Theory and performance of cavity

stabilised microwave oscillators

R. Knochel, K. Schiinemann and J-D. Biichs
Indexing terms:

Microwave oscillators, Frequency stability, Cavity resonators

Abstract: A comparison is made between transmission- , reaction- and reflection-cavity stabilised oscillators.
Their essential features, such as maximum loaded fi-factor and tunable frequency range are shown to differ
markedly. Based on these considerations, outlines are given for designing highly stable microwave oscillators.
1

Introduction

Microwave radio-relay and communication systems require

a highly stable oscillator with extremely low noise. At present, available microwave oscillators are derived from a
quartz crystal normally utilising high-order multiplication
by a factor of n, thereby increasing the f.m. noise power by
a factor of n2. A crystal-controlled oscillator is thus characterised by a good long-term stability (frequency drift with
temperature change) and a bad short-term stability. Replacing these crystal-controlled local oscillators in transmitters
and receivers by solid-state oscillators, which directly
generate the desired frequency, puts severe requirements on
their performance: the frequency stability against temperature changes must be high, and the f.m. noise level must be
low. In the case of a Gunn oscillator, e.g. that which inherently shows low a.m. noise, f.m. noise or short-term stability as
well as long-term stability are insufficient. The frequency
stability of an ordinary Gunn oscillator (with the Gunn
element postcoupled to a rectangular waveguide) arid the
required value of stability in the communication systems
are in the order of 10~3 and 10~5, respectively, over the
temperature range from 20C to +50C.
The most practical solution to this problem is the well
known cavity-stabilisation method. The oscillator is
coupled to a high-Q resonator, which, at the sacrifice of
output power, determines the stability of the total system
to a large extent. The essential advantages of this method
are simplicity, low cost and small size. However, several
practical problems, such as mode jumping during operation
as well as in the instant of switching, frequency adjustability,
decline of output power over the tunable-frequency range,
and presence of parasitics and harmonics in the output
spectrum, must be solved. Most of these problems are caused
by the presence of two resonant circuits: the oscillator resonant circuit and the stabilisation cavity. Therefore, these
oscillators may show two possible oscillation modes, and,
accordingly, frequency hysteresis over the pulling range.
Depending on the arrangement of oscillator, stabilising
cavity and coupling network, one must distinguish between
transmission- , reaction- , and reflection-cavity stabilised
oscillators (see Fig. 1). The transmission cavity is a 2-port
network that couples the oscillator to the load. The reaction
and reflection cavity are both 1-port networks that only
differ in their coupling to the oscillator. A reaction cavity is
inserted between oscillator and load, reflection cavity and
Paper T73M, first received 23rd February 1977
Dr. Knochel and Dr. Schiineman are with the Institut fur Hochfrequenztechnik, Technische Universit'at, Posfach 3329, 3300
Braunschweig, and Dr. Biichs is with AEG-Telefunken, Gerberstr.
34, 7150 Backnang.
MICROWA VES, OPTICS AND ACOUSTICS, JUL Y1977, Vol. 1, No. 4

load couple to opposite parts of the oscillator. In this latter

configuration, the waveguide short circuit of the oscillator
is replaced by the reflection cavity and its coupling line.
Hence, in the case of a postcoupled waveguide structure,
the oscillator resonant circuit is formed by the package of
the active device and the post alone, while for the other
stabilisation schemes the waveguide short circuit mainly
determines the oscillation frequency when the stabilising
cavity is detuned. As an import consequence, the loaded Qfactor of the unstabilised reflection cavity oscillator differs
from that of the two other oscillators. This must be
accounted for when making a comparison between the
three types of stabilised oscillators.
It is the aim of this paper to draw such a comparison,
and to show the partly considerable differences in maximum obtainable stabilisation factor, output power, and
tunable-frequency range. Based on these considerations,
outlines will be given for designing highly stable microwave
solid-state oscillators.
Although the technique of cavity stabilisation is well
known in the literature, the question of which type of a
cavity-stabilised oscillator meets the requirements of a
special application best is as yet unsolved. This is mainly
due to some contradictions in the work published so far on
this subject, and to erroneous conclusions that had partly
been drawn. Hence, it seems justified to start the investigation with a brief discussion of previous work.
2

Review of literature

The review of previous work on cavity-stabilised microwave

oscillators shall mainly be done in chronological sequence.
The first work that dealt extensively with cavity-stabilised
oscillators was that of Shelton.1 Based on a lumped-element
equivalent-circuit representation of the microwave network,
the essential features of a transmission- and a reflectioncavity stabilised oscillator have been worked out. Furthermore, it is pointed out that a damping resistor has to be inserted into the coupling line to the cavity to suppress unwanted modes of operation, which are due to the extra resonant circuit formed by the coupling line. The amount of
frequency stability is measured by the stabilisation factor

Fig. 1

Cavity stabilised oscillators. Qo is unloaded Q-factor

143

S, which relates the loaded 0-factor of the stabilised oscillator to that of the unstabilised one. This definition turned
out to be a source of erroneous comparisons made in later
investigations, because S is only a relative measure of stability, as it depends on the loaded 0-factor of the unstabilised
oscillator. Lowering this (?-factor leads to an increase in S
(the frequency stability of the original oscillator is then
deteriorated) without improving the absolute value of the
frequency stability in any way. Hence, comparing S of various oscillators can lead to wrong conclusions. Here, use of
the loaded Q-factor, instead of S, will be proposed as an
absolute measure of attained frequency stability.
A reaction-cavity stabilised oscillator has been introduced
in Reference 2. In spite of the fact that a damping resistor
is not necessary in case of a reaction cavity (for the load
itself fulfills its function), such a resistor has been used in
Reference 2. Hence, the output power is unnecessarily reduced. The damping resistor has been omitted in Reference
3, and the result is a well designed oscillator with reaction
cavity; the output power is maximum and the tuning
characteristic, i.e. the dependence of the oscillation frequency on the mechanical tuning of the stabilising cavity, is
single valued in the interesting frequency range.
In Reference 4 some design hints are given for a transmission cavity stabilised oscillator, but they will not lead to
an optimally designed system, because both the power loss
in the damping resistor and the circuit losses in the oscillator (the diode-mounting structure) have been neglected in
the calculations, although they amount to, typically, 2 dB.
Hence, the tuning condition of Reference 4, a zero-input
reflection coefficient of the stabilising cavity, will indeed
lead to a nonoptimally tuned oscillator. As will be shown
later, it reduces both output power and loaded Q-factor.
Furthermore, there are no hints in Reference 4 concerning
the minimum amount of damping necessary for a single
valued tuning characteristic over the frequency range of
stabilisation.
A reflection-cavity stabilised oscillator has first been
developed in Reference 5. As the tuning characteristic of
this system was double valued, two modes of oscillation
existed; one with a low stabilisation factor and one yielding
a high 5-value. This was the first time that such a characteristic had been established. Unfortunately, some doubtful
conclusions have been drawn in Reference 5, which should
be held mainly responsible for the existing contradictions in
the field of cavity stabilisation of oscillators. The most important were as follows:
(a) to get any stabilisation effect there should be a difference between the natural frequencies of the cavity and
the diode mounting structure
(b) stabilisation increases with increasing frequency difference
(c) a reflection-cavity stabilised oscillator is superior to
any other system, for it yields the same stabilisation at the
sacrifice of only a smaller fraction of the output power, as
compared to a transmission or reaction cavity stabilised
oscillator.
A brief discussion of these statements is necessary. As will
be shown later, the double-valued tuning characteristic of
Reference 5 is only due to both a nonoptimum length of
the coupling line between oscillator and cavity and the
absence of a damping resistor. Statements (a) and (b) above
are correct only for that special configuration. In the case
of a properly designed reflection-cavity stabilised oscillator
with a single-valued tuning characteristic the statements do
144

not hold any longer. Needless to say, a double-valued

tuning characteristic is not desirable in any respect.
Statement (c) above is incorrect, and it will be shown in
this contribution that differences do exist in the performance of reaction-, transmission-, and reflection-cavity stabilised oscillators, but they are not so weighty as stated in
Reference 5. The advantages are not uniquely confined to
the reflection-cavity stabilised oscillator, but are equally
distributed among the three systems, e.g. replacing a reaction or reflection by a transmission cavity means handling a
large loaded (^-factor for an increased frequency range of
stabilisation.
In the following, reflection-cavity stabilised oscillators
have been developed showing a single-valued tuning characteristic. This was mainly due to the introduction of a damping resistor6' 7 or to a proper design of the coupling network.8' 9 However, the statements cited previously have not
been criticised or even corrected in any of these contributions. Instead, statement (c) is believed to be of general
validity. Nevertheless, other stabilisation schemes have been
preferred for local oscillators in modern radio link systems.10""12 These oscillators have been developed based on
an analytical treatment of an equivalent circuit. Their performance is, however, not optimum, because their
equivalent circuits have been oversimplified.
A comparison between the performance of oscillators
with reaction and transmission cavities has recently been
drawn13 based on a common equivalent circuit. Although
this contribution shows a way to successfully attack the
problem of cavity stabilisation, it fails in some of the following respects:
(i) the equivalent circuit does not allow the description
of an oscillator with reflection cavity
(ii) the circuit losses of the diode mounting structure
have been neglected
(iii) a general tuning condition cannot be established because the resonant circuit of the unstabilised oscillator has
also been neglected
These disadvantages will be overcome in the following investigations.

Development of a common equivalent network

To compare the performance of reaction-, reflection-, and

transmission-cavity stabilised oscillators, a common equivalent circuit for the three structures will now be developed.
An inspection of Fig. 1 shows that a cavity stabilised oscillator consists of the following three components: the
original oscillator or the diode-mounting structure, the
high-Q cavity and the coupling line with damping resistor.
Each component is a resonant structure, which shall be
modelled by a lumped element .RZ,C-circuit over a limited
frequency range. This is a simple task for both the cavity
and the coupling line but might be ambigous in the case of
the diode-mounting structure. Hence, the latter case will be
treated in more detail.
An active element postcoupled to a rectangular waveguide is believed to represent the most general oscillator
structure. (Either coaxial or m.i.c. arrangements are composed of fewer single elements and can thus be modelled by
an equivalent more easily). Two realisations of waveguide
oscillators can be found in the literature. In the first, the
waveguide is short circuited at a distance half a wavelength
X from the post, and in the second, this distance is only a
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. l,No.4

quarter of a wavelength. The first version will be regarded

now.
The diode-mounting structure, together with plots of the
electric and magnetic field, are sketched in Fig. 2. In the
case of a short X/2 distant waveguide, the magnetic field
strength is small in the region between the post and the side
walls of the waveguide, and so is the surface current density
at this part of the post (the surface current could be calculated according to Reference 14). Looking from the active
element D into the circuit, the post approximately appears
as a TEM-line which is a quarter of a wavelength long.14
The characteristic impedance of this inverter can be varied
by both post length and post diameter.

MM
"

'jj^

44/ftrxmkW

3invertor

invertor

Fig. 2

Waveguide mounting

a Electric field strength

b Magnetic field strength
c Simplified equivalent circuit
E = electric field strength
H= magnetic field strength

structure
waveguide wavelength
line conductance
loss conductance
negative conductance of
active device
I = current on the waveguide
post

\=
Go =
gp =
gj) =

D = active device

The diode-mounting structure can intuitively be modelled

by the equivalent circuit of Fig. 2c. The active element is
represented by a negative conductance gD that depends on
the r.f. voltage amplitude. A voltage dependent susceptance
has been omitted to simplify the calculations. As good
agreement was found between theory and measurement,
which could be established later, this seems to be justified.
The same is true for the frequency dependence of g>,
which is weak as compared to that of the passive components and can thus be neglected. The active element is
loaded by a parallel resonant circuit of conductance gp,
which represents the circuit losses in the fields around the
post gap. This loss conductance is of extreme importance in
the analysis of cavity stabilised oscillators and can by no
means be neglected. (Otherwise, this will lead to a nonoptimum tuning of the whole structure.) In fact, the circuit
conductance can absorb a substantial part of the generated
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. l.No. 4

power, as has been proved in Reference 15. The circuit

losses range between 0-5 and 3 dB of the generated power,
the higher values being valid for coaxial or m.i.c. arrangements. It was mainly due to the neglect of these losses that
previous investigations of cavity stabilisation arrived at
incorrect conclusions.
For stabilisation, /f0m-mode circular waveguide cavities
are usually utilised. They can be represented by a single
resonant circuit near the natural frequency. A reflectiontype cavity (1-port terminating a transmission line) will be
modelled by a series resonant circuit, and a reaction-type
cavity (1-port discontinuity in parallel or in series to a
transmission line) by a parallel resonant circuit in series to
the coupling line, and a transmission-type cavity (2-port) by
a series resonant circuit in series to the coupling line. The
equivalent circuit for the reaction cavity is valid if the cavity
is hole coupled to the broad wall of the rectangular waveguide. Otherwise, coupling to the narrow wall must be
modelled by a series resonant circuit shunting the coupling
line.
In the equivalent circuits presented previously, the coupling reactances have been neglected owing to the high un
loaded Q-factors of the cavities. Then, the circuits are completely described by their unloaded Q-factors Qo and the
coupling coefficient /3 for the reaction and reflection cavity
and the input/output coupling coefficient /3i/j32 for the
transmission cavity. )3 describes the standing-wave ratios at
the respective ports.
We are in a position now to develop a common equivalent
network for the three types of cavity stabilised oscillators.
This will only be done for oscillators with a single-valued
tuning characteristic, for it is believed that oscillators with a
double-valued tuning characteristic will hardly be applied in
practical systems. To arrive at a single-valued tuning characteristic the coupling lines for the oscillator structures of
Fig. 1 must be a quarter of a wavelength in the case of a
reaction cavity, and half a wavelength long for the cases of
a reflection or transmission cavity. This holds as long as the
active element is loaded by a parallel circuit (thick post in
the case of a waveguide mounting structure). In the cases of
reflection- and transmission-cavity stabilised oscillators, a
damping resistor has to be inserted in the middle of the
coupling line. If the cavity is detuned, the X/2 coupling line
is terminated by an open circuit. Then the line input impedance behaves like a parallel resonant circuit, whose loss
conductance is mainly due to the damping resistor.
To construct the equivalent network of a cavity stabilised
oscillator the circuits for the various components as developed above have to be combined. Referring to the active
element, one arrives at the common equivalent circuit of
Fig. 3. According to the three components of the oscillator
structure it consists of the following three resonant circuits:
the first describing the diode-mounting structure, the
second for the cavity and the third for the energy storage in
the coupling line. The various circuit elements have the
following meanings (see Table 1):
(a) Transmission-cavity stabilised oscillator: The unloaded
(2-factor of the original oscillator is QH. The loss conductance gH represents the losses in the diode mounting structure gD and those of the X/2 waveguide short circuit. The
series circuit with Q-factor Qt is due to the coupling line,
and gDA represents losses in the coupling line and mainly
in the damping resistor. The unloaded (2-factor of the transmission cavity transforms to Qc = Qo/(l + j32) at the cavity
input port. Hence, gc = (1 + foVft contains the transformed
load j32/0i and the cavity loss conductance l/0i.
145

Table 1 . Circuit elements

Oscillator with specified cavity type

9H
9DA
9C
Qc

Reaction

Reflection

Transmission cavity

9p
9L
1

9L+9P

9p

9-DA
1

9DA
1 + 02

Qo

Qo

01
Qo

The diode mount, which was first introduced in Reference

11, together with its simplified equivalent circuit, is
sketched in Fig. 4, and also shown are plots of the electricand magnetic-field strength.
The diode mount shows three characteristic features: a
thin post, a quarter wavelength distant waveguide short and
a bias filter that has been removed from the top of the
waveguide leading to a lengthened post. As for the former

1 +02

(b) Reflection-cavity stabilised oscillator: QH is the

loaded Q-factor of the original oscillator. However, differing
from the previous case, gn represents the sum of the load
gL and the circuit loss conductance gp. In the series circuit
SDA > Qt represents the transmission line, the parallel circuit
is gc and Qc represents the stabilising cavity withgc = 1/0
and Qc = Qo.
(c) Reaction-cavity stabilised oscillator: QH and gH have
Fig. 3 Common equivalent circuit for cavity stabilised oscillators
the same meaning as for the oscillator with a transmission
The meaning of the circuit elements can be read from Table 1
cavity, Qc and gc represent the stabilising cavity as for the
yi, = load admittance
oscillator with reflection cavity. Now Qt -> 0 and gDA repgp = loss conductance
gL = load conductance
resents the load conductance that simultaneously fulfils the
gj) == negative conductance of active device
function of a damping resistor vnthgDA =SLgDA conductance of damping resistor
Regarding the common equivalent network for cavity
0, 0i 02 = cavity coupling coefficients
Qo= unloaded Q-factor of stabilising cavity
stabilised oscillators, the following conclusions can be
drawn: the whole oscillator structure is composed of three
parts, and hence the simplest equivalent network must represent these resonant structures by three suitably coupled
resonant circuits. As three positive conductances do
exist in the equivalent network, any output power extraction can be performed in three different ways. A special
oscillator configuration belongs to each method. If the
output power is extracted from the cavity, we have a
transmission-cavity stabilised oscillator. If it is extracted
from the original oscillator circuit ##, the oscillator is stabilised by a reflection cavity. In the case of an oscillator with
a reaction cavity, power extraction is performed through
the location of the damping resistor gDA. These generalisations are proposed here as a suitable means for defining
the three possible configurations of cavity stabilised oscilators. They are superior to the method used until now,
where the localisation of the cavity to the original oscillator
was used to distinguish between a reaction and a reflection
cavity, because this method may fail in the case of a more
complicated geometric configuration.
In the light of the new definition a reflection-cavity
stabilised oscillator that showed a single-valued tuning
characteristic has not been built before. Instead, the so-called
oscillators behaved, electrically, like oscillators with reaction
cavity. As will be seen later, an oscillator with reflection
cavity (output power extraction from the circuit of the
original oscillator) represents a compromise between the
electrical performance of a transmission- and a reactioncavity stabilised oscillator in that it yields intermediate
Fig. 4 Waveguide mounting structure with thin post
loaded Q-factors and tunable-frequency ranges. Hence, for
a Electric field strength
many applications as local oscillators the reflection-cavity
b Magnetic field strength
stabilised oscillator exhibits a useful alternative.
c Simplified equivalent circuit
Every cavity stabilised oscillator with single-valued
E electric field strength
tuning characteristics can be described by the common
H = magnetic field strength
D = active device
equivalent circuit of Fig. 3 or by its dual circuit. The latter
\ =. wavelength in waveguide
is suitable if the active element is driven in a series resonant
\ 0 = wavelength in coaxial guide
circuit. This description if often found in the Japanese
Go = line conductance of waveguide
Gl = line conductance of coaxial guide
literature. In a waveguide-mounting structure, the wavegj) = negative conductance of active device
guide short then is a quarter of a wavelength off the post.
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1, No. 4
146

structure, the post might be regarded as a TEM transmission

line whose length measures almost half a wavelength Xo
The distribution of the electric field lines suggests a simplified equivalent circuit, in which the quarter-wavelength
waveguide resonator is coupled to the centre of the half
wavelength TEM resonator (Fig. 4c). Hence, the active
element is loaded by a series resonant circuit.
Cavity stabilised oscillators with a diode-mounting structure, as shown in Fig. 4, can be modelled by the equivalent
network of Fig. 5, which is a dual circuit to that of Fig. 3.
The original oscillator is represented by the series resonant
circuit QH, which is gH of the X0/2-long TEM-line. A
stabilising cavity might be coupled to the maximum x or zcomponent of the magnetic field strength requiring a coupling line, which is a quarter or a half wavelength long,
respectively. The cavity can be represented by the series
resonant circuit Qc, with gc in parallel with the active
element. The short-circuited quarter-wavelength waveguide
stub is modelled by the parallel circuit with conductance
SDA due t 0 a damping resistor. This holds true for an
oscillator with transmission cavity, while, in case of a
reaction cavity, the function of the damptin resistor is
fulfilled by the load conductance gL, which is equivalent to
gDA. An oscillator with reflection cavity cannot be realised
because the output power cannot be extracted from gH.
Hence, the equivalent circuit of Fig. 3 is more general and
will be exclusively regarded in the following. (The dual
circuit of Fig. 5, of course, yields the same results
concerning the loaded 0-factor and the output power of
reaction and transmission cavity stabilised oscillators.)

mined from eqn. 1 by assuming a positive (negative) disturbance A# of the voltage amplitude. If a turns out to be
negative (positive) the operating point is stable, but otherwise it is unstable.
This procedure has to be applied to the equivalent network of Fig. 3. For simplicity of illustration (but without
any loss of generality) the device line will be assumed to be
real, so that ;>D(#)=#!>(#) Then, yo(v) lies on the positive part of the real axis in the admittance plane. It will be
further assumed to monotonically decrease against i>. The
load line shows one or two loops, depending on the relative
values of both, the unloaded ^-factors and the natural frequencies of the various resonant circuits. The unloaded
(2-factor Qo of the stabilising cavity dominates, and it
usually exceeds 10 4 ; the unloaded 0-factor Qt o f the coupling line is an order of magnitude smaller. It can be estimated by a formula for a TEM-transmission line resonator
yielding17 Qt = ir/(oc\) with X the wavelength and a the
damping constant. Assuming a = 0 - 04irf ! For brass yields
a frequency of 15 GHz, then Qt ^ 3 000. The Q-factor QH
of the diode-mounting structure is lowest in order to allow
a broadband matching of the active device, and it lies in the
range of 10 to 10 3 .
The natural frequencies of the three resonant circuits are
coH, cot and to c . Here, CJH = cot = a constant, while toc is
(mechanically) tunable. To get any noticeable stabilisation,
coc has to be tuned to about coH. In case of toc = uH, the
load line shows two loops within one another, the inner
loop being due to the stabilising cavity. Unfortunately, an
operating point will line on the outer loop, which is due to
the coupling line. Hence, the stabilisation effect is poor.
Slightly detuning the cavity to toc ^ ton, however, leads to
a double-valued tuning characteristic.5 A remedy in this
situation is a damping resistor in the middle of the coupling
line.1 If its loss conductance gDA (Fig. 3) is small enough,
the series resonant circuit can be neglected and the additional loop in the load line disappears. To get a design
formula, this statement has now to be formulated
quantitatively.
An inspection of the equivalent network of Fig. 3 shows
that

Fig. 5 Common equivalent circuit for cavity stabilised oscillators

The network is dual to that of Fig. 3

dej

Function of the damping resistor

The common equivalent network of cavity stabilised oscillators contains three resonant circuits, and, in the case of a
reaction cavity, it contains two resonant circuits. Hence,
several possible oscillation modes may exist. To check their
stability one should proceed as in Reference 16. Any operating point of an oscillator must fulfil the node equation
(Fig. 3):
= 0

0)

Here, yoiv) means the device admittance depending on the

r.f.-voltage z> and the load admittance .yi,(to), respectively
(which have been normalised to a wave impedance). The
frequency depedence of the device admittance is weak compared to that of the load admittance of a cavity stabilised
oscillator and may hence be neglected. Eqn. 1 is usually illustrated in the complex admittance plane showing the
'device line' 7 D ( # ) and the 'load line' J>L(W). TO check the
stability of a common point of both, the frequency is
replaced by a complex quantity to foe. Then a is deterMICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1, No. 4

LJt

(2)

must be fulfilled so that a loop due to the series resonant

circuit disappears. In eqn. 2 bH means the susceptance of
the .//-circuit, and bt that of the t-circuit. Approximating
these susceptances near the natural frequencies uH = to,
by
bH =

2QHgH

to

COH

and
gtgDA

(gt+gDA)

to,

(3)

and inserting into eqn. 2 yields

SDA < \/{gtgHQHlQt) = gent

(4)

In eqn. 4 use was made of gt

The influence of the damping resistor on the shape of
the load line is shown in Fig. 6. In the case of a detuned
stabilising cavity (Fig. 6a) the load line is almost parallel to
the imaginary axis for to =* CJH if gDA ^.gcrit has been
chosen. More important is the case of to// = to, = to c (Figs.
147

6b and 6c). A damping conductance less than the critical

value of eqn. 4 then leads to a 'susceptance gap' along the
real axis, giving rise to a single-valued tuning characteristic
of the stabilised oscillator around an operating point A.
Provided that the coupling line is sufficiently damped
according to eqn. 4, its resonant circuit might be neglected
in the equivalent network of Fig. 3. Hence, the common
equivalent network of cavity stabilised oscillators with
single-valued tuning characteristic consists of a parallel resonant circuit in parallel with the conductance of the active
device, which are both in parallel with the series connection
of a conductance gDA to the stabilising parallel resonant
circuit.
Concluding this Section, a few remarks will be given concerning cavity stabilised oscillators with double-valued
tuning characteristic. They can be deduced from oscillators
with single-valued tuning characteristic by omitting the
damping resistor and by lengthening or shortening the
coupling line by a quarter wavelength. Then a common
equivalent network can be found that consists of the parallel connection of the admittance of the active device with a
parallel resonant circuit (the //-circuit of Fig. 3) and with a
series resonant circuit (the C-circuit of Fig. 3). As above, gc
contains the load of a transmission cavity-stabilised oscillator. A reaction- and reflection-cavity stabilised oscillator,
however, cannot be distinguished any longer as the load is
contained in gH now for both configurations. Analysing the
equivalent circuit yields a loaded 0-factor of the stabilised
oscillator that monotonously increases with increasing frequency difference (coc COH) until mode jumping occurs.
More detail of this is given in Reference 5.
5

Analysis of equivalent network

The normalised load admittance of the common equivalent

network (Fig. 3) is given by

In eqn. 5 the normalised frequencies 2H and lc are

approximated by the linear term in the following Taylor
series expansion:
= 2QH

CO*

COH

(6)

The oscillation frequency co0 is given by ImyL(a>0) = 0.

Making use of
g A
^ <l
[gDA +gc

(7)

this yields
gDA

gH\gDA +gc

CO = C O ,

(8)

,_ I
gH\gDA +gc

Load and device lines are plotted approximately in the

admittance plane in Fig. 7. Here the shape of the admittance
loop is assumed to be circular. The stable operating point
sketched in this figure is point A. By detuning the stabilising
cavity, the oscillation frequency can be varied between
points B until mode jumping to an unstabilised operating
point occurs. Denoting the oscillation frequency at points B
by cos, an inspection of Fig. 7 yields
(
- T \gDA

2y

gcgDA \
gc+gDA)

(9)

Eqn. 9 can be solved for Acos, the hold-in range, as follows:

Acos

= 2

gDA

(10)'

gc+gDA

COH

gDA

= gH +gDA

gDA + gc

If the oscillation frequency is tuned from an operating

point on the linear branch of the load line towards the
operating point A, mode jumping occurs now at points C.
Denoting the oscillation frequency at points C by cof, the
lock-in range Aco/ = 2(cof co//) can be determined by
solving

gDA +gc
2

gDA

gDA +gc

1+

gA

gDA +gc

(5)

c?[Im,yL(co)]
dco

= 0

(11)

crit

y L (u)

Fig. 6

Plot of load line yi,((jS) in the admittance plane

a Cavity detuned
b Cavity tuned c Cavity tuned
8DA = conductance of damping resistor
gCrit = critical conductance, defined in eqn. 4

148

MICROWA VES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1, No. 4

. It yields approximately

for

2gDA

\ZigHQHgcQc)

The frequency stability of a cavity stabilised oscillator is

usually measured by the stabilisation factor S, which is
defined by

Inserting y^ from eqn. 5 and setting coc = COH yields

approximately
/
gc I
gc

(13)

V2
gDA

+gc

(18)

gDA

gH gDA + gc

It relates the stored energy W8tab of the stabilised oscillator

to that of the unstabilised ocillator Wunstab. Making use of
the generated power Pgenf eqn. 13 can be rewritten:

Pgen

(17)

W,
stab
S =
W,
unstab

S =

QL

(12)

gen

QL stab

tOo Wunstab

QL unstab

(14)

Hence, S turns out to be the ratio of the loaded 0-factors

for the two states. From eqn. 14 it can be seen that S is but

The analysis of the equivalent network is completed by a

calculation of the various dissipated powers. To proceed in
this, the r.f.-voltage amplitude-dependent device conductance has to be specified. For purposes of either generality
or simplicity, a van-der-Pol-type current/voltage characteristic will be assumed. Then the device conductance is described by
gD(v)

(19)

with k being a constant. To maximise the generated power,

the active device has to be matched according to
gD{V) =

(20)

leading to

v2 =

g D (0)-Re.y L (coo)

(21)

Then the generated power can be written as

(22)
Fig. 7 Plot of device line yo(v) and load line ^L(CJ)
Shown in the admittance plane for determining the hold-in and
lock-in ranges, where v is the r.f.-voltage amplitude

a relative measure of frequency stability, as it depends on

the loaded (^-factor of the unstabilised oscillator. Lowering
this <2-factor leads to an increase in S without improving
the absolute value of the frequency stability in any way.
The problem is solved if the loaded Q-factor of the
stabilised oscillator is used as an absolute measure of attained frequency stability. It can be defined by Reference
18:

The small signal conductance D ( 0 ) in eqn. 22 will be

expressed by parameters of the passive circuitry. For this
purpose, the generated power is written in terms of the
device conductance
Pgen
gen =

(23)

Now, setting the voltage amplitude derivative ofPgen, from

eqn. 23, equal to zero yields dmax, the voltage amplitude
that maximises the generated power. This yields
(24)

2k
QL

(15)
Then the resulting optimum load conductance is

In eqn. 15 CJO(W)represents the average power stored in the

stabilised oscillator and (dW/dh represents the average
value of the generated power. Setting

max

or

gd(0) = 2ReyL(coo)

(25)

8en

Substituting for^ D (0) in eqn. 22 from eqn. 25 yields

(16)
yields
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1, No. 4

Pgen

2 ReyL(uo)\max (26)
149

and hence for tke ratio pgem of the generated power to the
maximum generated power

a transmission cavity
Pout = Pc

Pgen

gen
p

', Pc.loss Pc

= 2

1+&'

(21)
Pgen of eqn. 27 will be used for normalising purposes.
For calculating the powers dissipated in the vammis conductances, eqn. 7 shall be assumed to he valid. Furthermore,
co0 = <*}c is t in order to aiiirce st simple expressions. The
power pH dissipated in tire conductance gjj is giveaby
gH

gen

gDA +gc

gc

+ gc
1

glgDA
{gDA

(29)

gDAgc

gH

gDA +gc
The power P dissipated in the conductance ge is
P

gX

gDA

+ Ic

gen

gcgD
DA

(gDA + gcf

<30)

SDAgc
gDA+gc

The powers are suitably normalised on the maximum

generated power of eqn. 27 as follows:

PN

- IK.

~ rp
Pgen
"""
gen

PDA

PDA = p Pgen
"gen

PC

Pc = p Pgen
rsen

(31)
Finally, the device conductance has to be matched. An
obviously favourable choice is to maximise the generated
power to the midband frequency ton = UJJ? = coc of the
jstafctifised frequency range. Then,
gDAgc

(32)

has to be inserted into eqn. 27.

The output power Pouu the loss in the stabilising cavity
Pc low and in the lass oondwetance of the original oscillator
Pa, ton different for the three oscillator configurations.
This yields, in the case of a reaction cavity,
Pout = PDA', Pc, loss Pc',

(33)

PfJ

a reflection cavity,
Pout = PH ' > PcXmr = Pc', Pu, loss = PH

fy
150

PH,

IOSSQH + Pc, IOSSQQ

(36)

(28)

gDAgc

Defining gx as the real part of the admittance right to tihe

port x x in the equivalent rretwork of Fig. 3, the power
PDA dissipated in the conductance gDA can be written as
gx

In concluding this section, a simple expression for the

loaded (Mactor of a cavity stabilised oscillator will be
given. It is of universal validity for the three oscillator
configurations.. Expressing the various conductances in eqn.
19 by normalised powers yields.

where Q^ is the unloaded (Mactor of the stabilising cavity.

gH

gen

PH

(35)

gen, max

nr

Performance of reaction-, reflection-, and

transmission-cavity stabilised oscillators

Sketches of oscillation frequency and output power against

cavity tuning are shown in Fig. 8. The tuning characteristic
oj0(ojcyis single-valued over the frequency range Aojf, and
deviations of the oscillation frequency from the natural frequency of the cavity are small. At the borders of the stabilised frequency range there exist two regions of hysteresis
due to Acos > Acjf. The shape of pout against CJC is different for a transmission (curve a) and for a reaction or
reflection-cavity (curve b) stabilised oscillator. In the case
of a transmission cavity stabilised oscillator, the output
power is zero beyond the tunable frequency range of stabilsation, and in the case of the other two oscillators it is
maximum here.
The curves of Fig. 8 are typical for a well designed
cavity stabilised oscillator. They do represent, however,
only a qualitative picture of oscillator behaviour. In fact,
the tunable frequency range of stabilisation differs greatly
for a transmission- and reaction-cavity stabilised oscillator,
for example.
For comparing the performance of reaction-, reflection-,
and transmission-cavity stabilised oscillators, the following
criteria will be used:
(i) the loaded (7-factor of the stabilised oscillator
(ii) the tunable frequency range of stabilisation
(iii) the output power
(iv) the maximum decline of the output power Apout in
a constant frequency range around the midband frequency
(v) the maximum decline of the loaded ^-factor AQL in
a constant frequency range around the midband frequency.
To guarantee a common basis for a comparison, the following points must be carefully fulfilled for the three oscillators:
(a) the active device has to be matched according to eqn.
32 so that the generated power is maximum
(fr) the unloaded Q-factor of the cavity must be the same
for aB oscillators. As a numerical example, Qo = 2 0 0 0 0
will be chosen here
(c) the damping resistor is chosen according to eqn. 4 in
ordeT to avoid an additional loop in the load line
(d) the diode mounting structiwe must be the same for
all oscillators.
Point (d) above must be regarded in more detail. The diodemounting structure (or the original oscillator) can be combined either with a- reaction or a transmission cavity without
changing its parameters. The unloaded Q-factor QH of the
//-circuit in Fig. 3 can then be determined from the loaded
MICROWA VES, OPTICS AND ACOUSTICS, JUL Y1977, Val / , No. 4

Q-factor of the unstabilised oscillator, because the

equivalent circuit of this oscillator consists of the //-circuit
in parallel to the active device conductance. Then, the
loaded Q-factor QLt umtab and the unloaded Q-factor QH
are related via
QH

, unstabjl + gp)

QL, unstabQ + gp)

gL+gp

(38)

as QH = 30. It greatly differs from the unloaded (^-factor

of the other two oscillators.
Concluding, it shall be stressed once more, that for a
realistic comparison of the performance of the various
cavity stabilised oscillators, different values for the unloaded Q-factor QH must be accounted for. For either a
reaction or a transmission-cavity stabilised oscillator QH has
the same value, but it has a different value in the case of a
reflection-cavity stabilised oscillator.
We are now in a position to evaluate the formulas of the
preceding Section and to compare the performance of the

(37)

A
12

10

out

\r> 6

02

06

08

10

Fig. 8 Variables plotted against cavity frequency

a Oscillation frequency
b Output power
curve a: stabilisation by a transmission cavity
curve )3: stabilisation by a reaction or a reflection cavity

Fig. 9 Loaded Q-factor against output power of cavity stabilised

oscillators
Parameters are coupling coefficients
reflection cavity QH = 30
~nnn
transmission cavity I _
reaction cavity
\QH = 2000

A typical value of a post coupled, rectangular waveguide

Gunn oscillator isgp =t)-l; QLi umtab n a s been measured by
injection locking. A special oscillator structure showed
QL, umtab = 180 leading to QH = 2000. This value will be
used in the following for the numerical examples. It is the
same for a reaction as for a transmission-cavity stabilised
oscillator.
In the case of a reflection-cavity stabilised oscillator,
however, the parameters of the original oscillator must be
changed as compared to the structure above, because the
waveguide short circuit is now replaced by the coupling line
with damping resistor and terminating cavity. The otignaT
oscillator (modelled by the /{-circuit) is now formed by
post and active-device aloae; the active element must oscillate with the post alone, the two waveguide ports being
terminated with matched loads. For this configurationQL, unstab = I $ h*s been measured. The unloaded Q-factor
Qn is then calculated via

three types of cavity-stabilised oscillators. The maximum

loaded Q-factor QL at GJ0 = uc = CJH is plotted against the
normalised output power pout in Fig. 9. Parameters are the
cavity coupling factors |3 (reaction and reflection cavity)
and /3 2 , /3r = |3i/(l +2) (transmission cavity), which increase in the direction of the arrows.
To complete the results, the tunable frequency range of
stabilisation Acjs/toH has been plotted against the output
power pout in Fig. 10. A striking res-ult is the low value in
case of an oscillator with reaction cavity as compared to the
other cases..
From the results in Figs. 9 and 10, the performance of
cavity stabilised oscillators can be characterised by the
following statements;
(i) Tnansmimon cavity stabilised oscillator
(a) large tunable frequency range of stabilisation)
(b) a high loaded ^-factor corresponding to low values
of output power, a medium loaded Q-factor at medium
output power levels

MICROWA VES. OPTICS AND ACOUSTICS, JUL Y1977, Vol. I, No. 4

151

(c)AQL and Apout are small because of the large

tunable frequency range.
Increasing /32 while holding 0 T constant yields higher values
of output power but lower values of the loaded Q-factor.
On the other hand, with 02 = a constant both pout and QL
show an optimum when j3y increases. Both optima lie,
however, close to one another, e.g. they are achieved at
PT - 2 in the example given in Fig. 9. This optimum value
for 0 T is independent of 02
This fact can easily be proven when pout from eqns.
35 and 30 is inserted into eqn. 18 for QL yielding

QL

gcgh
'DA
+gc)2
gDAgc
gDA +gc

gDAgc

gDA +gc

Pout

(39)

transmission cavity
(QH-2000)
/

reflection cavity
( O H =30)

A=01 05 10 20 50

reaction cavity
/(QH-2000)

02

0-4

0-6

08

10

Fig. 10 Hold-in range against output power of cavity stabilised

oscillators
Parameters are coupling coefficients

From eqn. 39 it can be seen that Q L O W ) is a straight line

with slope Qo, if $T = 1 lie a constant
(ii) Reflection-cavity stabilised oscillator
(a) large tunable frequency range of stabilisation
(b) relatively low values of QL corresponding to a high
output power level
(c) small decline of loaded Q-factor and output power.
Increasing the cavity-coupling coefficient 0 increases the
output power pout- The loaded (2-factor QL shows a maximum againt pout if |3 is increased.
(iii) Reaction-cavity stabilised oscillator
(a) small tunable frequency range of stabilisation
(b) high loaded Q-factor at high output power level
(c) large decline of loaded (2-factor and output power
due to the small tunable frequency range.
Increasing the cavity-coupling coefficient reduces the
output power but enhances the loaded (?-factor until a
152

maximum value at a very small output power level is

achieved.
These statements show that the three types of cavitystabilised oscillators differ markedly from one another. As
guidelines for the design of a frequency stable oscillator
some hints can be given as rules of thumb: 19
(a) A transmission cavity can be chosen if the value of
QL is crucial, but the value otpout is unimportant.
(b) If large QL and pout are required, and the amount of
Ao)s is unimportant, a reaction cavity meets these requirements best.
(c) A reflection cavity can be chosen if the value oipout
is the crucial point and if only a medium QL is needed.
Concluding this Section, some results will be given concerning cavity-stabilised oscillators with double-valued
tuning characteristic. Their common equivalent circuit,
mentioned in Section 4, has been analysed according to the
procedure of Section 5. The maximum loaded (2-factor
against the output power is shown in Fig. 11. Two results
are remarkable, in that an oscillator with reaction or reflection cavities yields a higher output power for constant QL
than an oscillator with transmission cavity. The latter
structure leads, however, to greater absolute values of the
loaded (2-factor. This is due to limitations in the coupling
coefficient |3, which cannot exceed a certain value as given
by the maximally possible diameter of the coupling hole.
Comparing the QL against pout curves of oscillators with
single-valued tuning characteristic to those with doublevalued tuning characteristic shows that for an equal loaded
(2-factor the output power of the latter is about 1 dB larger.
The difference in the output powers is dissipated in the
damping resistor of oscillators with single-valued tuning
characteristic. For a realistic comparison, however, one has
to account for the fact that the maximum loaded (?-factor
of oscillators with double-valued tuning characteristic is
achieved just before mode jumping occurs. Hence, it cannot be utilised. Accounting for a sufficiently great safety
distance of the. operating point from the point of mode
jumping, both oscillators with double- and single-valued
tuning characteristic yield equally loaded (2-factors at equal
output power levels. Hence, there are no advantages at all
(but some significant disadvantages) for the oscillator with
double-valued tuning characteristics as compared to the
other varieties.
7

Further optimisations of cavity stabilised oscillators

The common equivalent circuit of cavity stabilised oscillators contains three conductances gH, gDA and^ c in which
the output power and the losses of the original oscillator, in
the damping resistor and in the stabilising cavity, are
absorbed. In the case of an oscillator with a transmission
cavity, it was shown that an optimum value with respect to
loaded Q-factor and output power exists for the conductance gc = 1 /(iT. Although the three conductances are then
determined (gH by undesired but inevitable diode mounting
losses, gDA by a desired and well defined damping, gc by
optimising QL and pout), there exists, however, a further
parameter, the cavity output coupling coefficient |3 2 ,
through which the distribution of the generated power can
be controlled. It is thus possible to realise an effective load
conductance that is different from the transmission line
conductance (g, = 1).
From Fig. 9 it can be seen that the cavity input conductance of an oscillator with transmission cavity, which plays
the role of a transformed load, must differ from the transmission line conductance (|3T = 1) in order that output
MICROWA VES, OPTICS AND ACOUSTICS, JUL Y1977, Vol. l,No.4

power and loaded 0-factor are optimum. The situation is

quite different in the case of oscillators with reaction and
reflection cavity. Their normalised load conductance has so
far been assumed to equal one (gL = 1). Then, the distribution of the generated power is fixed a priori: gc is given
by the desired cavity coupling coefficient;gDA =gL = 1 in
the case of a reaction cavity stabilised oscillator of gDA =
gcrit (eqn. 4) for an oscillator with reflection cavity, gH is
given by the losses of the diode mounting structure and
additionally for an oscillator with reflection cavity by gL =
1. Taking the results of the transmission cavity stabilised
oscillator into account, one must suppose that the prefixed
load gi, = 1 does not lead to optimum values for output
power and loaded (^-factor.

performance of reaction and reflection-cavity stabilised

oscillators should be optimised by introducing an additional
degree of freedom through a tunable load (in the case of a
reflection cavity, QH changes when gL is varied according
to eqn. 38).
There does exist a further possibility of optimising the
performance of a transmission-cavity stabilised oscillator,
which shall now be discussed. The power loss in the
damping resistor can be minimised (gDA >gcru) without
introducing mode jumping, and this will lead to an enhancement in output power, typically 0-5 dB, and,simultaneously,
to a slight increase of the loaded (2-factor.

12

12
transmission cavity (QH=2000)

10

10

V-x\\\ \
o

\\ \\

cavity

2-

reaction cavity (QH = 2000)'

0-2

0-4

06

Q'=50 30
g\]=0-5 10
0-8

10

02

15
20
0-6

10

Pout

Fig.11 Loaded Q-factor against output power of cavity stabilised

oscillators with double-valued tuning characteristic
Parameters are coupling coefficients

To answer this question, the formulas of Section 5 have

been evaluated once more with gL as parameter. Possible
realisations of oscillators with reaction and reflection cavity
are, for example, a quarter-wavelength transformer in front
of the load, which is half a wavelength distant from the
reaction cavity, or from the diode mounting structure for
an oscillator with reflection cavity. Possible values for the
resultant load conductance are, however, restricted by the
shape of the load line. The aforementioned half-wavelength
long coupling line must not lead to an additional loop in
the load line ^LCCO), thereby again introducing modejumping problems. This can be avoided by using a reflection
coefficient of the resultant load that is not too large.
The results of the above sketched calculation are shown
in Fig. 12. Either for a reaction and a reflection cavity
stabilised oscillator output power and loaded Q-factor can
considerably be enchanced by choosing gL = 2 instead of
gL = 1 (This value does not yet lead to an additional loop
in the load line.) Moreover, the tunable frequency range of
stabilisation increases with gL in the case of a reactioncavity stabilised oscillator. This is of special importance
because such a structure is marked by a poor Acos value.
Changing gL from 1 to 2 nearly doubles Acos. Hence, the
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1, No. 4

Fig. 12 Loaded Q-factor against output power of reaction and

reflection cavity stabilised oscillators
Shown for the case of a mismatched load
reaction cavity
.reflection cavity

The idea of a damping resistor loss minimisation is most

easily illustrated by plotting device line and load line in the
admittance plane. This has been done for oscillators with
reflection and transmission cavity in Fig. 13. The value of
the damping conductance has been chosen according to
eqn. 4 so that an infinitesimally small susceptance gap is
formed in the loop of the load line (the solid lines in Fig.
13). Thus, a single-valued operation is achieved for CJ0 =
w c = LOH- If one wants to increase the frequency range
Aajf over which single-valued operation is possible, one
must enhance the damping resistor loss, i.e. gDA has to be*
decreased, and then the susceptance gap widens. This is,
however, only necessary for an oscillator with reflection
cavity, because in this case the diameter of the admittance
loop is less than the small-signal conductance gn(0) of the
active device. This is a consequence of the special conductance relations in the common equivalent circuit. In the
case of an oscillator with transmission cavity, however, the
diameter of the admittance loop is so large that go(0) lies
inside the loop. Hence, a susceptance gap is not necessary
153

to guarantee single-valued operation. Instead, the damping

resistor loss may be decreased (i.e. gDA increased) so far
that the crossover point P just touches D ( 0 ) (the dotted
line in Fig. \3b).
This is a new condition for determining the necessary
amount of damping. It reads as follows:
(40)
So far, a value of gDA according to eqn. 40 ensures singlevalued operating only for CJ0 = coc = coH- Plotting the load
line for various amounts of detuning, uci^<jjH showed,
however, that the single-valued tuning characteristic was
preserved owing to an increasing rotation and change of
shape of the admittance loop with increasing detuning of
the cavity.
This minimisation of damping resistor loss yields an
output power enhancement of typically 0-5 dB. Furthermore, any hysteresis effects at the borders of the tunable
frequency range of stabilisation vanish owing to the fact
that Aco/ = Acos, and the hold-in range Acofi no longer overlaps the lock-in range Aco/\

Fig. 13 Plots of load line JL(CJ) and device line )>D(V) in the
admittance plane
v is the r.f.-voltage amplitude
a Oscillator with reflection cavity
b Oscillator with transmission cavity
8

Conclusions

A comparison was made between transmission-, reaction-,

and reflection-cavity .stabilised oscillators. To compare theii
essential features, such as loaded Q-factor and tunablefrequency range, on a common basis, a common equivalent

154

network had been developed that is well suited to describe

the oscillator performance over a broad frequency range. It
was shown that the three variants of cavity stabilised oscillators differ markedly in attainable output power, loaded
(?-factor, and tunable-frequency range. Based on those differences, outlines could be given for designing highly stable
microwave oscillators.
To demonstrate the effectiveness of the theory developed
above, a transmission-cavity stabilised oscillator with a
Gunn element as an active device was built, with an operating frequency of about 15 GHz. An H012-mode circular
waveguide cavity with measured unloaded (?-factor of
27 000 had been used for frequency stabilisation. The
loaded ^-factor of the oscillator amounted to 5400 at an
output power of 250 mW. The overall power loss was
slightly in excess of 2dB (available power of the Gunn
element was 450 mW), and 0-5 dB was the power loss in
both the diode-mounting structure and the damping resistor
(due to a minimisation following the outlines of Section 7),
and 1 dB was dissipated in the stabilising cavity. Such a high
loaded (2-factor at the sacrifice of only 2 dB power loss has
never been reported before in the literature. The tunable
frequency range of single-valued operation was Acos/(27r) =
700 MHz, and the tuning bandwidth defined for an output
power decline of less than 1 dB was 300 MHz.
This excellent performance could be achieved theoretically by designing the single components of the whole oscillator structure. At first, the optimum value of/3T belonging
to a Qo = 27 000 was calculated to be slightly greater than
2. Then the cavity input reflection coefficient was imitated
with a quarter-wavelength transformer. A pin of absorbing
material was inserted in the middle of the half wavelength
coupling line, generating an insertion loss of 0-5 dB. Then
the diode-mounting structure, coupling line, and imitated
load were brought together. Using a variable waveguide
short circuit and posts of different diameters the diodemounting structure was tuned to maximum ouput power.
Then the imitated load was replaced by the transmission
cavity; the coupling coefficients follow from the optimum
|3T value and an additional condition concerning the tolerable cavity loss or desired output power. Finally, the
switching-on behaviour of the oscillator had to be checked.
Slight corrections of the damping resistor might then be
necessary to extend the single-valued operation over the
desired frequency range.
The theory of cavity stabilised oscillators presented in
the preceding paragraphs is mainly based on the consideration of the resonant behaviour of the diode-mounting
structure and on the circuit losses there (the //-circuit in the
common equivalent network of Fig. 3). This leads to the
tuning condition |3T = 1 and j i ^ l , respectively, which
greatly enhances both loaded (?-factor. and output power.
The greater the circuit losses are, the more the results so
achieved differ from those for (3T or gL = 1. Hence, the
actual amount of circuit losses decides whether or not the
new tuning condition is superior to that hitherto existing.
Owing to the investigations of Reference 15, however, the
circuit conductance can absorb a substantial part of the
generated power. The circuit losses range between 0-5 and
3 dB of the generated power, the higher values being valid
for coaxial or m.i.c. arrangements. Even losses- of 0'5-dB
greatly influence the optimum oscillator tuning, as could be
seen from the numerical results of Section 6. Hence,
neglecting circuit losses seems to be by no means justified.

MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol. 1,-No. 4

References

1 SHELTON, E.J.: 'Stabilisation of microwave oscillators', IRE

Trans., 1954, ED-1, pp. 30-40
2 GOLDSTEIN, I.: 'Frequency stabilisation of a microwave oscillator with an external cavity', ibid., 1957, MTT-5, pp.
57-62
3 MULLER, C : 'Unterdriickung des Frequenzrauschens von (BurmOszillatoren mit Hilfe ekies aufieren Resonators', Frequenz,
1969, 23, pp. 364-368
4 ASHLEY, J.R.,and SEARLES.C.B.: 'Microwave oscillator, noise
reduction by a transmission stabilising cavity', IEEE Trans.,
1968, MTT-16, pp. 743-748
5 NAGANO, S., and KONDO, H.: 'Highly stabilised half-watt
Impattoscillator', ibid, 1970,MTT-18,pp. 885-890
6 KOHTYAMA, K., and MOMMA, K.: 'A new type of frequencystabilised Gunn oscillator', Proc. IEEE, 1971, 59> pp.
1532-1533
7 KOMIZO, H., MEGURO, T., ITO, Y.,.ancr- SHINODA, M.:
'K-band' high power single tuned Impatt oscillator stabilised by
hybridKJoupled cavities', IEEE Thins., 1972, MTT-20,, pp-.
799-805
8" NAGANO, S., and OHNAKA, S.: 'Highly stabilise*Impatt oscillators at millimeter wavelength', ibid.,, 1973, MTT-21,
pp. 491-491
9 WILSON, K., TEBBY, A.J., and LANGDON, D.W.: 'A novel,
high- stability, high Impatt oscillator'. Presented at the 2nd

European microwave conference, Stoddioftw, T971

10 NAGANO, S., and OHNAXA, S.: 'A low noise 80 GHz Silicion
Irnpatt t)scillator highly stabilised with a transmission cavity',,
IEEE Trans., 1974, MTT-22, pp. 1152-1159
11 NAGANO, S., and OHNAKA, S.: frequency stabilisation of a
millimeter-value solid-state oscillator by a reaction cavity'r
Electron, and Commun. Jap., 1974, 57-B, (<6), pp. 73>-82
12 KDHIYAMA, KM and: NfOMMA, !C: 'Dynamic characterization
of band reflection type cawty stabilised Gunn oscillator', ibid.,
1974, 57-B, (9), pp. 70-78
13' VAN IPEKN,,BJ3..: 'Design and construction of an inexpensive
20 GHz stabilised Impatt oscillator in micr.ostrip technique'.
Presented at the 5th European nstciowave conference, Hamburg,
1975
V4 WHITE, J J \ : 'Simplified theory for post coupling Gunn diodes
to waveguide',IEEE Trans., 1972, MTT-20, pp. 372-278
15 TOZER, R'.C, CHARTON, B., and HOBSON, G.S.: 'Characterisation of microwave oscillator and amplifier circuits using an
Frnpatt diode biased below breakdown', ibid., 1974, MTT-22,
pp. 806-808
16 KUROKAWA, K.: 'Injection-locking of microwave solid-state
oscillators', Proc. IEEE, 1973,61, pp. 1386-1410
17 UNGER, H.G.: 'Theorie der Leitungen' (Viewer, 1967).
18 UNGEK, H.G.: 'Elektromagnetische WellenJ' CViavwg-, 1967)
19 KNOCHEL, R., and SHtJNEMANN, .: 'Design of cavitystabilised microwave osoillaBis\ Electron. Lett, 1975, 11, pp.
405-406

Radiation pattern distortion by sensor booms

K. Pontoppidan and L. Sdymar
Indexing terms: Antenna radiation patterns, Satellite antennas, Dipole antennas, Radiation distortion
Abstract: The distortion introduced.by-a sensor boom into the radiation pattern of a satellite-mounted dipole
antenna.is studied both experimentally and theoretically. It is shown that the theory developed by Kellerand
Ahltiwalia for diffraction bya thin wire gives good approximation to the. measured results.

For many of the measurements carried out by artificial

satellites it is necessary to use long sensor booms which will
generally introduce some distortion into the radiation
patterns of the antennas mounted on the satellite* body.
The aim of the present study was to investigate this distorting effect for the simple configuration shown in Fig. 1,
where a dipole antenna is mounted on the top of a cylindri-

cal- satellite parallel with the boom but off the axis so that a
larger part of the boom is directly illuminated. Note furtfiar
that the dipole .and the boom are in the same vertical'plane.

-10

CO
XT

-20
22 cm
90

Fig. 1 Relative geometrical arrangement of dipole and boom on

satellite body
Paper TS3M received 1st December 1976
Dr. Pdntoppidan and Df. Solymar are with the Department of
Engineering Science; University of Oxford, Parks Road, Oxford
OX1 3PJ, England.
MICROWAVES, OPTICS AND ACOUSTICS, JULY 1977, Vol.l, No. 4

135"

Fig. 2 Radiation patterns

1 Radiation pattern in the xy-plane
2 Difference of the patterns with and without the. boom

The best indication of the effect of the boom may be

obtained by measuring the radiation pattern in the horizontal xy-plane. In the absence of the boom, one measures
155