An automatic antenna matching
method for monostatic FMCW
radars
Professor: Prof. Dr.-Ing. Klaus Solbach
Supervisor: Dipl. -Ing. Michael Thiel
Student: Yan Shen
Outline
•
•
•
•
Introduction
System Development and Design
Impedance Tuner Design
Test Results
• Controller Algorithm
• Conclusions and Further Work
Introduction
Hardware Realization of the FMCW Monostatic Radar
d ≈ 1 / Δf
TX × RX = [ A1 cos(ω1 )][ A2 cos(ω2 )] = A1A2[cos(ω1 − ω2 ) + cos(ω1 + ω2 )] =
A1A2
[cos(Δω ) + cos(Δω + 2ω1 )]
2
If RX and TX are not well decoupled:
Reduced performance of
the mixer due to changed
DC operation.
RX × (TX + n × RX ) = RX × TX + n × RX × RX
DC offset
Decoupling
Diplexers
Rat-race coupler
if Zimage(V1,V2) = Zantenna, RX and TX are well decoupled.
Temperatures, radiation environments
Antenna impedance changes
Impedance tuner
System Design and Development
•
•
•
•
•
Rat-race Coupler
Wilkinson Power Divider
Gilbert Cell Mixer
Patch Antenna
System Modelling and Development
Rat-race Coupler
Schematic VS Momentum
ADS layout
P1
P4
P2
P3
Wilkinson Power Divider
ADS layout
Schematic VS Momentum
P2
P1
P3
Gilbert Cell Mixer
Mixer schematic
Power level test
R
R10
R=0.6 kOhm
R
R9
R=0.6 kOhm
R
R7
R=1200 Ohm
V_DC
SRC4
Vdc=5.6 V
IF output
C
C5
C=1 pF
R
R8
R=1200 Ohm
y2
y1
R
R2
R=10 Ohm
V_DC
SRC1
Vdc=2.9 V
LO input
Port
P2
Num=2
cmim
C1
c=3 pF
w=54.71 um
l=54.71 um
Port
P1
Num=1
cmim
C2
c=3 pF
w=54.71 um
l=54.71 um
RF input
rpnd
R6
R=3 kOhm
w=0.48 um
l=5.72 um
Imax=0.24 mA
rpnd
R5
R=3 kOhm
w=0.48 um
l=5.72 um
Imax=0.24 mA
npnH3shp4
Q1
Icmax=8 mA
npnH3shp4
Q2
Icmax=8 mA
npnH3shp4
Q3
Icmax=8 mA
npnH3shp4
Q5
Icmax=8 mA
rpnd
R3
R=3 kOhm
w=0.48 um
l=5.72 um
Imax=0.24 mA
npnH3shp4
Q4
Icmax=8 mA
npnH3shp4
Q6
Icmax=8 mA
npnH3shp8
Q7
Icmax=16 mA
cmim
C4
c=3 pF
w=54.71 um
l=54.71 um
cmim
C3
c=3 pF
rpnd w=54.71 um
l=54.71 um
R4
R=3 kOhm
w=0.48 um
l=5.72 um
Imax=0.24 mA
R
R1
R=10 Ohm
I_DC
SRC3
Idc=0.06 mA
V_DC
SRC2
Vdc=1.9 V
OpAmp
AMP1
Gain=60 dB
CMR=75 dB
Rout=100 Ohm
RDiff=15 kOhm
CDiff=1 pF
RCom=1 MOhm
CCom=1 pF
SlewRate=5e+8
IOS=0.2 uA
VOS=200 uV
BW=500 MHz
Pole1=
Zero1=
VEE=-5 V
VCC=5 V
Port
P3
Num=3
Patch Antenna
Single inset-fed patch antenna
Quad inset-fed patch antenna
Twin inset-fed patch antenna
larger bandwidth
largest bandwidth
System Modelling
Case 1:
R_image=50 Ohm, E_image=0
Case 2:
R_image=25 Ohm, E_image=80
Case 1
Case 2
IF_gain
=20log(1.015/0.066)
=24dB!
System Matching
DC 0.486V
10 MHz 1.613V
-40.97dBm
@10.01G
-26.23dBm
-21.10dBm
@10.00G
1.34dBm
Tuner Design
The traditional transmission tuner:
Additional induced losses on the
feed line due to multiple reflections
and losses in the ATU itself:
Transmission Tuner
The reflection tuner:
Losses on the tuner has no
influence to the system.
Reflection Tuner
Principle of our tuner
Tuner schematic:
Phase shifter
Term
Term1
Num=1
Z=50 Ohm
TLIN
TL1
Z=50.0 Ohm
E=C1
F=10 GHz
C1: 0~90°
tune the phase
Simulation result:
Variable resistor
R
R1
R=RX Ohm
RX: 33.3~75 Ohm
tune the amplitude
FET as Voltage-controlled Resistors
nonlinear Triquint MGF1402 package.
Rds~Ugs
MGF1402
Ugs:-0.5913 ~ -0.5101 V
Phase Shifter Design
Variable reactance reflection phase shifter
90°hybrid coupler:
Lange coupler
Branch-line coupler
Phase shifter schematic:
Branch-line coupler and Silicon tunning Varactor SMV 2019-108
P1
P2
Udiode: 0~20 V
Simulation result:
Phase shift: 218°
Tuner Schematic:
Udiode: 0~20V
Ufet: -0.6~0V
Simulation result:
PCB:
PCB of the Final Radar System
PCB VS Momentum
Rat-race coupler
Test Results
Power divider
NWA
Branch-line coupler
Antenna PCB VS Momentum
Tuner
Phase shift is not
enough;
FET works good.
Too high series
inductance
Two ways to improve
(a)
(b)
(c)
Controller System
Real control system
HP 4142B
Modular
DC
Source
SMU0
SMU1
Optimizer
Simulated control system
Udiode Radar
system
Ufet
Udc
Optimizer
DAQ
Instrument
Control algorithm
Aim:
Minimize Udc
ADS model
Original
data set
Udc=Interp(Udiode, Ufet)
Interpolation
in Matlab
Udc = f (Ufet , Udiode)
Starting points
Minisearch
function
in Matlab
Udcmin
Udiode, Ufet
Original data set, Column 1 is Udiode and
Column 2 is Ufet. Column 3 is Udc.
Three dimentional plotted graph
Examples
1. [x, fval, history, DC] = func2 ([1, 0])
Result: x = 4.7380 -0.0219
fval = 2.5215e-005
2. [x,fval,history,DC]=func2([3,-0.4])
Result: x = 4.9778 -0.2191
fval = 5.0413e-010
3. [x,fval,history,DC]=func2([2,-0.5])
Result: x = 2.0001 -0.7000
fval = 2.0723
Conclusions
This master thesis developed a dynamic method to minimize the DC offset at the
output of the mixer. A demonstrator was built on an RF grade circuit board (PCB)
working at an RF of 10 GHz and consisting of a voltage controlled oscillator (VCO),
a Rat- race coupler, a power divider, a tunable impedance network, a Gilbert cell
mixer. The hardware is shown below.
Further Work
•
There is a large space for the optimization of the tuner. Some methods can be found
out to reduce the series inductance in order to increase the phase shift, which will
lead to a larger range of realizable impedance values as shown in the ADS
simulation.
•
The performance of the dynamic method to minimize the DC offset can be
improved by using an I/Q mixer. An IQ-mixer consists of two balanced mixers and
two hybrids. It provides two IF signals with equal amplitudes which are in phase
quadrature. Two outputs provide two DC values which can be used better to control
the two control voltages for the tuner.
•
In the future, this work can be transferred into
an integrated circuit solution working at much
higher frequencies (e.g. 77) based on CMOS or
BICMOS technology, where resistors, capacitors,
diodes, transistors and multi level metals conductors
are available.
A 10-bit data multiplexor manufactured
in a SiGe BiCMOS process.
Appendix A
Patch Antenna
•
•
•
Let the substrate dielectric constant, thickness, patch length, patch width,
be denoted by ε r , h, L, W respectively.
In this experiment the patch will be fed by a microstrip transmission line,
which usually has a 50 Ohm impedance. The antenna is usually fed at the
radiating edge along the width (W) as it gives good polarisation, however
the disadvantages are the spurious radiation and the need for impedance
matching.
Here, an inset feed is used to match the antenna, because the resistance
varies as a cosine squared function along the length of the patch. A 50 Ohm
can be found in a distance from the edge of the patch. This distance is
called the inset distance.
1) Width of the patch
c
W=
2 f0
εr +1
2
Where c = the velocity of light
f 0 = operating frequency
2)
Because the electric field lines reside in the substrate and parts of some lines
in air. This transmission line cannot support pure transverse-electric-magnetic
(TEM) mode of transmission, since the phase velocities would be different in
the air and the substrate, an effective dielectric constant must be obtained in
order to account for the fringing and the wave propagation in the line.
Effective dielectric constant:
ε reff =
ε r +1 ε r −1
2
+
1
h −2
(1 + 12 )
2
W
3) The length may also be specified by calculating the half-wavelength value
and then subtracting a small length to take into account the fringing fields
as:
L = Leff − 2ΔL
ΔL = 0.412h
W
+ 0.264)
h
W
− 0.258)( + 0.8)
h
(ε reff + 0.3)(
(ε reff
4) For a given resonance frequency, the effective length is given as:
Leff =
c
2 f 0 ε reff
We get:
W=9.945mm, L=7.801mm
We use the curve fit formula
to find the exact inset length
to achieve 50 Ohm input
impedance for the commonly
used thin dielectric substrates.
7
6
5
4
3
⎫ L
⎧
ε
ε
ε
ε
ε
0
.
001669
0
.
1376
6
.
1783
93
.
187
682
.
69
+
−
+
−
⎪
r ⎪
r
r
r
r
−4
y0 = 10 ⎨
⎬×
2
⎪⎭ 2
⎪⎩+ 2561.9ε r − 4043ε r + 6697
we get:
y0 =2.22
Rat-race Coupler
Principle
Basic circuit
TLIN
TL1
Z=70.7 Ohm
E=270
F=10 GHz
Port
P1
Num=1
TLIN
TL5
Z=50 Ohm
E=90
F=10 GHz
TLIN
TL6
Z=50 Ohm
E=90
F=10 GHz
TLIN
TL2
Z=70.7 Ohm
E=90
F=10 GHz
TLIN
TL3
R
Z=70.7 Ohm
R1
E=90
R=R_image Ohm
F=10 GHz
Port
P2
Num=2
Real circuit schematic
Var
Eqn
ADS Layout
MLIN
TL3
Subst="MSub1"
W=_width mm
L=0.44 mm
VAR
VAR1
_radius=4.36
circle_width=0.62
_width=1.07
MCURVE
Curve2
Subst="MSub1"
W=circle_width mm
Angle=60
Radius=_radius mm
MCURVE
Curve1
Subst="MSub1"
W=circle_width mm
Angle=180
Radius=_radius mm
Port
P1
Num=1
Port
P4
Num=4
MLIN
TL1
Subst="MSub1"
W=_width mm
L=0.44 mm
MLIN
TL2
Subst="MSub1"
W=_width mm
L=0.44 mm
MSub
MCURVE
Curve4
Subst="MSub1"
W=circle_width mm
Angle=60
Radius=_radius mm
MSUB
MSub1
H=0.508 mm
Er=3.55
Mur=1
Cond=4.1e7
Hu=20 mm
T=0.070 mm
TanD=0.0027
Rough=20 um
Port
P2
Num=2
MLIN
TL4
Subst="MSub1"
W=_width mm
L=0.44 mm
MCURVE
Curve3
Subst="MSub1"
W=circle_width mm
Angle=60
Radius=_radius mm
Port
P3
Num=3
TLIN
TL4
Z=70.7 Ohm
E=90
F=10 GHz
Port
P3
Num=3
Port
P4
Num=4
Wilkinson power divider
Principle
Basic circuit
TLIN
TL1
Z=70.7 Ohm
E=90
F=10 GHz
TLIN
TL4
Z=50 Ohm
E=90
F=10 GHz
Port
P2
Num=2
R
R1
R=100 Ohm
Port
P1
Num=1
TLIN
TL3
Z=50 Ohm
E=90
F=10 GHz
TLIN
TL2
Z=70.7 Ohm
E=90
F=10 GHz
Real circuit schematic
MCURVE
Curve1
Subst="MSub1"
W=circle_width mm
Angle=90
Radius=_radius mm
Port
P1
Num=1
MLIN
TL7
Subst="MSub1"
W=1.07 mm
L=1 mm
MLIN
TL3
Subst="MSub1"
W=circle_width mm
L=_length1 mm
Var
Eqn
VAR
VAR1
circle_width=0.70009 {o}
_width2=0.400138 {o}
_width=0.533843 {o}
_radius=2.3
_length=_radius-circle_width/2+0.26
_length1=2.13804 {o}
_length2=1.04503 {o}
MTEE_ADS
Tee2
Subst="MSub1"
W1=circle_width mm
W2=_width2 mm
W3=_width mm
MLIN
TL5
Subst="MSub1"
W=_width2 mm
L=_length2 mm
MSTEP
Step1
Subst="MSub1"
W1=1.07 mm
W2=_width2 mm
MLIN
TL9
Subst="MSub1"
W=1.07 mm
L=0.5 mm
Port
P2
Num=2
MLIN
TL1
Subst="MSub1"
W=_width mm
L=_length mm
MTEE_ADS
Tee1
Subst="MSub1"
W1=circle_width mm
W2=circle_width mm
W3=1.07 mm
R_Pad1
R1
R=100 Ohm
W=0.4 mm
S=0.15 mm
L1=0.5 mm
MSub
MSUB
MSub1
H=0.508 mm
Er=3.55
Mur=1
Cond=4.1e7
Hu=20 mm
T=0.070 mm
TanD=0.0027
Rough=20 um
ADS Layout
MLIN
TL2
Subst="MSub1"
W=_width mm
L=_length mm
MCURVE
Curve2
Subst="MSub1"
W=circle_width mm
Angle=90
Radius=_radius mm
MLIN
TL4
Subst="MSub1"
W=circle_width mm
L=_length1 mm
MTEE_ADS
Tee3
Subst="MSub1"
W1=circle_width mm
W2=_width2 mm
W3=_width mm
MLIN
TL6
Subst="MSub1"
W=_width2 mm
L=_length2 mm
MSTEP
Step2
Subst="MSub1"
W1=1.07 mm
W2=_width2 mm
MLIN
TL8
Subst="MSub1"
W=1.07 mm
L=0.5 mm
Port
P3
Num=3
TLIN
TL7
Z=50 Ohm
E=90
F=10 GHz
Port
P3
Num=3
wave variable
U =U +U
in
Relative voltage and current:
Wave variables:
So
u = a + b; i = a − b
Power:
I = (U in − U re ) / Z L
re
u=
U
;[u ] = W
ZL
i = I × Z L ;[i ] = W
U in
U re
a=
,b =
ZL
ZL
a=
(u + i)
U
=(
+ I × ZL ) / 2
2
ZL
P in =
1 2 re 1 2
a ;P = b
2
2
b=
(u − i )
U
=(
− I × ZL ) / 2
2
ZL
Appendix B
Tuner with Branchline coupler
and SMV 2019-108
Tuner with Branchline coupler
and SMV 1245-011
Mixer testing circuit board
System circuit board
Appendix C
Tuner test results
Appendix D
Interpolation
Optimization
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
function v3=interpolation(v1,v2)
userdata = importdata('final.txt');
data = userdata.data;
Ufet=-0.7:0.05:0;
Udiode=0:1:6;
Udc1=data(1:15,3)';
Udc2=data(16:30,3)';
Udc3=data(31:45,3)';
Udc4=data(46:60,3)';
Udc5=data(61:75,3)';
Udc6=data(76:90,3)';
Udc7=data(91:105,3)';
Udc=[Udc1;Udc2;Udc3;Udc4;Udc5;Udc6;Udc7];
v3=interp2(Ufet,Udiode,Udc,v2,v1);
v3=abs(v3);
•
•
•
•
•
•
•
•
•
•
•
•
function [x fval history DC] = func2(x0)
history = [];
options = optimset('OutputFcn', @myoutput);
[x fval] = fminsearch(@(x)
interpolation(x(1),x(2)),x0,options);
function stop = myoutput(x,optimvalues,state);
stop = false;
if state == 'iter'
history = [history; x];
end
end
DC=interpolation(history(:,1),history(:,2));
plot3(history(:,1),history(:,2),DC,'-*')
xlabel('Udiode'),ylabel('Ufet'),zlabel('Udc');
grid on
axis ([0 6 -0.8 0 -2 6])
end