MWT Book

1
Microwave Tubes
G. Faillon, G. Kornfeld, E. Bosch, and M.K. Thumm
1.1 Introduction
1.1.1 Review of State-of-the-Art and Present Situation
For more than 60 years, microwave tubes are used in many applications as oscillators
and, in particular, as powerful ampliers of electromagnetic waves at frequencies
ranging from about 300 MHz to several hundreds of GHz, with some even reaching
the 1 to 2 THz range.
A number of microwave tube types exist, classied on the basis of their operating
frequency and output power. At the same time, they may be divided into two general
categories: pulse wave tubes and continuous wave (CW) tubes. The power/frequency
relationships of the most commonly used microwave tubes are shown in Fig. 1.1.
Microwave tubes may be broadly used in three main areas of applications:
radio, TV and telecommunications;
radars and military systems (especially airborne);
industrial, scientic, medical (ISM).
Even though microwave tube performance is generally described in terms of
power, efciency and gain, other factors of specic and professional interests to each
user should also be taken into consideration: bandwidth, linearity, signal/noise ratio,
mode of propagation, tube and transmitter weight, reliability, etc.
1.1.2 Historical Development
One century ago, in 1904, Sir J.A. Fleming discovered the valve, or the diode,
immediately used to detect the electromagnetic waves recently demonstrated by
H.R. Hertz. In fact, the diode was the rst vacuum tube, with a poor vacuum, al-
though vacuum. A few months later, in 1906, Lee de Forest has been inspired to
2 G. Faillon et al.
Fig. 1.1. Peak and CW power of various microwave tubes vs frequency
add a 3rd electrode, and in that way the triode was born. After about 10 years of
tests and discussions, this new vacuum tube has been used as a very attractive ampli-
er, which gave rise, in the 1920s, to a rapid expansion of the broadcasting and the
radio-links.
At the same time, research was carried not only to improve the triodes, but also
to consider other vacuum devices at higher frequencies. In 1920, H. Barkhausen in-
vented the retarding-eld tube (or reex triode), which can be regarded as the rst
transit time tube. The magnetron has been discovered by A.W. Hull in 1921, followed
by E. Habann and A. Zacek. Twelve years later, K. Posthumous understood the oper-
ating principles of such an oscillator, and in 1939 H.A.H. Boot and J.T. Randall used
klystron type resonators for conning the RF elds. The introduction of the oxide
cathode with a high secondary emission coefcient was also a major improvement.
During World War II, the urgent need for high power microwave generators for radar
transmitters led to the development of the magnetron to its present state.
In 1953 W.C. Brown introduced the crossed-eld amplier (CFA), a magnetron
with an interrupted RF circuit to provide input and output connections. At the same
1 Microwave Tubes 3
time the backward oscillator (MBWO), or carcinotron, derived from the CFA but
with an injected beam and a specic delay line came into light on both sides of the
Atlantic Ocean. This oscillator offered the advantage to be electrically frequency
tunable over a very large band.
Nevertheless, neither the triodes nor the magnetrons were high frequency and
large gain ampliers. In 193435, A. & O. Heil were the rst scientists to imag-
ine the use of a periodic and localized electron velocity variation in order to get
bunches. After the registration of patents in 1937, Hahn & Metclaff and especially
R. & H. Varian gave the exact description of the multicavity klystron. The rst proto-
type delivered 50 W at 3 GHz with a gain of 30 dB, and very soon 30 to 50 dB gain
1 to 10 GHz klystrons have been widely used, for instance, on radars and particle
accelerators. Born from the principles of this klystron and from previous studies on
electron reectors free of microwave elds, the reex klystrons made great strides
as local oscillators in radars.
The possibility of traveling wave interaction has been described in 1942 by
R. Kompfner, who developed the rst TWT, one year later, using a helix as the de-
lay line. But the real starting development of the TWTs took place after 1946, when
J.R. Pierce made the theory of these tubes and gave solutions to suppress many par-
asitic oscillations. The coupled cavity TWT appeared in 1950. The evolution of the
TWTs is impressive, when we know that they are still nowadays used in modern
versions for many purposes.
From that time, and because of important military and civilian needs, the mi-
crowave tubes entered into their industrial period. Of course the R&D was strongly
going on, not only to improve the present devices and to replace them, according
the new user requirements higher frequencies, longer lifetime, greater power and
bandwidth but also to compete with the transistors and the solid state devices or,
especially now, to collaborate with them.
The last 40 years period can be characterized by the following main events [1]:
Coaxial magnetron, frequency agility magnetron.
Mass production of magnetrons in the 1960s to provide the microwave ovens.
Brazed and pressed helix TWTs, mainly used up to the 1980s in radar transmit-
ters and ground radio-links.
Variable pitch helix TWTs improving efciency.
Gridded electron guns and introduction of the impregnated cathodes in the 1970s.
Then, development of TWT depressed collectors and radiating collectors.
From 1962 space TWTs are used on satellites.
Tunable klystrons or TV and communications. 10% instantaneous bandwidth
klystrons. Inductive output tubes (IOTs).
Multibeam medium power klystrons, mainly used in soviet countries from 1960
to 1980.
Vapor cooled collectors, and 300 kV pulsed electron guns.
High efciency (65%), pulsed power (tens of MW) and CW (1.3 MW) kly-
strons. Multibeam high power klystrons (MBKs) from 1995.
4 G. Faillon et al.
Recently born gyrotrons, which are the only tubes capable of delivering very high
power (MW) at very high frequencies (100 GHz).
In 1959, without knowledge of the astrophysicist R.Q. Twisss work, J. Schneider
in USA and A.V. Gaponov in Russia proposed an explanation for the amplifying
mechanism based on free electron gyro radiation. But the rst successful exper-
iments took place in the 1970s and became really attractive in the 1980s, thanks
to the support of the thermonuclear fusion plasma community, especially in Eu-
rope. And now after having solved many severe technological problems, such as
the superconducting electromagnets and the sapphire or diamond windows, the
gyrotrons and the associated low loss overmoded waveguides or quasi-optical
transmission lines are manufactured and their development is still progressing.
From about 20 years, the market conguration is slowly changing and the require-
ments are now more specic, more severe and exacting, to the detriment of mass
production. Nevertheless, after the merging and the reorganization of many tube
manufacturers, the applications are mainly oriented as shown in Sect 1.1.
At the same time new research and developments are conducted in the direction
of better performance [1] as higher frequencies and greater power, but also towards
advanced technologies, such as the eld emission and cold cathodes, nanotechnolo-
gies, compactness and optimization of the whole ampliers.
1.1.3 Basic Operating Principles and Denitions
Basic Operating Principles
A microwave tube may be dened as an evacuated envelope (vacuum), inside which
an electron beam interacts with an electromagnetic wave [2, 3]. This interaction
means that the electrons of the beam give up a part of their kinetic or potential energy
to the electromagnetic wave, thereby generating or amplifying this wave.
Since it is relatively known how to generate and then to accelerate or, in other
words, to give energy to an electron beam especially thanks to the high voltage
electron guns the objective of this book is to go further and to present the physics
behind the interaction and the transfer of the beam energy to the electromagnetic
wave. Such a process involves several separate or simultaneous physical phenomena
as follows.
1. Formation and acceleration of an electron beam.
2. Periodic bunching of the electrons at a frequency f . This bunching is started
up by the RF input or drive power P
d
in the case of an amplier, or by the
electromagnetic noise in the case of an oscillator.
3. Deceleration of the bunches (or reduction of their relativistic mass) in such a
way that their kinetic or potential energy is converted into an electromagnetic or
a microwave energy at the frequency f .
4. Forwarding of this microwave energy outside the tube, which yields the tubes
output power P
OUT
.
1 Microwave Tubes 5
Characteristics and Denitions
To operate a microwave tube [4], the rst step is to heat the electron-emitting cathode
by connecting the attached lament to a source of electric power P
F
= V
F
I
F
. Then a
power supply (P
0
= V
0
I
0
) is connected between the cathode and the anode in order
to generate the required energetic electron beam, which will travel at a velocity v
0
,
carrying a current I
0
, given by the two non-relativistic expressions (the relativistic
ones will be given in the following pages)
mv
2
0
/2 = eV
0
, (1.1)
I
0
= PV
3/2
0
, (1.2)
with P being a factor related to the geometry of the gun, called perveance.
A magnetic eld is used to focus the beam [5]. This eld is created either with
permanent magnet(s) or just by an electromagnet with a power consumption P
FOC
.
In the case of an oscillator, there is no RF input signal, and the tube directly converts
the beam energy to microwave energy at the frequency f . Let P
OUT
denote the tube
output power. Then the overall tube efciency is given by = P
OUT
/(P
F
+V
0
I
0
+
P
FOC
), while the interaction efciency is
interaction
= P
OUT
/V
0
I
0
.
The oscillation does not begin immediately with the application of the voltage
V
0
, but after a uctuating starting time
1
, as shown in Fig. 1.2. Then, during the rise
time
2
, the output power increases up to the nominal level P
OUT
. The frequency f
changes during this time
2
and then stabilizes.
The power P
OUT
is dependent on the anode current I
0
and on the load impedance
Z
L
, as shown in the Rieke diagram of Fig. 1.2. A pushing ratio f/I
0
and a pulling
ratio f/
L
, measured at constant reection coefcient from the load (or standing
wave ratio, SWR), characterize the frequency sensitivities of the oscillator tube.
L
is the phase of the load impedance Z
L
. It is sometimes useful to lock or to control the
oscillation frequency f by means of a controlling signal P
c
< P
OUT
at a frequency
f
c
(close to f ), which is injected into the tube via its output. For a given P
c
setting,
the Adler relationship gives the total range of frequencies over which this frequency
locking is possible, f
max
= 2f
c
/Q
x
P
c
/P
OUT
.
In the case of an amplier, a microwave signal of power P
d
, at frequency f , is
injected at the tube input and is then amplied up to P
OUT
. The amplier overall ef-
ciency is determined in practically the same way as for oscillators, = P
OUT
/(P
F
+
Fig. 1.2. Denitions: oscillators
6 G. Faillon et al.
Fig. 1.3. Denitions: ampliers
V
0
I
0
+P
FOC
+P
d
), and the interaction efciency again is
interaction
= P
OUT
/V
0
I
0
.
The ratio G = P
OUT
/P
d
is the gain, which is often written in decibels (dB) as
G = 10 log
10
(P
OUT
/P
d
). Usually, the P
OUT
/P
d
transfer curve (Fig. 1.3) shows an
almost linear relationship at low signal levels and a saturation part at large signals.
The bandwidth, generally measured at 1 dB, is dened on the basis of the vari-
ation of P
OUT
versus the frequency, with P
d
adjusted, at each frequency, to maxi-
mize P
OUT
. The bandwidth can also be dened with P
d
constant and, for instance,
adjusted to the value giving maximum P
OUT
(Fig. 1.3).
The group delay
g
= d/d represents the signal delay inside the tube amplier,
(f ) being the phase shift between the input and the output. The inuence of the
load impedance Z
L
is given by the Rieke diagram where the P
OUT
(Z
L
) curves are
plotted.
Basic Physical Laws of E-Beams
Equation of Motion and Relativistic Corrections
The only non-negligible forces which inuence electrons in microwave tubes dur-
ing their interaction with the RF wave are the electromagnetic Lorentz forces. All
other forces, e.g. due to its spin or the gravitation, are typically 11 or 16 orders of
magnitude lower, respectively.
From the relativistic equation of motion of a single electron
d(m
e
v
e
)/dt = e(E +v
e
B), (1.3)
m
e
=
m
0
1 (v
e
/c)
2
= m
0
(1 +V/V
n
), (1.4)
v
e
= c
1 1/(1 +(V/V
n
))
2
(1.5)
1 Microwave Tubes 7
(with the charge e = 1.60 10
19
C, the relativistic electron velocity v
e
, the rela-
tivistic mass m
e
, the rest mass of the electron m
0
= 9.11 10
31
kg, the velocity of
light c, the acceleration voltage V and V
n
= m
0
c
2
/e = 511 kV, the acceleration volt-
age to produce the equivalent of one electron rest mass as relativistic mass increase),
we can easily conclude that the kinetic energy of an electron (or another charged
particle) can only be changed by the electric eld E but not by the magnetic eld B,
because the magnetic Lorenz force vector F
L
= e(v B) is perpendicular to both,
the velocity v and B, and because the scalar product of two orthogonal vectors is
always zero. Therefore, the energy transfer from an electron beam to a RF wave or
vice versa employs always only the electric eld components of the RF wave and
space charge bunches
dE
kin
= Fds = e(E +v B)v dt = eEv dt, (1.6)
where ds is the line path element along the trajectory. This does not mean that the
magnetic elds B are not very important. Conversely, magnetic elds are required to
maintain the focused beam properties, as we will see in Sect. 1.3.
Maxwell Equations
Since in a beam the number of electrons is far beyond the computation power of
modern computers, the particle picture can be replaced for many considerations by
the continuum theoretical approach of the Maxwell equations. Here the elds E and
B are self-consistently determined by the local charge and current densities and
j, which themselves depend on a set of differential equations, boundary and start
conditions on the elds E and B. The Maxwell equations cannot describe localised
collision processes or even the passing of particles, which occurs in RF tubes close
to saturation. Nevertheless, they allow drawing helpful general conclusions and sym-
metry considerations and are also the platform for a lot of computer simulations and
modelling.
div E = /
0
; Charge density is a source of E-elds; (1.7)
rot E = B/t ; Induction law; (1.8)
div B = 0; Magnetic eld is source free, no magnetic
single poles; (1.9)
rot B =
0
j +
0
0
E/t (=
0
j for stationary beams). (1.10)
From the stationary Maxwell equations one can conclude that the electric and mag-
netic elds E and B can be derived from a scalar potential and a vector potential
A in the form
E = grad , (1.11)
B = rot A. (1.12)
8 G. Faillon et al.
Inserting these denitions into the Maxwell equations (1.7) and (1.10), one obtains
equations for those potentials dened by the space charge density and the beam
current density j, respectively:
div grad = /
0
; Poissons equation (the Laplace equation for = 0);
(1.13)
rot rot A =
0
j. (1.14)
Since the inuence of the beam current density j on the magnetic eld can be in
most applications completely neglected compared to the externally applied magnetic
elds, the fourth Maxwell equation (1.10) reduces for the stationary case simply to
rot B = 0; For negligible current contributions
to B-eld; (1.15)
B =
0
grad ; with B-eld dened by scalar potential ; (1.16)
div grad = 0; Magnetic equivalent to the electric
Laplace equation. (1.17)
In this case, like for the scalar electric potential , one can dene a magnetic scalar
potential , and (1.17) becomes the magnetic equivalent to the electric Laplace
equation (1.13).
Buschs Theorem
As a direct consequence of Maxwell equations in axially symmetric (rotational sym-
metric) systems, the Busch theorem is very useful for the analysis of magnetically
focused linear electron beams (Fig. 1.4). It states a surprising conservation law, valid
along an electron trajectory, by saying that the sum of its angular momentum m
e
v
r
with respect to the z-axis and the magnetic ux parallel to the symmetry axis
through a circle area with the trajectory radius r and multiplied by the factor e/2 is
a constant.
m
e
v
r +e/2 = constant; Buschs theorem, valid along trajectories

m
e
r
2
+
e
2
= constant; of charged particles in rotational
symmetric systems (Fig. 1.4).
(1.18)
Fig. 1.4. Illustration for Buschs theorem in rotational symmetric systems
1 Microwave Tubes 9
Since the angular velocity is zero at cathode emission, according to the Busch the-
orem (1.18) the angular velocity v
or

in S is just dened by the two ux values
K
, and the radius r (with

= angular velocity around z-axis). There are a lot
of other applications for the Busch theorem. Some of them will be introduced in the
chapter on beam focusing.
Scaling Laws
Geometrical scaling of microwave tubes is often appropriate when an approved ex-
isting design at a given frequency f
1
has to be transferred to a new frequency f
2
.
Since the tube dimensions scale with the RF wavelength or the reciprocal of
the frequency f , the tube scaling factor for all linear dimensions is
=
2
/
1
= f
1
/f
2
; Scaling with frequency. (1.19)
Now the question arises how the electromagnetic properties of the tube have to
be scaled in order to get identical beam trajectories in the scaled coordinate sys-
tem x
2
?
The answer is obtained with the help of the Maxwell equations. It can be easily
shown by insertion that the Maxwell equations remain invariant in the scaled, new
coordinate system x
2
when the following scaling laws are applied.
x
2
= x
1
; Linear geometrical scaling of all dimensions in system
x
1
with the factor in the
new coordinate system x
2
; (1.20)
E
2
=
1
E
1
; Electric eld; (1.21)
B
2
=
1
B
1
; Magnetic eld; (1.22)
2
=
0
1
; Electric potential; (1.23)
A
2
=
0
A
1
; Vector potential; (1.24)
j
2
=
2
j
1
; Current density; (1.25)
I
2
=
0
I
1
; Total current (from gun); (1.26)
P
2
=
0
P
1
; Perveance P = IV
3/2
. (1.27)
With the above scaling laws, we can easily understand why the power handling ca-
pability P
OUT
of a given microwave tube design scales roughly with frequency f
2
or P
OUT
f
2
= constant.
Let us assume a tube scaled with = 1/2 for the doubled frequency f
2
= 2f
1
.
If all the electrode potentials are kept constant, also the total currents and power con-
sumptions are kept constant. But since the current densities are scaling with
2
, we
get a 4 times higher thermal loading at the critical areas. Therefore, to maintain the
same material loading at the thermal material limits, the power handling capability
in the doubled frequency case is roughly only . A reference for the topic of scaling
is given in [6].
10 G. Faillon et al.
Beam Formation in the Electron Gun
First, we consider the emission of electrons at the cathode surface. Now mainly Ba-,
Ca-Aluminates impregnated dispenser cathodes are used. Since they are in detail de-
scribed in Chapter 10, we limit ourselves here to reproduce the emission laws of
this type of thermionic emitters described by workfunction and cathode tempera-
ture T .
Thermal Cathode Emission
In the temperature limited emission regime (all emitted electrons are drawn away
from the cathode surface), the saturated emission current density j
s
is given by the
RichardsonDushman equation (1911; Nobel Prize 1928)
j
s
= A (1 r)T
2
e
/kT
, (1.28)
where A = 4em
0
k
2
/h
3
= 120 A/(cm
2
K
2
), k is the Boltzmann constant and the
quantum-mechanical reection coefcient r at the metal/vacuum surface is set to 0.
A more rened elaboration by Schottky includes the electric eld E at the surface
and is called the RichardsonDushmanSchottky equation
j
s
= A T
2
e
/kT
e
KaE
1/2
with Ka = (e
3
/4
0
)
1/2
/kT.
(1.29)
Space Charge Limited Emission
Normally, due to space charge limitation the maximum emission current j
s
is not
obtained during operation of a linear beam microwave tube. The gyrotron cathodes
only are operating in the temperature limited mode. The electron charges, emitted
into a certain volume, dened by the boundary conditions for the electrostatic poten-
tial , modify the potential distribution inside this volume. As a consequence, this
volume is transparent or perveant only for a limited current, the space charge limited
emission current. The following gives the space charge limited emission for a planar
diode conguration (Fig. 1.5):
Fig. 1.5. Planar diode conguration
1 Microwave Tubes 11
j
0
= 2.335 10
6
V
3/2
a
/d
2
; ChildLangmuir law for space charge
(1.30)
limited current density in a diode;
I
0
= F/d
2
2.335 10
6
V
3/2
a
; ChildLangmuir law for total current;
P = F/d
2
2.335 10
6
; P = diode perveance;
F = diode area. (1.31)
It is important to note that this space charge limited current does not depend neither
on the cathode temperature nor the nature of the emitting cathode (workfunction ),
but only on the applied voltage and the gun geometry; here is the diode distance d
and F is the diode area (see Figs. 1.6 and 1.7). The result (1.30) can be generalised
as
I
0
= PV
3/2
a
; ChildLangmuir law for total current. (1.32)
We remind of the result of the scaling laws (1.27) that the perveance P remains
constant in case of a linear scaling in all dimensions.
Fig. 1.6. Cathode emission vs the square root of applied anode voltage. For sufciently high
voltages the current is limited by the RichardsonDushmannSchottky saturation current
Fig. 1.7. Cathode emission vs the temperature. For low temperatures of the cathode the emis-
sion is in the temperature limited regime. At the so-called knee temperature the emission
becomes space charge limited and independent of temperature
1.2 Klystrons
1.2.1 Klystron Ampliers
Introduction
Klystrons are amplier microwave tubes, designed mainly for use at high frequencies
from about 0.3 GHz to approximately 30 GHz. They are characterized by high gains
(around 50 dB) and high output powers:
3 kW CW tunable from 14 to 14.5 GHz with efciency = 40%,
60 MW with 4 s pulses at 3 GHz ( = 38%),
1.3 MW CW at 352 MHz with going up to 65%,
500 kW CW at 3.7 GHz with = 45%.
However, almost all klystron instantaneous bandwidths are fairly limited.
In a klystron (Fig. 1.8), the main functions are separated, which means that de-
sign and technology factors can be optimized for each function. These functions are
the electron emission, the magnetical beam focusing, the electron bunching or the
Fig. 1.8. Klystron cross-section
beam density modulation, the extraction of the microwave energy and, lastly, the
dissipation of residual energy.
Interaction in a Klystron: Modulations and Energy Extraction
Thanks to a high voltage V
0
applied to the cathode in regard the anode, a Pierce
type electron gun generates an electron beam which carries the current I
0
= PV
3/2
0
according to (1.2) and which is maintained approximately cylindrical thanks to a
magnetic eld B
z
(z) over the entire length of the tube.
After leaving the anode, the electrons travel across the rst cavity, or the input
cavity. This cavity is excited by the weak input signal P
d
which is to be amplied
and whose frequency f is about f
0
, the cavitys resonance frequency. This cavity
is designed to resonate at its fundamental mode TM
110
(rectangular) or TM
010
(cir-
cular) with a maximum electric eld E
1
at the center where the electrons pass [7].
Moreover, this electric eld is increased by the fact that the drift tubes, placed on
both sides of the cavity, are reentrant (Fig. 1.9).
Electrons traveling across this narrow gap of the cavity are under action of the pe-
riodic electric eld E
1
e
jt
. They are accelerated for a half of period and decelerated
for the other half (Fig. 1.10). Consequently, the fast moving electrons from a given
half-period catch up with slow moving electrons from the preceding half-period, re-
sulting in the creation of electron bunches. In other words, the velocity modulation
produced in the rst cavity creates a beam current modulation in the following drift
tube, which is expressed as I
b
(z, t ) [8, 9].
Fig. 1.9. Beam modulation along a klystron
Fig. 1.10. Applegates diagram z(t )
A second cavity (Fig. 1.9) is then added at the distance L
1
where the bunches
form. The beam current I
b
(z, t ) induces a current I
2
in this cavity which gives rise
to a voltage V
2
= Z
2
I
2
(or an electric eld E
2
) at the gap ends, Z
2
being the cavity
impedance. On its turn V
2
(or E
2
) gives rise to a new velocity modulation, generally
to an extent far exceeding the former modulation. Again, a third cavity is added at a
distance L
2
, etc.
The process is repeated at each drift tube and each cavity along the klystron, un-
til the last, or the output, cavity is reached where the bunches are very narrow and
dense. In this last cavity the very high induced current I
N
gives rise to the large cav-
ity voltage V
N
= Z
N
I
N
. Then the bunches are strongly decelerated by this voltage
V
N
(or E
N
) so that they are themselves generating. In that way, they give up a large
part of their kinetic energy to the electromagnetic eld which builds up in this cavity
and maintains V
N
(or E
N
). This energy stored in the cavity yields the output power
P
OUT
= V
2
N
/2Z
L
routed to the load, through an iris or a coupling loop and a window
(Figs. 1.8 and 1.11). P
OUT
represents the power P
d
which is amplied by the kly-
stron. Typically, klystrons have four to six cavities, gains (P
OUT
/P
d
) of about 50 dB,
and interaction efciencies (P
OUT
/V
0
I
0
) of about 35% to more than 65%.
From the First to the Second Cavity
Assuming several simplications, the relationship between V
1
and P
d
is given by
V
2
1
= 8
R
Q
Q(Q/Q
x
)P
d
1 +Q
2
f
f
0
f
0
f
1
(1.33)
with f
0
, the cavity resonance frequency, and f , the operating frequency [8, 9, 11].
Solving the equation for the movement of an electron entering the gap of the
cavity at a velocity v
0
(1.1) yields the velocity v(d
1
) at the output of the gap,
v(d
1
) = v
0
1 +
M
1
V
1
2V
0
e
jt
, (1.34)
where t is the time when the electrons pass the center of the gap and
Fig. 1.11. Equivalent circuit of a klystron cavity
M
1
= sin
d
1
2v
0
d
1
2v
0
.
M
1
is called the coupling coefcient. The electrons leave the cavity and travel in the
following drift tube at the velocity given by (1.34).
Due to the periodicity of the modulation (Fig. 1.10), their respective positions
change over time (some electrons bunch, other separate) and bunches form, once
per period. The velocity modulation caused by the rst cavity therefore results in
electron spatial density modulation, i.e. current modulation in the drift tube. This
current along the z-axis is given by
I
b
(z, t ) = I
0
+2I
0
1
J
n
(nX) cos n(t z/v
0
) (1.35)
with X = M
1
(V
1
/2V
0
)(z/v
0
), J
n
= Bessel function, I
0
= DC or average beam
current, and M
1
given just above. Equation (1.35) shows harmonics. The rst har-
monic (n = 1) maximum is 1.16I
0
and is located at a distance z = L
1
from the
center of the gap of the rst cavity such that X = 1.84. The second harmonic (n = 2)
maximum is 0.96I
0
and it is located at z = L
2
such that X = 1.54 [8, 9]. The sec-
ond cavity is usually located near z = L
1
where the electron bunches are strongly
formed.
Contrary to the rst cavity entrance, both the electron velocity v(L
1
, t ) from
(1.34) and the current I
b
(L
1
, t ) from (1.35) at the entrance of the second cavity
change over time and are time-modulated. The beam current I
b
(z, t ) travels inside
the gap (width d
2
) of the second cavity. I
b
(z, t ) is an instantaneous current along z
at time t .
Each electron induces a positive image charge in the walls at the ends of the drift
tubes and especially in the cavity. Thus, at time t , the current I
2
(t ) induced inside
the cavity is the sum of all the charges induced by the electrons in the gap at time t .
This can be expressed mathematically by stating that I
2
(t ) is the integral of I
b
(z, t )
from L
1
to L
1
+d
2
.
After calculations, I
2
(t ) comprises the three main terms:
I
0
, the average current, which does not interact with the cavity;
I
21
= M
2
I
b1
e
jt
(1.36)
with I
b1
e
jt
= the rst harmonic of the beam current I
b
(z, t ) at L
1
+ d
2
/2 and
M
2
the coupling coefcient given by (1.38) below;
I
22
= jG
0
(V
2
/D
2
)[exp(jD
2
/2)][cos(D
2
/2) sin(D
2
/2)/(D
2
/2)] (1.37)
with G
0
= I
0
/V
0
and D
2
= d
2
/v
0
. The voltage V
2
induced in this second
cavity (between the ends of gap d
2
) is given further in the Sect. 1.2.1.
Behaviour of the Second Cavity
The coupling coefcients, for instance M
2
, are always <1 and express the fact that
the eld E
2
= V
2
e
jt
/d
2
experienced by an electron is a function of time and
changes during the traveling of this electron through the gap.
Up to now, the above calculations assumed an ideal gap with imaginary grids [9].
If we now consider a real gap with the electric eld extending inside the adjacent
drift tubes on the left and right and varying in both the axial and radial directions,
the general expression of M
2
is given by
M
2
=
1
a
0
E
2
(z, 0)r dr dz
a
0
E
2
(z, r)e
jz/v
r dr dz, (1.38)
where
E
2
(z, 0) dz = V
2
.
The narrower the gap d and the smaller the radii a or b, the more M
2
ap-
proaches 1, and only when D
2
1.2 rad and b/v
0
1 rad, M
2
drops off to
any large extent. These factors determine the orders of magnitude of the gap d
2
,
the drift tube diameter 2b, and the electron beam width 2a, given that the lling
coefcient a/b 0.6 and that b and d
2
are also related to the R/Q values of the
cavity.
I
22
/V
2
looks like a conductance Y
bl
= G
bl
+ jB
bl
. This is known as the beam
loading conductance. A positive G
bl
value means that an energy is taken from the
modulated electron beam, whereas a negative one means that an energy is given
to the electron beam. G
bl
is at a maximum for D
2
= and becomes negative after
D
2
= 2. Thus, if a cavity has a gap such that D
2
= (d
2
/v
0
) > 2, G
bl
is negative,
G
bl
+G
c
can be also negative, and an oscillation (known as monotron oscillation)
can occur. G
c
is the conductance representing the cavity losses. = 2f is related
not only to the operating fundamental mode of the cavity, but also to anyone of the
numerous resonances at higher frequencies.
The expressions given above determine the equivalent schema for the second
cavity as for any of the other intermediate cavities. This equivalent schema is an
oscillating circuit, as represented in Fig. 1.11a, driven by the induced current I
21
and
loaded by the losses conductance G
c
and the beam loading conductance Y
bl
.
Space Charge
Up to now, we have neglected the space charge, i.e. the repelling force between
electrons which naturally tends to separate them, thereby hindering bunching. If we
disturb an electron beam in a state of equilibrium by moving an electron in relation
to the other ones, this electron is rstly repelled by its neighbours and then returns
to its position of equilibrium, which it overshoots, before turning back again towards
the equilibrium position, like a swinging pendulum. This movement takes place
within a beam moving at a velocity v
0
. The frequency of this oscillation phenom-
enon (known as the plasma frequency f
p
=
p
/2) is given by
2
p
= (e/m)(
0
/
0
)
with
0
= J
0
/v
0
, the beam charge density, and J
0
I
0
/a
2
, the current den-
sity [10]. Nevertheless, this expression is somewhat inexact because the true plasma
frequency f
q
must take into account the metal or conducting walls of the drift
tubes.
Taking the space charge into consideration, it changes the second term of v(d, t )
in the expression (1.34), which has to be multiplied by cos(
q
z/v
0
). Furthermore,
the expression (1.35) also changes when the space charge is included. X becomes
M
1
(V
1
/2V
0
)
q
sin
q
z
v
0
.
At
q
0, this yields the original expression.
From the Second Cavity to the Output Cavity
The current I
2
creates a voltage V
2
= Z
2
(f ). I
2
across the gap of the second cavity.
This voltage in turn modulates the velocity of the electrons, whereby this modulation
is not necessarily in phase with the preceding modulation but is considerably greater
in magnitude. In the drift tubes between cavities 2 and 3, the newvelocity modulation
is transformed on its turn into a density or a current modulation (Fig. 1.10). This
iterative process continues until the addition of another cavity has quite no effect on
the modulation [9].
This last cavity is the output cavity and is characterized by a high degree of
coupling to the external load, that is the users device to which the tube is connected.
In practical terms, a loop or an iris is used for this coupling (Fig. 1.8).
The equivalent circuit presented above applies to all cavities, including the last
cavity which is coupled to the load (Fig. 1.11b and Fig. 1.11c). If we now assume
that the beam loading expression has no imaginary part (Y
bl
= 1/Z
bl
= G
bl
) and the
load too (Y
L
= 1/Z
L
= G
L
= 1/(R/Q)Q
L
), this yields
Y
N
= (1 +2jQ
N
)/(R/Q)Q
N
(1.39)
with (
0
)/
0
= (f f
0
)/f
0
and 1/Q
N
= 1/Q
c
+1/Q
bl
+1/Q
L
. Then
V
N
= I
N
/Y
N
. (1.40)
Therefore, P
OUT
can be written as,
P
OUT
=
1
2
V
2
N
G
L
=
1
2
I
2
N
G
L
Q
2
N
(R/Q)
2
/
1 +4Q
2
N
. (1.41)
This expression can be used to yield the gain G = P
OUT
/P
d
and the interaction
efciency = P
OUT
/V
0
I
0
.
The methodology we have applied since the Sect. 1.2.1 shows how P
OUT
can
be deduced from P
d
throughout the formulae (1.33) up to (1.41). However, it is im-
portant to bear in mind that these expressions hold only at weak elds or signals.
As soon as P
d
increases in magnitude (i.e. modulation increases), we are no longer
dealing with linear problems and the equation for P
OUT
, as given by (1.41), becomes
more complex, and in fact, even ceases to be analytical. The nonlinearities are pri-
marily due to electron bunching and space charge forces.
This being the case, more complete and complex calculations are required, neces-
sarily drawing on computer processing power. In fact, computer permits us to analyze
considerably more than just the non-linear phenomena; it also allows us to incorpo-
rate radial movements (neglected in previous calculations), the 3D nature of certain
events, relativistic effects and reected electrons. Schematically speaking, this in-
volves solving the electron movement equation for varying electromagnetic elds
and space charge elds.
Output Power Optimization
Most of the time klystrons use 4, 5 or 6 cavities. These cavities gradually increase
the RF beam current I
b
, then I
N
and P
OUT
, despite the space charge forces, which
become very disturbing when I
b
is more and more large. At the same time, to op-
timize P
OUT
, the velocity dispersion must be kept low, (v v
0
)/v
0
10 to 20%,
in order to avoid reected electrons from the output cavity. Therefore, the interme-
diate cavities are carefully frequency tuned to adjust the gap voltages in amplitude
and in phase and, at the same time, to control the RF beam current and the velocity
dispersion.
Some klystrons use a second harmonic cavity which resonates at a frequency
slightly lower than 2f
0
. Such a cavity is excited by the second harmonic of the beam
current and generates a voltage which reinforces the bunching and the efciency.
This cavity is usually located after the second one, where the electrons are already
concentrated in a phase extension .
Due to the values of the impedance Z
L
, the absolute value of V
N
= Z
N
I
N
can ex-
ceed V
0
, which is the voltage corresponding to the average electron velocity. Were V
N
to exceed V
0
, the electrons would be decelerated, but some would even be reected
back towards the cavities and the cathode, which would generate beam interception
and lead to klystron malfunction or even permanent damage.
The output cavity behaves at large signals like a current generator with an inter-
nal impedance R
around the operating point. Therefore, P

OUT
is maximum when
Z
L
= R
. Since the coupling iris or the loop transformthe load impedance at the user
device (radar or transmitter antenna, accelerator structure) to the load impedance at
the cavity, these coupling devices have to be adjusted so that Z
L
= R
, in order to
maximize the output power.
Klystrons Engineering and Technologies
This presentation of different steps of the interaction in a klystron allows us to un-
derstand the klystron structure which is designed around the electron beam on one
hand and the cavities and the drift tubes on the other hand.
From a technological point of view, as shown in Fig. 1.8, the constitutive ele-
ments of the klystron tube are:
The Pierce type electron gun, especially with the high voltage insulator and the
cathode. The high voltage insulator is a cylinder made of alumina located be-
tween the anode and the cathode. Its dimensions are such that the DC electric
elds are smaller than the breakdown limit. The electron emissions from the
triple point are minimised thanks to a screen, called anticorona ring. The anode is
grounded for electrical safety reasons, explaining why the gun and the cathode
are usually raised to a negative voltage;
The modulating cavities (usually frequency tunable in a factory);
The last or output cavity coupled to the output waveguide thanks to a coupling
loop or iris;
The output window(s) which is located in the waveguide(s) and which separates
the external atmosphere from the internal vacuum (10
8
Torr) inside the tube.
The drive power to be amplied is usually injected inside the rst cavity through
a coaxial connection which is also vacuum tight thanks to a small input win-
dow;
The collector whose function is to collect the electrons after their interaction in
the last cavity and to dissipate their remaining energy. Moreover, it must be able
to dissipate the whole electron beam energy, when P
d
= 0 and then P
OUT
= 0.
Therefore, it has to be efciently cooled. The collector is usually at the same
potential than the body or the anode;
The magnetic circuit including an electromagnet, or permanent magnet(s), and
the two pole pieces located near the anode and between the collector and the out-
put cavity. These pole pieces concentrate the magnetic ux to get the required
focusing B
Z
parallel to the axis (B
R
/B
Z
better than several ).
The vacuum is usually maintained thanks to a small ion getter pump. However, this
pump is not always necessary, because the beam itself has also a pumping function.
The accelerated electrons ionize the residual gases. Then the generated ions are elec-
trically attracted by the beam and slowly drained toward the cathode.
1.2.2 Multibeam Klystrons
Limitations in the Increase of P
OUT
in a Klystron
As explained in Sect. 1.2.1, the space charge forces are opposed to a perfect electron
bunching. These repulsion forces appear in the form of the plasma frequencies
q
or
p
, which is proportional to the square root of the perveance P.
In a conventional klystron, the perveance of the beam is in the range 0.5 to
2.5 10
6
AV
3/2
. A perveance P = 2.5 10
6
AV
3/2
is the practical limit,
beyond which the electron beams are difcult to be maintained fairly cylindrical and
to be focused without notable body interception. When the perveance is high, the
efciency is low. At the same time the instantaneous bandwidth is enlarged to 5%
or more, instead of the usual 1 to 2%. This last point is explained by the high value
of the beam loading conductance G
bl
proportional to PV
1/2
0
, which damps all the
cavities.
On the contrary, a perveance P = 0.5 10
6
AV
3/2
is the practical lower
limit under which the smallness of the current I
0
dictates unacceptable high beam
voltages V
0
and involves many electrical insulation difculties on the tube itself and
on the whole transmitter or equipment. But the low perveances are favourable for
high interaction efciencies because a strong bunching can be achieved.
In conclusion, larger output powers are obtained by increasing either the beam
current I
0
or the high voltage V
0
. In the rst case the perveance is important; the
efciency notably decreases, but the instantaneous bandwidth is enlarged and the
limitations are the control and the focusing of the beam. In the second case the per-
veance is kept low; the efciency is therefore high, but the main limitations are the
breakdowns and the practical use of high voltages.
The Multibeam Klystrons (MBKs)
The MBKs give the possibility to reach larger RF output powers with high efcien-
cies and, at the same time, acceptable high beam voltages, despite very high per-
veances. Aconventional or one beamklystron (OBK) uses just a single electron beam
which travels along the common axis of the tube and the magnet. On the contrary,
the MBK, as shown in the Fig. 1.12, uses several (N) electron beams which travel
all together through the same cavities, but each one has its own drift tube. Therefore,
each electron beam has a low current i
0
giving rise to a high efciency . The whole
current Ni
0
is important; then, for the same RF output power P
OUT
= V
0
Ni
0
, the
cathode voltage is modest and even small. For example, the number of beams is 6
or 7. The perveance of a single beam is about 0.5 to 0.610
6
AV
3/2
and the total
perveance P = Ni
0
V
3/2
0
3 to 4 10
6
AV
3/2
. The interaction efciency is
around 65%.
In comparison with a OBK, the cathode voltage V
0
is usually reduced by 40%.
Consequently, the electrical elds E
0
are decreased and the product E
0
V
0
(kV
2
/mm),
Fig. 1.12. Multibeam klystron cross-section (4 beams)
which interprets the breakdown occurrence, is also reduced by more than 50%, giv-
ing rise to the possibility of much more safe and reliable operations in longer pulses
or DC voltages.
The cavities are cylindrical. The RF resonant mode is generally the conventional
one, the TM
O
010
, and the re-entrant drift tubes, where the beams are passing through,
are concentrated around the axis [7], as shown in Fig. 1.13. The RF eld pattern
allows, of course, an in phase interaction of each electron beam, but also high cou-
pling factors M between the electrons and the RF electric eld E, despite a non-
negligible variation of E versus the radius and the possible inuence of the RF mag-
netic eld.
The overall high perveance of the electron beams increases the conductance of
the beam loading of the cavities. Consequently, the instantaneous bandwidths be-
come larger, by a factor of 5. Finally, the cavities are designed in such a way that
Fig. 1.13. MBK cavity (7 beams)
the adjacent modes are very far from the operating one. The reason is to prevent
monotron oscillations (see Sect. 1.2.1), encouraged by the large whole beam cur-
rent Ni
0
.
The focusing of the electron beam is difcult because the beams axes are offset
and not identical with the tube axis, generating an important radial magnetic eld B
R
.
Therefore, the electromagnet is equipped with specic multiple pole pieces. Between
those the axial eld is nearly constant, and the relative radial components B
R
/B
Z
smaller than a few .
In short, the main characteristics of the MBKs are the signicantly lower cathode
high voltages and the larger efciencies and/or the broader bandwidths. The results
are:
Smaller dimensions and size and reduced electrical supplies;
Lower breakdown or arcing risks. HV insulation in air and less and less in oil;
Less X-rays parasitic radiations;
Better reliability.
However, the designer has to take into account the following two difculties:
Focusing and transmission of the N beams, N1 of which (at least) being offset
in respect to the axes of the tube and the magnetic eld;
Elimination or damping of parasitic oscillations in the cavities, which can present
many high order modes.
1.2.3 Inductive Output Tubes (IOT)
In the lower part of the frequency spectrum, for instance around 500 MHz in the
TV bands, the ground transmitters are now equipped with Si and SiC transistors in
parallel. But the RF output power is limited to several hundreds of W. To achieve
much higher output power in the range of tens of kW, vacuum tubes are still used
and for a long time thanks to their proven reliability and great efciency. The tubes
are triodes, tetrodes [11], klystrons and now mainly IOTs.
The gains of about 13 to 15 dB of triodes and tetrodes are very low. On the other
hand, the large size and the unpracticable frequency tuning of the TV klystron cav-
ities are not very attractive. Thats why a hybrid tube, the IOT, has been developed
about 20 years ago for the main use with transmitters operation in the 470830 MHz
Fig. 1.14. Inductive output tube (IOT)
range, at power of several dozen kW and gains of 20 to 23 dB compatible with solid-
state drivers.
IOTs represent a modication from conventional coaxial grid tubes towards a
linear beam structure close to the one of the klystrons, but with just one cavity
(Fig. 1.14). A Pierce type electron gun tted with a cathode and a spherical mod-
ulation grid G
1
creates a converging electron beam, with an average current I
0
, mod-
ulated by a periodic voltage V
G
at the microwave frequency f .
To avoid any malfunction, no microwave interaction must take place in the large
accelerating gap between the grid and the anode. This is why the gun is designed so
that no microwave energy enters this gap. Such an energy could come from the tube
outside or because of leakages from the coaxial cathodegrid resonator and from the
output cavity through the anode. Also, no oscillation may occur in this gap; therefore,
microwave resonances must be eliminated.
The electron bunches between the grid and the anode are accelerated by high volt-
ages V
A
going up to 30 kV or more, which is considerably higher that the 810 kV
limit of triodes. After the anode, the bunches are traveling across a klystron type
resonating cavity where much of their kinetic energy is transferred to a waveguide
or coaxial cable, in order to yield the output microwave power P
OUT
. The process
is exactly the same as in the output cavity of a conventional klystron. It should be
noted that a too high deceleration voltage in the cavity can reect electrons towards
the cathode and damage the tube.
Finally, the remaining energy of the electrons is dissipated in the collector. Be-
cause the beam is short and dont include any cylindrical part, the focusing is sim-
plied; just a permanent magnet around the anode is needed.
From a technological point of view, the most critical part is the grid gun, not only
because the grid is located very close (1 mm) to the hot cathode, but also because
the drive power P
d
must be injected in the gun assembly which is raised to a high
potential of V
A
30 kV. This is done thanks to a coupling loop and RF traps
that are difcult to design. By the way we note that, in most of the tubes, the anode
is grounded for electrical safety reasons, explaining why the gun and the cathode are
raised to V
A
.
Before to conclude it should be point out that:
the presence of a grid close to the hot cathode and the design of the coaxial
resonator limit the operating frequencies to 1.5 GHz maximum,
and the high voltage V
A
is practically limited by the necessary use of air and
not of oil to insulate the gun including the cathodegrid resonator.
Finally, IOTs are very attractive specic grid tubes operating below 1.5 GHz, at high
voltages of about 30 kV. The gains are around 2223 dBand efciencies around 55%,
for microwave output power of 10 to 100 kW.
1.3 Traveling Wave Tube (TWT)
1.3.1 Introduction
Looking back from the beginning of the 21st century to the history and evolution
of microwave tubes (as described in Sect. 1.1.2 of this book), we might wonder
why the TWT (Lindenblad, 1940; Kompfner, 1942), as the purest realization of
the microwave generation principle in electron tubes, was invented after the mag-
netron (A.W. Hull, 1921, to K. Posthumus, 1933) and after the klystron (R.H. &
S.E. Varian, 1937). The limited information exchange during the war might have
been the reason why the U.S. patent of Lindenblad [12] was not known by Dr.
Rudolph Kompfner, an Austrian refugee working on microwave tubes for the British
Admiralty, and why it took again years fromtheir documented inventions of the TWT
principle to Kompfners publication [13] in 1946.
In fact, the TWT shows all the necessary elements for RF generation in a tube
in a pure, geometrically separated and physically subsequent form. Those elements
are electron beam formation and acceleration in an electron gun, power transfer from
the electron beam to the amplied RF wave in an interaction section and collection
of the decelerated spent electron beam in a collector. Additional elements are the RF
input and output couplers at the beginning and the end of the interaction section and
the magnetic beam focusing system.
In contrast to its clear physical principle, the TWT employs the most challeng-
ing manufacturing technology amongst the microwave tubes. The necessity to focus
a high power density electron beam through a tiny helix with length more than 100
times the diameters ranging from 0.3 mm (60 GHz) to 4 mm (1.5 GHz) requests mas-
tering of very small tolerances between beam forming structures and the magnetic
focusing system. The fragile delay line structures limit the obtainable RF power com-
pared to the bulk resonator structures of magnetrons or klystrons. The difference in
technological maturity and power handling capability at time of TWT appearance is
shown in Table 1.1.
Though the basic difference in the power handling capability was remaining over
the years (see Fig. 1.1), TWTs have developed since that time, due to their excel-
lent linearity and broadband capability, to the major microwave amplier used in
terrestrial and especially in space telecommunication. Other TWT applications are:
coherent microwave sources for radar surveillance and guidance systems, satellite
based Earth observation radars and as broadband Electronic Counter Measurement
(ECM) ampliers.
1.3.2 TWT Design and Operation Principle
Figure 1.15 gives a cross-section through a permanent periodic magnet (PPM) fo-
cused electrostatic 4 stage collector TWT showing the essential functional elements.
Table 1.1. Output power and application of early C-band magnetrons, klystrons and TWTs
around the time of rst industrial production of TWTs. Data are from [1416] for magnetron,
klystron and TWT, respectively
Parameter Magnetron Klystron TWT
Year 1942 1949 1952
Manufacturer Western Electric Stanford University, Var-
ian
STC
Type WE 718 Mark III
Application Radar Linear Electron Accelera-
tor
TV, Ground Link
Communication
Frequency 2.72.9 GHz 2.85 GHz 3.64.4 GHz
Output power 193 kW, pulsed 12 MW, 1 s pulses 2 W, CW
Fig. 1.15. Cut through a PPM focused, electrostatic 4 stage collector TWT
Fig. 1.16. Electric circuit of a CW-TWT with double stage collector
Figure 1.16 provides the electrical circuit needed for its operation (only 2 collector
stages are shown). A vacuum envelope, comprising the electron gun, the RF inter-
action section and the collector is evacuated to the 10
9
mbar range. Metal/ceramic
brazing technology is used for the electrical feed-through to the electron gun and
collector and for the windows at RF in- and output. In the further description we
follow the electrons from emission to collection.
Electric Circuit and Electron Gun
To emit the electron current I
K
from the cathode surface, a thermionic cathode is
heated to its operating temperature T
C
of about 1000C brightness by a lament
power supply which provides lament voltage V
F
and lament current I
F
on a neg-
ative cathode potential V
K
= V
H
. Application of positive voltages V
G2
, V
H
, V
C1
,
V
C2
let the electrons ow through the vacuum to the respective electrodes and from
there in closed circuits through their respective power supplies back to the cathode.
Looking closer, they rst form an accelerated electron beam which is focused by a
PPM focusing system through the helix and is mainly collected by the collector elec-
trodes. The kinetic beam power P
beam
= I
K
V
H
of the accelerated electron beam en-
tering the delay line section at ground potential had to be spent as potential power by
the various electrode power supplies when pumping the negative charged electrons
from their electrode potentials V
G2
, V
H
, V
C1
, V
C2
to the negative cathode potential
V
K
= V
H
. Thus the major power supplies are the collector power supplies because
anode and helix supplies have to provide only very small powers P
G2
= I
G2
V
G2
and
P
H
= I
H
V
H
due to the small interception currents I
G2
and I
H
.
The electron beam entering the delay line is formed by the electron gun optics to
a small diameter electron beam characterized by the area beam convergence factor
CF as the ratio of cathode surface to beam cross-section area. For various TWT
applications CF can vary in the range 1100. For very good beam focusing properties
CF should be maintained below around 30.
Focusing System and Delay Line/RF Interaction Section
The delay line, as the structure with the largest interface to the TWT environment,
is kept on ground potential. This has the advantage that the interception current I
H
with the electron beam, which needs to be controlled in safe operational limits, can
simply be monitored on ground potential. Here the subscript H reminds of helix,
the RF delay line used in many TWT designs. To avoid the interception with the
delay line due to the repulsive radial space charge forces in the beam, compensating
magnetic forces have to be introduced by a focusing system. Constant axial elds
produced by solenoids or periodic elds created by alternating permanent magnets
can be used. Sufcient details of the focusing concepts will be discussed later in
Sect. 1.3.2.
The RF power P
1
is fed through the RF input window to the delay line. It travels
on the delay line with reduced phase velocity v
p
approximately equal to the velocity
of the electrons in the beam. Interacting with the electron beam, the RF wave gets
amplied towards the output to the RF power P
2
. Normally, the delay line is divided
into an input and output section to limit the gain in each section. At the centre sever
of the delay lines the RF wave is attenuated to avoid reections and oscillations in
the sections.
The maximum RF gain can be obtained when the electrons are slightly faster
than the RF wave. At that condition the phase velocity is equal to the moving elec-
tron density bunch v
bunch
which is forming in the decelerating phase of the RF wave.
It is the velocity of the slow space charge wave on the electron beam moving back-
wards with the velocity v
sc
or, looking from outside, with the velocity v
e
v
sc
.
Thus
v
p
= v
e
v
sc
= v
bunch
(1.42)
is the fundamental synchronism condition of a TWT. The understanding of this
interaction and RF amplication process can be enlightened by the following
everyday observation. Assume that electrons, moving across the decelerating and
accelerating phases of the RF eld, behave like a steady stream of cars moving over
a hilly highway. On the uphill side the cars become decelerated and the car density
becomes higher. The opposite occurs on the downhill side. Thus we create a bunch-
ing of cars on the uphill side, which can remain with time at the same hillside position
even when the single cars move over the hill. Back in our electron picture this means
that the electron bunch remains and growth for some time in the same decelerating
phase with respect to the RF wave. Due to deceleration of bunch electrons, resulting
in a growing of bunch space charge density and thus growing of RF current, a max-
imum amount of kinetic electron energy is at this synchronism condition transferred
to the amplied RF wave. Thus at ideal synchronism we get the maximum power
gain of the TWT. A quantitative approach to the beam/RF interaction process was
given by Pierce, see Sect. 1.3.3.
At the end of the delay line we nd: (a) an amplied RF wave with power P
2
cou-
pled via the RF output coupler to a load, and (b) a decelerated spent electron beam
with a wide electron velocity spectrum. Depending on the interaction efciency, the
beam has still about 65% to 95% of its initial kinetic power left, which can be recov-
ered up to approximately 80% in the electron collector.
To support the understanding of the interaction physics further, it is mentioned
here that a linear accelerator (LINAC), which transfers power from a propagating
RF wave to a stream of charged particles (e.g. electrons), can be understood as an
inversed TWT. To accelerate electrons in a growing bunch, it requests for its optimal
operation the coupling of the RF wave to the fast space charge wave on the particle
beam, and thus the LINAC synchronism condition is given by
v
p
= v
e
+v
sc
= v
bunch
. (1.43)
That means, in a LINAC the RF wave has to be faster than the charged particles
(e.g. electrons) in order to maintain synchronism with the growing bunch during
acceleration of single particles in the bunch. In other words, the accelerating phase
of the wave front densies the slower moving single particles in a growing bunch
moving with the RF phase velocity.
Having this inversion in mind, we might wonder again why the TWT was in-
vented only in 1942, because the LINAC principle was described already in 1924 by
Gustaf Ising, a Swedish physicist, and it was built already in 1928 by the Norwegian
engineer Rolf Widere.
Collector
The rst TWT collectors built were like klystron collectors, single stage collectors
on ground potential (V
C
= V
H
). At the grounded collector electrode the current loop
is closed for the dominant portion I
C
of the electron beam current I
beam
. Since the
spent electron beam entering the collector carries still a lot of kinetic power, this
power is thermally dissipated when the electrons hit the collector walls. Now we
want to know this dissipated power P
C,diss
.
Assuming that the generation of the RF power P
2
has slowed down only the
electrons entering the collector, we can write, because of conservation of energy for
the dissipated power in the single stage collector,
P
C,diss
= I
C
V
C
P
2
, (1.44)
where P
2
is the RF power (fundamental and harmonics) created in the delay line
circuit. The goal of TWT designers is to reduce all losses, but due to its dominance
especially the collector losses. According to (1.44) this can be done for a given out-
put power P
2
in two directions: rst, by reducing the beam current and thus the
collector current I
C
required to produce the output power P
2
, which needs improve-
ments of the beam to RF interaction in the delay line section; second, by reducing the
collector voltage V
C
with respect to the cathode. This collector voltage depression
is possible as long as the slowest electrons in the spent beam entering the collector
have sufcient kinetic energy E
kin,min
to arrive at the electric potential of the collec-
tor surface,
E
kin,min
> |e(V
H
V
C
)|. (1.45)
Here e is the elementary charge of the electron. In modern space application TWTs
a depression ratio up to (V
H
V
C
)/V
H
= 0.55 can be obtained. Since for a constant
pitch helix the beam power efciency
0
is about 13%, it was a breakthrough, when
H. Wolkstein [17] reached in 1958 with a single stage depressed collector TWT 30%
efciency.
Similar consideration can be made when i collector stages are introduced to fur-
ther decelerate the fast electrons in subsequently depressed collector voltage stages.
In the formula (1.44) the term I
C
V
C
, the electric power provided by the single stage
power supply, can be replaced by the much smaller sum over the electric power sup-
ply stages i, and the dissipated power in a multi stage collector becomes
P
C,diss
=
i
I
C
i
V
C
i
P
2
, (1.46)
E
i,kin
> |e(V
C
i
V
C
i+1
)|. (1.47)
In (1.47) the entrance condition for the residual kinetic energy of electrons in stage i
is related to the increase of potential energy of electrons entering the stage i +1.
An exotic extreme with 10 depressed collector stages was reached by Neuge-
bauer and Mihran [18] in 1972. It allowed increasing the efciency of a certain
klystron from 54% (single stage) to 70.9%. In modern space TWTs the best com-
promise between efciency improvement and complexity is found in the range of 3
to 5 depressed collector stages depending on the width of the spent beam velocity
spectrum. There, with 4 collector stages a total TWT efciency up to 74% can be
obtained.
1.3.3 TWT Physics
After the qualitative introduction to the TWT concept and its global DC-power op-
eration, we give in here a basic description of the TWT physics and a deeper re-
view of the TWT components. For a detailed study additional textbooks are required.
Amongst others, the three English language books [6, 19, 20] are recommended.
Pierce Electron Gun for Space Charge Limited Beams
The Pierce electron gun starts a beam from a cathode surface which is larger than
the nal beam cross-section in order to keep the cathode current density in the or-
der of a few A/cm
2
, and thus the required operating temperature as low as possible.
As Fig. 1.17 shows, this can be obtained with a spherical concave shaping of cath-
ode surface. The equipotential lines are concentric, and the electron trajectories start
perpendicular on those from cathode surface.
The perveance of such a Pierce gun can be analytically approached by the model
of a conical segment of a spherical diode as calculated by Langmuir and Blod-
gett [21]. Here we limit ourselves to show in Fig. 1.18 the dependence of the gun
perveance on the geometrical properties.
Fig. 1.17. Launch of an electron beam in a modied Pierce optic electron gun
Fig. 1.18. Perveance P of different electrode congurations measured in A/V
3/2
. The
Wehnelt (Pierce) electrode is kept on cathode potential in all cases. The perveance increases
with smaller cathode to anode distance and larger half-cone angle (width of cathode)
Now computer codes allow to calculate the perveance and the electron trajecto-
ries of all wanted gun geometries.
Beam Focusing Over Delay Line Length
Before dealing with the external measures to focus an electron beam, we look to the
self-forces acting on the beam due to its charge and current density.
Repulsive Space Charge and Attractive Magnetic Forces in a Beam
For a uniform cylindrical beam with uniform axial velocity v, it can be shown that
the radial forces at radius r on a charged particle are given by
Fig. 1.19. Electron distribution and space charge potential using XOOPIC, a 2.5 dimensional
particle in cell code developed by the Berkeley University in California. Simulation parame-
ters: I
beam
= 110 mA, V
beam
= 7.5 kV, r
beam
= 0.25 mm radius beam r
tube
= 0.5 mm radius
grounded tubing
F
sc
= Qe/(2r
0
); outward radial space charge force; (1.48)
F
m
= Qe
0
v
2
/(2r); inward radial magnetic force, (1.49)
where Q is the total charge per unit length inside the radius r.
Inserting the physical identity
0
0
= 1/c
2
into the ratio of both self forces one
simply obtains
F
sc
/F
m
= c
2
/v
2
. (1.50)
Above equations have two noteworthy consequences:
The radial outward space charge force is dominating the magnetic pinching effect
for all beam velocities, since v < c. But, there can be the exception that positive
ions are neutralising the electron beam space charge which then can lead to a
dominance of the contractive magnetic self force. (Example: vacuum switches,
there this destructive effect is counteracted by an applied axial magnetic eld.)
At the inner edge of a cylindrical hollow beam the inside charge Q is zero, thus
electrons on the inner radius of a hollow beam do not see a radial space charge
force!
Universal Beam Spread
In Fig. 1.19, we plot the beam expansion and potential depression of an unfocused,
typical beam required for a 200 W Ku-band TWT in a grounded tubing with 1 mm
diameter and 20 mm length. We see that within 1 cm length the beam would expand
to hit the inner diameter of a helix. The potential depression at the origin is 45.6 V
and reduces due to the beam expansion.
Magnetic Focusing
Inserting the same beam into an axially constant magnet eld with B
z
= 0.33 T, as
it could be produced by a solenoid, we get a beam with a small so-called scalloping
ripple, as seen in Fig. 1.20. The potential depression is now almost uniformly 45.8 V.
Fig. 1.20. Particle in cell simulation of a 110 mA, 7.5 kV, 0.25 mm radius beam in a 0.5 mm
radius grounded tubing as simulated with XOOPIC, in a homogeneous magnetic eld with
B
z
= 0.33 T
Fig. 1.21. Particle in cell simulation of a 110 mA, 7.5 kV, 0.25 mm radius beam in a 0.5 mm
radius grounded tubing as simulated with XOOPIC in a PPM magnetic eld with 0.33 T peak
eld
Figure 1.21 shows the overlay of scalloping and PPMripple period (7 mmmagnet
period) for the same beam and a periodic peak eld of 0.33 T. In comparison to
the uniform magnetic eld case, we recognise a much stronger beam ripple due to
the overlay of scalloping and magnetic eld periodicity with a pronounced potential
variation on the axis of 16 V compared to 3 V, respectively. Note: The potential traps
along the axis can cause ion oscillation phenomena by ionisation of residual gas in a
TWT. In practical life, small disadvantages of a PPM magnet system are more than
compensated by large savings with respect to mass, power, reliability and interface
complexity.
Brillouin Flow
There is only one special case, known as the Brillouin ow, without any beam and
potential ripples. Astraightforward derivation fromthe Busch theoremgives the Bril-
louin ow eld B
zBr
as function of beam current I, beam voltage V or electron ve-
locity v
e
and beam radius a:
B
zBr
=
2 I
0
a
2
v
e
m
e
e
,
B
zBr
Vs/cm
2
= 8.3 10
6
(I/A)
1/2
a
cm
V
V
1/4
.
(1.51)
Fig. 1.22. Envelope of an electron beam with a golden edge trajectory from cathode sur-
face into the PPM focused interaction region of a typical Ku-band space TWT. Parameters
of the modied Pierce gun are: I = 150 mA, V = 7.5 kV into a = 1 mm, grounded
tunnel, B
peak
= 0.325 T. (Simulated with 2D-gun program and visualised with virtual re-
ality shareware code by W. Schwertfeger, TED, Ulm. Different relative scaling, r = 8 z
direction)
Fig. 1.23. PPM magnet stack with Fe-pole pieces and nonmagnetic spacers braced to a tight
vacuum envelope
Equation (1.51) is used as a design rule for TWT magnet systems: the peak PPM
magnetic eld should be 1.2 to 2 times the Brillouin eld to guarantee a save focusing
of an electron beam over the interaction length of a TWT.
Another design rule, the focusing stability requirement, is valid for the length of
the magnetic eld period l
m
in a PPM system:
l
m
418
V
B
2
zBr
m
e
e
. (1.52)
Figure 1.22 visualises the helical movement of the electrons in the PPM focused
beam. A technical realisation of a PPM focusing system is sketched in Fig. 1.23.
Delay Line System and RF Interaction
Types of Delay Lines
Depending on the application and power level, several types of delay lines are used:
Helix; low power, very broadband (up to several octaves), needs ceramic sup-
porting rods.
Fig. 1.24. a Helix delay line: the pitch angle denes the RF phase velocity v
p
= c tan =
c2a/p, c = speed of light, p = axial pitch; b Double helix delay line
Fig. 1.25. a Ring and bar line; b Interdigital or comb line; c Coupled cavity line
Counter wound double helix; suppresses backward wave oscillations (due to me-
chanical complexity scarcely used), needs ceramic supporting rods.
Ring and bar line; suppresses backward wave oscillations, needs ceramic sup-
porting rods.
Interdigital line; rugged and compact delay line for high power high frequency
TWTs, reduced bandwidth.
Coupled cavity; low bandwidth, high power capability.
Figures 1.24 and 1.25 give a survey on the major types.
RF Input and Output
Depending on the delay line concept, several congurations for the RF input and
output are known. We give in Fig. 1.26 only two examples for waveguide windows,
as used, e.g. for a coupled cavity TWT input, and a coaxial feed through, as often
used for low power helix TWT input and output.
Small Signal Theory (Pierce)
At the time Pierce developed his TWT theory, the Eulerian approach was the only
success promising approach. It treats the electron beam as a charged uid, charac-
terised by a current I with beam velocity v
0
or beam voltage V
0
and the space charge
Fig. 1.26. Left: waveguide input transition with a Chebyshev step transformer for matching.
Right: coaxial output transition from a helix with coaxial matching and /4 ceramic window
for the vacuum seal
Fig. 1.27. Equivalent transmission line model for a helix delay line
bunches as an AC current i propagating with the same frequency as the RF wave.
As indicated before, this approach is limited to the small signal behaviour of the
TWT because saturation effects cannot be treated. The ingenious concept of Pierce
combined this Eulerian view of the electron beam with the replacement of the delay
line as propagation structure by an equivalent transmission line model consisting of
distributed inductances L and capacitances C, as shown in Fig. 1.27.
As a result of his derivation, he obtained a fourth degree equation for the wave
propagation constant which is known as the so-called determinantal equation
0 = 1 +

e
(
e
)
2

2
c
(
2
2
c
)
2C
3
, (1.53)
C
3
=
I Z
4V
0
, (1.54)
where C is the Pierce parameter. It has four roots corresponding to (1.55)(1.57),
1,2
=
e
+
e
C/2 j
3
e
C/2; Solutions for exponentially
increasing and decreasing
slow waves; (1.55)
Fig. 1.28. The four waves propagating along the delay line of a TWT
3
=
e
(1 C); Solution for a fast forward
wave with constant amplitude; (1.56)
4
=
e
(C
3
/4 1); Solution for a fast backward
wave with constant amplitude. (1.57)
The meaning of the solutions becomes clear by reminding that the waves vary pro-
portional to e
j(t z)
. Figure 1.28 sketches the four solutions. Considering only the
exponentially increasing solution, we get from (1.55) the linear power gain Gover N
wavelengths of the TWT in dB
G = 10 log 10e
3CN
= 47.3 CN. (1.58)
To satisfy the boundary conditions when launching the four waves at the tube en-
trance, one gets an additional so-called distribution loss, and (1.58) becomes
G = 9.54 +47.3 CN, (1.59)
which is the total linear power gain of the tube with distribution loss. It should be
mentioned here that above solutions are not limited to helices, but are valid for
all type of delay lines when introducing their respective equivalent circuit proper-
ties.
Cutlers Beam and RF Signal Characterization up to Saturation
A great improvement in the understanding of the nonlinear behaviour of the TWT
was reached when Cutler [22] published his test results on RF circuit voltage (dashed
line), beam velocity (solid line) and beam current (vertical width of shaded area) at
the TWT output relative to the undistorted signal phase as function of drive level
(see Fig. 1.29). His ndings for the 15 drive conditions from 22 dB input back off
to +9 dB overdrive level are interpreted as follows. With respect to the circuit volt-
age, one can dene an acceleration and a deceleration zone for the beam electrons.
Fig. 1.29. Results of Cutler experiment in 1956
Due to the acceleration or deceleration we nd the relative velocity v
r
of the elec-
trons to be increased between acceleration and deceleration zone and smaller than
zero in the other regions. Also, we see that v
r
starts to become double-valued for
drive levels larger than 14 dB input back off because some electrons are reected
back from the deceleration zone into the acceleration zone. The amount of beam
current (shaded area) is larger in the deceleration zone till saturation, where the elec-
tron bunch in acceleration zone becomes about equal due to decelerated electrons
moving into this zone. Two bunches in one period means that a harmonic frequency
wave has been created on the circuit. This reaches its maximum after saturation.
At higher overdrive level, the electrons are smeared out again more uniformly over
the relative phase. If one makes a vertical cut through the electron velocity body
and integrates that over various relative phases, one gets the electron velocity spec-
trum at the tube exit as a function of drive. This exit velocity spectrum is entering
into the collector and plays an important role for the total efciency of the TWT as
we will see in the next section. Here, it can be summarised that most of the elec-
trons are decelerated, but few of them are accelerated compared to the start veloc-
ity v
0
.
Large Signal Theory (Rowe) and Simulation Results
It is not space enough to report on the numerous theoretical approaches which deal
with the nonlinear effects occurring closer to saturation drive of a TWT or even in
overdrive. It should be sufcient to refer the reader to the compendium of Rowe [23]
and to declare that no analytical solutions exist for the large signal operation of
the TWT. The theoretical approaches are mainly based on the Lagrangian concept
which treats the electrons as particles or as representative particles and which allows
multi-valued velocity distributions of electrons at a given space or phase location.
Still today, fully 3-dimensional numerical simulations are too heavy for useful de-
sign work. A lot of 2.5-dimensional codes (2-dimensional axially symmetric elec-
tric and magnetic elds and 3-dimensional movement of representative electrons)
(layers and shells of electrons, or statistically distributed heavy electrons), are
providing reliable design information on gain, output power, nonlinear phase shift
and even intermodulation distortions as a function of input drive and frequency as
well as the trajectories of the representative electrons in the real space and phase
space. As an input information to the code, they need from the delay line side only
the coupling impedance K(f, r, z) as a function of frequency, radius and position,
the phase velocity v
p
(f, z) as a function of the frequency f and the axial posi-
tion, the RF attenuation (f, z) as a function of the frequency f and the axial posi-
tion.
For the beam, the representative charge (r, z
0
) and current density j(r, z
0
) in-
formation is required at the entrance z
0
of the delay line. Figure 1.30 gives for a
modern Ku-band space TWT the simulated relative phase of the electrons, being
uniformly distributed over 2 at the delay line entrance, the outer electron beam di-
ameter and the small signal and saturation drive gain over the length of the delay
line.
For Fig. 1.30a we note the crossing of electrons approaching saturation at the
very last centimetres before the delay line end. From Fig. 1.30b we see again the
magnetic ripple on the beam and the moderate increase of the beam radius as the
electron velocity slows down close to the exit. From Fig. 1.30c we recognise the in-
Fig. 1.30. a Applegate diagram of electrons (relative phase with respect to RF wave); b Outer
diameter of electron beam inside delay line; c Small signal (crosses) and saturation (circles)
gain along tube axis
sertion loss at the delay line entrance, the extended sever region where RF power is
transported by the electron beam and the difference between small signal and satura-
tion gain of less than 5 dB, which is important for excellent linearity characteristics
of the tube.
Electron Beam Collector
Collection of an Electron Beam with Distributed Velocities
Figure 1.31 shows a typical velocity spectrum of a highly efcient Ku-band space
TWT at saturation drive as a cumulative distribution function. It gives on the ordinate
the amount of the current ratio I/I
0
which has a higher axial velocity than that cor-
responding to the reference acceleration voltage given on the abscissa. The original
zero drive velocity distribution corresponds to the depicted rectangular distribution
(all electrons have uniform velocity corresponding to helix voltage of 6250 V). The
shaded areas correspond to the recovered power P
recover
in the collector if the re-
spective current contributions are gathered at their corresponding reduced collector
potentials of a multistage collector.
Potential Depression in an Ideally Focused Cylindrical Beam
Assume an ideally focused cylindrical beam with I = 110 mA in a metallic tube
at potential (b) = 0 with the mono-energetic beam voltage corresponding to the
negative cathode potential
c
= 7500 V, as shown in Fig. 1.32. With the beam
velocity
v
z
= (2(
b
c
))
1/2
, (1.60)
Fig. 1.31. Typical velocity/energy spectrum of the spent beam at the collector entrance
and recovered power proportions (shaded area) of a Ku-band TWT with V
H
= 6250 V,
V
C1
= 3250 V, V
C2
= 2500 V, V
C3
= 1750 V and V
C4
= 450 V with respect to cathode.
The collector efciency is represented by the ratio of the shaded area to the total area below
the spent beam distribution curve
Fig. 1.32. Cylindrical electron beam of radius a in a tube with radius b and potentials (0) at
the beam centre, (a) at the beam edge and (b) at the grounded tube
Fig. 1.33. Focused beam with voltage
beam
entering a collector iris at a reduced collector
potential V
C
we write the approximately valid analytical formulas for the depressed beam poten-
tial:
(a) = (b) I ln(b/a)/2
0
v
z
; Potential at beam edge;
(1.61)
(0) = (b) I(ln(b/a) +1/2)/2
0
v
z
; Potential at beam centre.
(1.62)
It is quite satisfying that the results for the potential depression at the beam edge and
the beam centre, (a) = 26.7 V and (0) = 45.9 V, is in agreement with those
from the particle in cell code XOOPIC simulation (see Figs. 1.191.21), where we
used the same parameters I = 110 mA, (b) = 0,
c
= 7500 V and b = 2a =
0.5 mm.
Collector Current Limitation by Space Charge Effects
Assume the beam of Fig. 1.32 approaching a collector entrance iris at a reduced po-
tential, as shown in Fig. 1.33. The question arises about the limited beam current
I
limit
which can pass without reections into the collector tunnel at a depressed col-
lector potential V
C
. Note: V
C
is measured with respect to the cathode potential.
As we see from (1.63) below, the maximum current depends on the ratio of the
collector tunnel radius to the beam radius b/a pressed to avoid reection of parts of
the beam for a given beam current and the ratio of the collector entrance radius b to
the beam radius a (Fig. 1.34),
Fig. 1.34. Transmission current limit as a function of beam (collector) voltage for 3 values of
tunnel to the beam radius ratio b/a = 1, 2, and 3. As we see for the example of a 110 mA
beam, the lowest beam voltage (depressed collector voltage) can be achieved if the beam is
uniformly lling the collector tunnel (b/a = 1)
Table 1.2. Minimum collector voltage for a 110 mA electron beam as function of the tunnel-
beam radius ratio
Ratio b/a 1 2 3
Minimum collector voltage for a 110 mA beam 260 V 500 V 620 V
I
lim
(b/a) = 4
0
e/m
0
1
1 +2 ln(b/a)
2
3
(
beam
V
C
)
3/2
. (1.63)
Equation (1.63) has again the form of a perveance or here acceptance, I
lim
=
A(b/a)(V
C
)
3/2
, which means that the scaling laws apply (see Table 1.2).
The ndings imply a theoretical limit to the collector efciency. Depending on
the width of the electron velocity spectrum, 3 to 5 collector stages are found to give
an optimal compromise between the total tube efciency and system complexity.
Amongst the MDCs the two categories, electrostatic collectors and magnetically fo-
cused collectors, are distinguished.
Electrostatic Collectors
Due to the deceleration of the beam at the collector 1 entrance, the radial space
charge force in the beam increases and the beam expands radially. This beam expan-
sion affects the slower electrons more than the faster ones and is therefore used to
separate the slower electrons from the faster ones which land further down-
stream.
Figure 1.35 sketches the electron trajectories in a 6 stage (5 stages +1 spike on
cathode potential) collector as it was used in the 1980s for the 250 W space TWT TL
Fig. 1.35. Electrostatic 5 stage +spike collector used for radiation cooled space TWT
Fig. 1.36. Distribution of primary reected and secondary electrons in an electrostatic 4 stage
collector simulated with the Thales Electron Devices PIC program Collect 3D for a 150 W
Ku-band TWT
Fig. 1.37. Collector geometry, axial magnetic eld and simulation mesh of a magnetically
focused 4 stage collector for a 300 W Ku-band TWT (the transverse magnetic elds are not
shown)
12250. It achieved typically 48% total efciency. Figure 1.36 shows the design of a
modern electrostatic 4 stage collector used in some Thales Ku-band TWTs with the
simulated distribution of primary (green), reected (blue) and secondary electrons
(red) in the collector volume, which achieves about 68% total efciency.
Magnetically Focused Collectors
These collectors minimise the radial dimensions of the collector by continuing the
magnetic PPM focusing of the electron beam with an adapted periodicity and
strength of the magnetic eld into the various collector stages. Figure 1.37 shows
the principle for a modern 4 stage collector used for a Thales 300 W Ku-band TWT.
Fig. 1.38. PIC-simulation of electron distribution in the magnetically focused 4 stage tilted
eld collector at zero drive (left) and saturation. Primary electrons (green), reected (blue)
and secondary (red) simulated with the Thales Electron Devices PIC program Collect 3D
The magnetically focused collector allows an optimised beam lling of the col-
lector at saturation drive as shown on the right side of Fig. 1.38 and thus a maximum
efciency up to 74%. To minimise backstreaming of electrons, the most advanced
magnetically focused collectors are using tilted electric and magnetic elds produc-
ing compensating deection effects in downstream direction to focus the fast elec-
trons into the collector 4 tunnel and enhanced deection in backstreaming direction.
Relation between TWT Total-, Beam- and Collector Efciency
The TWT total efciency
tot
and the beam power efciency
beam
(or basic ef-
ciency
0
) can be simply dened and rewritten as
tot
=
P
fund
P
el
=

beam
P
beam
P
RF
+P
losses
=

beam
P
beam
P
beam
(P
beam
P
RF
P
losses
)
, (1.64)
where P
fund
is the fundamental RF power at the TWT output, P
RF
is the total RF
power (including fundamental and harmonic RF power and RF losses), P
beam
=
V
H
I
K
is the electron beam power and P
losses
are the thermal losses produced in the
tube. A short analysis of the thermal TWT losses P
losses
provides
P
losses
= P
Coll,losses
+P
H,therm
+P
A,therm
+(P
lament
). (1.65)
Neglecting the small quantities (lament power, helix and anode losses), the term
P
recover
= (P
beam
P
RF
P
losses
) in (1.66) becomes approximately equivalent to
the recovered kinetic electron beam power by reduced collector voltages. With the
denition of the collector efciency
C
as the ratio of the recovered power to the
entering beam power
C
=
P
recover
P
enter
=
P
beam
P
RF
P
losses
P
beam
P
RF
, (1.66)
we get a simplied relation between the total beam power and collector efciency by
the further neglecting the harmonic power and the RF losses in P
RF
and by division
with P
beam
tot
=
P
fund
P
el
=

beam
P
beam
P
beam
C
(P
beam
P
RF
)
=

beam
1
C
(1
beam
)
. (1.67)
Fig. 1.39. Total efciency as function of the two variables beam efciency and collector ef-
ciency. The historical path from single stage collector tubes (station 1, 10%) to 4 stage tilted
eld collector tubes with enhanced beam efciency (station 7, 70%) is sketched for Ku-band
TWTs. The gure is adapted from Kornfeld et al. [24]
Fig. 1.40. Total distribution of thermal losses in a modern TWT
The failure in making these simplications can be estimated from Fig. 1.39. In
Fig. 1.39 the relation (1.67) is plotted. It indicates further the historical development
axes for TWT efciency. First this occurred via the collector efciency improvement
and later by essentially improving the beam efciency with tapered helix concepts.
Though Fig. 1.39 seem to imply that beam and collector efciency are independent,
this is physically not the case. The larger the beam efciency becomes, the wider is
the electron velocity spectrum entering the collector with detrimental effects on the
collector efciency.
Loss Analysis of a Typical Ku-Band Space TWT
From Fig. 1.40 we can see, that further efciency improvements need to reduce
mainly the collector losses (45%) and the skin effect losses (35%). The other losses
in a developed space TWT are well below 10%: harmonic power 8%, helix intercep-
tion losses 6%, cathode heater 5%, and reected RF power 1%.
1.3.4 TWT Applications
Table 1.3 gives a survey on the different application elds of TWTs. We note that the
helix is the most used type of delay line. For communications the reason is its large
bandwidth and very good linearity. For military radar and ECM, again its extreme
bandwidth, which for some type of ECM tubes can achieve up to 3 octaves. In the
following we provide application examples based on the Thales Electron Devices
product spectrum.
Communication TWTs
TWTs for Ground Station, Airborne and Shipboard Communication
Following an internal TED communication of Francis Payen, we give a review of
these TWT applications.
Depending on the system architecture, the microwave tubes used in ground-
based, airborne or shipboard communications system transmit signals to a satellite
(up-link) or to a ground based receiver (point to point or point to multipoint commu-
nication). Though in some communication systems, due to power requirements, also
klystrons are used, TWTs are becoming predominant, because the relative bandwidth
requirements are becoming more and more demanding (usually more than 10% are
required). Table 1.4 presents the relative bandwidth and the output power capability,
as a function of frequency, for the TWTs made by TED for up-link communications
systems.
Some examples of ground-based or airborne communication systems using TED
microwave tubes are DirecTV, Echostar and Astra for DBS, Iridium for commercial
systems and Milstar, Syracuse, Stentor, SBIRS for military systems.
Table 1.3. Application elds for different types of TWT delay lines
Application Communication Radar and ECM
Sub- Ground Earth Ground Airborne Missile ECM
application station Space Observation & Ship Radars Radars Seekers
Helix
Ring & Bar
Coupled Cavity
Interdigital Line
Table 1.4. Survey on up-link TWTs
Frequency band C X Ku Ku Ka Q
GHz 5.857.1 7.98.4 12.7514.5 17.318.4 25.531.5 43.545.5
Relative bandwidth Up to 20% 6% Up to 13% 6.2% Up to 20% 4.5%
CW output power 3 kW 2.5 kW 1 kW 500 W 350 W 250 W
Table 1.5. Comparison of historic C-band tubes
First TWT First First European Modern Space TWT
in Use Space TWT Space TWT
Program TV-Ground Link Telstar 1 Symphonie Measat
Manufacturer STC Bell Lab AEG TED
Year 1952 1962 1973 2003
Frequency 3.64.4 GHz 3.74.2 GHz 3.74.2 GHz 3.44.2 GHz
Output Power 2 W 2 W 13 W 70 W
Gain 25 dB 40 dB 46 dB 56 dB
Efciency 1% <10% 34% 71%
Nonlinear Phase ? 50
50
38
Mass >5000 g >1000 g 640 g 800 g

Collector 1 stage 1 stage 1 stage depressed 4 stage depressed
Focusing Syst. Solenoid PPM PtCo PPM PtCo PPM SmCo
Cathode Oxide Oxide Oxide MM Dispenser
TWTs for Space Communication
The major requirements for microwave tubes used in satellite transponders for trans-
mitting down-link signals are:
long useful operating life and high reliability (>15 years, <100 FIT, respectively;
more than 35 000.000 h accumulated in orbit),
high total electrical efciency (>60%),
high linearity (nonlinear phase shift <48
),
low mass (depending on output power and radiation or conduction cooling).
TWTs meet all these demanding requirements and beat by far the competition from
solid state ampliers especially with respect to life, reliability and efciency. The
driving force for the impressive improvements in all those parameters was in the past
and is still today the economical pressure to save power by improving the efciency
and to reduce the mass of space ampliers. Two gures are best characterizing this
environment: the savings in launch and system costs per satellite are for DC power
saving 3.000 /W and for mass reduction 20.000 /kg. Therefore, it might be of
interest to have a short look on the history of some of the most important space TWT
parameters as reviewed by Kornfeld et al. [24].
Table 1.5 shows the performance characteristics of historic C-band space tubes
and compares them with the rst operational TWT built as TV relay tube by STC in
1952, D.C. Rogers [25]. Though the output power was increased from 2 to 70 W, the
mass was reduced.
Figure 1.41 gives the efciency improvements vs time of the leading space tube
manufacturer. The signicant improvements obtained in the rst 30 years after the
TWT invention led in the late 1970s to the impression that the TWT technology
might be completely mature and will provide only little space for further improve-
ments. In contrast to this, a steady and for the 1990s even accelerated efciency
Fig. 1.41. Efciency improvement of space TWTs (courtesy of TED)
improvement was found for the space TWTs of more or less all commercial man-
ufacturers. The main reason for this performance explosion was the availability of
codes to simulate the large signal beam to RF wave interaction in improved tapered
helices. It is expected that efciency might nally approach a level between 75% and
80%. Table 1.6 presents the frequency band, the output power, efciency and future
expected trends in the respective applications.
As seen from Table 1.6, there are many different applications for space TWTs
ranging from standard telecommunication systems, TV- and digital radio broadcast-
ing to modern digital internet and multimedia services. Also Earth observation with
pulsed radar TWTs becomes increasingly important. New applications where the
satellite TWT technology might be used are local multimedia services from small
satellites or stratospheric balloons over big cities. Also, for the low Earth orbit satel-
lite eet of global positioning systems (GPS, Galileo) and their continuous upgrade,
powerful and highly efcient TWTs might be used.
Radar and ECM TWTs
Several types of microwave tubes are used for radars: magnetrons, crossed-eld am-
pliers (CFAs), klystrons and TWTs.
From a historical point of view, the magnetrons were the rst microwave tubes
to be used in radar transmitters, more than sixty years ago. But they are oscillators,
and most of the radars, since several tens of years, require transmitters using a coher-
ent amplication chain. Among the three types of ampliers (CFAs, klystrons and
TWTs), the TWTs are most widely used, thanks to their wide instantaneous band-
width, high gain and noise free coherent operation.
Surface Radars
Surface radars are ground based or are used in naval systems (shipboard radars). The
main types of surface radars are:
Table 1.6. Space TWT applications for communication and Earth observation
Application Band Power efciency Future trend power
frequency/GHz efciency
Direct Digital Radio L-band 50 to 150 W 250 W
Navigation/GPS & Galileo 1.11.5 55% 65%
Communication/TV- S-band 7090 W 120 W
broadcast 2.32.6 59% 65%
Direct digital radio for S-band 200240 W 200250 W
automotive 2.32.6 61% 68%
Telecommunication & C-band 20130 W 150 W
broadcasting 3.44.2 6069% >73%
SAR, for Earth observation, C-band 5 kW >5 kW
Radar TWT, pulsed 5 to 6 40% 45%
Scientic applications & deep X-band 25/120170 W 25/120170 W
space missions 78.5 60% 65%
Earth observation, radar TWT, X-band 4 kW >4 kW
pulsed 78 40% 45%
Telecommunication and Ku-band 25200 W 25300 W
broadcasting 10.712.75 6268% 6875%
Internet Multimedia services
Altimeter; radar application Ku-band Up to 100 W 150 W
for Earth observation, pulsed 1315 or 1218 55% 60%
Telecommunication and Ka-band 15130 W 15220 W
multimedia Services 1722 5566% 5570%
Deep Space & Scientic Ka-band 2030 20100
Mission 2732 54% 58%
Multimedia Services for low Q-band 40 W 40100 W
orbit Satellites or Stratosphere 4045 40% 4045%
Balloons
Inter satellite links for V-band 20 W 20100 W
multimedia Services 5864 35% 3540%
long range surveillance radars for Air Trafc Control (ATC),
air defence radars,
tracking radars,
re control radars; those are most time integrated into weapon systems,
trajectography radars.
Table 1.7 presents the relative bandwidth and the output power capability, as a func-
tion of frequency, for the TWTs (Ku- and Ka-band) and klystrons (up to and includ-
ing X-band) made by TED for surface radars.
Table 1.7. Survey on surface radars
Frequency band L S C X Ku Ka
GHz 1.261.36 2.73.5 5.45.9 8.510.5 1518 3338
Relative bandwidth 3% 3 to 15% 5 to 10% 10% 10 to 20% 3 to 10%
Peak output power 4 MW 20 MW 1 MW 120 kW 2.5 kW 1 kW
Average 12 kW 20 kW 20 kW 5 kW 200 W 200 W
Table 1.8. Survey on missile radars
Frequency band X X X Ku Ka W
Type of tube Helix TWT CC TWT Magnetron TWT TWT TWT
Relative bandwidth (%) 2 3 Tunable in 600 MHz 20 3 1
Peak output power (kW) 20 120 220 2 1 0.15
Average output power (W) 800 1500 200 400 200 15
The Ka- and W-band TWT use interdigital delay lines; W
under development
Airborne Radars
Microwave tubes (magnetrons and TWTs) are used in airborne radar transmitters in
two categories:
multimode and multifunction radars; TWTs are widely used, either with coupled
cavity slow wave structure or Helix;
Terrain following radars; generally, TWTs are used.
Missile Seekers
The requirements for microwave tubes (magnetrons, klystrons and TWTs) used in
active RF missile seekers are small size and weight, high electrical efciency, very
short start-up time, capability to withstand very severe environmental conditions and
high reliability after long storage periods.
Some examples are new generation MICA, ASTER and PAC3 missile seekers.
The main performances are presented in Table 1.8. It shows the relative bandwidth
and the output power capability for magnetrons and TWTs made by TEDfor airborne
radars or missile seeker applications, as a function of frequency.
ECM Applications
The requirements for microwave tubes used in ECM Systems are very wide instan-
taneous frequency bandwidth (more than one octave), small size and low weight and
high electrical efciency.
The only microwave tube which can meet a specication with more than one
octave bandwidth is the Helix TWT. Table 1.9 presents TEDs helix TWTs made for
ECM systems.
Table 1.9. Survey on broadband ECM tubes
Frequency band 6 to 18 GHz 18 to 40 GHz
Pulsed TWTs Peak output power 2 kW
Average output power 80 W
CW TWTs CW output power 200 W 80 W
1.4 Extended Interaction Klystron EIK
1.4.1 Introduction
Similar to klystron and TWT, the extended interaction klystron (EIK) is a linear
electron beam device which tries to combine the advantages of both, the ruggedness
and high power capability of a klystron and the larger bandwidth of a TWT. There-
fore, the EIK is especially similar to the rugged coupled cavity and interdigital line
TWT.
Since the EIK can be considered as a renement of both microwave devices, it is
not easy to mention an inventor. The main work on EIKs seems to go back to Tore
Wessel-Berg [26] and his A General Theory of Klystrons with Arbitrary, Extended
Interaction Fields issued in 1957. He was working at SLAC where he improved
efciency and power capability of klystrons for the linear accelerators. Further pio-
neering work was done by Chodorow and Kulke [27] in the 1960s. It was recognised
by those groups that the circuit impedance is enhanced proportional to the extended
interaction region and that larger gain bandwidth product and higher efciency could
be obtained relative to the conventional klystron circuit, particularly in the submil-
limetres/millimetres region. These characteristics suit today requirements for Ka- to
W-band air and space borne radars (see [28] and [29], respectively) and also ECM
applications.
1.4.2 Extended Interaction Circuit Design
Figure 1.42 compares the principle cross-sections of a conventional two cavity kly-
stron with an extended interaction klystron. The number of cavities and interaction
gaps can differ depending on the required application. A 3-dimensional impression
of a multicavity multigap design of an EIK is given in Fig. 1.43. This and the previ-
ous gure are taken from [28].
The extended interaction oscillator (EIO) is a single cavity device with interac-
tion gaps (segmented drift tube) that function like a coupled cavity TWT structure
with extremely strong cavity-to-cavity coupling. At sufciently high beam currents,
oscillations are sustained. Variation of the beam voltage allows 0.4% frequency tun-
ing.
1.4.3 Typical Performance and Applications
The today leading manufacturer of EIKs, CPI-Canada, describes the typical
frequency and power range as follows (Table 1.10) [30]. The instantaneous band-
Fig. 1.42. Cross section of a conventional and extended interaction klystron
Fig. 1.43. Left: Rising sun type tunable EIK ladder circuit. Right: a 3-dimensional cross-
section; b top-view; and c side-view of an EIK ve gap structure in a cavity
Table 1.10. Typical power levels of pulsed and CW EIKs at various frequencies
Pulsed EIK CW EIK
3000 W at 30 to 95 GHz 1500 W at 30 GHz
400 W at 140 GHz 100 W at 95 GHz
50 W at 220 GHz 30 W at 140 GHz
5 W at 280 GHz 1 W at 220 GHz
width of those EIKs is about 1%. Because of these device characteristics, EIKs have
been used for the following military, scientic and commercial applications:
re control radar radio astronomy
seeker satellite communication
illuminator cloud radar
tracking radar radar modelling
low noise cw radar fusion diagnostics
surveillance radar plasma heating
1.5 Backward Wave Oscillator (BWO)
1.5.1 Introduction
To nd the roots of the backward wave oscillator (BWO), one has probably to go
back to Rudolf Kompfner, one of the TWT inventors (see Sect. 1.3.1) and the pe-
riod around 1950 to 1952. In 1951 S. Millmann [31] published on a spatial har-
monics amplier for 6 mm wavelength where the energy transport in backward di-
rection with respect to the electron beam is clearly a device feature. Kompfner,
working in this period intensely on microwave sources capable of electronic fre-
quency tuning, mentions Millmanns work in his US patent led in 1952 [32] on
Backward Wave Tubes. How closely the work on the oscillator type of tubes was
related to the amplier devices is indicated by the patent led by Kompfner and
Williams [33] at the same date on Backward Wave Ampliers. His work published
together with N.T. Williams in 1953 [34] belongs also to this period of early in-
vestigations on Backward Wave Tubes. The investigated devices were based on lin-
ear electron beam/slow wave circuit interaction, using periodically disturbed wave
guides but also helices. H. Heffner contributed in this period with his Analysis on
Backward Wave Traveling Wave Tubes [35].
The application, these pioneers had in mind for the oscillator type of devices, was
frequency modulated signal generation. Interesting information on this period can be
found in J.R. Pierces review article on R. Kompfner and his work [36]. To avoid
confusion, it should be mentioned that in some literature the BWO is also called the
carcinotron.
1.5.2 BWO Operation Principle
The basic idea of the BWO is to use the space charge bunches of an electron beam
interacting along their path with a periodic delay line structure (the rst space har-
monic). The delay line can be of all types (helices, periodically corrugated wave-
guides, folded waveguides, interdigital lines, etc.). The preferred type depends, as in
TWTs, on the frequency range and power level of interest. When passing the periodic
structure, the space charge bunch initiates a wave propagating in backward direction.
If the phase velocity and the bunch velocity are such that the total phase delay in the
loops is
n
= n2, a backward wave oscillation can start.
In Fig. 1.44 the situation is illustrated by an electron beam passing along a tape
helix. The phase delay
1
in a closed loop between two adjacent gaps in the delay
line is given by the sum of the phase lag
e
=
e
p during the time the electron bunch
Fig. 1.44. Electrons adjacent to a tape helix interacting with the elds in the helix gaps.
1
,
2
,
and
3
denote loop phase shifts for one, two and three periods, respectively
needs to travel with v
e
to the next gap and the phase lag
w
=
w
p during the time
the wave travels back with its phase velocity v
p
. Here p is the pitch of the periodic
structure,
e
and
w
are the propagation constants of the electron bunch and the wave,
respectively. Thus, if
1
=
e
p +p = 2, and according to Fig. 1.44
2
= 22,
3
= 32, etc., one gets a feedback loop system with total gain eventually above 1
depending on the number of sub-loops or the length of the delay line, the coupling
strength between electron bunch and electric eld of the wave in the gap and the
current in the electron beam.
Thus, the synchronism condition for BW oscillation is given by
e
p = 2 p. (1.68)
This compares simply to the quite different synchronism condition for the forward
traveling wave amplication
e
p = p. (1.69)
It is clear that in a delay line with large dispersion, the synchronism condition (1.68)
is met optimally only by one frequency, which depends strongly on the propagation
constant
e
and thus the beam voltage. To oscillate, the gain loop condition must also
be met. Thus, a BWO oscillates only at a current higher than a certain start current
I
st
. Above this value, the summation of all gains in the loops yields >1. Within
the framework of the small signal theory, the condition for the start current is given
by
LC
st
= 2(2
5/3
) = 1.97 (1.70)
with C
st
being the Pierce parameter (see Sect. 1.3),
C
st
= (KI
st
/4V)
1/3
, (1.71)
at the start current for oscillation. Below this current no oscillation can occur. I
st
is
of the order of a few mA for a voltage V of a few thousand volts for conventional
O-type BWOs as sketched in Fig. 1.45.
To distinguish the O-type BWO where the magnetic eld is parallel to the elec-
tron beam axis (used for focusing), in Fig. 1.46 a M-type BWO (see also Sect. 1.6.3)
Fig. 1.45. O-type backward wave oscillator (O-carcinotron) with bilar helix as delay line
Fig. 1.46. M-type backward wave oscillator (M-carcinotron) with planar periodic delay line
Fig. 1.47. Folded waveguide type of BWO
is sketched, which is an E B device. Though the basic interaction physics and
synchronism condition is the same for O-type and M-type BWOs, the M-type BWO
is an E B device with the magnetic eld perpendicular to the beam axis and the
electric eld between delay line ground and sole.
A further example of an O-type BWO with a folded waveguide line is shown
in Fig. 1.47. These types or interdigital lines are the most common BWOs. Further
examples and a compact description of BWOs is, e.g. given by J. Arnaud in The
Encyclopedia of Electronics issued by Charles Susskind [37].
1.5.3 BWO Applications
The typical applications for O-type BWOs are local oscillators and frequency tune-
able microwave sources up to the THz range. There, 10 mW of CW power can be
obtained in the frequency range from 0.1 to 1.5 THz with a tuning range of 200 GHz.
The further development of THz BWO sources is a topic of increasing commercial
interest.
1.6 Magnetrons and Cross-Field Ampliers
1.6.1 Magnetrons
Introduction
Magnetrons are oscillators where all functions are grouped in one given volume be-
tween cathode and anode. Magnetron tubes are robust, compact and efcient. They
are used as the nal stage in ordinary transmitters (radars) but their main utilization
is in industrial microwave heating applications (Table 1.11).
Microwave Circuit
Magnetrons consist of two coaxial electrodes: an inner cathode of radius r
K
placed at
the centre, and an outer anode of radius r
A
(Figs. 1.48 and 1.49). A permanent mag-
net (at the origin of the designation of the magnetron) or an electromagnet is used
to create a magnetic eld B
Z
parallel to the axis [9]. The anode is machined with
an even number N of slots or resonating cavities which are coupled between them-
selves by their overow or leakage eld in the cathode/anode space. Consequently,
we can distinguish N/2 + 1 distinct resonance frequencies corresponding to phase
differences = 0, 2/N, 2/(N/2), 2/(N/3), between two adjacent slots or
cavities.
The modes are identied by the number n of times that the eld pattern is re-
peated in going around the anode once [7]. In that way, becomes 2n/N. In
Table 1.11. Survey on industrial magnetrons
8.59.6 GHz 200 kW peak/200 W average 22 kV 27.5 A 6.5 kg
2.45 GHz 5 kWCW 7.2 kV 0.95 A 4.3 kg
Fig. 1.48. Magnetron (interaction)
Fig. 1.49. Magnetron
practice, the most common mode is the mode, with n = N/2, that is, a phase
difference of between adjacent cavities. The distance between two adjacent cavi-
ties is l = n/N, which is equal to /2 if is the operating mode.
For a given resonance mode of frequency f = /2, the electromagnetic eld
and, for example, its E
component, varies along the anode (with a so-called ab-

scissa s) as described by the usual equation of a stationary wave: E
(s, t ) = A
cos(2s/) cos t . But, since a stationary wave can be considered as the sum of two
progressive waves moving in opposite directions, each one with an amplitude equal
to half of the original wave, the stationary wave equation can be rewritten as
E
= (A/2)[cos(2s/ +t ) +cos(2s/ t )] (1.72)

or
E
= (A/2)[cos (t +s/v
) +cos (t s/v
)], (1.73)
where
v
= f = /2. (1.74)
Finally, an observer (or an electron, for example) moving along s at a velocity v
=
v
will be in phase with one of the two waves, for example A/2 cos (t s/v
),
and will always see the same eld.
Magnetrons are therefore multi-resonating microwave structures offering a num-
ber of resonance modes, each with the equivalent of two progressive waves, only one
of which being taken into consideration for any given mode.
We will now consider the interaction process that takes place within magnetrons,
i.e. the transformation of electron movement energy into electromagnetic energy.
Interaction
The electrons emitted by the cathode would move radially towards the anode, but
thanks to the magnetic eld B
Z
, their trajectories are curved and somewhere they
become concentric (v
r
= 0) to the cathode and the anode [3]. At that point, the
electrons are subjected to an electrostatic force F
e
= eE
Anode
and a magnetic
force F
m
= e(v
xB
Z
), acting in opposing directions (Fig. 1.48). Then a radius r
exists, between r
K
and r
A
, for which the amplitudes of these two elds are equal
and characterized by v
r
= 0 and v
= E
A
/B
Z
. Theoretically, the electron can rotate
indenitely around the magnetron axis at this radius.
The combined forces of F
e
and F
m
also apply albeit very schematically to
all the electrons leaving the cathode at varying velocities. Also, all the electrons
experience, to varying extents, the effects of the electromagnetic eld and the space
charge of the other electrons. In any case it can be said that most of the electrons
rotate around in the space between cathode and anode about once, after which they
terminate either on the anode or the cathode. Naturally, these electrons are replaced
by those emitted non-stop by the cathode. The end result is a real cloud of electrons
rotating between the anode and the cathode at a velocity approximately v
[9].
V
A
(or E
A
) and B
Z
are then adjusted so that v
= v
, with v
= (lN/n)/2
the phase velocity, deducted from the previous expressions of l and of v
(1.74). As
explained above the rotating electrons always see the same microwave eld E
. If E
is an accelerating eld, they are continuously accelerated, and if E
is a decelerat-
ing eld, they are continuously slowed down. This electron accelerating/decelerating
is comparable to the TWT process, with the difference that, in the magnetron, the
electrons follow a closed circular path and electron bunches are now replaced by n
rotating electron arms or spokes, where n = N/2 if the mode is used, that is the
most common case.
The energy transfer process in the magnetron is a little bit more complicated than
that in the tubes examined in the previous sections. This takes place as follows: the
electrons in the spokes are attracted to the anode but, at the same time, repulsed by
the RF eld. In other words, they are continuously transferring their potential energy,
acquired via V
A
(or E
A
) to the RF eld.
The whole microwave structure, including N resonators, can be regarded as a
unique resonator, represented by an equivalent circuit similar to the one of the kly-
stron cavity. In Fig. 1.48, we notice not only the 3 elements, L, C and the loss conduc-
tance G
c
, but also the load G
L
. The load G
L
is directly experienced by the electrons
and converted from the nal load for example, the antenna or the microwaves oven
through the waveguides and the coupling loop or iris located in the bottom wall
of one of the resonators. The beam loading is neglected. The voltage V in Fig. 1.48
is the microwave voltage between two anode vanes and is related to the electromag-
netic eld E
(r = r
A
). Almost all the electrons are supposed to arrive at the anode,
that means on the top of the vanes, with a phase such that E
(r = r
A
) is maximum.
Therefore, their radial velocity is v
r
= E
(r = r
A
) max /B
Z
, while the azimuthal
velocity is given by v
= v
.
Moreover, we can write
E
(r = r
A
) = MV/l, (1.75)
where l is the distance between two vanes and M is the coupling factor between V
and the electrons, M sin(l/(2v
))/(l/(2v
)).
The kinetic power which denitively is lost of these electrons landing on the
anode, is given by
P
A
= I
o
(m/2e)[v
2
r
(r = r
A
) +v
2
] = I
o
(m/2e)
E
2
(r = r
A
) max
B
2
Z
+v
2
. (1.76)
According to the equivalent circuit, the output power P
OUT
delivered to the load G
L
is
P
OUT
=
1
2
NV
2
G
L
=
N
2
l
2
C
M
2
Q
L
E
2
(max) (1.77)
with Q
L
= C/G
L
= 1/(R/Q)G
L
and (1.75).
Thanks to the obvious expression P
OUT
= V
0
I
0
P
A
, we get the approximate
expression of the magnetron efciency [9]
= P
OUT
/V
0
I
0
=
1 mv
2
/2eV
0
1 +I
0
mM
2
Q
L
/B
2
Z
eNl
2
C
. (1.78)
The efciencies of magnetrons usually are quite large and very attractive, 30% up to
70% and even 80%.
A magnetron presents very few control or optimisation parameters: the high volt-
age V
A
and the magnetic eld B
Z
. This is why an operating zone is dened in the
plane (V
A
; B
Z
), where the magnetron correctly oscillates. Such a zone is located
between two limits [3, 9]:
the Hull parabola, below which (for B
Z
increasing and/or V
A
decreasing) the
electrons are no more intercepted by the anode and begin to rotate around the
cathode-anode space,
the Hartree line that determines the synchronism condition (for a given mode).
Typically, the operating point is quite close to the Hartree line.
From a technological and engineering point of view, the main parts and subassem-
blies look like the parts of the other microwave tubes, even if the geometries are
notably different (Fig. 1.49).
However, let us notice the frequent use of cold cathodes without direct heating
and based on the secondary emission. In this case, the primary electrons are the elec-
trons, whose emission conditions are such that they are coming back and bombard
the cathode (self-heating). Also quite all the magnetrons include straps, i.e. metallic
wires that connect the vanes or the resonators, which must be at the same electro-
magnetic instantaneous potential. At the mode, for instance, straps connect the
vanes 0, 2, 4, N 2, N. Their objective is to minimize the risk of oscillation and the
rise of unwanted oscillations or parasitic modes.
1.6.2 Cross-Field Ampliers (CFA)
In the magnetrons and, generally speaking, in all the cross-eld devices also called
M-type tubes the electrons move under the inuence of the perpendicular static
electric eld E V
0
/(r
A
r
K
) and magnetic eld B
Z
(Fig. 1.50). As a result, the
Fig. 1.50. Cross-eld amplier (CFA)
motion is perpendicular to both elds and, when the values of V
0
and B
Z
are carefully
adjusted, the electron beam or clouds rotate in the narrow space KA between the
cathode and the anode. If now the microwave circuit at the same time the anode
is interrupted to provide input and output connections, we deal no more with a
magnetron oscillator but with a cross-eld amplier (CFA).
In CFAs, resonators are preferably replaced by a specic slow wave structure,
whose ends are terminated by the input and output couplers, which must be matched
at least in the expected frequency bandwidth [3]. Between these two ends, a specic
short circular drift space is located in order to minimize any coupling between input
and output, but also to provide a necessary electron bunching debunching process
(Fig. 1.50).
As soon as the drive power P
d
is injected, the RF elds give rise to bunches
or spokes which rotate in the KA space. This process is nearly the same as in the
TWTs, where the beam is coupled to a growing forward or to a backward wave.
At the same time, the energy transfer inside the spokes is the same as in the mag-
netrons [9].
In the case of a forward wave, the electrons and the microwave energy ow in the
same direction, and the spokes travel in synchronism with the circuit wave, causing
it to grow. In the case of a backward wave, the electrons and the microwave energy
ow in opposite directions, and the amplication is obtained when P
d
is injected at
the circuit output. The slow forward wave circuits are often helix structures, when
the backward circuits use strapped bar lines or interdigital structures.
It should be remembered that in CFAs and in magnetrons, the slowwave structure
acts not only as the anode but also as the collector, and in spite of a usual high
efciency, a lot of power is dissipated in the structure, which must present excellent
thermal capabilities.
The CFAs, also known as amplitrons, are characterized by impressive output
powers (for example, 700 kW peak and 10 kW average), efciencies around 70%
and instantaneous bandwidths in the order of 5 to 10%. The cathode voltages V
0
are very acceptable (50 kV) with perveances of P 2 10
6
. At the same time,
the compactness and the lightness are other impressive advantages. The weak points
remain not only the risk of oscillations and the possible noisy behaviour but also the
low gains of 11 dB [3, 9].
As an important precaution, the RF drive pulse P
d
must always be applied be-
fore V
0
, for two main reasons. First, if the CFA employs a cold cathode now the
most frequent situation the emission process may fail to build up. As a result, the
modulator will be unloaded and so excessive voltage, arcing and damage. Secondly,
the RF drive anterior to V
0
and strictly present throughout the voltage pulse helps to
control the space charge and prevents oscillations at band edge and the generation of
broadband noise.
As for the ends of the slow wave circuit, the input VSWR must be very low
(1.2:1), not only in the frequency band but also outside, in order to avoid any start-
ing of oscillation during the voltage pulse and during the rise and the fall of V
0
.
The resulting spurious output is referred to as rabbit ears because of the way they
appear on a time display of P
OUT
.
In addition, it should be pointed out that the transmission loss of the signal
through the CFA in absence of beam current is very low. Then, the CFA can be
considered transparent, and this advantageous characteristic is used in many system
applications.
1.6.3 Cross-Field Backward Wave Oscillator (MBWO)
The M-type backward wave oscillators, sometimes called carcinotrons, are voltage
tunable oscillators, mainly used as ECM (electronic counter measures) noise gener-
ators but also swept signal sources and drivers for high power transmitters. They are
able to deliver, for example, 50 to 100 W with efciencies of 50% from the L to the
X band. They look like the former CFAs, but they use an injected thin beam, and
the principle of their operation is close to the one of the backward oscillator (BWO)
derived from the conventional TWTs [3].
The electron emission is not distributed over the total surface of the inner elec-
trode, but uses a complete gun structure with a cathode of limited length, a grid and
a controllable anode, this gun structure being at one end of the circular interaction
space. The electrons are emitted approximately radially from the cathode towards
the controllable anode but, because of the strong axial magnetic eld B
Z
, they are
curved at 90
and enter the interaction region between the sole and the slow wave
structure (Fig. 1.51). The sole voltage is a little bit lower than the anode potential in
order to provide a radial electric eld E
r0
. Thanks to both E
r0
and B
Z
, the velocity of
the electrons becomes E
r0
/B
Z
and is adjusted to get a thin electron beam following
a circular path.
Usually the circuits are folded waveguides or interdigital delay lines, where the
signal grows as it travels from the collector end to the electron gun end. The mi-
crowave output connector is located at this electron gun end. At the collector end,
Fig. 1.51. Cross-eld backward wave oscillator (MBWO)
the MBWOs require a relatively large attenuator, well bonded to the heat sink of the
tube, to absorb the unavoidable microwave reections.
The electrons interact with a backward-wave space harmonic of the circuit, the
energy on the circuit owing opposite to the direction of the electron motion. This
easily provides the feedback necessary for oscillation. The circuit is terminated at
the collector end, but the microwave output is removed at the gun end.
Despite this short presentation, it is understandable that the design and the tech-
nology of MBWOs are critical, but the above mentioned performances remain very
attractive.
1.7 Fast-Wave Devices
Conventional microwave tubes operate with longitudinal phase bunching of the in-
teracting electrons and are commonly referred to as slow-wave devices, since in
some way, the RF structure is congured so that the phase velocity of the electromag-
netic (EM) eld is slowed down to be a little bit lower than the electron beamvelocity
and is thus less than the velocity of light. This is most evident in the helix-type TWT
and BWO where the transverse RF circuit dimensions are typically a fraction of the
wavelength. The dimensions of the interaction circuits of klystrons, extended inter-
action tubes, magnetrons and cross-eld ampliers also have dimensions in the order
of the wavelength or smaller. The inherent power limitation with these conventional
microwave tubes is the tremendous decrease in the dimensions of the interaction
structure with increasing frequency (that is, decreasing wavelength, see Fig. 1.52).
Therefore, the possibility of extracting high average power in the millimeter (mm)
Fig. 1.52. Chart showing the limitation of conventional microwave tubes for the generation of
higher power levels at elevated frequencies. Here, FWDs are fast-wave devices
wavelength range has been ruled out. In the case of two-level far-infrared LASERs,
the possible average power, which can be generated, decreases with increasing wave-
length since the energy between the interacting quantum levels (hf = 0.41 meV at
100 GHz) becomes smaller than the thermal energy quantum at room temperature
(kT = 25 meV). To close the gap for the achievable average output power in the
mm- and sub-mm wavelength range (Fig. 1.1), new interaction mechanisms had to
be discovered, which operate in highly overmoded interaction circuits, i.e. in fast
wave circuits where the electron beam is placed well away from the RF structure.
With larger dimensions, the power-handling capability in enlarged. Since fast waves
have a phase velocity larger than the velocity of light, the phase bunching mechanism
in such fast-wave devices must be generated by a transverse interaction, e.g. (1) the
electroncyclotron interaction in a longitudinal magnetic eld (electron cyclotron
maser: ECM) or (2) the electron undulation in a wiggler eld (free-electron laser:
FEL).
The origin of the ECMs traces back to the late 1950s, when three investigators
began to examine theoretically the generation of microwaves by the ECM interac-
tion: R. Twiss in Australia [38], J. Schneider in the US [39] and A. Gaponov in
Russia [40]. A short note on the possibility to use the rotational energy of a helical
electron beam for microwave generation was published by German H. Kleinwchter
in 1950 [41]. In early experiments with devices of this type, there was some de-
bate about the generation mechanism and the relative roles of fast-wave interactions
mainly producing azimuthal electron bunching and slow-wave interactions mainly
producing axial bunching. The predominance of the fast-wave ECM resonance with
its azimuthal bunching in producing microwaves was experimentally veried in the
mid-1960s in the US [42] (where the term electron cyclotron maser was apparently
coined) and in Russia [43] (where the term gyrotron was introduced).
FELs exploit relativistic electron beam technology to upshift the electron wiggle
frequency. In this respect, perhaps a more descriptive name is that introduced by
R.M. Phillips: UBITRON for an undulated beam interaction electron tube. Long
wavelength FELs ( 0.5 mm) are called free-electron masers (FEMs).
1.7.1 Interaction Principles
Fast-wave devices in which the phase velocity v
ph
of the EM wave is greater than the
speed of light c, generate or amplify coherent EMradiation by stimulated emission of
bremsstrahlung from a beam of relativistic electrons. The electrons radiate because
they undergo oscillations transverse to the direction of beam motion by the action of
an external force (eld). For such waves the electric eld is mainly transverse to the
propagation direction (TE or HE modes).
The condition for coherent radiation is that the contribution from the electrons
reinforce the original emitted radiation in the oscillator or the incident EM wave in
the amplier. This condition is satised if a bunching mechanism exists to create
electron density variations of a size comparable to the wavelength of the imposed
EM wave. To achieve such a mechanism, a resonance condition must be satised
between the periodic motion of the electrons and the EM wave in the interaction
region [4447]
k
z
v
z

= s, s = 1, 2, . . . (k
z
v
z
= Doppler term), (1.79)
where and k
z
are the wave angular frequency and characteristic axial wavenumber,
respectively, v
z
is the translational electron drift velocity, is an effective frequency,
which is associated with macroscopic oscillatory motion of the electrons, and s is the
harmonic number.
1. In ECMs, the EM energy is radiated by relativistic electrons gyrating in an ex-
ternal longitudinal magnetic eld. In this case, the effective frequency corre-
sponds to the relativistic electron cyclotron frequency
c
=
co
/ with
co
= eB
o
/m
o
and = [1 (v/c)
2
]
1/2
= 1 +eV
o
/m
o
c
2
, (1.80)
where e and m
o
are the charge and rest mass of an electron, is the relativistic
factor, B
o
is the magnitude of the guide magnetic eld and V
o
is the acceleration
voltage. The nonrelativistic electron cyclotron frequency is f
o
/GHz = 28B
o
/T.
A group of relativistic electrons gyrating in a strongmagnetic eld will radiate
coherently due to bunching caused by the relativistic mass dependence of their
gyration frequency. Bunching is achieved because, as an electron loses energy,
its relativistic mass decreases and it thus gyrates faster. The consequence is that a
small amplitude waves electric eld, while extracting energy from the particles,
causes them to become bunched in gyration phase and reinforces the existing
wave electric eld. The strength of the magnetic eld determines the radiation
frequency.
The phase bunching process can be most easily understood in a reference frame
in which the axial velocity vanishes. In Fig. 1.53 an annular electron beam with
radius R
e
is depicted in this frame. The electrons arranged around this annu-
lus execute circular orbits with the Larmor radius r
L
= v
/
0
. Typically
r
L
R
e
. Initially, the phase of the electrons in their cyclotron orbits is ran-
dom, as shown in Fig. 1.53 (left).
Fig. 1.53. Principle of azimuthal phase bunching in an annular electron beam with initial
random phasing of electrons in their cyclotron orbits (left) and with electron bunched in phase
in their cyclotron orbits (right)
In the presence of a transverse RF electric eld characteristic of a microwave
cavity TE
mn
mode, the electrons will be accelerated or decelerated. As a specic
example, Fig. 1.53 depicts an electric eld with only an azimuthal component,
as is characteristic of TE
0n
modes in circular waveguide. With random phasing,
there is no net energy exchange. Electron 1 will be decelerated by the azimuthal
electric eld and thus lose energy, while electron 2 will be accelerated and hence
gain an equal amount of energy.
Phase bunching and net transfer can occur if the wave frequency is slightly larger
than the initial value of the cyclotron frequency
eB
0
m
e
0
= ()
> 0, (1.81)
where the subscript 0 denotes the initial value and the prime denotes the ref-
erence frame in which the electron axial velocity vanishes. Then, as electron 2
gains energy, its cyclotron frequency decreases; this electron falls farther from
resonance gaining less energy on each successive cycle. On the other hand, elec-
tron 1, which initially loses energy, experiences an increasing value of
c
and
moves closer to exact resonance with the electric eld, thereby using an increas-
ing amount of energy on each successive cycle. An instability develops in which
the wave energy grows in time and the electrons bunch in phase within their
cyclotron orbits, as shown in Fig. 1.53 (right).
2. In the case of a spatially periodic magnetic or electric eld (undulator/wiggler),
the transverse oscillation frequency
b
(bounce frequency) of the moving charges
is proportional to the ratio of the electron beam velocity v
z
to the wiggler eld
spatial period
w
. Thus, the operating frequency of such devices, an example
of which is the FEM [4851], is determined by the condition that an electron
in its rest frame observes both the radiation and the periodic external force at
the same frequency. If the electron beam is highly relativistic (v
ph

= v
z

= c),
the radiation will have a much shorter wavelength than the external force in the
laboratory frame (

=
w
/2
2
so that

= 2
2
b
). Therefore, FEMs are ca-
pable of generating EM radiation of very short wavelength determined by the
relativistic Doppler effect. The bunching of the electrons in FEMs is due to the
perturbation of the beam electrons by the ponderomotive potential well which is
caused by beating of the EM wave with the spatially periodic wiggler eld. It
is this bunching that enforces the coherence of the emitted radiation
b
= k
w
v
z
, k
w
= 2/
w
. (1.82)
In the case of the ECMs and FEMs, unlike most conventional microwave sources
and lasers, the radiation wavelength is not determined by the characteristic size of
the interaction region. Such fast-wave devices require no periodically rippled walls
or dielectric loading and can instead use a simple hollow-pipe oversized waveguide
as a circuit. These devices are capable of producing very high power radiation at
cm-, mm-, and sub-mm wavelengths since the use of large waveguide or cavity cross
sections reduces wall losses and breakdown restrictions, permitting the passage of
larger, higher power electron beams. It also relaxes the constraint that the electron
beam in a single cavity can only remain in a favourable RF phase for half of a RF
period (as in klystrons and other devices employing transition radiation). In con-
trast with klystrons, the reference phase for the waves in fast-wave devices is the
phase of the electron oscillations. Therefore, the departure from the synchronism
condition, which is given by the transit angle = ( k
z
v
z
s)L/v
z
, can now
be of order 2 or less, even in cavities or waveguides that are many wavelengths
long.
1.7.2 Dispersion Diagrams of Fast-Wave Interactions
Many congurations can be used to produce coherent radiation based on the ECM
instability. The departure point for designs based on a particular concept is the wave
particle interaction. Dispersion diagrams, also called k
z
plots or Brillouin dia-
grams [4447], show the region of cyclotron interaction (maximum gain of the in-
stability) between an EM mode and a fast electron cyclotron mode (fundamental or
harmonic) as an intersection of the waveguide mode dispersion curve (hyperbola)
2
= k
2
z
c
2
+k
2
c
2
(1.83)
with the beam-wave resonance line (straight) given by (1.79). In the case of a device
with cylindrical resonator, the perpendicular wavenumber is given by k
= X
mn
/R
o
,
where X
mn
is the nth root of the derivative of the corresponding Bessel function
(TE
mn
modes) and R
o
is the waveguide radius. Phase velocity synchronism of the
two waves is given in the intersection region. The interaction can result in a de-
vice that is either an oscillator or an amplier. In the following subsections, dif-
ferent ECM devices and the FEM are classied according to their dispersion di-
agrams. Table 1.12 presents a review of ECM devices (also called gyro-devices)
and a comparison with the corresponding conventional linear beam (O-type) de-
vices.
Table 1.12. Review of the interaction circuits of gyro-devices and comparison with corre-
sponding conventional linear-beam (O-type) devices
Gyrotron Oscillator and Gyro-Klystron Amplier
Gyrotron oscillators and gyro-klystrons are devices which usually utilize only
weakly relativistic electron beams (V
o
< 100 kV, < 1.2) with high transverse
momentum (pitch angle = v
/v
z
> 1) [46, 47]. The wavevector of the radia-
tion in the cavity is almost transverse to the direction of the external magnetic eld
(k
k
z
, and the Doppler shift is small) resulting, according to (1.79) and (1.80),
in radiation near the electron cyclotron frequency or one of its harmonics
= s
c
, s = 1, 2, . . . . (1.84)
In the case of cylindrical cavity tubes, the operating mode is close to cutoff (v
ph
=
/k
z
c), and the frequency mismatch s
c
is small but positive in order to
achieve correct phasing, i.e. keeping the electron bunches in the retarding phase. The
Doppler termk
z
v
z
is of the order of the gain width and is small compared with the ra-
diation frequency. The dispersion diagrams of fundamental and harmonic gyrotrons
are illustrated in Figs. 1.54 and 1.55, respectively. The velocity of light line is deter-
mined by = ck
z
. For given values of and R
o
, a mode represented by X
mn
and
oscillating at frequency is only excited over a narrow range of B
o
. Quasi-optical
gyrotrons employ a FabryPerot mirror resonator perpendicular to the electron beam,
also providing k
k
z
[46].
Cyclotron harmonic operation reduces the required magnetic eld for a given
frequency by the factor s. At low voltages, the number of electron orbits required for
efcient bunching and deceleration of electrons can be large, which means that the
resonant interaction has a narrow bandwidth and that the RF eld may have moderate
amplitudes. In contrast with this, at high voltages, electrons should execute only
about one orbit. This requires correspondingly strong RF elds, possibly leading
to RF breakdown, and greatly broadens the cyclotron resonance band, thus making
possible an interaction with many parasitic modes.
Fig. 1.54. Dispersion diagram of gyrotron oscillator (fundamental resonance)
Fig. 1.55. Dispersion diagram of harmonic frequency gyrotron oscillator
Cyclotron Autoresonance Maser (CARM)
In a gyrotron with a highly relativistic beam (1 MeV), an efcient interaction will
lead to an average energy loss in the order of the initial electron energy. As a result,
the change in the gyrofrequency is much greater than in the mildly relativistic case. It
is therefore desirable to identify the condition under which such a highly relativistic
electron beam remains in synchronism with the RF eld. A possibility for achieving
synchronism is to utilize the interaction of electrons with EM waves propagating
with a phase velocity close to the speed of light in the direction of the magnetic
eld. In this case, the Doppler shift term k
z
v
z
is large, and the appropriate resonance
condition is
= k
z
v
z
+s
c
. (1.85)
If v
ph

= c, the increase in cyclotron frequency due to extraction of beam energy
(decrease of ) nearly compensates the decrease in the Doppler shifted term. There-
fore, if the resonance condition is initially fullled, it will continue to be satis-
ed during the interaction. This phenomenon is called autoresonance, and the cy-
clotron maser devices operating in the relativistic Doppler-shifted regime are called
cyclotron autoresonance masers [52]. Figure 1.56 shows how the Brillouin diagram
of the fast cyclotron wave changes during the autoresonance interaction such that the
working frequency remains constant even though both
c
and v
z
are based on the
same instability mechanism as that of the gyrotron but changing. The CARM inter-
action corresponds to the upper intersection and is operated far above cutoff. The
instability is convective, so a feedback, e.g. by a Bragg resonator [52], is required
Fig. 1.56. Dispersion diagram of the cyclotron autoresonance maser (CARM)
for an oscillator; and it is necessary to carefully discriminate against the other inter-
actions corresponding to the lower frequency intersection in the dispersion diagram
(Fig. 1.56). The problem can be alleviated by employing the fundamental TE
11
or
(HE
11
hybrid mode) and properly choosing system parameters to be within the sta-
bility limit. Compared to a gyrotron, there is a large Doppler frequency upshift of
the output (

=
2
c
) permitting a considerably reduced magnetic eld B
o
. Since
the axial bunching mechanism can substantially offset the azimuthal bunching, the
total energy of the beam and not only the transverse component is available for RF
conversion.
In contrast to the gyrotron, the CARM has an electron beamwith low to moderate
pitch angle ( < 0.7). The efciency of CARMs is extremely sensitive to spread in
the parallel beam velocity. The velocity spread v
z
/v
z
must be lower than 1% to
achieve the full theoretically expected efciency of 40% [52].
Gyro-Travelling Wave Tube (Gyro-TWT)
From the theoretical point of view, the gyro-TWT differs from the CARM only in
regimes of operation. The gyro-TWT utilizes a moderately relativistic electron beam
to interact with a fast waveguide mode in an oversized structure near the grazing
intersection of the frequency versus wavenumber plot (see Fig. 1.57) where the
resonance line is tangent to the EM mode. This produces high gain and efciency
because the phase velocities of the two modes are nearly matched and the group
velocity of the waveguide mode is nearly equal to v
z
. In the gyro-TWT regime
(/k
z
c), the axial bunching mechanism is too weak to be of any signicance.
To benet from autoresonance, the cutoff frequency should be reduced relative to the
cyclotron frequency.
Gyro-Backward Wave Oscillator (Gyro-BWO)
If the electron beam and/or magnetic eld is adjusted so that the straight fast-wave
beam line crosses the negative k
z
-branch of the waveguide mode hyperbola (see
Fig. 1.57. Dispersion diagram and scheme of interaction circuit of gyro-TWT amplier
Fig. 1.58. Dispersion diagram and scheme of interaction circuit of gyro-BWO
Fig. 1.58), then an absolute instability (internal feedback) with a backward wave
occurs. In the gyro-BWO the frequency of operation is now governed by the slope
of the line, which is a function of v
z
, and thus of the beam acceleration voltage V
o
.
Consequently, just as in the case of other BWOs (e.g. carcinotron), the frequency
of oscillations can be continuously changed very fast over a broad range, using
V
o
in place of B
o
. However, there is a Doppler down shift in frequency (
c
/2 <
<
c
), so that very high magnetic elds are required for high frequency opera-
tion.
Fig. 1.59. Dispersion diagram and scheme of interaction of FEM
Free-Electron Maser (FEM)
The resonance condition in FEMs is given by
k
z
v
z
= k
w
v
z
with k
w
= 2/
w
, (1.86)
where
w
is the spatial period of the wiggler. The generated radiation has a much
shorter wavelength than the external force in the laboratory frame:

=
w
/2
2
.
The basis FEM conguration and the corresponding dispersion diagram is shown in
Fig. 1.59. Electrons in the injected electron beam undulate in the periodic magnetic
eld B
w
of the wiggler.
1.7.3 Gyrotron Oscillator
Gyrotron oscillators (gyromonotrons or simply gyrotrons) were the rst ECMs to
undergo major development. Increases in device power were the result of Russian
developments from the early 1970s in magnetron injection guns, which produce elec-
tron beams with the necessary transverse energy (while minimizing the spread in
transverse energies) and in tapered, open-ended waveguide cavities that maximize
efciency by tailoring the electric eld distribution in the resonator [46]. Typical
conventional gyrotrons are built as shown schematically in Fig. 1.60. A gyrotron can
be described as follows. The magnetron injection gun produces an annular electron
beam with the desired beam parameters. The beam is transported to the interaction
region, where the interaction cavity converts a fraction of the beam power to RF
power. In case of axial output coupling, the spent beam will be collected on the uni-
form output waveguide section after the uptaper, and the RF power in the TE
mn
mode
Fig. 1.60. Schematic of gyrotron with solenoids, magnetron injection gun with annular emitter
ring, beam tunnel, cylindrical interaction cavity with cutoff section and output taper, cylindri-
cal output waveguide and RF window: a with axial output coupling; b with radial output
coupling [47]
is coupled through the axial output vacuum window. In the case of radial output cou-
pling, a quasi-optical mode converter is connected to the output waveguide and it
transforms the rotating TE
mn
mode with an axial power ow to a Gaussian mode
with a radial power ow. The power is then transmitted through a radially located
vacuum window, and the spent beam is dissipated on the collector.
A strong externally applied magnetic eld in the interaction region is chosen
such that the cyclotron frequency or one of its harmonics is close to the frequency
of the RF eld in the beam frame of reference. The interaction region consists of
an open-ended waveguide cavity, usually with a circular transverse cross-section.
The electrons in the beam must have a substantial transverse velocity v
and the
usual longitudinal velocity v
z
. Most of this transverse velocity comes as a result of
adiabatic compression resulting from the increasing magnetic eld leading to the in-
teraction region. The nal ratio of transverse to longitudinal velocity = v
/v
z
in
the interaction region is typically between 1 and 2 for gyrotrons that use magnetron
injection guns with thermionic cathodes, mostly in temperature limited operation.
The electrons follow helical paths around the lines of force of the external eld. At
high frequencies the use of superconductive magnets is mandatory. By variation of
the magnetic eld, a sequence of discrete modes can be excited. The frequency scal-
ing is determined by the value of B
o
/ . Modern high power high order volume mode
gyrotron oscillators for fusion plasma applications employ an internal quasi-optical
mode converter with lateral microwave output, a single-stage depressed collectors
(SDC) for energy recovery and a chemical vapour deposition (CVD) diamond win-
dow.
Gyrotrons for Fusion Plasma Heating
At present, gyrotron oscillators are mainly used as high power mm-wave sources
for electron cyclotron heating (ECH) applications and for diagnostics of magneti-
cally conned plasmas in controlled thermonuclear fusion research [53]. Long-pulse
(a few sec) gyrotrons utilizing open-ended cylindrical resonators which generate out-
put powers of 100500 kW per unit, at frequencies in the range 28160 GHz, have
been used very successfully for plasma formation, ECH and local current density
prole control by noninductive electron cyclotron current drive (ECCD) in tokamaks
and stellarators. Gyrotron complexes with total power of up to 4.5 MW have been
installed. As experimental devices become larger and operate at higher magnetic
elds (B
o
= 6 T) and higher plasma densities (n
eo
= 12 10
20
/m
3
) in steady
state, present and forthcoming ECRH requirements call for gyrotron output pow-
ers of at least 1 MW, CW at frequencies ranging from 100170 GHz. Since efcient
ECH&CD need axisymmetric, narrow, pencil-like mm-wave beams with well de-
ned polarization, single-mode emission is necessary in order to generate a TEM
00
Gaussian beam mode at the plasma torus launching antenna. Single-mode mm-wave
gyrotron oscillators capable of high average power, 0.51 MWper tube, in long-pulse
or CW operation, are currently under development in several scientic and industrial
laboratories [46, 47, 53, 54].
Table 1.13 and Fig. 1.61 summarize the present status of long-pulse gyrotrons
for EC H&CD applications at 110170 GHz [54]. The maximum pulse length of
commercially available 140 GHz megawatt-class gyrotrons employing synthetic di-
amond output windows is 30 minutes (CPI and European FZK-CRPP-CEA-TED
collaboration). The world record parameters of the European 140 GHz gyrotron are:
0.92 MW output power at 30 min pulse duration, 97.5% Gaussian mode purity and
44% efciency, employing a single-stage depressed collector for energy recovery.
A maximum output power of 1.2 MW in 4.1 s pulses was generated with the JAEA-
TOSHIBA 110 GHz gyrotron. The Japan 170 GHz ITER gyrotron holds the energy
world record of 2.16 GJ (0.6 MW, 60 min) for tubes with an output power of more
than 0.5 MW. The Russian 170 GHz ITER gyrotron achieved 0.5 MW with a pulse
duration of 300 s. Figure 1.62 shows the 118 GHz and 140 GHz long-pulse gyrotrons
developed by the EU team.
To achieve output powers in excess of 2 MW at frequencies around 170 GHz and
long pulses, it is necessary to employ a coaxial-cavity geometry. A maximum out-
put power of 2.2 MW (1 ms pulse length) was obtained at FZK with an efciency of
28%. At the nominal output power of 1.5 MW the efciency increases from 30% to
48% in operation with an SDC [47]. Two MW synthetic diamond windows are feasi-
ble. At power levels around 1 MW the coaxial-cavity gyrotron can probably operate
at frequencies >300 GHz. This means that the gyrotron oscillator is the dominant
mm-wave source for high power ECH and ECCD. Even in the case of local non-
inductive current drive for suppression of plasma instabilities in future tokamak fu-
sion reactors, the gyrotron is a good option, since ultra broadband Brewster windows
and specic magnets allow stepwise frequency tuning in the seconds time-scale in
the full D-band (110170 GHz) [47]. Diagnostic gyrotrons deliver P
OUT
= 40 kW
Table 1.13. Development status of long pulse gyrotron oscillators for fusion plasma applica-
tions at 110170 GHz [54]
Institution Frequency Cavity Output Power Efciency Pulse Fusion device
(GHz) mode mode (MW) (%) length(s)
CPI, 110 TE
22,6
TEM
00
1.05 31 5.0 D III-D
Palo Alto 0.6 31 10.0 D III-D
140 TE
28.7
TEM
00
0.9 33 (SDC) 1800 W7-X
GYCOM-M 110 TE
19,5
TEM
00
0.93 36 2.0 D III-D
(TORIY, IAP), 0.5 35 5.0 D III-D
Moscow, 0.35 33 10.0 D III-D
Nizhny 140 TE
22,6
TEM
00
0.96 36 1.2 ASDEX-U
Novgorod 0.54 36 3.0 W7-AS
170 TE
25,10
TEM
00
0.9 44 (SDC) 21 ITER
0.5 40 (SDC) 300 ITER
GYCOM-N 140 TE
22,6
TEM
00
0.8 32 0.8 W7-AS
(SALUT, IAP), 0.88 50.5 (SDC) 1.0 W7-AS
N. Novgorod 158.5 TE
24,7
TEM
00
0.5 30 0.7 T 10
JAEA, 110 TE
22,6
TEM
00
1.2 38 (SDC) 4.1 JT 60-U
TOSHIBA, 1.0 36 (SDC) 5.0 JT 60-U
Naka, 0.5 34 (SDC) 16.0 JT 60-U
Otawara 170 TE
31,8
TEM
00
1.0 43.4 (SDC) 800 ITER
0.6 45.5 (SDC) 3600 ITER
THALES, 118 TE
22,6
TEM
00
0.53 32 5.0 TORE SUPRA
CEA, CRPP, 0.35 23 111 TORE SUPRA
FZK, 140 TE
28,8
TEM
00
1.0 49 (SDC) 12 W7-X
EUROPE 0.92 44 (SDC) 1800 W7-X
Fig. 1.61. State-of-the-art of long-pulse gyrotrons for EC H&CD applications at 110170 GHz
with = 40 s at frequencies up to 650 GHz ( 4%) [53]. Operating at the fun-
damental or the second harmonic of the electron cyclotron frequency enables the
gyrotron to act as such a medium power step tunable, mm-wave, sub-mm wave and
Fig. 1.62. CEA/CRPP/FZK/TED-gyrotrons. Left: 118 GHz, 0.35 MW, 11 s with liquid nitro-
gen Cryo-Window. Right: 140 GHz, 0.92 MW, 1900 s with CVD-diamond window and SDC
THz source in the frequency range from 38 GHz (fundamental) to 1005 GHz (TE
6,11
mode, second harmonic) [54].
Gyrotrons for Industrial Applications in Materials Processing
Recently, CW gyrotrons have also been successfully utilized in materials processing
(e.g. sintering of advanced nanocrystalline- and piezo-ceramics, surface hardening,
joining or dielectric coating of metals and alloys) and in plasma chemistry [53].
Such technological applications require sources with the following parameters: f
24 GHz, P
OUT
= 1030 kW, 30%. The present state-of-the-art of industrial CW
gyrotrons for technological applications is summarized in Table 1.14 [54].
The use of gyrotrons appears to be of great interest if one can realize a relatively
simple, low cost device which is easy to use (such as a magnetron). Gyrotrons with
low magnetic eld (operating at the second harmonic of the electron cyclotron fre-
quency) which can be provided by a permanent magnet system, low anode voltage,
high efciency and long lifetime are under development (see Table 1.14).
1.7.4 Gyro-Ampliers
Bunching of electrons in gyro-devices has much in common with that in conven-
tional linear electron beam devices, namely, klystron, TWT, twystron and BWO. In
both cases the primary energy modulation of electrons gives rise to bunching (az-
imuthal or longitudinal) which is inertial. The bunching continues even after the
primary modulation eld is switched off (at the drift sections of klystron-type and
twystron-type devices). This analogy suggests the correspondence between linear-
beam (O-type) devices and various types of gyro-devices (Table 1.12).
Table 1.14. Performance parameters of industrial CW gyrotron oscillators for technological
applications [54]
Institution Frequency Cavity Output Power Effciency V
o
Magnet
(GHz) mode mode (kW) (%) (kV)
CPI, 28 TE
02
TE
02
15 38 40 room temp.
Palo Alto 28(2
c
) TE
02
TE
02
10.8 33.6 30 room temp.
60 TE
02
TE
02
30 38 40 cryo. mag.
CPI, NIFS 84 TE
15,3
TEM
00
50 14 80 cryo. mag.
GYCOM/ 24.15(2
c
) TE
11
TE
11
3.5 23 12 room temp.
PM, 116 kg
IAP, 24.15 TE
32
TE
32
36 50 33 room temp.
Nizhny 23(2
c
) TE
12
TE
12
13 50 25 room temp.
Novgorod 28 32 25 room temp.
28.3(2
c
) TE
12
TE
12
12 20 25 PM, 68 kg
30(2
c
) TE
02
TE
02
10 42 26 room temp.
30 35 26 room temp.
37.5 TE
62
TEM
00
20 35 30 cryo. mag.
83 TE
93
TEM
00
1040 3040 2530 cryo. mag.
MITSUBISHI, 28(2
c
) TE
02
TE
02
15 38.7 21 PM, 600 kg
Amagasaki tapered B
UNIV. Fukui 300 TE
22,8
TEM
00
2.0 11 15 cryo. mag.
Figure 1.63 shows the cross-section and the photograph of a two cavity gyro-
klystron amplier. Advanced devices use several staggered cavities (up to 5) and
optimized shape of magnetic eld. The state-of-the-art of weakly relativistic gyro-
klystrons is given in Table 1.15 [54].
Table 1.16 summarizes the status of the high-power relativistic gyro-klystron
development at the University of Maryland [54]. For comparison, the experimen-
tal results of the SLAC periodic permanent magnet 11.4 GHz klystron are: 75 MW
output power with 48% efciency and 55 dB gain at a pulse duration of 1.5 s.
The circuit employed in a gyro-TWT consists simply of a modestly oversized
waveguide. Since no resonant structures are present, the gyro-TWT is potentially ca-
pable of much larger bandwidth than a gyro-klystron. Recent devices employ tapered
magnetic eld, interaction circuit and two partially loaded stages in order to optimize
the beam-wave interaction along the waveguide [55].
The sensitivity to velocity spread can be strongly reduced by coupling between
the second harmonic cyclotron mode of a gyrating electron beam and the radiation
eld in the region of near innite phase velocity over a broad bandwidth by using a
cylindrical waveguide with a helical corrugation on its inner surface [56]. The state-
of-the-art of weakly relativistic pulse gyro-TWTs is given in Table 1.17.
The gyro-twystron, a hybrid device, is derived from the gyro-klystron by extend-
ing the length of the drift section and replacing the output cavity with a sligthly ta-
pered waveguide section like in a gyro-TWT. The output waveguide section is excited
Fig. 1.63. Schematic cross-section and photograph of a pulsed gyro-klystron amplier (CPI,
formerly Varian)
Table 1.15. Weakly relativistic gyro-klystron experimental results
Institution Frequency Mode No. of Power Effciency Gain BW
(GHz) cavities (kW) (%) (dB) (%)
CPI, Palo Alto 10 (2
c
) TE
01
3 20 8.2 10 0.2
28 TE
01/02
2 76 9 30 0.2
CPI, Litton, NRL, 93.8 TE
01
4 118 (10 av.) 29.5 24.7 0.64
UNIV. 5 130 (10 av.) 33 39.5 0.75
MARYLAND
GYCOM- M 35.2 TE
02
2 750 (5 av.) 24 20 0.6
(TORIY), 2 350 32 19 0.9
Moscow 35.0 TE
01
3 250 (1.2 av) 35 40 1.4
IAP, 15.8 TE
02
3 160 40 30 0.5
Nizhny 32.4 (2
c
) TE
02
3 300 22 22 0.1 PM, 350 kg
Novgorod 35.1 (2
c
) TE
02
2 258 18 17 0.3
93.5 TE
02
2 207 30 21 0.2
93.2 TE
02
3 340 24.5 23 0.3
IAP, ISTOK 91.6 TE
01
4 2.5 (CW) 25 31 0.36
NRL, 4.5 TE
10
3 54 30 30 0.4
Washington, D.C. 34.9 TE
01
3 225 31 30 0.82
93.4 TE
01
4 60 25 27 0.69
84 34 42 0.37
5 72 27 48 0.44
Table 1.16. Relativistic pulse gyro-klystron experimental results [54]
Institution Frequency Mode No. of Power Effciency Gain Type
(GHz) cavities (MW) (%) (dB)
IAP, Nizhny Novgorod 30 TE
5,3
2 5 25 27 TE
5,2
/TE
5,3
UNIV. MARYLAND 8.57 TE
01
3 75 32 30 coaxial
9.87 2 24 30 33 max. power
9.87 TE
01
3 27 32 36 max. efciency
3 16 37 33 max. gain
3 20 28 50 coaxial
17.14 (2
c
) TE
02
3 27 13 25 coaxial
4 18.5 7.0 23.3
19.76 (2
c
) TE
02
2 32 29 27
29.57 (3
c
) TE
03
2 1.8 2.0 14
Table 1.17. Development status of weakly relativistic gyro-TWTs (short pulse) [54]
Institution Frequency Mode Power Efciency Gain Bandwidth
(GHz) (kW) (%) (dB) (%)
CPI, 5.18 TE
11
120 26 20 7.3
Palo Alto 93.7 TE
11
28 7.8 31 2
NRL, Washington, 32.3 TE
10
50 28 25 11
D.C. 34.0 TE
01
137 17 47 3.3
35.6 TE
11
70 17 60 17
IAP, 36.3 TE
21
/TE
11
180 27 27 10
Nizhny Novgorod
MIT, Cambridge 140 HE
()
061
30 12 29 1.6
UNIV. HSINCHU 34.2 TE
11
62 21 33 12
33.6 TE
11
93 26.5 70 8.6
UC LOS ANGELES 9.3 TE
10
55 11 27 11
15.7 (2
c
) TE
21
207 12.9 16 2.1
93.4 TE
01
75 22 60 4.5
by the beam of electrons that are bunched because of modulation in the input cavity.
The gyro-twystron conguration can mitigate the problem of microwave breakdown
at high power levels, since the microwave energy density in the output waveguide
can be much smaller than in an output cavity.
The development of high-power gyro-ampliers has opened up a number of pos-
sible applications to advanced mm-wave radars for high resolution ranging and imag-
ing in atmospheric and planetary science (e.g. cloud and space-debris monitoring) as
well as for deep-space and specialized satellite communications.
A 120-element phased-array 34 GHz radar system (4
beam width) using two

0.5 MW gyro-klystrons with 50 MHz bandwidth, 100 s pulse duration and 0.01
duty factor is operating in Russia. In the USA, a W-band radar system employing
a 92 kW, 94 GHz gyro-klystron with 420 MHz bandwidth and a duty factor of 0.11
(10 kW average power) is operating.
The cost of a future linear electronpositron collider with centre of mass energy
in the TeV range will depend both on the number of microwave phase-coherent am-
pliers required to drive the collider and on the cost of each amplier. In the past,
the microwave amplier of choice for driving the highest energy linear electron ac-
celerators has been the klystron (e.g. S-band). In considering microwave amplier
requirements for future supercolliders there has been a widespread perception that
higher frequency and higher peak power will be required. Because of inherent lim-
itations when simultaneously handling high power and high frequencies in klystron
circuits, other types of ampliers as relativistic gyro-klystrons and gyro-twystrons
have been explored.
The klystron development programs will probably result in producing
microwave ampliers that are suitable for linear colliders in the energy range up
to 1 TeV. However, for multi-TeV supercolliders, 34 GHz gyro-klystrons producing
about 100 MW of peak power in 1 s are needed [53].
Due to their larger bandwidth, gyro-TWTs can be used as output ampliers in
mm-wave communication systems.
1.7.5 Gyro-BWO
Gyro-BWOs are being developed as fast and continously frequency tunable drivers
of FEMs and for materials processing applicators with extremely homogeneous EM-
eld distribution (by frequency sweeping). The status of weakly relativistic devices
is summarized in Table 1.18.
1.7.6 Free-Electron Maser (FEM)
Free-electron lasers (FELs) differ from the other high-power microwave sources con-
sidered in this chapter in that they have demonstrated output over a range of frequen-
cies extending far beyond the microwave spectrum, well into the visible and ultravi-
olet range [4851]. To achieve this spectral versatility, FELs exploit relativistic beam
technology to upshift the electron wiggle frequency by an amount roughly propor-
tional to
2
. The magnetostatic wiggler is the most common, but not the sole means,
for providing electron undulation. An electrostatic wiggler or the oscillatory eld of
a strong electromagnetic wave can also play this role. The distinction between long
wavelength free-electron maser (FEM) ( 0.5 mm) and short wavelength FELs
is natural because higher current and lower energy beams are typically employed in
this regime and space-charge effects are more important. In particular, the dominant
interaction mechansim is often coherent Raman scattering. Also, while short wave-
length FELs excite optical modes, dispersion due to the beam dielectric effects and
nite transverse dimensions in the drift tubes and cavities are important effects at
longer wavelengths. A low power (3 W, 2 ms pulses) FEL operating at radio frequen-
cies (FER) employing a 420 V, 0.2 A electron beam holds the world record for long
Table 1.18. Development status of weakly relativistic pulse gyro-BWOs [54]. Only the
24.7 GHz device is operating in CW
Institution Frequency Mode Power Efciency Bandwidth
(GHz) (kW) (%) (%)
IAP, Nizhny Novgorod, 24.7 TE
21
/TE
11
7 15 5
FZK, Karlsruhe 23 (SDC)
IAP, Nizhny Novgorod 3538 TE
21
/TE
11
34 7 15
MIT, Cambridge, 140 TE
c
12
2 2 9
LLNL, Livermore
NRL, Washington, D.C. 27.8 TE
r
10
2 9 3
29.2 TE
r
10
6 15 13
UNIV. Hsinchu 33.5 TE
c
11
2067 6.521.7 5
100 25 8.5
164 41 1
UNIV. Strathclyde 8.5 TE
21
/TE
11
60 16 17
UNIV. Utah 10 TE
r
10
0.72 10 8
r: rectangular waveguide; c: circular waveguide
wavelength (f = 266 MHz, = 1.1 m,
w
= 0.04 m, B
w
= 0.04 T). The highest
CW power generated by a FEM is 36 W (15 GHz) whereas the IR (3.1 m)-FEL at
Thomas Jefferson National Accelerator Facility obtained a record average power of
2.13 kW at 3.5% efciency employing beam energy recovery ([54], and references
therein).
The FEM appears to be potentially capable of fullling all the requirements for
a frequency tunable high-power mm-wave source. Coverage of the entire frequency
range of 130260 GHz presents no severe problem, and even higher frequencies are
quite feasible. Rapid tunability over more than 5% could be obtained by variation
of the beam energy. The interaction occurs in a cavity operating in low-order modes,
which have very good coupling to a Gaussian beam output. The relatively low RF
wall loading and the use of high electron beam energy (>0.5 MeV) and a multi-stage
depressed collector are compatible with a high unit power at high efciencies if the
electron beam interception could be maintained at an acceptable level. A survey of
the development status of short pulse FEM oscillators and ampliers is presented in
Table 1.19 [54]. Electrostatic, pulseline/modulation and induction linac accelerators
are employed.
The induction-linac based 140 GHz, 2 GW short-pulse (2030 ns) LLNL-FEM
was successfully employed in high-power microwave transmission measurements in
fusion plasmas where the nonlinear absorption (higher transmission) at ECH fre-
quencies was observed which is only possible at such high power levels. The UCSB-
FEL was used to study nonlinear dynamics in semiconductor nanostructures and
devices.
Table 1.19. Development status of pulse free-electron masers
Institution FrequencyMode Power Efciency VoltageCurrent Type
(GHz) (MW) (%) (MV) (kA)
FOM, 206 HE
r
11
0.73(0.5) 5.7(3.9) 1.77 0.0072 oscillator
Nieuwegein 167 HE
r
11
0.36(0.26) 3.1(2.3) 1.61 0.0071 oscillator
169 HE
r
11
0.1 0.9 1.60 0.007 oscillator
(14 with MDC)
IAP/INP 75 TEM 100 4.2 0.8 3.0 oscillator
Novosibirsk
JINR 30.74 TM
c
11
/TE
c
11
48 35(30) 0.8 0.17(0.02) oscillator
Dubna/IAP 38.2 TM
c
12
/TE
c
11
3 3(3) 0.8 0.15(0.2) oscillator
Nizhny 35 TE
c
11
30 10 1.5 0.2 amplier
Novgorod
LLNL, 34.6 TE
r
01
1000 34(7.2) 3.5 0.85(4.0) amplier
Livermore 140 TE
c
11
2000 13.3(10) 6.0 2.5(3.0) amplier
5001000 in up to 50
MIT, 27.5 TE
c
11
1 10.3(6.3) 0.32 0.03(0.05) oscillator
Cambridge 35.2 TE
c
11
0.8 8.6(5.2) 0.31 0.03(0.05) amplier
NRL, 13.2 TE
c
11
4.2 18 0.245 0.094 amplier
Washington, 16.6
D.C.
RI, 625 TE
c
11
/TM
c
01
10 1.7 0.6 1 spon. emiss.
Moscow
TRW, 35 TE
r
01
0.1 9.2 0.3 0.004 oscillator
Redondo 35 TE
r
01
0.002 6.9 0.29 0.0001 amplier
Beach
UC Santa 120880 0.027 0.5 2.6 0.002 oscillator
Barbara
r: rectangular waveguide; c: circular waveguide
1.8 Future Trends and Applications
After the structural reorganization of most of the microwave tube manufacturers, as
explained in Sect. 1.1.2, the market is still performing evolution and the needs of the
customers change [1]. Certainly the three main applications listed in Sect. 1.1.1 will
be always with us, but in different ways.
The use of TWTs in space and especially on communication satellites is impor-
tant and still increases, probably because of the robustness and reliability of these
tubes, long lifetime, high efciency, and low weight of the whole associated trans-
mitter. Also they are able to provide the large required power, 200 W/tube and sev-
eral kW in total. As a matter of fact, on powerful transmitters, tubes are preferred
compared to solid state devices, because the size and the cost of a tube do not at
all increase proportionally to its RF output power. On the contrary, this is also why
the tubes are not seen in our everyday life, where the small power transistors are
prevalent (mobile phones, computers).
The extraordinary success of microwave ovens using cheap magnetrons is also
keeping with this presentation: magnetrons are robust, compact, reliable in a do-
mestic and hostile industrial environment and capable of delivering approximately
800 W, in the S-band.
According to Fig. 1.1, we point out that the tubes become more attractive than
the solid state devices when the operating frequency increases, what is the general
tendency. Except some conventional ones, most of the modern radars use phased
array antennas, working with a great number of low power modules and therefore
solid state devices. However, in military aircrafts and especially missiles, tubes are
still used (50 to 100 W, X-band, 2 to 3 octaves instantaneous bandwidths) again,
because of their compactness and robustness with respect to vibrations, shocks and
temperature variations.
As for the ISM applications, developments and improvements are foreseen in
several directions: lower costs, fewer risks of arcing and parasitic oscillations, higher
frequencies, better linearity, larger bandwidths, capabilities to deliver custom de-
signed specic components, such as windows, waveguides and safety devices. De-
velopments of new tubes may be extrapolated from the IOTs or the MBKs. In med-
ical applications new low cost and compact MW magnetrons or klystrons will be
requested. In the future there will be the possibility to provide a whole equipment
including not only the microwave tube, the supplies and the cooling system, but also
the electron linear accelerator. In that way the interface problems will disappear and
the maintenance time, the overall cost, the number of failures will be reduced.
At higher frequencies, the impressive development of the present gyrotron oscil-
lators is being followed by the one of gyroampliers, capable to be amplitude and
frequency modulated, with the intention to be used not only on plasma fusion reactors
but also for deep space communications, in some specic radars or in power trans-
mission systems. Therefore, provided that theoretical developments, breakthroughs,
and technological progress should come available on time, microwave tubes still have
a condent future.
The R&D activity on microwave tubes, which requires a deep knowledge of
many scientic areas and the mastery of many technologies, induces and encour-
ages the tube physicists and technicians to be interested in the development of other
electron vacuum devices, for instance:
specic high efciency multistage plasma (HEMP
@
) thrusters, whose design is
based on the TWT principle and, more precisely, on PPM focusing and depressed
collectors;
Cs atomic clock, where the RF cavity design and machining, the control of RF
leakages, the screening of the magnetic elds, the quality of the vacuum and the
reliability are familiar to microwave tube experts.
The presently rising micro- and nanotechnologies will also be important, especially
regarding the combination of solid state and vacuum devices. For instance, 1 m
long needles (tips) with a diameter of 20 to 50 nm can be not only manufactured
but also arranged in arrays of one or several mm
2
, giving rise to very small cold
cathodes. Associated to a grid for beam modulation at the cathode and to RF cavities,
they can be the master parts of very compact microwave triodes, TWTs or IOTs,
which can be used to boost the solid state ampliers. The state-of-the-art of such
vacuum microelectronic TWTs employing so-called Spindt cathodes is P = 55 W
at f = 4.5 GHz (NEC Corporation: 50 000 tips giving 90 mA) and P = 28 W at
f = 11.5 GHz (Northrop Grumman: 540 000 tips giving 56 mA).
In conclusion, microwave tubes are currently progressing and will progress in the
future in the two main directions:
the improvements of their basic principles and technologies;
the evolution towards totally new devices using, at least partly, their concepts and
these technologies.
Therefore, the future of microwave tubes is really encouraging and the microwave
tube community is engaged in a long term program.
References
[1] R.H. Abrams, B. Levush, et al., Vacuum electronics in the 21st centuries, IEEE Mi-
crowave Mag. 2, 61 (Sept.) (2001)
[2] M. Sedlacek, Electron Physics of Vacuum Gaseous Devices (Wiley, Hoboken, 1996)
[3] A.S. Gilmour, Microwave Tubes (Artech House, London, 1986)
[4] J.F. Gittins, Power Travelling Wave Tube (The English Universities Press, Ltd., 1964)
[5] A. Septier, Focusing of Charged Particles, vol. 1 & 2 (Academic, 1967)
[6] P.T. Kirstein, G.S. Kino, Waters WE: Space Charge Flow (McGraw-Hill, New York,
1967)
[7] S. Ramo, J.R. Whinnery, T. Van Duzer, Fields and Waves in Communications Electronics
(Wiley, New York, 1984)
[8] R. Warnecke, P. Guenard, Les Tubes lectroniques Commande par Modulation de
Vitesse (Gauthier-Villars, 1951)
[9] J.W. Gewartowski, H.A. Watson, Principles of Electron Tubes (Van Nostrand, Princeton,
1965)
[10] A.H.W. Beck, Space Charge Waves & Slow Electromagnetic Waves (Pergamon Press,
1958)
[11] A. Leblond, Les Tubes Hyperfrquences (Masson, 1972)
[12] N.E. Lindenblad, U.S. Patent 2,300,052, led May 4, 1940, issued October 27, 1942
[13] R. Kompfner, The traveling wave valve, Wirel. World 53, 369 (Nov.) (1946)
[14] H.-T. Schmidt, Homepage. www.hts-homepage.de/WesternElectric/WesternElectric700.
html
[15] http://accelconf.web.cem.ch/accelconf/p95/ARTICLES/RPA17.PDF
[16] J.R. Pierce, Traveling Wave Tubes (D. Van Nostrand, Princeton, 1950)
[17] H.J. Wolkstein, Effect of collector potential depression on the efciency of traveling wave
tubes, RCA Rev. 19, 259 (June) (1958)
[18] W. Neugebauer, T.G. Mihran, A ten stage electrostatic depressed collector for improving
klystron efciency, IEEE Trans. Electron Devices ED-19(1) (Jan.) (1972)
[19] A.S. Gilmour Jr., Principles of Traveling Wave Tubes (Artech House, London, 1994 &
1999)
[20] J.R. Pierce, Theory and Design of Electron Guns (D. Van Nostrand, Princeton, 1954)
[21] I.L. Langmuir, K. Blodgett, Currents limited by space charge between concentric spheres,
Phys. Rev. 24, 53 (July) (1924)
[22] C.C. Cutler, The nature of power saturation in traveling wave tubes, Bell Sys. Tech. J. 35,
841 (1956)
[23] J.E. Rowe, Nonlinear Electron Wave Interaction Phenomena (Academic, New York,
1965)
[24] G. Kornfeld, E. Bosch, From History to Future of Satellite TWT Ampliers, Frequenz
Zeitschrift fr Telekommunikation, Band 55, 9-10/2001
[25] D.C. Rogers, Proc. IEE 100(Pt. III), 151 (May) (1953)
[26] Tore Wessel-Berg, A General Theory of Klystrons with Arbitrary, Extended Interaction
Fields, M.L. Report No. 376 (1957)
[27] M. Chodorow, B. Kulke, IEEE Trans. Electron Devices ED-13, 439 (1966)
[28] Y.M. Shin, G.S. Park, G.P. Scheitrum, G. Caryotakis, Circuit analysis of an extended
interaction klystron, J. Korean Phys. Soc. 44(5), 1239 (May) (2004)
[29] B. Steer, State of the art W-band EIK for Cloudsat, in International Vacuum Electronics
Conference, IVEC 2004, Monterey, 2004
[30] B. Steer, Wide bandwidth, high average power eiks drive new radar concepts, in Interna-
tional Vacuum Electronics Conference, IVEC 2000, Monterey, 2000
[31] S. Millmann, A spatial harmonics amplier for 6 mm wavelength, Proc. IRE 39, 1035
(Sept.) (1951)
[32] R. Kompfner, The backward wave tube, U.S. Patent 2,985,790, led 17 May 1952
[33] R. Kompfner, N.T. Williams, Backward-wave amplier, U.S. Patent 2,916,657, led
17 May 1952
[34] R. Kompfner, N.T. Williams, Backward-wave tubes, Proc. IRE 41, 1602 (November)
(1953)
[35] H. Heffner, Analysis of the backward-wave traveling-wave tube, Proc. IRE 42, 930 (June)
(1954)
[36] J.R. Pierce, R. Kompfner, Natl Academies Press, Biographical Memoirs, pp. 156181
(1983). Use also: http://books.nap.edu/books/0309033918/html/156.html
[37] J. Arnaud, Backward wave tubes, in The Encyclopedia of Electronics, ed. by C. Susskind
(Reinhold Publishing Corporation, New York, 1962)
[38] R.Q. Twiss, Radiation transfer and the possibility of negative absorption in radio astron-
omy, Aust. J. Phys. 11, 424, 564 (1958)
[39] J. Schneider, Stimulated emission of radiation by relativistic electrons in a magnetic eld,
Phys. Rev. Lett. 2, 504 (1959)
[40] A.V. Gaponov, Izv. VUZ Radioz. 2, 450, 837 (1959)
[41] H. Kleinwchter, Zur Wanderfeldrhre, Elektrotechnik 4, 245 (1950) (in German)
[42] J.L. Hirsheld, J.M. Wachtel, Electron cyclotron masses, Phys. Rev. Lett. 12, 533 (1964)
[43] A.V. Gaponov, M.I. Petelin, V.K. Yulpatov, The induced radiation of excited classical
oscillators and its use in high frequency electronics, Izv. VUZ Radioz. 10, 1414 (1967)
[44] V.L. Granatstein, I. Alexeff, et al., High-power Microwave Sources (Artech House,
Boston, London, 1987)
[45] J. Benford, J. Swegle, High-power Microwave Sources (Artech House, Boston, London,
1992)
[46] C.J. Edgcombe (ed.), Gyrotron Oscillators their Principles and Practice (Taylor &
Francis, London, 1993)
[47] M.V. Kartikeyan, E. Borie, M.K.A. Thumm, Gyrotrons High Power Microwave and
Millimeter Wave Technology (Springer, Berlin, 2003)
[48] T.C. Marshall, Free Electron Lasers (MacMillan, New York, 1985)
[49] P. Sprangle, T. Coffey, New high power coherent radiation sources, Infrared Millim.
Waves 13, 19 (1985)
[50] H.P. Freund, T.M. Antonsen Jr., Principles of Free-electron Lasers, 2d edn. (Chapman &
Hall, London, 1996)
[51] H.P. Freund, G.R. Neil, Free-electron lasers: Vacuum electronic generators of coherent
radisation, Proc. IEEE 87, 782 (1999)
[52] M.I. Petelin, One century of cyclotron radiation, IEEE Trans. Plasma Sci. PS-27, 294
(1999)
[53] A.V. Gaponov-Grekhov, V.L. Granatstein, Application of High-Power microwaves
(Artech House, Boston, London, 1994)
[54] M. Thumm, State-of-the-art of high power gyro-devices and free electron masers, Update
2006, FZKA 7289 (2007)
[55] K.R. Chu, Overview of research on the gyrotron traveling-wave amplier, IEEE Trans.
Plasma Sci. PS-30, 903 (2002)
[56] G.G. Denisov, V.L. Bratman, A.D.R. Phelps, S.V. Samsonov, Gyro-TWT with a helical
operating wavequide: new possibilities to enhance efciency and frequency bandwidth,
IEEE Trans. Plasma Sci. PS-26, 508 (1998)

MWT Book

Uploaded by

Copyright:

Available Formats

MWT Book

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MWT Book

Uploaded by

Copyright:

Available Formats

1

r +e/2 = constant; Buschs theorem, valid along trajectories

around the operating point. Therefore, P

Mass >5000 g >1000 g 640 g 800 g

The Ka- and W-band TWT use interdigital delay lines; W

component, varies along the anode (with a so-called ab-

= (A/2)[cos(2s/ +t ) +cos(2s/ t )] (1.72)

is an accelerating eld, they are continuously accelerated, and if E

beam width) using two

You might also like