping and

thermal issues
S-parameters measure differential of

state functions
Trapping and thermal issues

I1(t)
750
Q1RF(V1(t),V2(t))
ids ( t ) = I d ( v gs (t ),v ds (t )) +
"Q g ( v gs (t ),v ds (t ))
"t
"Q d ( v gs (t ),v ds (t ))
"t
Q2RF(V1(t),V2(t))
I2(t)
600
Id
450
300
V1(t)
150
0
0.6
0.2
-0.2
Vgs
-0.6
-1.0
Vds
V2(t)
I1RF(V1 (t),V2(t))
I2RF(V1 (t),V2(t))
Ideally require direct dynamic measurement of state functions
Non-Linear State Function CAD Models:
- indirect extraction from bias dependent s-parameters

Small Signal Equivalent Circuit Model
igs ( t ) = I g ( v gs (t ),v ds (t )) +
ygd
"t
"Q d ( v gs (t ),v ds (t ))
ygs
"t
v.ygm
yds
Linearize to get
s-parameters
1000
!
800
600
400
200
0
0
1
2
3
4
Drain Voltage [V]
Integrate bias dependent

linear s-parameters to get
non-linear parameters
Provides data for Root model
or for analytical curve-fitting
igs = y11 .v gs + y12 .v ds
y11 = y gs + y gd
y12 = - y gd
ids = y 21 .v gs + y 22 .v ds
y 21 = y gm - y gd
y 22 = y ds + y gd
Non-linear State
Functions
!
Qg =
Id =
# "(y ).v
11
# $(y ).v
21
gs
gs
# "(y ).v
12
ds
+ # $( y 22 ).v ds
Gate Charge [pC/mm]
Drain Current [mA/mm]
ids ( t ) = I d ( v gs (t ),v ds (t )) +
"Q g ( v gs (t ),v ds (t ))
6
4
2
0
-2
-4
0
1
2
3
4
Drain Voltage [V]
0
-2
-4
-5
-6
-10
-8
Q1 (V1(t),V2(t))
I2(t)
1.0
0.8
0.6
0.4
-5
0.2
-10
0.0
12
80
40
-6
-4
-2
0
Gate Voltage V
[V]
gs
V2(t)
I1RF(V1 (t),V2(t))
I2RF(V1 (t),V2(t))
If we have measured the terminal current flow

resulting from an applied and measured terminal
voltage and we reverse process and determine
state functions?
160
0.0
120
-0.2
-0.4
80
-0.6
40
-0.8
-1.0
-8
-6
-4
-2
0
Gate Voltage V[V]
gs
YES: Solutions in both the time and frequency domain.
Trans Capacitance [pF]
120
Drain Voltage V[mA]

ds
16
180 360 540 720

Phase []
10
-8
160
d [mA]
Drain Current I
Q2 (V1(t),V2(t))
180 360 540 720

Phase []
V1(t)
RF
DrainCurrent I [mA]
ds
I1(t)
RF
Gate Current I[mA]

gs
Model uses state functions to describe the

arbitrary time dependent terminal current flow
resulting from the applied arbitrary time
dependent terminal voltages
Gate Capacitance [pF]
10
[mA]
Gate Voltage
gsV
[mA]
d
Gate Current I
- direct extraction from RF I-V Waveforms

0.9
Qg(Vg(t))
One-Port Problem
Vg(t)
Select
Vgs(t)
Vgs(t1) Vgs(t2)
0.8
Vg(t)
Ig(t)
Two State functions

Depend on one
variable
0.7
0.6
0.5
0.4
igs ( t ) = Ig (v gs ( t )) +
200
300
400
300
400
300
400
60
"t
40
g
IgRF(Vg(t))
"Q g (v gs ( t ))
I (t)
100
I g(t2)
I g(t1)
20
0
"v
igs ( t 2 ) = Ig (v gs ( t 2 )) + C g . gs "t ( t 2 )
0
20x10
dVg(t)/dt
"v
igs ( t 1 )!= Ig (v gs ( t 1 )) + C g . gs "t ( t 1 )
Two equations with two unknowns,

so solve for Ig and Cg
200
dVgs(t2)
dt
10
dVgs(t1)
dt
0
-10
100
-3
t1
100
t2
200
Schreurs et al, EuMC 1997
Ig
RF
60
Ig
Ig(t)
40
Ig
disp
(t)
20
0
-20
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Gate Voltage, Vgs(t) [V]
Gate Capacitance [pF/mm]
Cg
1
0
-1
-2
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Gate Voltage, Vgs(t) [V]
Extraction of state functions for all measured values of Vgs(t)

Only one large signal measurement needed
Model extraction or model validation
Schreurs et al, EuMC 1997
-4
-5
-6
-10
-8
0
Waveform Measurement
Power Sweep @ 1.8 GHz
16
120
12
80
40
0
0
180 360 540 720

Phase []
10 Input Characteristic
0.8
0.6
0.4
-5
0.2
-10
0.0
-8
-10
-6
-4
-2
0
Gate Voltage V[V]
gs
Performance Understanding
Reactive Non-Linearity's
Transfer Characteristic
160
120
160
0.0
120
-0.2
-0.4
80
-0.6
40
-0.8
-1.0
-8
80
-6
-4
-2
0
Gate Voltage V[V]
gs
Extraction State Functions I & Q

Reactive strongly second order
-5
-6
-4
-2
0
Gate Voltage V[V]
gs
40
0
-8
Transistor Behavior
1.0
-8
10
Trans Capacitance [pF]
160
Drain Voltageds
V[mA]
Drain Current
I
d [mA]
180 360 540 720

Phase []
Gate Current[mA]
I
gs
Drain CurrentdsI[mA]
-2
Gate Current I[mA]

gs
Drain CurrentdsI[mA]
Gate Capacitance [pF]
10
Gate Voltagegs
V[mA]
Gate Current
I
d [mA]
-6
-4
-2
0
Gate Voltage V[V]
gs
State Functions
10

Stimulate with appropriate engineered
waveforms V1(t) and V2(t) voltages and
perform measurements
INPUT
Reverse Modelling Process
I1(t)
Q1RF(V1(t),V2(t))
Q2RF(V1(t),V2(t))
I2(t)
V2(t)
V1(t)
I1RF(V1(t),V2(t))
I2RF(V1(t),V2(t))
OUTPUT
Extract I and Q state functions: i.e. their

dependence on voltages V1 and V2
Extracted Fully Dynamic Intrinsic I and Q Surfaces of a pHEMT transistor
Morgan et al, IMS 2001
11
Waveform Measurement and Engineering

- are we looking at the device or the system?
v1
i1
measure
&
engineer
b1/a1 or a1/b1
frequency domain
measure
t
a1
b2
DUT
b1
a2
v2
measure
i2
measure
&
engineer
a2/b2 or b2/a2
time domain
t
12

Forward and Reverse Looking Measurements
separation in the frequency domain
fundamental
harmonics
400
source Impedance (a1/b1)
20
10
0
-2
-10
-20
-4
0
200 400
Phase
600
300
200
Stimulus
100
Gate Voltage [V]
Gate Current [mA]
Input Impedance S11(b1/a1)
a1 wave spectra
500
0
0
10
10
500
400
Responce
300
200
100
0
0
b1 wave spectra
13

Forward and Reverse Looking
Measurements
separation in the time domain
20
80
15
40
10
5
0
0
200
400 600
Phase
Drain Voltage [V]
Drain Current [mA]
Reverse Looking
Load Impedance (a2/b2) when current generator is active

Device Impedance (b2/a2) when current generator is in-active
Drain Current [mA]
80
40
10
15
20
Drain Voltage [V]
Extract Output Capacitance Cds = 0.4 pF/mm
14

15
Transistor RF I-V Waveforms
Verification of non-linear CAD models
Control mode of excitation
Similar to circuit operation

Maximize coverage of output I-V
space, state variable space.
Measure: V1(t), V2(t) and I1(t), I2(t)

I1(t)
Simulated
V1(t)
Measured
I2(t)
Simulated
V2(t)
Measured
Compare Simulated I1(t), I2(t) with

Measured I1(t), I2(t)
Import Measured V1(t) and V2(t) into the simulator
16
Verification/Optimization of non-linear CAD models

Provides insight to why and where the model is failing to
accurately predict non-linear behavior
Some
divergence
near knee
More robust and useful than what is typically done: simply

comparing simulated and measured Power performance
Some divergence at
pinch-off
17
Verification/Optimization of non-linear CAD models

0.4
0.3
Current [A]
Total Simulated Drain Current

Real Current Generator
Drain Capacitance
Measured Drain Current
0.2
0.1
0.0
-0.1
-0.2
-0.3
1
Real Current
Component
4
5
Gate Voltage [V]
Separate the measured

currents into their individual
components
Displacement and real
contributions
Results can be presented as
a function of input or output
voltage
Check model
formulations
Validity as a function of
bias
Displacement
Current Component
Result, in this case, show that the LDMOS model used is not
accurately modelling the variation of output capacitance as a function
of gate bias.
18

Behavioural Black Box Non-Linear Models

19
Review Linear Design Situation
- back to basics: s-parameters behavioral models

utilize measured s-parameter
data tables in RF CAD Tools
DUT
Vector Network Analyzer
ts
e
s
ata
ling
e
d
mo
Can simply transform sparameter to any arbitrary

impedance environment
Amplifier
Gain, Bandwidth,
Stability, Matching
Characterization and
CAD Design Enabling Tool
Can also measure say S21

and S11 as function on
input drive to get a very
basic non-linear behavioral
model
20
Consider Non-Linear Design Situation
- large signals: waveform based behavioral models

DUT
Waveform
Measurement and
Engineering
utilize measured waveform data

tables in RF CAD Tools
ts
e
s
ata
ling
e
d
mo
Cannot simply transform

waveforms to any arbitrary
impedance environment
Amplifier
Power, Efficiency,
Linearity
Characterization and
CAD Design Enabling Tool
Can measure as a function of

input and/or output
fundamental and harmonic
load impedances: CAD
interpolation and extrapolation
21
CAD Enabled Waveform Engineering
- Direct Waveform Look-up (DWLU) Data Model

Measure Waveforms
High Power
Bias-tee
Current /A
100 Watt LDMOS

Transistor
Sweep:
250W
20A
Klopfenstein
Impedance
Transformers
Pin, LOAD, Bias
6
4
2
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
time, nsec
High Power
Test Fixture
Formulate in frequency domain
Populate multi-dimensional datasets

FFT
Phase Ref
i1 (n." ) = M n ( v1 (" ) , Z L , V1 , V2 ).v1 (" )

i2 (n." ) = N n ( v1 (" ) , Z L , V1 , V2 ).v1 (" )
Scale
Magnitude(Iout) /dB
20
-20
-40
-60
0
10
12
freq, GHz
22
!

- DWLU Data Model Implementation
Data Import Unit constructed in Agilent ADS using FDD & DAC
I1(t)
I2(t)
FFT
V1(t)
IFFT
v1 (n" )
Black-Box
i2 (n" )
i1 (n" )
Non-Linear Model
v 2 (n" )
o io g v
r
ic ers in
M v be
R
ly
W
t
A
n
e
s
e
pr
FFT
IFFT
Requires ZL determination
e
DAC component
c
i
f Generic
fthe
links to
O
dV ,
V2(t)
e
e
MDIF
data
file:
i
if in
avZL, Biaserlook-up
w n
23

- DWLU Data Model Utilization
I/P Voltage Waveform
Output Power vs Pin
15
0
-5
-10
20
10
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
3rd Harmonic
16
15
28 29 30 31 32 33 34 35 36 37
Fundamental Input Power/dBm
Output Phase vs Pin
Efficiency vs Pin
400
40
30
20
10
0
200
0
-200
55
Fundamental
3rd Harmonic
2nd Harmonic
-400
1.0
0.7
0.9
0.8
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Time/nsec
17
Phase /degree
Voltage /V
2nd Harmonic
18
28 29 30 31 32 33 34 35 36 37
Time/nsec
50
----- Simulated
Fundamental
30
60
----- Measured
40
Gain /dB
Power /dBm
Voltage /V
O/P Voltage Waveform
Simulate on Data
Look-up Grid
19
50
10
ZREF=7.1
Gain vs Pin
28 29 30 31 32 33 34 35 36 37
Fundamental Input Power /dBm
Drain Efficiency /%
100W
LDMOS
50
45
40
35
28 29 30 31 32 33 34 35 36 37
DWLU Accurately regenerates RF waveforms

24

- DWLU Data Model Utilization
Output Power vs Pin

50
10
40
30
Power /dBm
15
0
-5
-10
10
0
-10
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
ZREF=7.1
20
Gain vs Pin
16.5
Fundamental
2nd Harmonic
16.0
Gain /dB
I/P Voltage Waveform
Voltage /V
100W
LDMOS
3rd Harmonic
15.5
15.0
14.5
28 29 30 31 32 33 34 35 36 37
28 29 30 31 32 33 34 35 36 37
in
h
t
i
O/P Voltage Waveform
Output Phase vs Pin
s Efficiency
w vs Pin
m tly
r
fo rec s
Fundamental
e
i ol
v
d
a
3
Harmonic
Simulate off Data
o
d
W
T
Look-up Grid
e
2 Harmonic s
u AD
----- Measured
----- Simulated
be C
n
a
c
DWLU Accurately interpolates RF waveforms
400
45
50
300
40
40
200
30
20
10
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Time/nsec
100
rd
nd
-100
-200
28 29 30 31 32 33 34 35 36 37
Fundamental Input Power /dBm
Drain Efficiency /%
60
Phase /degree
Voltage /V
Time/nsec
35
30
25
20
15
28 29 30 31 32 33 34 35 36 37
25
CAD Based Waveform Engineering
- Parameter Based Data Models: Formulation Concepts
Non-linear Data look-up
Direct looks up measured waveform data
Stored in the frequency domain
Non-linear Data Formulation: Parameter look-up
Transform waveform data into circuit parameters
Equivalent functionality to linear data formulation: s-parameters

Circuit analysis and design formulation
Travelling wave a-b rather than I-V formulations
Agilent Solution: X-parameters
Natural extension of linear s-parameters data-set to non-linear dataset
Cardiff Formulations
Natural extension of X-parameters. Cardiff Mixing Formulation for

load-pull contours: contains higher order mixing terms
26
CAD Based Waveform Engineering
- Formulated Based Data Lookup Models FDLU

n -1
2
100W
LDMOS
# Q &m
bk = " Ck,m % ( a1 +
$P'
m=0
Output power
contours
Gain
16
15
14
30.0
32.0
34.0
Q=phase(a1)
P=phase(a2)
Good simulation accuracy can be kept for

quite a large area on Smith Chart
Step size between adjacent lines is 0.2dB
13
28.0
#P&
U
( a2
"
k,m %
$Q'
m=0
Ck,m, = f(|a1(fo)|, |a2(fo)|) & Uk,m, = f(|a1(fo)|, |a2(fo)|)

!
17
n -1
2
36.0
Available Input Power /dBm
38.0
Fast and robust simulation

implementation
6
4
2
0
-2
-4
-6
-8
28.0
AM to PM
30.0
32.0
34.0
36.0
38.0
Available Input Power /dBm
Good accuracy for different drive power levels
27
CAD based Waveform Engineering

- Parameter Based Data Models: Formulation Concept
Circuit Formulation Requirement: remove direct reference to load

Component dependency: f(|a1|,|a2|,(Q/P))
Input
Power
2
a1 ! b1
a1 Stimulus
2
b1 Response
a1 = a1 e
j!1
"
"
%%
b1 = P. f $ a1 , a 2 , $Q P ''
#
&&
#
Parameter Model
"
"
%%
b2 = P. f $ a1 , a 2 , $Q P ''
#
&&
#
b2 Response
Output
Power
2
b2 ! a2
Load
a2 Stimulus
a2 / b2
a 2 = a 2 e j! 2 = a 2 Q
= a1 P
Linear System uses!s-parameters: 1st order system
Constant Parameters
b1 = {S11 " a1 " P + S12 " a 2 " Q}

b 2 = {S21 " a1 " P + S22 " a 2 " Q}
28

Use of s-parameters in non-linear design: Hot S-parameters
Wrong functionality:
model circular function

measurement elliptical functionality
0.70
0.65
_
# Q &0
# Q &1 -+
" % ( + S12 " a 2 " % ( .
$P'
$ P ' +/
# Q &0
# Q &1 -+
" % ( + S 22 " a 2 " % ( .
$P'
$ P ' +/
Imaginary (W )
)+
b1 = P.*S11 " a1
+,
)+
b 2 = P.*S 21 " a1
+,
0.60
0.55
0.50
0.45
0.40
Parameter
dependency:
Sm,n(|a1|,|a2|)
-0.5
-0.4
Real (W )
measured b2
-0.3
-0.2
modelled b2
29

Non-Linear System: include mixing components
Weakly Non-Linear System: 3rd order: relates to S&T Parameters (X-parameters)

1
X
X
2.1 2
2 &
# T
2
S
S
T
Q
P
b1 = $ X12 " a 2 " Q + X11 " a1 " P + X12 " a 2 " Q + X11 " a1 "
P '(
%
2 &
# T
2
S
S
T
Q
P
P
b 2 = $ X 22 " a 2 " Q + X 21 " a1 " P + X 22 " a 2 " Q + X 21 " a1 "
'
P
P.P/Q
%
(
For small perturbation reduces to three parameters: X-parameters
2
2.2 1
Q
Q.Q/P
Allow 1 = 2
*,
# Q &)1
# Q &0
# Q &1
# Q & 2 .,
T
S
S
T
b1 = P.+ X12 " a 2 " % ( + X11 " a1 " % ( + X12 " a 2 " % ( + X11 " a1 " % ( /
$P'
$P'
$P'
$ P ' ,0
,*,
# Q &-1
# Q &0
# Q &1
# Q & 2 .,
T
S
S
T
b 2 = P.+ X 22 " a 2 " % ( + X 21 " a1 " % ( + X 22 " a 2 " % ( + X 21 " a1 " % ( /
$P'
$P'
$P'
$ P ' ,0
,-
Parameter
dependency:
Xm,n(|a1|)
30

3rd Order Mixing Model: S&T-parameters (X-parameters)
Significantly improved functionality:
model is now an elliptical function

measurement elliptical functionality
0.65
0.60
Next Step
Compute local X-parameters
Imaginary (W )
0.55
function of load
Allow for full amplitudes dependence

Increase order of mixing
0.50
0.45
0.40
-0.5
-0.4
measured b2
*,
T
b1 = P.+ X12
" a2
,*,
b 2 = P.+ X T22 " a 2
,-
# Q &)1
S
" % ( + X11
" a1
$P'
# Q &0
S
" % ( + X12
" a2
$P'
# Q &1
T
" % ( + X11
" a1
$P'
# Q &2 .,
"% ( /
$ P ' ,0
# Q &-1
" % ( + XS21 " a1
$P'
# Q &0
" % ( + XS22 " a 2
$P'
# Q &1
" % ( + X T21 " a1
$P'
# Q &2 .,
"% ( /
$ P ' ,0
Real (W )
-0.3
-0.2
modelled b2
Parameter
dependency:
Xm,n(|a1|,|a2|)
31

Non-Linear System: include mixing components
Strongly Non-Linear System: nth order: relates to C&U Parameters (R-parameters)

)+
b1 = P.*S11 " a1
+,
)+
b 2 = P.*S 21 " a1
+,
# Q &0
" % ( + S12 " a 2
$P'
# Q &1 -+
"% ( .
$ P ' +/
1
2.1 2
3.1 2.2
# Q &0
# Q &1 -+
" % ( + S 22 " a 2 " % ( .
$P'
$ P ' +/
P.(P/Q)2
P
*,
#Q &
#Q &
#Q &
#Q & ,
!T
S
S
T
P.P/Q
b1 = P.+ X12 " a 2 " % ( + X11 " a1 " % ( + X12 " a 2 " % ( + X11 " a1 " % ( /
0
)1
,*,
b 2 = P.+ X T22 " a 2
,*,
b1 = P.+ R!1,-2 " a 2
,*,
b 2 = P.+ R 2,-2 " a 2
,-
$P'
$P'
2.
$ P ' ,0
# Q &-1
# Q &0
# Q &1
# Q &2 .,
S
S
T
" % ( + X 21 " a1 " % ( + X 22 " a 2 " % ( + X 21 " a1 " % ( /
$P'
$P'
$P'
$ P ' ,0
# Q &)2
" % ( + R1,-1 " a 2
$P'
# Q &)1
" % ( + R1,0 " a1
$P'
# Q &0
" % ( + R1,1 " a 2
$P'
# Q &-2
" % ( + R 2,-1 " a 2
$P'
# Q &-1
" % ( + R 2,0 " a1
$P'
# Q &0
" % ( + R 2,1 " a 2
$P'
2
2.2 1
3.2 2.1
Q
Q.(Q/P)2
Q.Q/P
$P'
# Q &1
" % ( + R1,2 " a1
$P'
# Q &2
" % ( + R1,3 " a1
$P'
# Q &1
" % ( + R 2,2 " a1
$P'
Allow 1 = 2
# Q &3 .,
"% ( /
$ P ' ,0
# Q &2
" % ( + R 2,3 " a1
$P'
# Q &3 .,
"% ( /
$ P ' ,0
32

- Parameter Based Data Models: Cardiff Formulation
Modelling with Extracted Fundamental Rm,n components: g(|a1|,|a2|)
bm
Accurate reproduction of measured b2 contours (load-pull contours) with 7th order

(Q/P) phase model
Avoids any implicit load based lookup
"
" Q %%
= P. f $ a1 , a 2 , $ '' = P.
# P &&
#
n= N +1 2
0
(
7th order model: R-parameters
)(
*
n=- N /1 2
" Q %n ,*
) Rm,n $ ' # P & *.
*+
0.65
Imaginary (W )
0.60
Parameter
dependency:
Rm,n(|a1|, |a2|)
0.55
0.50
0.45
0.40
-0.5
-0.4
measured b2
Collapse large data lookup to small (6*12 or 8x12) Rm,n parameter lookup
Real (W )
-0.3
-0.2
modelled b2
33

Cardiff Circuit Parameter Formulation
Generalized to nth order in terms of the relative phase component (Q/P)

bm
"
" Q %%
= P. f $ a1 , a 2 , $ '' = P.
# P &&
#
( ) (
*
n= N 2
0
(
n=- N 2 /1
" Q %n ,*
) Rm,n $ ' - where Rm,n = g( a1 , a 2 )
# P & *.
*+
Determination of parameters Rm,n requires measurements at constant |a1| and

|a2| while sweeping relative phase component (Q/P), normalized to optimum
load: easy to achieve with active load-pull
0.6
b2
0
-20
imaginary
mag(dB)
0.4
0.2
Extract
model
coefficients
0.0
a1
-40
-60
-80
a2
-100
-120
-30
-20
-10
0
10
20
coefficient number
30
b1 obtained
Example
of resulting set of coefficients
-0.2
from a measured device b2 response
-0.4
-0.2
real
0.0
0.2
34

Circuit Formulation that is an extension of linear s-parameters

remove direct reference to load
Formulation dependency: f(|a1|,|a2|,(Q/P))
Function dependency: f(|a1|,|a2|)
Input
Power
2
a1 ! b1
a1 Stimulus
2
b1 Response
a1 = a1 e
bm
j!1
"
"
%%
b1 = P. f $ a1 , a 2 , $Q P ''
#
&&
#
Parameter Model
"
"
%%
b2 = P. f $ a1 , a 2 , $Q P ''
#
&&
#
= a1 P
!
n= ( N )
2 (
"
*
" Q %%
" Q % n ,*
= P. f $ a1 , a 2 ,$ '' = P. 0 ) Rm,n $ ' # P &&
# P & *.
*+
#
n= -( N 2 /1)
b2 Response
Output
Power
2
b2 ! a2
Load
a2 Stimulus
a2 / b2
a 2 = a 2 e j! 2 = a 2 Q
Parameter dependency:
Rm,n(|a1|,|a2|)
35

Magnitude Function Fitting to Extracted Fundamental Rm,n components: g(|a1|,|a2|)

Rm,n = 0 + 1.|a2| + 2.|a2|2 + 3.|a2|3 + 4.|a2|4 + 5.|a2|5 + 6.|a2|6 + 7.|a2|7
Only 20 relevant coefficients
Accurate reproduction of measured b2 contours (load-pull contours) with 7th order model
0.1
7th order model: R-parameters
0.0
-0.1
R2,0 (Xf)
R2,1 (Xs)
R2,-1 (Xt)
-0.2
0.65
-0.3
_
Imaginary (W )
0.60
-0.4
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.6
0.55
0.50
0.4
0.2
R2,0 (Xf)
R2,1 (Xs)
R2,-1 (Xt)
0.45
0.40
0.0
-0.5
-0.2
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
-0.4
Real (W )
measured b2
-0.3
-0.2
modelled b2
36
- Parameter Based Data Models: CAD Implementation

Schematic of ADS Simulation
The model can be directly
imported into CAD software,
after processing the
measurement data.
t
i
m
i
L
t
n
s
e
d
m
a
e
o
r
l
u
w
s
e
a
N
e
M
t
n
e s
l
va ion
i
u at
q
Simulated load-pull
l
e
u
Contours (Left)
r m ng
e
et for rgi
Load-pull
am ear me
r
Power Contours a
p -lin e e
S n ar
o
Measured load-pull
Measured load-pull n
Contours
1.0
p
_
s
r
u
o
t
n
o
c
_
t
u
o
p
21.4
21.4
21.4
21.9
21.4
0.5
22.9
22.4
22.9
23.4
22.9
21.9
Simulated load-pull
Contours
21.4
22.4
23.4
21.4
21.9
22.4
0.0
-0.5
Contours (Right)
-1.0
Ig (-0.300 to 0.300)
freq (0.0000Hz to 4.500GHz)
A2phase (0.000 to 360.000)
-1.0
-0.5
0.0
0.5
1.0
37

Analytical Model validation and optimization

Behavioral Black Box Non-Linear Models

38

RF IV Waveform Measurement and Engineering: - Role in Supporting Non-Linear CAD Design

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RF IV Waveform Measurement and Engineering: - Role in Supporting Non-Linear CAD Design

Uploaded by

Copyright:

Available Formats

RF IV Waveform Measurement and Engineering

- Role in Supporting Non-Linear CAD Design -

Centre for High Frequency

Non-Linear CAD Models:

- state function based formulation

Physics Based State function

Four quasi-static I and Q surface

Drain Current [mA/mm]

Advanced formulations include time

Gate Charge [pC/mm]

Non-Linear CAD Models:

- behavioral black box: based formulation

Frequency or Time domain

Behavioral based formulation

b2 (% ) = f ( a1 (% ),a1( 2.% ),.....,a1( n.% ),,a 2 (% ),a 2 ( 2.% ),.....a 2 ( n.% ))

Many different formulations

i ds ( t ) = " 0 + "1v gs ( t # $ 1) + " 2 v gs ( t # $ 2 ) + " 3 v gs ( t # $ 3 ) + ......

Analytical or experimental data based

Output power [dBm]

Generally focus on describing a specific behavior

RF I-V Waveform Measurement & Engineering

State Function I(V) - Q(V) Non-Linear Models

Directly Measures Model related parameters I & V

I-Q function Extraction

Data Lookup Model Generation

Analytical Model validation and Optimization

Behavioural Black Box Non-Linear Models

Directly Measures Non-Linear Behaviour

Directly Import into CAD Tool

Data Lookup behavioural model

Indirectly Import into CAD Tool

Formulated behavioural models (Volterra)

Non-Linear CAD Models:

- state function based formulation

Requires measurement of the state

DC I-V provides Current State Function

Static measurements: trapping and

S-parameters measure differential of

Trapping and thermal issues

Ideally require direct dynamic measurement of state functions

Non-Linear State Function CAD Models:

- indirect extraction from bias dependent s-parameters

Integrate bias dependent

igs = y11 .v gs + y12 .v ds

Gate Charge [pC/mm]

Drain Current [mA/mm]

Non-Linear State Function CAD Models:

If we have measured the terminal current flow

YES: Solutions in both the time and frequency domain.

Trans Capacitance [pF]

Drain Voltage V[mA]

180 360 540 720

180 360 540 720

Gate Current I[mA]

Model uses state functions to describe the

Gate Capacitance [pF]

- direct extraction from RF I-V Waveforms

Non-Linear State Function CAD Models:

Two State functions

Two equations with two unknowns,

Schreurs et al, EuMC 1997

Non-Linear State Function CAD Models: