a small mount loss and large bandwidth at frequencies of the

order of 3 GHz and above.

Assuming a d.c. resistance of 200Q. (the value generally
used in power measurements), the values of R^ in eqn. 1 are
125 Q. and 30 Q at frequencies of 3 GHz and 10 GHz, respec-
tively. If the resistance of the supporting wires connected to
the bead is about 3 Q, as suggested by the work of Collard,
Nicoll and Lines,
2
we would expect power measurements
made in terms of a d.c. calibration of the thermistor to be in
error by about 2-5% at 3GHz and 10% at 10GHz. These
figures agree well with the results of earlier experiments in
which thermistor milliwattmeters made in the UK were
compared with reference standards (water calorimeters,
1
and force-operated wattmeters).
6
An improved design of
thermistor with a smaller self-capacitance is obviously
desirable.
The work described was carried out at the Radio & Space
Research Station of the UK Science Research Council and is
published with the permission of the Director.
The author acknowledges the interest shown by H. M.
Barlow in these experiments. He is also indebted to J. E.
Pearson who supervised the experimental measurements.
J. A. LANE 11th April 1968
Radio & Space Research Station
Ditton Park, Slough, Bucks., England
References
1 LANE, J. A.: 'The measurement of power at a wavelength of 3cm by
thermistors and bolometers', Proc. IEE, 1955, 102B, pp. 819-824
2 COLLARD, J., NICOLL, G. R., and LINES, A. w.: 'Discrepancies in the
measurement of microwave power at wavelengths below 3 cm,
Proc. Phys., Soc, 1950, 63, pp. 215-216
3 LANE, J. A., and EVANS, D. M.: 'The design and performance of trans-
verse-film bolometers in rectangular waveguides', Proc. IEE, 1961,
108B, pp. 133-135
4 SOUTHWORTH, G. : 'Principles and applications of waveguide trans-
mission' (Van Nostrand, 1950), p. 104
5 BROWN, J. : 'Reactive effects in transverse-film bolometers in rect-
angular waveguides', Proc. IEE, 1963, 110, pp. 77-78
6 CULLEN, A. L. : 'A general method for the absolute measurement of
microwave power', ibid., 1952, 99, pt IV, pp. 112-120
CLARKE COMPONENTS FOR A SYSTEM
OF BUSBARS WITH PRONOUNCED
PROXIMITY EFFECT
The equations governing multiple straight transmission lines
are applied to a configuration of three lines in the same plane.
By a transformation with a matrix T, corresponding to the
Clarke components, the propagation coefficients and modes
can easily be found. A numerical example is given of three
busbars showing pronounced proximity effect. The example
is calculated by analogue simulation on a capacitor-resistor
network.
In a very comprehensive paper by Snow,
1
the Maxwell
equations were applied to multiple straight transmission lines
and solved for the steady state by integral equations. It was
shown that, under certain assumptions, the transmission-line
equations are
(1)
Z and Y are quadratic matrices of impedance and admittance
per unit length, / and U are column matrices of current and
voltage, and z is the co-ordinate parallel to the axes of the
lines. Snow proved further that the eigenvalues of YZ are the
squares of the propagation coefficients y, of the waves, and
that the general solution for /can be built up by superposition
of the eigenvectors of YZ corresponding to these eigenvalues.
In this letter, an application of the theory of Snow will be
188
made for the case of three solid busbars encircled by a
perfect cylindrical earth of very large diameter (Fig. 1). The
earth has zero voltage and conducts the earth-return currents.
Fig. 1 might represent a vertical section through a 3-phase
busbar system for heavy currents and low voltages.
From the symmetrical configuration relative to the middle
conductor in Fig. 1, it is recognised that the Z and Y matrices
contain only four different complex elements each and may
be written in the form
Z =
'Zxx
Z
X
2
i2
^-12
Y
l2
Y22
Y
l2
^ 1 3
ZxxJ
Y
l3
Y
i2
YnA
(2)
If we assume a perfect dielectric (air), each element of Z will
be complex, and each element of Y will be purely imaginary:
Z
ik
=
R
,k
Y
ik
= jwC
ik
ik "11
J
(3)
If the conductors are at small mutual distances, as in Fig. 1,
the elements of R
ik
form a full quadratic matrix. When
c > a, it reduces to a diagonal matrix with all R
n
equal.
Because, in busbar installations, the conductors are often very
close together, we consider the influence of the current
distributions in the different conductors on each other, i.e.
the proximity effect. The proximity effect also has its in-
fluence on the inductances, but there it is rather smaller.
The equations of Snow
1
for the Z
jk
can be solved by
analogue simulation on a capacitor-resistor network.
2
-
3
Because of the symmetries of the configuration of Fig. 1 with
respect to the x and .yaxis, only one quadrant of the x-.yplane
needs to be simulated. The following combinations of Z
ik
and
Y
ik
are found to be the most practical for evaluation:
z
c
= z
22
Z
d
= Z
l{
]3
13
(4)
2Z
22
- 4Z
12
and
Y
b
= Y
U
- y
13
Y
c
= Y
22
Y
cl
=2Y
n
+2Y
l3
+Y
22
(5)
The analogue quantities can be directly measured in the net-
work. It is observed that the expression for Y
d
shows a slight
difference from the corresponding expression for Z
d
. If the
matrices Y and Z are known from eqns. 4 and 5, the eigen-
values and eigenvectors of YZ and ZY, respectively, can be
directly calculated for the case considered.
4
But, as Wedepohl
has shown,
5
the eigenvectors lie, for transmission lines in a
horizontal plane, near to the following normalised vectors
( = columns of T):
T =
1
V3
1
LV3
A/6
2
V6j
(6)
The elements of one row represent the voltage or current for
one conductor, and the elements of one column represent the
voltage or current for one mode of propagation.
1
'
4
When
treating the transmission-line equations of a horizontal-line
configuration, it is therefore advisable to transform eqn. 1
by the substitution:
(7)
U= Tu
1= Ti
ELECTRONICS LETTERS 17th May 1968 Vol. 4 No. 10
Authorized licensed use limited to: Green Hills Engineering College. Downloaded on April 29, 2009 at 21:58 from IEEE Xplore. Restrictions apply.
where u and / are the (slightly modified) Clarke or Concordia
components. If the equations are brought into the form of
eqns. 1, the following transformed impedance z and admit-
tance y matrices result:
From eqns. 4 and 5, the Z and Ymatrices can be calculated,
and the z and y matrices are calculated from eqn. 8. Their
product zy becomes:
= T~
l
ZT =
y=
2Z
n
+ 4Z
I 2
+ 2Z
)3
+ Z
22
3
0
\ / 2( Z, , Z
12
+ Z,
3
Z
22
)
1
u
- t - Hi ,
2
- t - Z J ,
3
- t - I
22
3
0
V2(Y
U
-Y
n
+Y
u
-Y
22
)
V
2
(Z
1
1 - Z
12
+ Z
13
- Z
22
)
3
0
^ii - 4Z
12
+ Z
13
+ 2Z
22
V2(Y
U
-Y
n
+Y
l3
-Y
22
)
3
0
Y
u
- 4Y
i2
2F
22
(8)
It is seen that z
22
= Z
b
, z
33
= (1/3) Z
d
, y
22
= Y
b
and_y
n
=
(1/3) Y
d
. The other elements have to be calculated from linear
equations. The matrices are symmetrical and have four
elements each that disappear. If the other nondiagonal
elements had also disappeared, T would be the exact matrix
of the eigenvectors. (This is a sufficient, but not necessary,
condition.) In order to find the approximation that T gives
for the case of Fig. 1, we consider a numerical example.
shield
radius
1
Fi g. 1 Vertical section through 3-phase busbar system with
centres of bars all in same plane
Three busbars of section 6-4 x 0-8cm and spacing
c = 0 8cm are simulated on the analogue network
2
-
3
with mesh width A = 0-1 cm. Y and Z are calculated
for the frequency 340Hz. The d.c. resistance of one bar
/?(0) = 317 x 10~
5
Q/m. The following results were
obtained: Z
a
= 2 20R(0) + jajfji
0
1-90, Z
b
= 2-59/?(0) +
0 121, Z
c
= 2-48/?(0) + ya^
0
1 00, Z
d
= 4-42/?(0)
0
0 147, Y
a
= jwe
0
9-82, Y
b
= ja>e
0
11-45, Y
c
=
joj
Q
18-81, Y
d
= ju)
0
1 071.
ELECTRONICS LETTERS 17th May 1968 Vol. 4 No. 10
P = Zy =
-f- j
"1013 0 - 0 0 5 2
0 1-385 0
a
.0 002 0 1-381 J J
1-227 0 13-30] I
0 29-66 0 -i
0-425 0 41-64J 1
The numerical values of the common factors are aj
2
e
O
ju.
o
\
5-06 x 10- " m-
2
and jcoe
0
R(0) = j5-96 X 10-
13
m~
2
3
The propagation coefficients then become y
x
= 5 16 x 10~
8
+
y'7-16 x 10-
6
m-
1
, y
2
= 105 x 10~
6
+y8-44 X l O^m-
1
,
y
3
= 1-46 X 10-
6
+y8-48 X 10~
6
m-
1
.
The matrix of the eigenvectors of p is
t
u
=
i
where /x and v are the necessary correction terms for the easel
when T is not the exact matrix of the eigenvectors of ZY. In j
our example, they become /x = p^xKPw P33) = 8-61 x i
IO-
3
+y"2-45 x 10~
3
andv= -PUKPH - P33) = 0 1 5 3 - }
y'0-228. The transformation T
u
= Tt
u
finally gives the eigen-1
vectors of ZY. These eigenvectors of voltage become <
'1
0
n,
0
1
0
V
0
1_
J
u =
1
73
1
73
1
. 73
-
+
[X
76
76
fj.
76
The corresponding solution
which
ti =
results
1
0
v
in
0
1
0
-fl
0
1
1
72
0
1
"72
for the
1
76
H
1
76
H
current
V
73
V
"73
V
\ / 3
is
1 V 1 1 /X
\/3 \/6 \/2 \/6 \/3
J-4-^L 0 - -
J
.73
1 1 _ jx.
72 76 ~ 73 J
(12)
From the considered numerical example, it can be concluded
that the eigenvalues are, within the accuracy of the calcula-
tion, the square roots of the diagonal elements of p =
T~
l
Z YT, but the vectors of T must be corrected in the
manner described by eqns. 10 and 12. This conclusion is valid
189
Authorized licensed use limited to: Green Hills Engineering College. Downloaded on April 29, 2009 at 21:58 from IEEE Xplore. Restrictions apply.
for the general case (including proximity effect) of 3-conductor
transmission lines in a plane.
The application of the z and y matrices to the case of
a balanced 3-phase current system simplifies the calcula-
tions because only two modes i
2
and /
3
of the three possibilities
appear. The result for | /
t
| = |/
2
| = |/
3
| = 7
0
is
so that
/= T-
l
I=I
0
0
/3 A/3
/
3
V2
(13)
The series losses per unit length become, in general,
S
L
= I*ZI = (7i)*Z(77) = i*(T*ZT)i = i*zi (14)
where T* means the Hermitian transpose of T etc. Because
T is real and orthonormal, the relation T* = T' = T~
l
,
which was used in eqn. 14, holds. For the balanced 3-phase
current distribution, it finally gives
H.-J. SIGG 17th April 1968
M. J. O. STRUTT
Department of Advanced Electrical Engineering
Swiss Federal Institute of Technology
Zurich, Switzerland
References
1 SNOW, CH. : 'Alternating current distribution in cylindrical con-
ductors', 5c/. Pap. US Bur. Stand., 1924-26, 20, pp. 277-338
2 STRUTT, M. j . o., and VUILLEUMIER, R. : 'Analogieverfahren fur die
Bestimmung elektromagnetischer Wechselfelder in Leitern und
Halbleitern', Arch. Elektrotech., 1961, 46, pp. 259-276
3 SIGG, H.-J., and STRUTT, M. J. o.: 'Solutions of the Laplace equation
in an infinite plane by means of a double-layer resistance network',
Electronics Letters, 1966, 2, pp. 454-455
4 WEDEPOHL, L. M. : 'Application of matrix methods to the solution of
travelling-wave phenomena in polyphase systems', Proc. IEE, 1963,
110, pp. 2200-2212
5 WEDEPOHL, L. M. : 'Electrical characteristics of polyphase trans-
mission systems with special reference to boundary-value calculations
at power-line carrier frequencies', ibid., 1965, 112, pp. 2103-2112
1 f
+ co
Aw) = 0{pj)C*ico - AUJ) t-J
ATT J _ O T
ATT J _ O T
On differentiating the last equation with respect to CJ, we
obtain, after several simple operations,
Here ip(T,a>) is the measured spectrum of the f.m.-noise
correlation function, U
m
(t) is the amplitude of the signal,
G(w) is its power spectrum, T is the delay of the correlated
signal and co = l-rrfis the angular frequency.
Example: Let the measured spectrum have a shape
By the described method
~ 1-75
This result is the same as that which was derived by Hofstetter
3
for impulse signals.
j . POiivKA 17th April 1968
Paraha 3, Jagellonskd 25
Czechoslovakia
References
1 ZINOVEV, A. L., and FILIPPOV, L. 1.: 'Methods of analytic expression of
radio signals', Vysshaya Shkola, Moscow, 1966 (in Russian)
2 POUVKA, J. : B.S. thesis, CVUT Prague, 1966 (in Czech.)
3 HOFSTETTER, J. M. : 'Construction of time-limited functions with
specified autocorrelation functions', IEEE Trans., 1964, IT-10,
pp. 127-133
CALIBRATION OF AN OSCILLATOR-NOISE-
MEASURING SYSTEM
The calibration of an oscillator-noise-measuring system has
been a problem, chiefly when the f.m. component of oscillator
noise was measured. One of the best methods for such a
measurement is the correlation method. Using a spectrum
analyser, the correlation-function spectrum of the f.m. noise
component can be obtained. By the described procedure, the
original spectrum of the f.m.-noise component can be cal-
culated.
The envelope of a twin-dimensional correlation function is
given by
dt
\\U
m
(t)\
2
dt
Using the Fourier transform to describe the relation between
the signal and its spectrum, we can write
1 r
ZTT J
1 r
+0

m
(t - T) = C(co) e-*
2TT J_OO
1 r
m(
f
~
T
) = =-
+ 0

ZTT JOO
+ 0

+ 0

j'daj
PROPOSED GUNN-EFFECT SWITCH*
The possibility of using a Gunn device as a switch similar to
a tunnel diode is investigated theoretically. Three switching
types are obtained, one of which leads to the homogeneous
high-field state with predicted switching times of a few \R
N
\C
(corresponding to the homogeneous sample).
For Gunn-effect semiconductors, various kinds of application
possibilities have been published. However, to use the Gunn
device according to its typical current/voltage characteristic
as a switch similar to a tunnel diode has not yet been pro-
posed, and this will be investigated theoretically in this letter.
The problem is illustrated using the familiar static
characteristic () according to Fig. 1. A load line is assumed
which intersects the characteristic at three points (i), (ii), (iii).
Now, is it possible to switch the device from the stable point
(i) to point (iii) by applying a short trigger pulse of amplitude
{E'
B
E
B
)L1 Such a transition is expected, since point (iii),
lying again in a range of positive differential mobility, should
be stable (or at least quasistable) against domain formation.
During this process, however, the sample field has to go
through the range of negative slope, and hence a high-field
domain should build up. When such a domain has fully
formed, the sample is described by its dynamic characteristic
1
and the working point is given by point (iv), E
x
= E
xaa
being
the field strength outside the domain. After reaching the
anode, the domain will be quenched, and E
x
increases until
the initial working point (i) has been re-established.
In order to investigate the detailed behaviour of the device,
one has to start with equations which describe the dynamical
properties of the high-field domains :
2> 3
190
* Supported in part by the Ministry of Defence, Federal Republic of Germany
ELECTRONICS LETTERS 17th May 1968 Vol. 4 No. 10
Authorized licensed use limited to: Green Hills Engineering College. Downloaded on April 29, 2009 at 21:58 from IEEE Xplore. Restrictions apply.