1

Matching Network Design Using

Non-Foster Impedances
Stephen E. Sussman-Fort, Ph.D.
Antenna Products and Technologies
EDO Electronic Systems Group
Bohemia, New York USA 11716
Dept. Electrical and Computer Engineering
State University of NY at Stony Brook
[email protected]
2
Outline
Introduction
Stability
Laboratory Measurements: High-Q Negative
Elements for Receive
Laboratory Measurements: Non-Foster
Monopole and Dipole
Technology Development
3
Introduction
4
Motivation
Requirement: broadband, efficient, electrically-small antennas (l << )
Electrically-short monopole: characterized by large reactance and low
radiation resistance (varies rapidly with frequency)
Wheeler and Chu (1947):
Electrically-small antennas have almost the same directivity as larger
antennas and can in principle perform as well as the larger ones
Problem is in transferring power to and from small antennas, which are very
difficult to impedance match
With conventional matching technology, small antennas suffer from:
Poor gain due to mismatch loss or lossy impedance matching (when
attempting a broadband match)
Narrow bandwidth due to the high Q of the antenna (when attempting to
maximize gain)
5
Noise Consequences of Low Antenna Gain
Consider:
An electrically-small broadband VHF antenna with typical gain
20-30dB below isotropic
A MIL receiver, designed for best compromise in sensitivity &
dynamic range (noise figures of 6-8 dB)
In such a system:
+
Greater sensitivity will result from increasing antenna
gain via non-Foster matching
Passive matching limited by gain-bandwidth constraints
It is receiver noise not external noise that limits sensitivity
6
1971: First mention of using negative inductance for bandwidth
extension of dipole antennas (Poggio and Mayes)
1977: First use of an active coupling network with negative resistance
to improve noise figure (Bahr)
Early Work on Negative Elements with Small Antennas
Active
coupling
network
Receiver Electrically-small
antenna
Z
C
= |r| + jx
Non-Foster impedance matching employs negative
reactive elements (L, C)
7
Fundamental Limits of Passive Matching
The Fano-Youla gain-bandwidth theory:
for matching networks containing positive RLC or distributed
elements
gives limits on the achievable bandwidth
implies that certain sources and loads (e.g. electrically-small
antennas) cannot achieve a good match, regardless of circuit
complexity
Circuits containing negative elements ( non-Foster
networks):
are not constrained by gain-bandwidth theory
can achieve wide matching bandwidths with difficultloads
arising from electrically-short antennas
8
Conventional vs. Negative Impedance Matching
matching reactance +L
capacitive reactance -1/(C)
(antenna)
freq
CONVENTIONAL
r
e
a
c
t
a
n
c
e
0
+
-
freq
r
e
a
c
t
a
n
c
e
0
+
-
net reactance = 0
at all frequencies
NEGATIVE
IMPEDANCE
net reactance = 0
at one frequency
Resonate +C with a positive inductor
Resonate +C with a negative capacitor
capacitive reactance -1/(C)
(antenna)
matching reactance +1/(C)
negative reactance-slope; violates
Fosters reactance theorem
9
Canonical Approach to Negative-Element Matching
Inductive-T completes
the match to 50
Antenna Model
Antenna model with L, C, and C
canceled by L, C, and C
10
Dualizer
For the tee and pi L networks shown
Z
in
=
2
L
o
2
/ Z
L
If Z
L
is the frequency-dependent
resistance
Z
L
=k
2
and if we want the input impedance Z
in
to
be the real and constant value R
o
, we
choose
L
o
2
= kR
o
(also used in coupled-resonator filter design)
Inductor-T Dualizer
Inductor-Pi Dualizer
11
Realizing Negative Elements
A negative element is produced by
terminating a negative impedance converter
(NIC) with a corresponding positive element
Grounded negative resistance
(Linvill, 1953)
Floating negative impedance
(Linvill, 1953)
Z
in
= Z
L
Z
L
R
in
= (R
2
/R
1
)R
L
12
How a Voltage-Inversion NIC Works
V
RL
Input current I
in
Input current flows through Q
1
producing V
RL
across R
L
V
RL
fed back through CE stage Q
2
producing 180 phase inversion at B
1
Voltage at E
1
, V
in
, appears in phase with voltage at B
1
R
in
= V
in
/ I
in
seen to be negative of R
L
: because current is same, but
voltages are inverted
Q
1
Q
2
B
1
E
1
V
in
13
Practical NICs - 1
Of the many NICs that
have been proposed,
only the Linvill and
Yanigisawa circuits have
been built and tested
Some other circuits can
be shown to possess
inconsistent phasing with
practical devices
14
Practical NICs - 2
Both the Linvill and Yanagisawa NICs are derived from the
same terminated or augmented network
Augmented network
Augmented network, redrawn, but
otherwise identical
Linvill OCS: Zin
A
= -(Z
b
/ Z
c
) Z
d
Linvill SCS: Zin
D
= -(Z
c
/ Z
b
) Z
a
Yana OCS: Zin
A
= -(Z
d
/ Z
c
) Z
b
Yana SCS: Zin
B
= -(Z
c
/ Z
d
) Z
a
The Linvill and Yanagisawa OCS configurations are, in fact, the same circuit
Z
i
n
B
Z
i
n
D
15
Early Use of NICs
G. Crisson (1931): Developed negative impedance repeaters to reduce
loss on telephone lines
vacuum tubes
The transfer function of any LC filter is realizable via an RC-NIC-RC structure
J . G. Linvill (1954): First active-RC filter
16
Modern Uses of Negative Elements and NICs
Active RC filters (NICs, gyrators, FDNRs are
fundamental elements)
Q-enhancement of passive resonators in active
filters
Broadband matching of electrically-small
antennas
17
Stability
18
Stability of NICs - 1
Theorem: Negative Impedance Converters are open circuit stable at one
port and short-circuit stable at the other port
Brownlie, 1965; Hoskins, 1966
Open-circuit stable:
For any passive impedance Z
L
at port 2, the network defined
by open-circuiting port 1 is
stable
Short-circuit stable:
For any passive impedance Z
L
at port 1, the network defined
by short-circuiting port 2 is
stable
Z
in1
= Z
L
Z
in2
= Z
L
19
Stability of NICs - 2
The inherent conditional stability of an NIC constrains the magnitude of the
impedances that can be connected to the open-circuit-stable port and to the
short-circuit-stable port
Open-circuit-stable port:
requires |Z
L1
| > |Z
in1
|
Short-circuit-stable port:
requires |Z
L2
| < |Z
in2
|
By what margins? depends upon nature of Z
L1
and Z
L2
20
NICs must be terminated properly for stability
In addition, the natural frequencies of any network
containing NICs must reside in left-half s-plane
In practice, the natural frequencies cannot be allowed to get
very close to the j-axis
Stability of Non-Foster Networks
antenna model Matching Network
C < 0 C > 0
C
net
For network stability, loop
impedances must be positive
e.g. C is positive : C is negative
but C
net
= (C in series with C)
must be positive
21
Predicting Stability in NIC Circuits
Transfer function of an NIC: T = A / (1 + A)
The feedback loop in an NIC must provide gain
and/or phase margin for stability:
|A| < 1 A < 180
o
Middlebrooks technique permits accurate
evaluation of A with all loading effects
Idea: break feedback loop; perform current-gain
and voltage-gain analyses, combine results to
yield A
With adequate component simulation models, the
technique is an excellent predictor of stability for
both the NICs and the overall network
A

Linvill Grounded NIC

22
Laboratory Measurements:
High-Q Negative Elements
for Receive
23
Design of High-Q Negative Elements
Historical results for negative-R and active filters: good
Results for negative L,C: poor (low-Q elements)
EDO has developed broadband, stabilized NICs and
high-Q negative L, C elements
Experimental results follow for representative circuits
24
Grounded Negative Capacitors and Inductors
Capacitor modeled as an ideal C
in
in parallel with a conductance G
Capacitor Q: magnitude of
C
in
is negative
G may be positive or negative
C
in
G
Inductor modeled as an ideal
L
in
in series with a resistance R
Inductor Q: magnitude of
L
in
is negative
R may be positive or negative
L
in
R
Linvill OCS
Negative
Capacitor
C
in
= -(R
1
/R
2
)C
L
Linvill SCS
Negative
Inductor
w/capacitive inversion
L
in
=-R
1
R
2
C
L
25
-0.001
0.000
0.001
10 30 50 70 90 110
Frequency (MHz)
C
o
n
d
u
c
t
a
n
c
e


(
S
i
e
m
e
n
s
)
simulated
measured
Experimental Results for Negative Capacitor
0
100
200
300
400
500
10 30 50 70 90 110
Frequency (MHz)
Q

measured
(smoothed)
simulated
-70
-60
-50
-40
10 30 50 70 90 110
Frequency (MHz)
C
a
p
a
c
i
t
a
n
c
e


(
p
F
)
measured
simulated
Capacitance C
in
Q
Conductance G
-100
-80
-60
-40
-20
0
0 500 1000 1500 2000
Frequency (MHz)
P
o
w
e
r


(
d
B
m
)
Spectrum Analyzer Meas. of Noise Power
Note: transistor model
valid only above 50MHz
26
0
100
200
300
400
500
10 30 50 70 90 110
Frequency (MHz)
Q

=

|
I
m

Z

/

R
e

Z
|
simulated
measured
(after tuning)
Experimental Results for Negative Inductor
-400
-300
-200
-100
10 30 50 70 90 110
Frequency (MHz)
I
n
d
u
c
t
a
n
c
e


(
n
H
)
measured
(after tuning)
simulated
-1.00
-0.50
0.00
0.50
1.00
10 30 50 70 90 110
Frequency (MHz)
R
e

Z


(
o
h
m
s
)
simulated
measured
(after tuning)
Inductance L
in
Q
Resistance R
-100
-80
-60
-40
-20
0
0 500 1000 1500 2000
Frequency (MHz)
P
o
w
e
r


(
d
B
m
)
Spectrum Analyzer Meas. of Noise Power
Tuned for best agreement
with simulation at low freq.
Spectrum Analyzer Noise Floor
27
Floating Negative Elements
Z Z
Linvill Floating Negative
Impedance Converter
Terminated in Z
Linvill Floating NIC Terminated in Z used as
Series Negative Element
Z
Series negative capacitor used in impedance matching of
electrically-short monopole and dipole
28
Laboratory Measurements:
Non-Foster Monopole and
Dipole
*
*
work performed for US Army I2WD (CECOM)
29
Monopole Experimental Results
Experimental demonstration of partial non-Foster
impedance matching with a monopole antenna:
Negative capacitor cancels the reactance of a
electrically-short 6monopole (partial matching)
Measured improvement in signal-to-noise ratio with
non-Foster-matched electrically-short monopole: up
to 9 dB at 30 MHz (as compared to lossy-matched
blade antenna of twice the size; receiver NF 8 dB)
30
Measurement of Signal-to-Noise Ratio on the
Antenna Range
Receiver
8dB NF
50
negative-C
50
6monopole
12lossy-matched
blade (EDO CNI24-3)
6 and lossy 12 reference antennas: behaved almost identically
Receiver
8dB NF
6monopole
50
Receiver
8dB NF
negative impedance converter
6 monopole with non-Foster matching improves signal-to-noise ratio
50
r + jx
Z
in
= r + jx
|x| << |x|
Z
in
31
Measured Improvement in Horizon Gain
Horizon Gain: Non-Foster Monopole compared to CNI24-3
-35
-30
-25
-20
-15
-10
-5
0
20 30 40 50 60 70 80 90 100 110 120
Frequency (MHz)
G
a
i
n


(
d
B
i
)
CNI-24 Gain (dBi) Monopole Gain (dBi)
Non-Foster
monopole
CNI24-3 Lossy-
Matched Blade
32
Measured Improvement in Signal-to-Noise Ratio
dB advantage in S/N ratio: for CECOM monopole antenna with negative-C as
compared to antenna without negative-C
0
2
4
6
8
10
20 30 40 50 60 70 80 90 100
Frequency (MHz)
d
B
Low noise receiver: 8 dB noise figure
up to 9 dB S/N advantage
Measurements taken at discrete10 MHz
intervals; Excel curve-fit produces plot
Improvement in S/N ratio: 6monopole with non-Foster matching over 6monopole alone
33
Dipole Experimental Results
Experimental demonstration of partial non-Foster
impedance matching with a dipole antenna:
Negative capacitor cancels the large portion of
reactance of an electrically-short dipole, 12total length
Measured improvement in signal-to-noise ratio with non-
Foster-matched electrically-short dipole: up to 20 dB
(see graphs) as compared to 12lossy-matched blade
monopole antenna
34
Measurement of Signal-to-Noise Ratio on the
Antenna Range
50
negative-C
negative-C
50
12lossy-matched
blade (EDO CNI24-3)
Receiver
8dB NF
12 monopole
reference antenna
Balun
Receiver
8dB NF
negative impedance converter
negative impedance converter
12 dipole antenna (6 per arm)
with non-Foster matching
improves signal-to-noise ratio
12dipole
6

Swept-frequency measurements
35
Measured Improvement in Horizon Gain
Horizon Gain: Non-Foster Dipole compared to CNI24-3
-35
-30
-25
-20
-15
-10
-5
0
20 30 40 50 60 70 80 90 100 110 120
Frequency (MHz)
G
a
i
n


(
d
B
i
)
CNI-24 Gain (dBi) Dipole Gain (dBi)
CNI24-3 Lossy-
Matched Blade
Non-Foster
dipole
36
Blade Antenna: Noise Floor and Received Signal
Blade Antenna: Noise Floor and Received Signal Level
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
20 30 40 50 60 70 80 90 100 110 120
Frequency (MHz)
d
B
m
Blade Antenna Noise Floor
Blade Antenna w/Swept Signal
Commercial FM stations
S
blade
N
blade
37
Negative-C Dipole: Noise Floor and Received Signal
Negative-C Dipole: Noise Floor and Received Signal Level
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
20 30 40 50 60 70 80 90 100 110 120
Frequency (MHz)
d
B
m
Negative-C Dipole Noise Floor
Negative-C Dipole w/Swept Signal
Commercial FM stations
S
dipole
N
dipole
-C
-C
38
Signal-to-Noise Advantage, 12 Negative-C Dipole
over 12 Lossy-Matched Blade Monopole Antenna
S/N Advantage, Negative-C Dipole over Lossy Matched Blade
0
5
10
15
20
25
30
20 30 40 50 60 70 80 90 100 110 120
Frequency (MHz)
d
B
S/N Advantage
6 per. Mov. Avg. (S/N Advantage)
raw data
smoothed
data
Improvement in S/N ratio: 12dipole with non-Foster matching over 12lossy-matched blade
Noise peaks from extraneous RF
sources cause loss of S/N advantage
J aggedness of curves results from
subtraction (S/N)
dipole
(S/N)
blade
39
0
5
10
15
20
25
30
20 30 40 50 60 70 80 90 100 110 120
Freq. (MHz)
d
B
-90
-85
-80
-75
-70
-65
-60
20 30 40 50 60 70 80 90 100 110 120
Freq. (MHz)
d
B
m
No S/N Advantage When External Noise Dominates
S/N advantage
Noise Floor
Noise peaks from extraneous RF
sources cause loss of S/N advantage
Commercial FM Stations
40
CECOM Dipole on Test Range
Side view
Head-on
view
41
CECOM Dipole Compared to CNI-24 Monopole Blade
on 8 ft. Ground Plane
Blade on ground plane
42
Technology Development
43
Technology Development
1. Investigate using additional negative elements in a non-Foster
matching circuits
2. Determine the optimal tradeoff among the design parameters to
obtain the largest improvement in signal-to-noise ratio over the
broadest bandwidth
3. Develop additional types of negative circuit elements, especially
negative inductors for electrically-small loop and flush cavity
antennas
4. Acquire or develop accurate device models to design low-noise FET
NICs
5. Auto-tuning / self-adjusting circuitry
6. Investigate alternative matching network topologies NIC
bracketing vs. individual negation of elements
7. Transmit applications a special problem