Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
2003 DURIP Equipment Program
Non-Linear RF Research and Educational Laboratory
PI: Patrick Roblin (MISES, SSEP)
CoPI: Roberto Rojas (ESL)
CoPI: Mohammed Ismail (MISES)
CoPI: Len Brillson (SSEP)
CoPI: Wu Lu (SSEP)
CoPI: Lee Potter (IPS)
CoPI: Oscar Takeshita (IPS)
Department of Electrical Engineering
The Ohio State University
Agency:
Air Force O�ce of Scienti�c Research
Dr. Witt
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Presentation
� Overview of DURIP equipment proposal - Patrick Roblin
{ DURIP team
{ DURIP equipment requested
{ Need for equipment and examples of applications
� Impact of Equipment: from systems, to devices, to materials (a top down approach)
{ Signal processing research - Lee Potter
{ Wireless testbed - Oscar Takeshita:
{ RFIC design - Mohammed Ismail
{ Mixed signals & UWB radar-on-chip - Steve Bibyk
{ Phase array antenna - Roberto Rojas
{ Projects relevant to non-linear RF - John Volakis
{ Designing GaN devices - Wu Lu
{ Defects and traps in GaN material/devices - Len Brillson
{ Electronic traps in III-Nitride- Steve Ringel
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Non-Linear RF DURIP Team
Electromagnetism
(SSEP)Materials and Devices
(IPS)
Signal Processing
(MISES)
Microwave, RFIC
Antenna
(ESL)
Communication
RFLaboratory
Non−LinearMixed Signals
Group Full Name Number of Faculty Members
ESL ElectroScience Laboratory 11
IPS Information Processing System Lab 8
MISES Mixed Signal Electronic Laboratory 4
SSEP Solid State Electronics and Photonics 8
CEA Postdoc coming from Sept. 2003-2004.
EE Department committed release time for center development.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Equipment Requested
IsothermalDevice Models
Non−linear RFBehavioral
System Models
Device ModelsIsothermal
Non−linear RF
Thermal Modeling
LSNAPulsed−RF
Pulsed−IV
Probe Station
Infrared Camera
DUT
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
DURIP Equipment Budget
Equipment: Source Cost Revised Cost
Large-signal network analyzer: Agilent $ 367,246 $ 300,000
Pulsed RF system: Agilent $ 506,025 $ 358,000
Infrared&Cascade probe station: Cascade&Indigo $ 156,530 $ 50,000
Total Equipment $1,029,801 $ 708,000
OSU cost-share $ 514,878 $ 354,000
Requested DoD share $ 514,923 $ 354,000
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Location of LNSAs in the World
13 LSNAs are in use around the world now.
Country Institution Number of LSNA
US Agilent (demo, design, modeling) 3 LSNA
NIST 1 LSNA
Belgium Katholieke Universiteit Leuven, Dept. TELEMIC 1 LSNA
Vrije Universiteit Brussel, Dept. Electrical Eng. 1 LSNA
Agilent-Europe 3 LSNAs
France IRCOM - Universite de Limoges 2 LSNAs� + 1 pulsed-RF
Ireland National Micro-Electronics Research centre 1 LSNA
Singapore National University of Singapore 1 LSNA
� IRCOM has just acquired a second LSNA dedicated to system measurement. Their �rst LSNA is
dedicated to non-linear pulsed-RF measurement.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Wireless Technology Trends
Continuous push for:
� Higher frequencies of operation
� Wide and ultra-wide bandwith communication and radar systems
� Highly linear electronics to conserve spectrum resource
� High power handling (base-station, satellite)
� E�ciency (talk time, satellite, linearization)
� New modulation and code techniques (constant envelope, vector modulation)
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Non-Linearities in RF Systems
Non-linearities arise from the complex dependence of the device current sources (IVs) and
charges (QVs) upon their state-variables such as terminal voltages, traps state, internal device
temperature Tdev, body voltages Vb and so on.
−5 −4 −3 −2 −1 0 1 2 3 4 5−140
−120
−100
−80
−60
−40
−20
0
20
40IP5 Theoretical CDMA Spectrum × Bandwidth B
Pow
er L
evel
(dB
m)
Normalized Frequency (f−f0)/B
Pout=30 dBmPout=20 dBmPout=10 dBm
Non-linearities leads to intermodulation and spectral regrowth
Spectral regrowth depends on the 3rd and 5th derivatives of the current sources
and charges!
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Need for Higher Derivatives in Behavioral Modeling
−20 −15 −10 −5 0 5 10 15 20 25 30−120
−100
−80
−60
−40
−20
0
20
Pin
(a1)
Pou
t(a1)
IMD Internal Output Power
−20 −15 −10 −5 0 5 10 15 20 25 30−120
−100
−80
−60
−40
−20
0
20IMD Internal Output Power
Pin
(a1)
Pou
t(a1)
The LSNA provides the required phase and amplitude of the in-band
intermodulation products to account for the higher derivatives.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Need for Higher Derivatives in Device Modeling
−10 −5 0 5 10 15 20 25−40
−30
−20
−10
0
10
20
30
Input Power (dBm)
IMD
3 (d
Bm
)
Comparison of IMD3 between Different Models
OSUFETOSUMET
−10 −5 0 5 10 15 20−70
−60
−50
−40
−30
−20
−10
0
10
20
Input Power (dBm)
IMD
(dB
m)
Comparison of IMD between Different Models
MET OSUFET − k=4, it=30 OSUFET − k=5, it=1 OSUFET − k=5, it=300OSUMET
(a) (b)
(a): Absence of higher derivatives leads to a 10 dB error in IMD.
(b) Using 3rd order derivative continuity gives a closer prediction of the IMD.
The LSNA will provide the missing higher order derivatives.
MISES/SSEP Microwave Laboratory The Ohio State University
Large Signal Network Analyzer
Characteristics:RF bandwidth: 0.6 to 20 GHzmax RF power: 10 WattModulation bandwidth: 8 MHzModulation: must be periodic
Building blocks:(1) test-set(2) down-converter(3) data acquisition(4) CW source
or modulation source(4)
(3)
(2)
(1)
Measurement Examples
CW in frequency domaina1, b1, a2, b2 (FET @ 2.4 GHz)
CW in time domainv1, i1, v2, i2 (FET @ 2.4 GHz)
-50-40-30-20-10
010
0 1 2 3 4 5 6 7 8
b2 /
dBm
Harmonic Index
-50-40-30-20-10
010
0 1 2 3 4 5 6 7 8
a2 /
dBm
-50-40-30-20-10
010
0 1 2 3 4 5 6 7 8
b1 /
dBm
Harmonic Index
-50-40-30-20-10
010
0 1 2 3 4 5 6 7 8
a1 /
dBm
0
8
16
24
32
0 0.5 1 1.5 2
Ids
/ mA
Time / period
0
0.5
1
1.5
2
0 0.5 1 1.5 2
Vds
/ V
-2
-1
0
1
2
0 0.5 1 1.5 2
Igs
/ mA
Time / period
0
0.5
1
1.5
2
0 0.5 1 1.5 2
Vgs
/ V
ModulationTime Domain Frequency Domain
Fundamental envelope3rd harmonic envelope
Spectral regrowth
2
What is large-signal analysis? S-parameters are commonly used to describe small signal device behavior. They are extremely valuable as they offer a mathematical representation of linear device behavior. The application of S-parameters allows circuit designers and modelers to achieve excellent correlation between simulation and measurement. However, S-parameters only apply to linear devices and systems, where the superposition principle is valid. Large-signal environments usually push devices into their nonlinear operating regions, and S-parameters and the superposition principle no longer apply. Researchers and design engineers face a significant problem when trying to translate high- level system specifications such as spectral regrowth, adjacent-channel-power-ratio (ACPR) and third-order-intercept (IP3) into concrete device and process parameters. There is an urgent need for a framework which deals with large-signal behavior from device to
system level in a coherent way, that can be applied to measurements, modeling and simulation. Imagine the increase of productivity in the design flow, if process engineers could work with circuit designers and system level engineers based on the same measurement tools and methods. A long-term cooperative effort between Agilent and several research institutions is under way with the goal of developing a new analysis framework for large-signal RF and microwave design process. With its strong presence in measurement, modeling and simulation, Agilent is uniquely positioned to lead this effort. Agilent’s large-signal network analyzer has allowed participating researchers to gain new
insight into their device’s behavior under large signal conditions. Now, for the first time, this measurement system is being made available to Agilent’s customers. Voltage and current relationships are the fundamental information used by computer-aided engineering (CAE) tools and device-level models. What has been missing is a way to accurately measure voltages and currents at the device under test’s (DUT’s) ports under realistic stimulus and load conditions. The large-signal network analyzer takes a first step towards bridging the gap between measurement, modeling and simulation, from device level to system level.
The large-signal network analyzer measures incident and reflected waves at fundamental, harmonic and modulation
frequencies.
ReferencePlanes
5 0 Ohm
Acquis it ionStimulus
Response
D UT
8
The concept of dynamic loadline is very useful for microwave power amplifier designers. The dynamic loadline represents current versus voltage at the output of a transistor for the fundamental and harmonics, under given bias conditions and input and output match. Once the dynamic loadline is
determined, the designer can easily tune bias parameters, input power and output impedance to optimize output power or power-added efficiency (PAE). Up to now, dynamic loadline information was only available from advanced simulators, and depended therefore on the accuracy of the large-signal models of
the device. The large-signal network analyzer makes it possible for the first time to accurately measure the dynamic loadline directly, without the dependence on accurate models. This allows faster and more efficient design cycles.
Time domain waveforms Frequency domain
gate drain
voltage
current
gate drain
Voltage - Current State Space
A GaAs HEMT is shown in voltage and current time domain view, as well as in voltage – current state space and frequency domain
For additional application information and technical papers about Large-Signal Network Analysis, please go to http:// www.agilent.com/find/lsna
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Memory E�ects
Memory e�ects are low-frequency dispersion e�ects (DC to MHz) associated with
various physical e�ects:
� Deep level traps (GaN, SiC)
� Self-heating (thermal distributed network) (HBT, LDMOSFET, SOI-MOSFET)
� Parasitics bipolar transistor & impact ionization (SOI-MOSFET, HEMT)
Because they respond to the RF modulation envelope (DC-MHz), memory e�ects
are hampering the development of wide-bandwidth modulation (WCDMA, UWB
radar).
� They introduce time-varying non-linearities leading to further performance degradation.
� Present linearization schemes used in power ampli�ers are not e�ective for wide-bandwith
modulation due to memory e�ects.
� Present device models are not capable of accurately accounting for non-linearities and
memory e�ects, limiting their usefulness in circuit design.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Modeling of Memory E�ects in Devices
Internal Device
Q
T
thCinstP
TrapsSelf−heating
dev
Tsub
B
+
−
+_
+
Parasitic bipolar
Intrinsic FET
Drain
A
GI
Gate
DQG
IC
I ii
EmitterBase(Body)
Collector
Source
−
+
bd
Vds
Cbd
Cbs
Rbs
RTemperature
Rd
bs
Rth,dev
instP
−
+
V
Rth,sub
I wii−
� Both analytic and table (B-spline) representations can be used
� Models parameters extracted from pulsed-IV and pulsed-RF measurements
� Models implemented in ADS
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Parasitic Bipolar E�ect in SOI-MOSFET
0 0.5 1 1.5 2 2.5 3 3.5 40
100
200
300
400
500
600
700
Drain−Source Voltage (V)
Dra
in C
urre
nt (
uA)
Vgs = 1 to 5 V (0.5 V steps)
Circle: Iso-thermal DC-IV, Plain line: Pulsed-IV.
Both plots are isothermal! Traps and parastic bipolar-transistor remain frozen in pulsed-IV
measurements but not in isothermal DC-IV.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Self-Heating in a LDMOSFET
0 5 10 15 20 25 300
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
VDS
(V)
I D (
A)
Vgs
= 4 to 6.5 V (0.25 V steps)
60 60 60 61 61 61 62 62 63 63 64 64 656061 62 63 64 65 66 68 69 71 73 75
78
60
62 65 67 70 72 7679 83 88
93 99106
61
64 68 7378 83 89
96 104113 122
134 147
61
66 7280 88
96 106116 128
140 155170 186
61
67 7787 98
110 123137 151
167 184
61
6981 95
109 124140 157
174 193
61
69
86 103 120138 157 176
60
70
89 110 130 151 172
61
70
92 116 140 163 186
60
70
95 122 149 175
Electro-thermal DC IV characteristics Tsub = 60�C.
The device surface temperature Tdev is indicated.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Pulsed IV Measurements
0 5 10 15 20 25 300
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Dra
in C
urre
nt in
A
Drain Voltage Vds
in V
Pulsed I−V Characteristics for VGS
=4.75 V and VDS
=13.7 V and Tsub
=35 C
0 10 20 30 40 50 600
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1IV MEASUREMENT−DEV TEMP=45C to 105C
Vds
Ids
device temp=45 device temp=60 device temp=75 device temp=90 device temp=105
Pulsed-IVs showing the starting bias point Pulsed-IVs from 45 to 105 oC
Repetition rate 10 kHz, Pulsed duration 1�s, Duty rate: 1% Pulsed-IV have a double bias
dependence!
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Pulsed RF system
� OSU system (same principle as IRCOM's system, but 2-6 GHz)
� Agilent Pulsed RF system (2-20 GHz) (dedicated system, high-performance)
� Agilent LSNA can be used for doing pulsed RF (same dynamic range but slower). LSNA
acquires non-linear pulse response as well.
HP8592A
RF RFout in
A B
port1 port2
DUT
2GHz
test port1 test port2
HP85047A
pulse
S−parameter set
modulatorHP11720A
network analyzer
HP8753E
AV−1011−C−OP1Bpulse generator
spectrum analyzer
MISES/SSEP Microwave Laboratory The Ohio State University
I. R. C. O. M. University of LIMOGES (FRANCE) PULSED LSNA 16/31IRCOM
LSNA WITH PULSED Fo SIGNAL (3/5)LSNA WITH PULSED Fo SIGNAL (3/5)
t
FiF
Pulses
The first sample point during the second pulse (B) should be at the phase of the first point outside (A), in order to obtain an array of samples suitable for FFT.Consequences :
- the LSNA calibration procedure from Agilent NMDG is suitable,- AdcF, Fo, sF, and To parameters are not independent,
- The AdcF clock must be phase-locked with RF synthesizer 10 MHz to avoid a large phase shift between A-1 and B point.
A : Next sample point without pulses B : Next sample point with pulses
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Need for Infrared Thermal Imaging of Transistors
Modeling of distributed thermal e�ects in FETs
Au
P2 P3 P4 P5 P6 P7P1Chip
z
y
x
L
W
(Xs,Ys)
ZsZ1
k
k
1
2
deviceactive
columns
Package
20mm
Adiabatic Plane
R13
R12
C1R1 C2
R2
LDMOSFET
R23
C3R3
Thermal Model. . .
Electrical Model
Image method is applied to a 140mm 7 �nger FET
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Lateral Temperature Distribution Across the FET
−1000 −500 0 500 10000
10
20
30
40
50
60
70
Y axis (µm)
Tem
pera
ture
Ris
e (o
C)
Temperature Distribution of a Simulated 140mm LDMOSFET
Z=0Z=50µmZ=200µmZ=500µm
10 oC drop for a 70 oC rise.
MISES/SSEP Microwave Laboratory The Ohio State University
Fig. 2 Image magnification sequence, from left to right, 1/5X, 1X, 5X and 25X, including visible overlay
Fig. 3 Quadrant display showing unpowered radiance image (upper left), powered radiance image (upper right), emissivity matrix(lower left) and “true temperature” thermogram (lower right)
. . .
Fig. 4 Typical InfraScope II screen display illustratingline scan and statistical display (upper left)
Fig. 5 Navigation image showing superposition of thermalinformation (color) over visible image (black and white)
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Need for Higher-Order Distributed Thermal Network
Transient and steady-state responses of the device average temperature.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
300
310
320
330D
evic
e te
mpe
ratu
re in
(K
)
(A)
0.094 0.095 0.096 0.097 0.098 0.099 0.1325
326
327
328
329
330
331
(B)
Time in (Sec)
Dev
ice
tem
pera
ture
in (
K)
A single RC thermal model is su�cient for DC and RF but NOT for modulated RF.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Example of Linearization Accounting for Thermal Memory
Thermal Memory Prediction
bb
IQ Modulator
dev
P
inst
sub
+
inst T sub
T
PP
bbbb dev dev
T
devT
I Q
Square−law detector
FilterLow pass
I (P ,T ) Q(P ,T )
PA
−
RF predistortion System
The real device temperature Tdev is estimated and used to update the temperature dependent
linearization while accounting for memory e�ect.
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Applications of DURIP equipment
The DURIP equipment requested will permit to
� qualify device/system linearity for wide-bandwidth system applications � (LSNA: Lu,
Takeshita)
� measure time constants associated with memory e�ects� (Pulsed RF: Brillson, Ringel, SSEP)
� develop DC to RF device models accounting for memory e�ects (Pulsed RF: Lu, Roblin)
� develop device models accurately accounting for non-linearities (LSNA+pulsed RF, Roblin)
� develop and test improved ampli�er linearization schemes accounting for memory e�ects
(LSNA: Serrani, MISES, IPS)
� develop and test improved non-linear system models (LSNA: IPS, MISES)
� develop and test non-linear RFIC (LSNA: Ismail, Rojas, Volakis)
� improved DAQ and RF signal processing (LSNA: Bibyk, Potter )
� test impact of non-linearities on wide-bandwidth modulation (LSNA: Takeshita, IPS)
� Application to GaN devices: Correlate memory e�ects observed in RF circuits with traps and
defects independently identi�ed in our SSEP labs (Brillson, Ringel) for improved material growth
(Ringel) and device fabrication (Lu).
MISES/SSEP Microwave Laboratory The Ohio State University
'&
$%
OHIOSTATE
T . H . E
UNIVERSITY
Presentation
� Overview of DURIP equipment proposal - Patrick Roblin
{ DURIP team
{ DURIP equipment requested
{ Need for equipment and examples of applications
� Impact of Equipment: from systems, to devices, to materials (a top down approach)
{ Signal processing research - Lee Potter
{ Wireless testbed - Oscar Takeshita:
{ RFIC design - Mohammed Ismail
{ Mixed signals & UWB radar-on-chip - Steve Bibyk
{ Phase array antenna - Roberto Rojas
{ Projects relevant to non-linear RF - John Volakis
{ Designing GaN devices - Wu Lu
{ Defects and traps in GaN material/devices - Len Brillson
{ Electronic traps in III-Nitride- Steve Ringel
MISES/SSEP Microwave Laboratory The Ohio State University
loadline.pdfattachmentOpen.pdfAgilentHardware OverviewWhat is large-signal analysis?
Voltage and current relationships are the fundamental information used by computer-aided engineering (CAE) tools and device-level models. What has been missing is a way to accurately measure voltages and currents at the device under test’s (DUT’s) portsThe large-signal network analyzer
CalibrationSoftware
Application examples
IRtransistor.pdfattachmentOpen.pdfAgilentHardware OverviewWhat is large-signal analysis?
Voltage and current relationships are the fundamental information used by computer-aided engineering (CAE) tools and device-level models. What has been missing is a way to accurately measure voltages and currents at the device under test’s (DUT’s) portsThe large-signal network analyzer
CalibrationSoftware
Application examples
IR2transistor.pdfattachmentOpen.pdfAgilentHardware OverviewWhat is large-signal analysis?
Voltage and current relationships are the fundamental information used by computer-aided engineering (CAE) tools and device-level models. What has been missing is a way to accurately measure voltages and currents at the device under test’s (DUT’s) portsThe large-signal network analyzer
CalibrationSoftware
Application examples