Development of SAR-based UWB S ee-Through-Wall Radar

Development of SAR-based UWB See-Through-Wall Radar

Yunqiang Yang Song Lin

Alex Zhang

Department of Electrical and Computer Engineering

University of Tennessee, Knoxville

Outline

Background Information Electromagnetic/Antenna Aspects UWB Components Design/DAQ Aspects Imaging Processing Aspects See-Thru-Wall Experiment Future Work

See-Thru-Wall Goals

Search Operation

Tactical Operation

provide dismounted and remote users with the capability to detect, locate and “see” personnel with concealed weapons/explosives behind obstructions from a standoff distance

Increased force protection and survivability of soldier in during operations, combat search and rescue, and hostage recovery operations.

Provide initial information on building layout and enemy personnel locations

Why Microwave UWB Radar?

Optical Quality Images at Microwave Frequencies Active System – Day and Night Imaging Adverse Weather Long Stand-off Ability (fine resolution imaging independent

of range) Both Broad and Spot Coverage Coherent Imaging Bi-static and Multi-static Configurations (transmitter

separate from receiver provides stealth) Penetration of Materials and Particulates (frequency

dependent) Detection of Ground Moving Target

Microwave Imaging

AdvantagesDay/night, all weatherPenetration (e.g. buildings)

Good scene recognition Poor object recognition

DisadvantagesNon-literal imagery

Imaging Fundamentals

Optical ImagesAngle vs. Angle

Microwave ImagesRange vs. Angle

CrossrangeAngle

RangeAngle

Optical Quality at Radar Frequency

Interior Image of Mannequin

Mannequin Only Mannequin Behind WallPhotograph

Resolution vs. Frequency

What controls the resolution of these

systems?

Downrange resolution is solely based on bandwidth in conventional RADAR (i.e. CW, FMCW)

UWB range resolution is based on the pulse width

meanwhile cross range timing resolution in a single antenna setup is a function of the antenna beamwidth (θ), where R is range

Multiple element or SAR system cross range resolution is a function of their effective aperture (L) and wavelength (λ)

B

cR

2

RAr

L

RAr

See-Through Wall Radar Prototype

RF Transceiver DAC/Control

Image Processing

Wall

Radar Rage: 20 m Radar PRF: 5 MHzPulse Width: 0.5 ns Center Frequency: 10 GHzHand-held portable/Ground Vehicle-Based System

Wave-propagation through the wall, and characterization of various Walls: Dielectric Constant, conductivity, attenuation Loss

Efficient EM modeling of scattering from objects inside a room

Wall parameter effects

Role of polarization in image enhancement Low-profile printed antennas/arrays for the system

Electromagnetic/Antenna Aspects of the System

UWB components design: power amplifier, low noise amplifier, power divider, SP16T switch, mixer, pulse generator.

Sampling of UWB signal: equivalent time sampling technique

UWB Transceiver Design and Data Acquisition Aspects

Image Processing Issues

Improve two-dimensional imaging resolution

Reduce antenna size

Mitigate the effects of the wall

Imaging quality depends on: Bandwidth, Baseline range, Wall distortions,

Wall uniformity, Wall absorption, Positioning errors

RF Attenuation in Different Wall Materials

N.C. Currie, D.D. Ferris, and al, “New law enforcement application of millimeter wave radar”, SPIE Vol. 3066, pp2-10, 1997

Propagation Modeling

Frequency domain measurement VNA for insertion transfer function.

Advanced Design System (ADS) models

UWB Antenna Consideration

Wide band-width Good impedance match Minimum waveform ringing Minimum pulse dispersion Small size Low cost

Types of UWB Antennas

Tapered slot: Two dimensional microstrip TEM horn: Most commonly used Bow-tie: Relatively high input impedance Requires a matching balun Resister loaded dipole Low gain and low efficiency Discone: High performance, Difficult to manufacture 3-D structure Bicone: High performance, Difficult to manufacture 3-D structure Log-periodic: Dispersive Spiral: Dispersive

Antipodal Vivaldi Antenna

Tapered flares on different layersDimension: 2.15cm x 5.52cm

Substrate: Roger 4003C, 10 mil-thick

Developed by Gibson in 1979 Wide band performance Fabricated on dielectric substrates Great potential to low cost and weight Small size

Vivaldi Sub-array

16 Element sub-array Dim: 18 cm x 40 cm Wilkinson power divider Element spacing: 2.15 cm

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

4 6 8 10 12 14 16f (GHz)

S11

(dB

)

7.5 GHz – 12.5 GHz

Pattern: Simulation Versus Measurement

@ 10 GHz

Measurement: 13dB Gain, 4° BeamwidthSimulation: 15dB Gain, 3° Beamwidth

Measured Radiation Pattern

E Plane

H Plane

Transmitter/Receiver Structure

Switch

1 2 3 4 16

........

System Block Diagram

UWB See-Through-Wall Imaging RadarSimulation (in ADS)

UWB See-Through-Wall Imaging Radar

VAR

VAR5

PulseEnergy_joule=.535e-10

Tstep=10 psec

EqnVar

VAR

VAR7Tmax=100 psec

EqnVar

VARglobal VAR6

LO_Freq=10 GHzPower_LO=-13 _dBm

EqnVar

VARVAR4

Tstop=(0.1*n) us

n=2PulseWidth=1 nsec

EqnVar

Tran

Tran1

ImpMaxPts=40960ImpMaxFreq=10 GHz

TimeStepControl=FixedMinTimeStep=Tstep

MaxTimeStep=TmaxStopTime=0.2 usec

StartTime=0 psec

TRANSIENT

DFDF

OutVar=DefaultTimeStop=Tstop

DefaultTimeStart=0 secDefaultNumericStop=Tstop/Tstep

DefaultNumericStart=0

uwb_antennaX4

freq=8-12 GHz

R X _ O U T

T X _ I N

RXTX

Rx_rX9

LOinput Rxintput

Ioutput

Qou

tput

2

43

1

Tx_rX8

Pulseinput

LOo

utput

Calibra

tion

Txoutput2

43

1

RES

R1R=50 Ohm

SpectrumAnalyzerFilter_Input_Spectrum3

RESR2R=50 Ohm

SpectrumAnalyzerFilter_Input_Spectrum2

TimedSinkFilterInput_Time1

Stop=0.2 usecStart=0 sec

SpectrumAnalyzer

Filter_Input_Spectrum1

TimedSinkFilterInput_Time2

TimedSinkFilterInput_Time3

RESR3R=50 Ohm

ImpulseFloatI1

Delay=100

Period=Tstop/(n*Tstep)Level=1.0

PULSE_SHAPE_GENERATORX1

TStep=TstepDoubletSeparation=350 psec

PulseEnergy_joule=PulseEnergy_joulePulseWidth=PulseWidth

Pulse ShapeGenerator

PulseO utputTriggerI nput

SpectrumAnalyzerFilter_Input_Spectrum

FloatToTimedF1

TStep=Tstep

TimedSink

FilterInput_Time

UWB_SubHarmonic_Mixer

Why SubHarmonic_Mixer? 1. Easy to implement in a PCB technology using coplanar

lines. 2. LO frequency can be lowered 3. Provides very high isolation between the RF port , LO

port and IF port. Specially the RF and LO have more than 40 dB isolation in the 8-12 GHz frequency range.

UWB_SubHarmonic_Mixer Simulation

IFIF

RF

RF

LOLO

MCLINCLin1

L=122 milS=6 milW=34 milSubst="MSub1"

MLOCTL1

L=LSW=WSSubst="MSub1"

TermTerm2

Z=50Num=2

MLINTL38

L=l21W=w

MTEETee2

W3=10 milW2=wW1=w

MLINTL32

L=l24W=w

MRSTUBStub2

Angle=60L=RFstub1Wi=10 milSubst="MSub1"

MLINTL40

L=100.0 milW=45.0 milSubst="MSub1"

MCORNCorn1

W=WSTEPSubst="MSub1"

MLOCTL2

L=1 milW=34 milSubst="MSub1"

bfp_layout_1bfp_layout_1_1ModelType=MW

Ref

mrstub7ghz2newmrstub7ghz2new_1ModelType=MW

Ref

MRSTUBStub1

Angle=60L=LOstubWi=10.0 milSubst="MSub1"

MTEETee4

W3=10 milW2=wW1=w31

MSUBMSub1

Rough=0 milTanD=0.0027T=0.1 milHu=3.9e+034 milCond=5.88E+7Mur=1Er=3.38H=20 mil

MSub

HarmonicBalanceHB2

Order[3]=3Order[2]=3Order[1]=5Freq[3]=IFfreq2Freq[2]=IFfreq1Freq[1]=LOfreq

HARMONIC BALANCE

LOBFPLOBFP_1ModelType=MW

RefP_nHarmPORT3

P[1]=polar(dbmtow(P_LO),0)Freq=LOfreqZ=50 OhmNum=3

MLINTL35

L=100 milW=w

MLINTL33

L=l32W=w32

VIAV1

W=10.0 milT=0.125 milH=20 milD2=15 milD1=20 mil

MTEETee1

W3=w32W2=w31W1=ww

MTEETee7

W3=wW2=wW1=w

MTEETee6

W3=10 milW2=wwW1=w

P_nTonePORT4

P[2]=polar(dbmtow(P_IF),0)P[1]=polar(dbmtow(P_IF),0)Freq[2]=IFfreq2Freq[1]=IFfreq1Z=50 OhmNum=4

MLINTL37

L=50 milW=w

MRSTUBStub4

Angle=60L=IFstubWi=10 milSubst="MSub1"

MLINTL23

L=lIFW=w

MTEETee5

W3=10 milW2=wW1=w

MRSTUBStub3

Angle=60L=matchWi=10.0 milSubst="MSub1"

MLINTL39

L=l22W=w

MLINTL24

L=l11W=ww

di_hp_HSMS8202_20000301D10

MLINTL31

L=l31W=w31

MLINTL30

L=l33W=ww

Harmonic Mixer

Frequency Range, RF: 8 - 12GHz

Frequency Range, LO: 8 - 12GHz

Frequency Range, IF: 0.1- 2.5GHz

Conversion loss <13dB

RF to LO isolation > 45dB

RF to IF isolation > 45dB

LO to IF isolation > 45dB

IP3 (Input) 14dBm

LO input power : 7dBm

Parallel-Feedback Dielectric-Resonator Oscillator

Why DRO? DROs are attractive microwave sources because

of their high Q, low phase noise, good output power and their high stability versus temperature.

They represent a good compromise of costs, size, and performance compared to alternative signal sources such as cavity oscillators, microstrip oscillators or multiplied crystal oscillators.

The parallel-feedback with BJT DRO can achieve the highest performance in some frequency range.

DRO Simulation

vout

RR7R=41 kOhm noopt{ 10 kOhm to 45 kOhm }

MLINTL19

L=20 milW=10 milSubst="MSub1"

MRSTUBStub1

Angle=70L=lstubWi=8 milSubst="MSub1"

MTEETee6

W3=10 milW2=10 milW1=10 milSubst="MSub1"

MRSTUBStub2

Angle=70L=lstubWi=8 milSubst="MSub1"

RR2R=50 Ohm noopt{ 120 Ohm to 350 Ohm }

V_DCSRC1Vdc=5 V

CC2C=1.0 pF

MLINTL7

L=20 milW=10 milSubst="MSub1"

MLINTL18

L=lbasebW=10 milSubst="MSub1"

MTEETee5

W3=10 milW2=10 milW1=10 milSubst="MSub1"

MLINTL17

L=lbasecW=10 milSubst="MSub1"

OscPortOsc1

MaxLoopGainStep=FundIndex=1Steps=10NumOctaves=2Z=1.1 OhmV=

MLINTL12

Mod=KirschningL=l9W=30 milSubst="MSub1"

MTEETee2

W3=wbiasW2=w1W1=w1Subst="MSub1"

MCURVECurve1

Radius=80 milAngle=90W=w1Subst="MSub1"

MSUBMSub1

Rough=0 milTanD=0.0001T=0.125 milHu=3.9e+10 milCond=1.0E+40Mur=1Er=3.38H=20 mil

MSubHarmonicBalanceHB3

OscPortName="Osc1"OscMode=yesSortNoise=Sort by valueNoiseNode[1]="vout"FM_Noise=yesPhaseNoise=yesNLNoiseDec=5NLNoiseStop=50 MHzNLNoiseStart=1k HzOversample[1]=5StatusLevel=3Order[1]=7Freq[1]=10 GHz

HARMONIC BALANCE

MLINTL14

Mod=KirschningL=l6W=w1Subst="MSub1"

MLINTL13

Mod=KirschningL=l8W=w1Subst="MSub1"

MCURVECurve2

Radius=80 milAngle=90W=w1Subst="MSub1"

RR6R=500000 Ohm

TFTF2T=-0.707

TFTF1T=0.707

PRLCPRLC1

C=3.18 nFL=0.08 pHR=35 Ohm

MICAP1C3

Wf=46.0 milWt=0.05 milNp=1L=205 milGe=8 milG=6 milW=8 milSubst="MSub1"

MLINTL8

Mod=KirschningL=95 milW=w1Subst="MSub1"

MLINTL16

Mod=KirschningL=l7W=30 milSubst="MSub1"

RR4R=1000000 Ohm

MLOCTL2

Mod=KirschningL=l1W=w1Subst="MSub1"

MLINTL4

Mod=KirschningL=l3W=w1Subst="MSub1"

MLINTL6

Mod=KirschningL=l5W=w1Subst="MSub1"

MTEETee4

W3=wbiasW2=w1W1=w1Subst="MSub1"R

R5R=5000000 Ohm

BFP640_MODELB1

MLINTL3

Mod=KirschningL=l2W=w1Subst="MSub1"

MLINTL5

Mod=KirschningL=l4W=w1Subst="MSub1"

MTEETee1

W3=w1W2=w1W1=w1Subst="MSub1"

MLOCTL1

Mod=KirschningL=l1W=w1Subst="MSub1"

RR3R=50 Ohm

DRO Oscillator

Operating Frequency Range:

9.9-10.1GHz

Phase noise:

-95dBc @ 10KHz

-120dBc @ 1 MHz

Output power: 7 dBm

Harmonics: -40 dBc min

Spurious: - 80 dBc min

Temperature stability: +/- 1MHz

Narrow Band Low Noise Amplifier

Freq range: 9.9-10.1 GHz

Gain: >11.5 dB

Gain Flatness: +/- 0.5 dB

Noise figure: 1.2 dB

P1dB: 16 dBm

IP3out: 24 dBm

UWB Power Amplifier

Freq range: 2-18 GHz

Gain: >12 dB

Gain Flatness: +/- 0.5 dB

Psat: 26 dBm

P1dB: 25 dBm

IP3out: 27 dBm

UWB System TopologyUWB System Topology

SP16T With Antenna ArraySP16T With Antenna Array

SP16T Using SPDT in SeriesSP16T Using SPDT in Series

Hittite SPDT (SMT))

DC - 14.0 GHz

SP4T MeasurementsSP4T Measurements

Frequency Range: 7 to 13 GHz

IL: - 4dB with flatness: +/-1dB

Isolation : <- 40dB

Test Fixture DesignTest Fixture Design

Top Side Bottom Side

RF LayoutRF Layout

Frequency Range: 9 to 13 GHz

IL: - 8dB with flatness: +/-2dB

Isolation : <-45dB

Switching Time: < 50ns

Driver LogicDriver Logic

Pulse GeneratorPulse Generator

Simulation & Measurement Simulation & Measurement Results of Pulse GeneratorResults of Pulse Generator

77 78 79 80 81 82 83 84 85 8676 87

0

2

4

6

8

10

-2

12

time, nsec

var(

"TR

AN

.V!"), V

0.5 1.0 1.5 2.00.0 2.5

-30

-20

-10

-40

0

freq, GHz

dBm

(fs(

var(

"TR

AN

.V!"))

)

Pulse Width:

Adjustable 400ps - 1ns

Rise Time: 50ps

Fall Time: 50ps

Bandwidth: up to 2GHz

Solutions for DAQ System

UWB Sampler: for hand-held portable model

PCI Digitizer: for ground vehicle based system

Oscilloscope: for experimental system

ADC Chip: for hand-held portable model

See-Trough-Wall Radar Experiment

Measurements without Wall

Measurements with Drywall

Targets Location

12cm X 24cm

20cm X 24cm

Concrete Wall

Metal-covered Door Targets

Radar Position

Top View -- Hallway Geometry and UWB Radar Setup

2.85m

9.30m

1.02m

Side Wall

Door 1 Door 2

Cylindrical Target

Door 1

Gas Tank

Door 2

Side Wall

Non-through-Wall Image

Image of Water Cup ----- Position 1

Water Cup

Door 1

Door 2

Side Wall

10cm X 12cm

Water Cup

Door 1

Door 2

Side Wall

Image of Water Cup ----- Position 2

10cm X 12cm

CFDTD Simulation

120cm

CFDTD Simulation ParametersMesh SizeNx = 330, Ny = 430, Nz = 330Cell Sizedx = dy = dz = 1.0cmTime resolutiondt = 19.15 ps

Local point source

Drywall boards thickness = 2cmEpson=2.4, Sigma=0.003

Free space gap 6 cm

Concrete @ f = 2 GHzEpson=7.0, Sigma=0.005

z

yx

240cm

Side View

Current simulation Problems

At f=2 GHz =15 cm requiring step size of 1cm.

To increase Mesh Resolution, we needed higher frequency Operation i.e. more mesh points.

Currently with a 4-processor server it requires 5 hours @ 2 GHz-at 4 GHz, it is anticipated 5x23 hours !!!-at 8 GHz it will be 5x26 hours.

1

16

Local point source

16-Elementreceiver array

250cm

350cm

12cm

Z = 120 cm

x

yz

30cm conducting cubic box at (x=70cm,y=195cm,z=120cm)

30cm conducting cubic box at (x=145cm,y=355cm,z=120cm)

55cm

Top View

Radiated UWB Pulse

Baseband signal is Gaussian with 0.8 GHz bandwidthCarrier is 2 GHz Sine Wave.

Recorded Response at 16 Receivers

Reflection from 1st Target

Direct Transmission from source to receivers

Reflection from 2nd Target

(m)

Without Gating

Direct Transmission from Source to Receivers

12 cm: Receiver Spacing

(m)

Direct Coupling Due to the Isotropic Point Source

Reflection from Targets

Reflection from 1st Target

Gating Direct Transmission

Reflection from 2nd Target

(m)

Reflection from far wall

After Gating of receivers response due to direct coupling

Extracted I/Q Channel

I Channel

Q Channel

1st Target

2nd Target(m)

Far wall

Image Recovered from Simulation Data

16

30cm conducting cubic box at (x=70cm,y=195cm,z=120cm)

30cm conducting cubic box at (x=145cm,y=350cm,z=120cm)

Future Work

Digital Signal Processing

Comparison of 2-D Spectral Estimation Techniques for Imaging Synthetic Point Scatterers

Image-Domain TCR is 13 dB

True Points

MVM

ARLP (2 quad)

Pisarenko

PML Estimates

MUSIC

TKARLP (2 quad)

TKARLP (all pred)

Taylor –35 dB n = 5

EV

SVA

2 Super SVA, Taylor

Sinc

RRMVM

ASR

2 Super SVA, SVA

0 dBRelative dB scale –60 dBNote: S.R. DeGraaf, “SAR Imaging…,”

IEEE T-IP, Vol. 7, No. 5, 1998