Thesis On TWT
Thesis On TWT
Thesis On TWT
January 2015
This work was supported by the Air Force Office of Scientific Research Program on
Plasma and Electroenergetics under Grant FA9550-09-1-0363. Reproduction,
translation, publication, use and disposal, in whole or in part, by or for the United States
government is permitted.
Design and Test of a 94 GHz Overmoded
Traveling Wave Tube Amplifier
by
Elizabeth J. Kowalski
B.S. Electrical Engineering, the Pennsylvania State University (2008)
S.M. Electrical Engineering and Computer Science, Massachusetts
Institute of Technology (2010)
Submitted to the
Department of Electrical Engineering and Computer Science
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2015
c 2015 Massachusetts Institute of Technology. All rights reserved.
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Electrical Engineering and Computer Science
December 31, 2014
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Richard J. Temkin
Senior Research Scientist, Department of Physics
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Professor Leslie A. Kolodziejski
Chairman, Committee on Graduate Students
Department of Electrical Engineering and Computer Science
2
Design and Test of a 94 GHz Overmoded Traveling Wave
Tube Amplifier
by
Elizabeth J. Kowalski
Abstract
This thesis discusses the design and test of an overmoded W-band Traveling Wave
Tube (TWT). The TWT was designed to operate in the rectangular TM31 cavity
mode at 94 GHz. The unwanted lower order, TM11 and TM21 , modes were suppressed
using selectively placed aluminum nitride dielectric loading. Simulations in 3-D CST
Particle Studio confirmed suppression of unwanted modes due to dielectric loading
and operation in the TM31 mode. The TWT was designed to operate at 31 kV with
310 mA and a 2.5 kG solenoid magnet. Simulations in both 1-D Latte and 3-D CST
predicted 32 dB of gain, 200 MHz bandwidth, and 300 W peak output power for the
TWT at 94 GHz. Test structures of 9- and 19- cavities were made via CNC direct
machining. Cold test measurements showed suppression of the unwanted modes and
transmission of the TM31 mode, which correlated well with HFSS simulations. Two
final 87-cavity structures were built and cold tested.
The experiment was designed and built in-house at MIT (with exception of the
electron gun cathode, manufactured by industry). It was operated with a 3 microsec-
ond pulsed power supply. A beam test was implemented which confirmed operation of
the TWT set up and electron gun. The electron gun operated at 31 kV with 3066
mA of current detected at the collector and 88 % transmission of current. Initial
operation of the TWT showed zero-drive stable operation and demonstrated 8 dB of
device gain and 10 W peak output power at 95.5 GHz. Following these first tests,
the magnetic field alignment was improved and the second structure, which showed
better circuit transmission in cold test, was installed. The overmoded TWT produced
212 dB device gain (defined as Pout /Pin ) at 94.3 GHz and 27 W of saturated output
power in zero-drive stable operation. The TWT was estimated to have about 6 dB
of additional loss due to coupling into and out of the circuit. Taking that loss into
account, the gain on the TWT circuit itself was estimated to be 272 dB circuit gain.
CST simulations for the experimental current and voltage predict 28 dB circuit gain,
in good agreement with measurements.
This experiment demonstrated the first successful operation of an overmoded
TWT. The overmoded TWT is a promising approach to high power TWT opera-
3
tion at W-Band and to the extension of the TWT to terahertz frequencies.
4
Acknowledgments
This thesis would not have been possible without my advisor, Dr. Richard J. Temkin.
Always willing to help, offer advice, and teach the finer points of vacuum electronics,
he guided me through this project and helped to make my PhD successful. Every
member of the Waves and Beams Group in the PSFC also helped me in my research.
Particularly, my officemates Dr. Emilio Nanni and XueYing Lu dealt with my dis-
tracting conversation, pungent teas, and research woes. Ivan Mastovskys expertise
helped my experiment to be operational, while Dr. Michael Shapiro ensured that my
theory and simulations were correct. In addition, Dr. Sudheer Jawla, Dr. David Tax,
Dr. Brian Munroe, Jason Hummelt, JeiXi Zhang, Sam Schuab, Alexander Soane,
and Haoron Xu all offered their advice and help, from Friday evening brainstorming
to unexpected company in the lab on Saturday.
Graduate Women at MIT, GWAMIT, helped me to realize I was not alone in my
endeavors and introduced me to some amazing women at MIT who encouraged me
and became great friends. It was amazing to be a part of shaping GWAMIT, and I
hope that the organization continues to grow in the future.
Of course, my family and friends helped me through the long nights, weeks,
months, and years of research. My parents taught me how to learn, and my sis-
ters taught me how to have a life. My husband, Edward Loveall, provided endless
support. He listened to me practice countless presentations, so he may even under-
stand 31 % of this thesis.
Elizabeth Kowalski
Cambridge, MA
December 22, 2014
5
6
to Edward
7
8
Contents
1 Introduction 21
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2 A Brief History of TWTs . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.1 Vacuum Tubes and Radar . . . . . . . . . . . . . . . . . . . . 29
1.2.2 Invention of the TWT . . . . . . . . . . . . . . . . . . . . . . 32
1.2.3 More Vacuum Devices . . . . . . . . . . . . . . . . . . . . . . 35
1.3 Modern Day W-Band Vacuum Tubes . . . . . . . . . . . . . . . . . . 36
1.3.1 W-Band TWTs . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.3.2 Other W-Band Devices . . . . . . . . . . . . . . . . . . . . . . 38
1.3.3 Overmoded TWTs . . . . . . . . . . . . . . . . . . . . . . . . 39
1.4 Overview of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
9
2.4.1 Helical TWTs . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.4.2 Coupled-Cavity TWTs . . . . . . . . . . . . . . . . . . . . . . 65
10
5.5 RF Vacuum Windows . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.6 High Power Pulse Modulator . . . . . . . . . . . . . . . . . . . . . . . 138
5.7 Safety Interlock System and Controls . . . . . . . . . . . . . . . . . . 141
5.8 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7 Conclusions 171
7.1 Summary of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.2 Discussion of Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . 173
7.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
11
12
List of Figures
13
3-1 An illustration of the electric and magnetic fields in the TM110 cavity
mode. (Figure modified from [12].) . . . . . . . . . . . . . . . . . . . 72
3-2 Contour plots of the electric field, Ez , for the lowest order cavity modes. 74
3-3 Two cavities of the TWT structure linked via staggered coupling slots,
as seen with an isometric view and from the top of the structure. . . 75
3-4 The electric field patterns of the three lowest order TM modes in a
rectangular cavity tuned to 94 GHz for the TM31 mode. . . . . . . . 77
3-5 The dispersion relation for the lowest order TM modes of the cavity. . 78
3-6 Placement of dielectric on the top and bottom of the cavity. . . . . . 79
3-7 The electric field patterns of the three lowest order TM modes in a
rectangular cavity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3-8 The transmission through a single cavity with and without dielectric
loading for the lowest four TM modes. . . . . . . . . . . . . . . . . . 81
3-9 Simulation results showing the variation of current, K, and C as de-
pendent on cavity parameters. . . . . . . . . . . . . . . . . . . . . . . 82
3-10 The effect of beam tunnel size on calculated coupling impedance through
the middle of the circuit. . . . . . . . . . . . . . . . . . . . . . . . . . 83
3-11 I0 /V0 ratio vs. operation voltage, V0 , for different electron beam radii. 85
3-12 The final cavity design showing the dielectric loading and manufactur-
ing fillets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3-13 The dispersion relation for the first three cavity modes of Design A, as
calculated by HFSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3-14 The coupling impedance and dispersion relation for the TM31 mode of
Designs A and B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3-15 The input coupler from standard waveguide into the first cavity of the
TWT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3-16 Linear gain vs. operation voltage for 94 GHz operation and design
parameters discussed in Table 3.2. . . . . . . . . . . . . . . . . . . . . 93
3-17 Linear gain vs. frequency for LATTE simulations. . . . . . . . . . . . 93
3-18 The full 87 cavity structure with dielectric loading as simulated in CST. 95
14
3-19 Power out vs. time for a 90 mW input signal along with an FFT of the
output signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3-20 The energy of the particles in the beam tunnel at the end of the circuit,
and the particle density along the length of the cavity. . . . . . . . . 96
3-21 Output power vs. input power for the simulated TM31 coupled-cavity
TWT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4-1 The assembly of 4-plate and 2-plate 9-cavity cold test structures. . . . 101
4-2 HFSS simulations showing a WR-08 waveguide with gaps at the top,
mid-plane, and bottom of the waveguide. . . . . . . . . . . . . . . . . 102
4-3 CAD drawings and machined 9- and 19-cavity cold test structures. . . 104
4-4 The measured transmission through (a) 9- and (b) 19-cavity OFHC
copper structures compared to HFSS simulations. . . . . . . . . . . . 105
4-5 The measured transmission through (a) 9- and (b) 19-cavity glidcop
structures compared to HFSS simulations. . . . . . . . . . . . . . . . 106
4-6 The measured transmission for copper cold test structures. . . . . . . 108
4-7 The measured transmission for glidcop cold test structures. . . . . . . 109
4-8 Detailed pictures of the final 87-cavity structure. . . . . . . . . . . . . 113
4-9 Transmission measurements for the first 87-cavity structure (A). . . . 114
4-10 Transmission measurements for the first 87-cavity structure (A) with
dielectric loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4-11 Transmission measurements for Structures A and B with no dielectric
loading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4-12 Transmission measurements for the assembled structures that under-
went hot test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
15
5-6 The 2-D cylindrical electron gun geometry. . . . . . . . . . . . . . . . 126
5-7 The beam envelope calculated via Michelle. . . . . . . . . . . . . . . . 127
5-8 The equipotential lines and particle trajectories as calculated by Michelle.128
5-9 The electric field in the electron gun for operation at 31 kV. . . . . . 129
5-10 A cross-sectional view of the electron gun modeled in Autodesk Inven-
tor with cold dimensions. . . . . . . . . . . . . . . . . . . . . . . . . . 131
5-11 Pictures of the electron gun and beam test assembly. . . . . . . . . . 132
5-12 A picture of the 94 GHz EIO. . . . . . . . . . . . . . . . . . . . . . . 134
5-13 Frequency vs. power for the Millitech AMC. . . . . . . . . . . . . . . 135
5-14 Measured and calculated transmission through the RF windows. . . . 137
5-15 One of the fused silica RF windows in a 2-3/4 con flat flange, installed
on the TWT. The input WR-28 waveguide can be seen . . . . . . . . 138
5-16 A block-diagram of the PFN and high voltage pulse transformer set-up
with dual-experiment access. . . . . . . . . . . . . . . . . . . . . . . . 139
5-17 The modulator set-up of the TWT experiment. . . . . . . . . . . . . 140
5-18 An electronic schematic of the high voltage pulse transformer and TWT
experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5-19 The control system for the experiment and safety interlock system. . 142
5-20 Installation of the TWT structure onto the anode. . . . . . . . . . . . 143
5-21 Installation of the WR-10 90 degree waveguide bends and support
structures into the tube. . . . . . . . . . . . . . . . . . . . . . . . . . 144
5-22 Completion of installation for the TWT experiment. . . . . . . . . . . 145
5-23 The fully assembled TWT experiment. . . . . . . . . . . . . . . . . . 146
16
6-5 Sample traces of the collector current and output power measurements
for Structure A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6-6 Device gain vs. bandwidth for operation at V0 =27 kV. . . . . . . . . 155
6-7 Device gain and output power vs. input power, showing 20.9 dB linear
gain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6-8 Circuit gain in the TWT as compared to simulated gain. . . . . . . . 160
6-9 Device gain vs. frequency in the TWT for two different operation points.161
6-10 The diagnostic measurements for high-gain operation at 94.27 GHz. . 163
6-11 The diagnostic measurements for saturated power operation at 94.26
GHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6-12 The diagnostic measurements for an oscillation observed during the
start of operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
17
18
List of Tables
19
20
Chapter 1
Introduction
1.1 Motivation
21
(a) (b)
Figure 1-1: The Haystack Observatory 94 GHz antenna. Two sets of simulated data
show a satellite that has been imaged with (a) 2 GHz or (b) 4 GHz bandwidth [11].
occur. Operation at higher frequencies means that larger bandwidths are available to
transmit data, leading to higher resolution imaging. For example, Haystack Radar,
shown in Figure 1-1 is a radar system located in Westford, MA that is used for space
tracking, communications, and astronomy. The system originally operated at 10 GHz,
and was recently upgraded to have a second operation point at 94 GHz. The band-
width available at 94 GHz can be as large as 4 GHz. The figure shows the simulated
detection of a satellite by the Haystack radar via two different operation bandwidths
for comparison [11]. Its easily observed that more details can be seen with a larger
bandwidth. Operation at 94 GHz allows for this large bandwidth to be achieved and
for details to be resolved which would not otherwise be detected.
Space communications also benefit from large bandwidths due to large distances
requiring long transmission times. NASA has built most of their communications
systems around 10 GHz, like those on board the Mars Curiosity Mission, because it
has a relatively high bandwidth for communication and reliable, high power devices
are available to operate at that frequency. Some NASA missions, like the Mars
Reconnaissance Orbiter, operate at higher frequencies, about 32 GHz, in order to
support higher bandwidths and higher data rates [47]. It is likely that future NASA
22
Figure 1-2: Average output power vs. frequency for vacuum and solid state devices.
This graph demonstrates the THz gap, from 0.310 THz, a frequency range for which
no or limited devices exist. Figure adapted from [23].
missions will be at even higher frequencies for better communications; but reliable,
high power, devices must exist at those frequencies first.
Many more applications for high power, high frequency devices exist, but all
electromagnetic applications are limited by our abilities to produce and amplify elec-
tromagnetic signals at the frequency of operation. At low frequencies, high power
sources and amplifiers are readily available. However, at higher frequencies there
is an undeniable need for reliable, high-power, and cost-effective power sources and
amplifiers.
23
devices are less desirable because they can be large, complex systems. However, their
ability to provide high power makes them viable options for certain applications.
For all electronic devices, the relative size of interaction circuits inversely scales with
frequency. At high frequencies, above 300 GHz, the small wavelengths of operation
severely limit the types of devices that can perform at high powers. At even higher
frequencies, above 10 THz, laser photonic devices are able to provide high power for
applications in the visible light spectrum, x-rays, and gamma-rays. These devices
have wavelengths less than 30 micrometers. The output power for devices such as
quantum-cascade lasers drops off rapidly at frequencies lower than 100 THz due to
physical limitations at relatively large wavelengths of operation [63]. In addition to
vacuum devices, quantum-cascade lasers in the THz range of frequencies are a large
area of interest.
At present, the two opposing frequency limits between electronic and photonic
devices leads to the range of frequencies for which there are no high power devices.
The range is between 300 GHz and 10 THz and is referred to as the terahertz gap. In
this gap there is very limited power output from all types of devices. Though, many
vacuum electron devices have recently been able to output relatively high power in
this frequency range. Recent experimental achievements with gyrotrons, klystrons,
and Traveling Wave Tubes (TWTs) have demonstrated that vacuum electronics are
able to perform well in the Terahertz Gap, though limitations in power output still
exist in all types of devices [5]. The characteristics and benefits of these devices will
be described in the next section.
24
100
60
40
20
0
50 100 150 200 250 300 350
Frequency (GHz)
Figure 1-3: Transmission (%/km) vs. frequency for sea-level conditions. The black
dotted line indicates the atmospheric window at 94 GHz.
were assumed in calculation. Several frequency windows can be seen in Figure 1-3
where there are transmission peaks in the atmosphere. Resonances in the atmosphere,
which absorb large quantities of electromagnetic waves leading to transmission dips at
certain frequencies, are due to molecules in the air. In the frequency range shown in
the figure (0350 GHz), oxygen and water particles dominate the loss effects. Atmo-
spheric losses are dependent on a wide variety of conditions that affect the molecules
in the air, most importantly weather conditions and altitude, but the windows of
low losses shown in the figure are still the best transmission frequencies to use in
communication applications [59]. The low loss windows can also be referred to as
high transmission bands, i.e. it takes less power to transmit electromagnetic waves at
those frequencies for longer distances than at the resonance frequencies. In addition,
higher frequencies have larger windows, allowing for larger operation bandwidths at
those frequencies.
25
Figure 1-4: Examples of electronic warfare usages [1].
that is possible. These high frequency windows are also in use for weather research
due to atmospheric transmission characteristics, the small wavelength of operation,
and the large bandwidth available. These two bands at 94 and 220 GHz are below
the defined Terahertz Gap in frequencies, and vacuum tubes exist which provide
useful output powers. However, high power devices are still in development at these
frequencies due to a need for more power than is currently accessible and a desire
for cheaper alternatives. It is also relevant that development of devices at these
frequencies will lead to more robust devices that can be modified for use at even
higher frequencies in the following years.
Just below the Terahertz Gap is a band of frequencies known as the W-band,
between 75110 GHz, for which there are many applications. This range includes
the 94 GHz transmission window in atmosphere. The possibility of high data rates
and the crowding of communications at lower frequencies make the W-band a good
communications channel. Though no commercial systems exist in the W-band yet,
many research and military applications are already implemented. The Haystack
Radar, shown in Figure 1-1 takes advantage of the transmission window in order to
image objects in space. Gyrotron devices are used to power the system since they
26
Figure 1-5: The U. S. Military Active Denial System, which operates at 95 GHz. A
humvee is necessary to move the equipment.
can provide the necessary high powers and bandwidths that are needed for operation.
Figure 1-4 shows electronic warfare applications, such as a frequency jammer, which
involves disrupting electronic systems with high power microwave and millimeter wave
signals. These applications require very high power devices to disrupt appropriately
and from a safe distance. There are many other communications capabilities that
exist at 94 GHz, many involve devices installed on satellites and airplanes. As such,
it is often necessary that the devices are robust, reliable, light weight, and have long
operation life-times. Traveling Wave Tubes (TWTs), which will be discussed in the
next section, satisfy these requirements.
27
perimeter security; it is depicted in Figure 1-5. If the ADS is pointed at a human, the
directed energy at about 95 GHz heats up water molecules that are present in skin and
creates a near-instantaneous sensation of burning without causing any physical harm
or injury to the subject. The subject will retreat from the area of directed energy,
relieving oneself of the burning sensation. This is effective in creating an area of
denied access, where people are unable to enter. It may also be used in riot control to
disperse a crowd without causing physical harm. Unlike the rubber bullets or pepper
spray that are used in these situations today, the ADS causes no physical harm to the
subject. In its current iteration, the ADS system is large and cumbersome, requiring
gyrotrons to power the high-energy beam and a large transport vehicle. In addition
to being difficult to transport, it is currently incapable of covering a significant area
due to power restraints of the gyrotron and the need for extremely directed energy.
This thesis focuses on the W-band, but there are a wide array of applications for
higher frequencies. Communications channels, radar, and imaging at higher frequen-
cies have increasing benefits in bandwidths, data rates, and resolutions. For example,
there is another peak in the transmission through atmosphere at 220 GHz, which
could have even larger bandwidths and data rates for communications. By focusing
this thesis on the W-band, we are creating an increase in our capabilities to produce
high power at high frequencies which can easily be converted to even higher frequen-
cies in a future experiment. In this regard, the ability to scale the experiment to
higher frequencies was kept under consideration. It is not enough to just design in
the W-band, one must also think toward the future of our applications and the need
for power within the Terahertz gap.
28
are one of the best amplifiers available in the W-band. Experimentally, they have
reached average powers up to 100 W at 94 GHz. The only other amplifiers which
outperform TWTs in the W-band frequency range are gyrotron devices, but they
require superconducting magnets to operate. These devices are generally large and
cumbersome making them impractical for many applications. TWTs offer a cheaper
and more portable alternative.
Therefore, the development of a high power TWT at 94 GHz is of immediate
practical use. In addition, there is a motivation to develop a new design for TWTs
which has a size that is not as limited by the wavelength of operation as the traditional
TWT design. A TWT of such a design would be able to easily be scaled to higher
frequencies and bridge the Terahertz gap. An experimental TWT at 94 GHz with a
novel cavity design is a profound scientific advancement.
The first and most generic vacuum tube, the thermionic tube, was invented in 1907.
These devices were the simplest form of an electronic switch and were the drive behind
the first commercial electronic devices. In its most generic form, a vacuum tube
consists of a cathode and anode which emit and accelerate electrons, respectively.
29
Output
Anode
Grid- Plate
Resonator
Cathode
Grid- Cathode
Resonator
Filaments
Input
(a) (b) (c)
Figure 1-6: Early vacuum tubes diagrams of (a) a tetrode tube with resonators, (b)
a klystron, and (c) a cavity magnetron [57], [25].
When a voltage is applied, the cathode emits and turns on the vacuum tube.
Vacuum tubes are easily used to oscillate at certain frequencies, act as relays, or
amplify signals due to a complex interaction between the electron beam in the system
and an electromagnetic wave. An early tetrode tube is shown in Figure 1-6(a), which
is designed to amplify a certain frequency.
As World War II approached, Britain began development of the first radar system
in 1935. The first military demonstration used a BBC (British Broadcasting Corpo-
ration) transmitter in order to see airplanes passing near. The reflected radio signal
from the planes was easily detected. The first radar system used triode thermionic
tubes with large antenna towers. It operated at 11 MHz (but the frequency was
quickly increased to 30 MHz) and able to detect airplanes up to 100 miles away de-
spite weather conditions for low visibility [55]. These systems were installed along the
coast of the UK to detect incoming airplanes, but it was quickly apparent that range
and accuracy for detection would be a key component in developing a better radar
system. It was also apparent that the installation of such a system on an airplane
would lead to superior air defenses. These requirements led to a search for higher
frequency components with high power signals.
The klystron was developed shortly after in 1937, by Russel and Sigurd Varian
at Stanford. They were developing a small tube for radar which had enough power
30
to detect oncoming airplanes. The klystron, a powerful tube in its own right, is also
important because it directed the military radar research that was being developed
prior to World War II and led to the development of countless other vacuum devices.
In a klystron, the electron beam travels from the cathode through a series of cavities
which oscillate at specific frequencies when the beam passes through, as shown in
Figure 1-6(b). The electron beam transfers energy into the electromagnetic wave in
the klystron, creating an oscillation of the cavity frequency. If the klystron is driven
by an input signal, it is possible to cause an amplifying device which creates a high
power output signal in the electromagnetic wave at the input frequency.
During World War II, Germany developed radar systems with klystrons, but the
U.K. developed radar systems which used a different device, the magnetron, that was
able to provide higher power outputs. The magnetron was developed in 1939 by John
Randall and Harry Boot in the UK based on previous similar devices; it is shown
in Figure 1-6(c). This is the same device which, years later, would go on to power
microwave ovens. (You probably have one in your kitchen.) Electrons from the center
cathode are directed circularly with a magnetic field around the cavities and to the
anode. The circular motion creates bunching and resonances that develop within the
cavities of the magnetron. These resonances are able to be combined, leading to a
high-power device. While klystrons were able to give decent power (about 500 W) at
500 MHz, the cavity magnetron provided about 50 kW at nearly 2-5 GHz, providing
a phenomenal improvement in both power and frequency.
In a key collaboration between the U.S. and U.K. in 1940, the Tizard Mission
sent one of the first magnetrons to the United States. The mission led to a key
international exchange of radar technology. Once in the U.S., the MIT Radiation
Laboratory and Bell Laboratories quickly optimized the magnetron output and put it
into mass production as part of the war effort. The magnetron was a large component
of the superior British air force in World War II, and was a large driving factor in
winning the war. The success of tubes in radar systems led to the development of
many known types of vacuum tubes and providing the scientific knowledge to pursue
new tube designs after the war.
31
Figure 1-7: Drawings from Kompfners notebook in 1942, when he first began work
on the helical TWT. First, the TWT concept is shown with a hollow electron beam,
traveling outside of the helix. Kompfner supposed it would be a completely untuned
amplifier and questioned, Would it work? Second, a diagram of the first TWT
experiment is depicted, which has a solid pencil beam traveling through the center of
the helix [39].
While working on the magnetron and related radar technology in the U.K., Rudolf
Kompfner developed the initial idea for the Traveling Wave Tube Amplifier (Abbre-
viated TWAT or TWTA in some publications, but more commonly simply called
a TWT). Figure 1-7 shows drawings from his 1942 notebook. The design involves
sending an electron beam through a slow-wave circuit; in this case, the circuit is a
helical wire which acts as the center of a coaxial input line. The helical wire slows
down the phase and group velocity of the electromagnetic wave traveling through the
device, such that the phase velocity of the wave matches the velocity of the beam.
32
Figure 1-8: A diagram of the first TWT experiment from Kompfners laboratory and
a picture of John R. Pierce holding one of the first successful TWTs [39] [55].
Energy can transfer from the beam into the wave and amplification can occur. This
device was proposed as an untuned amplifier, meaning that the device would have
a significant bandwidth in comparison to its klystron and magnetron counterparts.
A detailed drawing of the TWT is shown in Figure 1-8. A focusing magnetic coil
along with other magnets (not shown) direct the electron beam through the device,
while a matching input and output circuit couple to the electromagnetic wave that is
amplified. A detailed theory of the TWT will be provided in Chapter 2.
Near the end of World War II, Kompfner and John Pierce worked together on
developing the theory and optimizing the output power and bandwidths of a working
TWT experiment. The goal was to develop a low-noise device that could be used in
radar and communications applications. John Pierce is shown in Figure 1-8 holding
one of the first successful TWTs. The helical TWT was shown to be a highly efficient
and high bandwidth device with low noise, up to 80 % efficiency, and the capability
33
for an octave of bandwidth (50%) [55]. Though Kompfner invented the TWT, Pierce
is credited with developing the theory which ultimately led to a successful device.
The development of other types of TWTs followed. Though the helical TWT
is a high bandwidth device, it has low output powers and gain characteristics. For
radar transmitting, high power and gain is often necessary. In the basic theory of a
TWT, a slow-wave device must be used; this device can be anything which causes
the electromagnetic wave to travel at a slow velocity (less than the speed of light)
such that it can be phase-matched with the electromagnetic wave. In each device,
the theory remains the same, and the phase velocity of the wave can be calculated
via electromagnetic field theory. Different types of TWTs led to different ranges and
design trade-offs for bandwidth, gain, and power.
Using concepts from accelerator physics, high power TWTs were developed. The
folded waveguide TWT is a design which uses a meandering (or serpentine) waveg-
uide structure and a beam tunnel that cuts through the waveguide. The wave is
slowed down relative to the electron beam by traveling a longer distance through a
rectangular waveguide structure that folds repeatedly across the beam tunnel axis.
When the beam passes through the meandering structure, it interacts with the wave.
A cooling system on the folded-waveguide TWT allows for it to be more powerful
than the helical TWT, while maintaining a relatively high bandwidth. Similarly, in a
coupled-cavity TWT, the electromagnetic wave couples through subsequent cavities
along the electron beam line, in much the same way as the folded-waveguide struc-
ture. The coupling from one cavity to the next slows down the phase velocity of the
wave to match the electron beam. The coupled cavities allow for a high power device
due to the strong resonances that occur in the circuit. However, the cavities are
frequency specific, and the highest achievable powers are only obtained by limiting
the bandwidth capabilities of the circuit. The coupled-cavity TWT will be discussed
further in Chapter 2. The varied array of possibilities with TWTs mean that a device
can easily have an optimized, gain, power, and bandwidth which best suits the needs
of the application at hand.
34
1.2.3 More Vacuum Devices
The development of the klystron, led to an array of vacuum devices that all used very
similar concepts of phase-matching between the electromagnetic wave and the electron
beam. These devices can be split into slow-wave devices (like magnetrons and TWTs),
where the phase velocity of the wave is less than the speed of light such that the
velocity matching occurs, or fast-wave devices (gyrotrons), where the phase velocity
of the electromagnetic waves travel faster than the speed of light and RF interaction
occurs via the angular velocity of the electron beam. It is also worth mentioning
backward-wave oscillators, which are slow-wave devices that use resonances within
a circuit similar to a TWT structure. The resonances cause oscillations (instead of
amplifications) with a negative group velocity wave in the structure. This leads to a
high power output device with a tuned frequency.
When it comes to high power in millimeter waves, no other device can match the
achievements of the gyrotron. Gyrotrons are fast-wave devices which can operate as
either oscillator or amplifier circuits. In gyrotrons, an annular (or hollow) electron
beam is bunched azimuthally with the electric field. Gyrotron oscillators are capable
of achieving extremely high output powers (more than megawatts of power) at the
highest frequencies of any vacuum device, up to 100 GHz and above. Gyrotron
amplifiers are capable of producing high gain at those frequencies, as well. Gyrotrons
at even higher frequencies exist, and many designs are being pursued to bring gyrotron
devices into the Terahertz gap. In practical applications however, gyrotrons are often
seen as cumbersome; they often require high magnetic fields (10 T) which must be
achieved with a superconducting magnet and are not robust to alignment issues within
the magnetic field, making them hard to transport and maintain. Though gyrotrons
have many applications, gyrotron devices placed in airplanes or satellites are not a
practical implementation.
35
1.3 Modern Day W-Band Vacuum Tubes
Many advancements in vacuum tubes have occurred since the development of the
device in the 1940s and 1950s. With the advent of highly accurate computer-driven
simulations and optimized designs, many devices which could not be completely de-
rived and understood with direct theoretical calculations can be explored without
much experimental cost. In addition, machining techniques have improved, making
practical implementation in the W-band easier to achieve. As such, tubes at high-
frequencies have been rapidly developed in recent years.
After extensive development in the 1960s, helical TWTs have been demonstrated
to provide about 1 kW of power at 10 GHz. Helical TWTs also have large bandwidths,
providing up to two octaves of bandwidth at lower frequencies [21]. Folded waveguide
and coupled cavity TWTs operating in the fundamental mode were developed shortly
after and shown to provide approximately 100 kW of power with 10 % bandwidth up
to 10 GHz in frequency [20], [37]. Scaling to higher frequencies is difficult due to small
wavelengths and proportionally small structures that are difficult to manufacture.
In the W-band, vacuum tube designs must deal with smaller wavelengths than the
original TWT designs. (Wavelengths in this frequency range are from 4 mm to
2.7 mm.) Small operation wavelengths lead to mechanical difficulties in experimental
implementation. At 94 GHz, the wavelength of about 3.2 mm means that components
that interact with this wavelength will be much smaller in size. When dealing with a
TWT, this scaling also means that the electron beam must be subsequently smaller
in size and will be traveling through a small beam tunnel, too. Beam compression
is incredibly difficult at small sizes, and leads to extremely large magnetic fields
necessary for TWT operation. (The concept of Brillouin flow will be discussed in
Chapter 2.) Due to these limitations, development of high power TWTs in the W-
band is not straightforward.
One of the most successful W-band TWTs was a cylindrical coupled-cavity ladder
36
circuit known as the Bill James Tube developed in 1986 at Varian. With a 20 GHz
bandwidth, it achieved up to 100 W continuous (CW) power and 1 kW peak output
pulsed power [34], [33]. In the 25 years since, W-band TWTs have only achieved
50-100 W average power with less than 5 GHz bandwidth.
37
omniguide photonic band gap TWT was in development with a 10 % bandwidth ex-
pected [64], but achieved limited success. None of these designs have been successfully
implemented in experiment and all would require small electron beam radii with large
currents and significantly larger magnetic fields than lower frequency TWTs in order
to achieve high gain conditions.
Success has come using UV-LIGA fabrication techniques. A 50 W, 220 GHz
folded waveguide TWT was successfully built at the Naval Research Lab [36]. The
small device, with a beam tunnel of about 100 microns, uses a patented UV-LIGA
method which is susceptible to small manufacturing and alignment errors. With
the development of a manufacturing process for small components, devices at high
frequencies will see larger possible gains in the coming years. However, the sizes of
the components at these frequencies puts a very strict limit on the peak output power
that can be safely handled by the devices.
Development in TWTs has also focused on complex electron beam formation.
Sheet beams, which are elliptical, as opposed to circular, solid electron beams have
been successfully used in experiments [53] [18]. In addition, multiple-beam TWTs
have up to 8 circular beams in a single device [52]. These devices are successful,
but they often require extremely precise magnetic fields and complex beam focusing
mechanisms.
Gyrotron oscillators are capable of producing extremely high output powers in the
W-band. Fueled by ITER, 1-2 MW gyrotron tubes are in development at 170 GHz.
W-band devices are also achievable with over 1 MW of output power. These designs
often have interactions in over-sized cavities and operate in an extremely high-order
mode in order to maximize interaction with an azimuthal beam [30], [38], [46], [7].
They are phenomenal devices, but require significant magnetic fields and extremely
high voltages to operate.
Extended Interaction Klystrons and Oscillators (EIKs and EIOs) in the W-band
have produced extremely high power as well. Commercially, an EIO can be purchased
38
which gives up to 1 kW of power in the W-band. However, the instantaneous band-
width of operation is limited, at only 250 MHz for high power devices [58]. Larger
bandwidths, up to 2 GHz are available for devices with less output power [31].
Solid-state devices are also not to be ignored. With advances, Amplifier Multi-
plier Chains (AMCs) are capable of producing significant power in the W-band and
are extremely efficient. Though expensive, purchasing a solid state device which pro-
duces 50 mW of power across 10 GHz in the W-band is reasonable [48]. AMCs have a
distinct scale between power and bandwidth. Devices are available with large band-
width, 20 mW over the entire entire W-band (75-110 GHz), or high power, 3 W over
4 GHz [49] [61]. Though 3 W is hardly considered high power for vacuum devices,
it is a great achievement for solid-state amplifiers. AMCs can be cascaded together,
but reaching high powers is often prohibitively expensive, especially as the available
power in solid-state devices rapidly drops off at higher frequencies.
The TWT has been shown to be a robust and capable device for numerous applica-
tions. It is seen as one of the most practical vacuum tubes for modern-day endeavors,
capable of producing high power and high frequency. It is also an extremely reliable
device, already used for many space applications. Though devices with better output
power characteristics exist at 94 GHz, it is the goal of this thesis to develop a TWT
for these reasons.
Overmoded W-band TWTs have been proposed which would allow the possibility of
larger beam tunnels and the need for less magnetic field overall [9], [10], [22]. An
overmoded design, meaning a design which operates in a cavity mode which is at
a higher frequency than the fundamental cavity mode, is oversized for its intended
frequency of operation. By operating in a higher order mode, the overmoded design
offers the possibility of creating a larger device than the fundamental coupled cavity
equivalent at the same frequency. An oversized device offers many benefits that can
lead to higher gain and power in the TWT; these benefits will be explored in this
thesis.
39
Sheet beam EIKs are oversized devices that have been shown to be successful at
high frequencies and the technology could be applied to TWTs. By nature, these
devices are oversized for their intended frequency of operation, though they operate
in the fundamental device mode. The elliptical beam structure takes advantage of
peaks in the electric fields of the mode in the cavity or folded waveguide [53] [44].
Though moderately successful, strong and precise magnetic fields must be used to
successfully implement these devices due to the complex electron beam.
40
Chapter 2
Theory of Traveling Wave Tubes
The equations that govern Traveling Wave Tube (TWT) amplification and operation
are easily described using theory derived from Pierce, and further refined over the
decades, to describe different slow-wave structure interactions. This chapter will first
consider an overview of the TWT interaction. A discussion on the electron beam
compression with the magnetic field considerations will follow. Finally, there will be
a detailed analysis of the slow-wave structure and beam-wave interaction theory.
First, consider the generic TWT circuit, as shown in Figure 2-1. In this figure,
an electron gun produces a beam of a specific quality. The electron beam is solid,
typically round, and it is often described as a pencil beam. It is confined and
directed by a magnetic field along the z-axis of the system and travels through a
TWT interaction structure, as described in the next section. An electromagnetic
wave is fed into the interaction structure, which interacts with the electron beam
and gains energy along the length of the structure before exiting. The structure
and beam-wave interaction will be described in Section 2.3 along with interaction
equations. After the interaction structure, the beam exits the confining magnetic
field, expands, and is deposited on a collector. All of these components are outlined
in Figure 2-1 for simplicity. The beam, shown in pink, travels through the slow wave
41
Electron Gun Solenoid Magnet Electron Beam Collector
Figure 2-1: A block diagram overview of a TWT amplifier circuit. The electron beam
(pink) travels from the electron gun (gray), through the slow wave structure (cyan)
which is surrounded by a solenoid magnet, and to the collector (orange).
structure, shown in blue. A solenoid magnet surrounds the structure and confines
the beam. The electron beam, magnet, and slow-wave structure are the three main
parts of a TWT and will be described in detail in the subsequent sections.
For now, lets simply understand that the electron beam must be formed to a small
diameter and confined over a length of space via the magnetic field. The velocity of
the electron beam must be in synchronous operation with the electric field in the
slow-wave structure. This means that the phase of the electric field must travel at
the same velocity as the electron beam. This concept is illustrated in Figure 2-2,
which shows a generic plot of the electric field along the longitudinal length, or z-
axis, of the TWT device. Since the electric field and electrons are traveling at the
same velocity, some of the electrons will perpetually see a +
z -directed accelerating
field and some will see a
z -directed decelerating field, as shown in the cut-out from
the figure. This interaction leads to electron bunches forming along the longitudinal
axis of the interaction. After the bunch is formed, the electron beam transfers energy
to the electromagnetic wave over a length of space, and the wave sees exponential
growth in power. At the end, the field reaches saturation once the electron beam
begins to lose energy and de-bunches after falling into the decelerating field of the
42
Exponential growth
Saturation
Bunching
Electric Field
z-axis
Decelerating
Field
Accelerating
Field
Electron
Bunch Forms
Figure 2-2: A generic graph of the z-directed electric field in the TWT along the beam
tunnel of the TWT. In the bunching region, electrons in positive field are accelerated,
and electrons in negative field are decelerated, forming bunches.
RF wave.
The complex equations to describe this interaction will be explained in Section 2.3,
but the key take-away is that the beam and wave should travel at the same speed,
causing a consistent interaction over the length of the structure, and transferring
energy from the electron beam into the electromagnetic wave. This energy transfer
causes the amplification of the electromagnetic wave in the circuit. The rest of this
chapter will be dedicated to explaining this interaction in detail.
43
2.2 Electron Beams in Vacuum Tubes
A key component of a TWT is the electron beam that will be used for the interaction.
In high frequency tubes, a well-formed electron beam is necessary to travel through
the small beam tunnel of the circuit. Though there are several ways to go about
making an electron beam, this project utilizes a simple, single-beam Pierce electron
gun with a thermionic cathode, which will be described in this section.
In its simplest form, an electron gun consists of a cathode and an anode. In general,
the cathode has a reservoir of electrons and is biased by a negative voltage relative
to the anode, which is grounded. The cathodes considered here are thermionic cath-
odes, meaning that they are heated to a certain temperature and that a reservoir of
electrons, typically tungsten and barium based, is activated. The electrons at the
cathode have a low work function due to the barium. Therefore, when a voltage is
applied to the cathode, electrons easily emit from the surface. Figure 2-3(a) shows
the generic set-up. In this instance, the cathode releases electrons that accelerate
over the gap between the anode and cathode, this space is often referred to as the
A-K gap. When the electrons reach the anode, their speed, ve , as determined by
conservation of energy is
p
ve = c 1 1/ 2 (2.1)
and
V0 [kV]
=1+ (2.2)
511 [kV]
where V0 is the voltage between the electrodes (in kV), and c is the speed of light.
The current emitted from the cathode has two possible limitations: tempera-
ture and space-charge. In general, the larger factor to consider is temperature. A
thermionic cathode has a maximum current density that it can emit for a given tem-
perature of operation. This limit ranges between 1 A/cm2 to 100 A/cm2 at the surface,
depending on the type of cathode, operation temperature, and lifetime considerations
44
of the device. A longer lifetime circuit, will have a smaller acceptable current density
on the cathode. Pierce guns, however, are intended to operate below the temperature
limited regime, and in the space-charge limited regime. At this operating condition,
the cathode voltage, V0 , limits the current that can be extracted from the cathode
due to space-charge forces within the beam. That is, the repulsion of the electrons
between themselves is the limiting factor in extracting current from the cathode. The
limit can be determined via Poissons Law.
2 V 0 = (2.3)
0
The charge density is related to the current density, such that J0 = ve , with ve =
2V , where is the charge-to-mass ratio of electrons (note that this is the same
as eq. (2.1) if V0 511 kV). Consider that the electrons are traveling only along the
z-axis, then the equation can be reduced to,
d2 V 0 J0
2
= (2.4)
dz 2V 0
This equation can easily be solved for the Child-Langmuir equation, which provides
a relationship between current density and voltage in a space-charge limited electron
gun, such that [37]
3/2 3/2
4 V V
J0 = 0 (2)1/2 0 2 = 2.33 106 0 2 (2.5)
9 d d
3/2
I0 = P V0 (2.6)
Where I0 is the total current in the electron gun, and P is the perveance of the gun,
which is defined as
4 A0 A
P = 2
(2)1/2 = 2.33 106 2 (2.7)
9 d d
45
where A is the area of the cathode and d is the distance between the anode and
cathode [37]. As such, the perveance (typically with units of micropervs) is a defining
factor of the electron gun which describes the current limitations of the gun, and the
relationship between current and voltage in the gun.
3/2
In essence, current is proportional to V0 . At operating voltages, a Pierce elec-
tron gun should follow the Child-Langmuir limitation of current density. For higher
voltages, the gun will reach a maximum, temperature limited, current output. For
a Pierce electron gun, a space-charge limited operation is ideal. This allows for the
beam to be self-modified, and leads to a better confined beam within the circuit that
can easily be directed with appropriate magnetic fields.
Though simple, modeling the electron gun as two parallel plates, as shown in
Figure 2-3(a), has many practical problems. To begin, it requires two infinitely large
planes, which cannot be practically realized. The electron beam, also, must have
some space in which to interact with the slow-wave structure, requiring some sort
of beam tunnel to propagate through the anode. In addition, the set-up offers no
compression of the beam. Space-charge limited, thermionic cathodes will only emit
small amounts of electrons over their surfaces, typically 10 A/cm2 is the upper limit
of emitted current density for a thermionic cathode with a decent lifetime. However
many applications require a large density of electrons that is difficult to achieve with
thermionic emission. These two properties lead to a compression of the electron beam
between the cathode and anode being necessary. Figure 2-3(b) shows a simple set-
up of two spherical electrodes, which would have a larger density of electrons at the
anode than at the cathode. In this case, the velocity of the electrons is still defined
as in equation (2.1), and the Child-Langmuir limit is still applicable.
An even more practical set-up is shown in Figure 2-3(c), where the spheres have
been reduced to smaller surfaces. This allows for a circular spot of electrons to form at
the anode. The velocity is the same, but the edges of the beam need to be contained.
In general, electrons will travel along lines parallel to the equipotential lines formed by
the anode-cathode configuration. In Figure 2-3(a), the equipotential lines are parallel
to the anode and cathode, so the flow of electrons is simply straight across the gap.
46
(a) (b) (c) (d)
-V -V -V
-V
In Figure 2-3(b), the equipotential line are concentric circles between the anode and
cathode, leading to another simple flow of electrons directly between the cathode
and anode. However, in Figure 2-3(c), the equipotential lines are more complicated,
straitening out and curling around the two electrodes. Due to this configuration,
electrons along the edge of the beam will not be confined.
If the outer emitting sphere were simply cut like in Figure 2-3(c), the electrons
would repulse each other and not form a coherent beam. Figure 2-3(d) solves this
problem. Focusing electrodes are added to the edges of the partial spherical surface of
the cathode. These electrodes are at the same voltage as the cathode, but they do not
emit electrons; they provide an electric field outside the partial sphere of the cathode
to help to focus and maintain the electron beam shape. This addition changes the
shape of the equipotential lines, such that the electrons forming at the edge of the
cathode see a parallel equipotential, and move perpendicular to the cathode surface.
The equipotentials direct the beam to a confined spot at the center of the anode. The
angle of the focusing electrodes can be calculated by assuming planar flow of electrons
near the edge of the beam. That is, if we zoom into the edge of the cylindrical electron
beam, we can assume that the electrons are traveling locally only in the z-direction.
In this case, Poissons equation leads to boundary conditions at the space along the
edge of the beam to produce a straight electron flow [37],
V V
= =0 (2.8)
y x
47
This condition defines linear equipotentials along the beam line between the elec-
trodes. Equation (2.5) can be rewritten, such that
2/3
J0
V0 = d4/3 (2.9)
2.33 106
where d is the distance between the anode and cathode. Taking the anode to be at
+V0 , the cathode to be at ground, and solving for the boundary conditions in polar
coordinates, the line at which V=0 is defined as:
4
cos = 0 (2.10)
3
where is the angle between the focus electrode and the edge of the electron beam.
This equation defines the line of the focusing electrode, which requires that =
(3/8), or = 67.5 . In other words, the focusing electrode placed at 67.5 away
from the edge of the beam, will cause equipotentials between the cathode and anode
that appear as concentric circles for the beam. This means that the electrons will
travel along straight lines between the anode and cathode [37].
Figure 2-4 shows the general set-up for the cathode and anode of an electron gun,
using the modified spherical geometry. A hole in the anode has been introduced to
allow the beam to pass through the anode. After the anode, it will interact with
the slow wave structure. At the location of the anode, the electrons will be traveling
with velocity, ve . The compression of the gun is determined by the radii of the two
electrodes and the spacing between them. Ideally, the electrode spheres are centered
at the same point, such that the A-K gap is equal to Rc Ra , but in practice, these
are often offset in order to get a better compression of the beam.
48
Emitter Surface Focus Electrode Electron Beam
(Cathode)
Anode
67.5o
Ra
Rc
rc rb ra
Figure 2-4: A diagram showing the ideal design of a Pierce electron gun, with the
emitter surface highlighted in red, and the electron beam shown in pink.
b(z=0)
b(z=zmin)=bmin 2rb
(a) (b)
Figure 2-5: (a) The electron beam compression and minimum beam radius. (b)
Magnetic field confinement of the beam after the minimum beam radius has been
formed.
after the minimum diameter is reached due to repulsion forces. This figure consid-
ers no magnetic fields, and the beam is outside of the cathode-anode configuration,
meaning that the only forces on the beam are a result of space charge. As such, the
universal beam spread equation defines compression and divergence of the beam.
d2 b I0
=0 (2.11)
dz 2 2b0 ve3
49
where b is the diameter of the electron beam and z is the axial direction of propagation.
The equation can be simplified by introducing a constant parameter, D, such that
I0 1 I0 2
D2 = = = (174 P) (2.12)
20 ve3 0 2 V03/2
and
d2 b D 2
=0 (2.13)
dz 2 2b
By substituting for the parameters in Figure 2-5(a), we can obtain an equation for
the beam diameter.
2 (dB 2
B = e(dB/dZ) 0 /dZ)
(2.14)
where B = b/b0 , Z = Dz/b0 , and b0 = b(z = 0). At the minimum beam diameter,
dB/dZ = 0, so the minimum beam diameter, bmin , is determined as
bmin 2 2
Bmin = = e(dB0 /dZ) = e(tan /D) (2.15)
b0
with D defined above, and the angle defined in Figure 2-5 [37]. In reality, thermal
velocities of the electrons will disperse the beam, and achieving this predicted mini-
mum beam diameter is impossible. However, by changing the shape of the cathode
radius and other dimensions of the electron gun, the beam can be shaped to achieve
whatever beam diameter is necessary for the application.
Note that beams which are compressing sharply, with a large angle, , will have
a smaller beam diameter than those which are compressing more gradually. This
analysis is a way to easily manipulate the electron beam and design an electron gun
with the appropriate beam diameter for use in the slow-wave structure. Next, a
magnetic field will be introduced that prevents the divergence of the beam, so that
the electron beam can be used with the small radius. With an appropriately matched
field, the beam will stay compressed due to matched forces, as shown in Figure 2-5(b).
50
2.2.2 Magnetic Field Confinement
Magnetic field was not mentioned in the role of the electron gun, but it is a major
influence on the formation of the beam and direction of the beam after the anode.
The magnetic field influences the beam in two parts. First, the peak magnetic field
keeps the beam confined to the minimum beam radius over a distance. Secondly, the
compression of the beam is influenced by a matched magnetic field, which is near zero
at the cathode, and grows to the peak field at about the location of the minimum
beam radius. The combination of these two fields, if perfectly matched to the beam,
results in a compression which looks like Figure 2-5(b). In this section, we will discuss
solenoid magnetic field focusing with Brillouin flow, however other types of focusing
are often used, such as Periodic Permanent Magnet (PPM) focusing, which results in
a laminar flow of the beam.
For solenoid focusing, the peak magnetic field must be large enough to counter-
act the space-charge forces of the electron beam and keep the beam confined to
the minimum beam radius. A solenoid magnet is placed around the beam tunnel,
such that the magnetic field is z-directed and is near zero at the cathode. The field
limitation for confined flow is called the Brillouin field, and can be solved by analyzing
the forces on the beam and the equation of motion for the electrons. The result is that
the beam is confined when the Larmor frequency, L = c /2, satisfies the equation:
2L = p (2.16)
where p is the plasma frequency of the beam. Substituting for these two frequencies,
and solving for the Brillouin magnetic field, BB ,
s
2mI0
BB = (2.17)
0 eve rb2
where rb is the radius of the beam and all of the parameters are already known
quantities of the electron beam. This equation can be rewritten using the Alven
51
current, IA = 17 kA. s
8m2 c3 I0 /IA
BB = (2.18)
e2 ve rb2
Note that the ability of the Brillouin field to cause confined flow is the case for
Brillouin flow, where the field at the cathode is zero. For immersed flow, where the
magnetic field at the cathode is not zero, the magnetic field required for balanced
flow is slightly larger. If the field is not matched to the beam, a ripple in the radius
of the beam will develop as the beam travels along the z-axis. Insufficient magnetic
field will result in a ripple that is larger than the minimum beam radius, whereas
excessive magnetic field will lead to an over-compressed beam with a ripple smaller
than the ideal radius.
Secondly, the compression of the beam is influenced by the magnetic field at that
location. In a thermionic, space-charge limited cathode, the magnetic field is near-
zero at the cathode for Brillouin flow. Alternatively, for immersed flow, there is a
field at the cathode. In either case, as the beam is compressed, the magnetic field
should grow in a matched method. The addition of an magnetic pole piece, a piece
of ferrous material, such as iron, and/or an opposing-field gun coil magnet can help
to match the beam and the field. A pole piece disturbs the magnetic field in a very
predictable way. For instance, if a thin pole piece with an aperture of size radius rpole
is placed at z=0 next to a solenoid magnetic which provides Bz (z > 0) = B0 , then
the magnetic field can be described as
1 1 1 u
Bz = B0 + tan u+ (2.19)
2 1 + u2
where u = z/rpole [6]. It can be seen that a smaller aperture will cause a sharper rise
in the magnetic field.
As with the peak magnetic field, a mismatch between the compression magnetic
field and the beam can cause rippling due to over-compression or under-compression of
the beam. In general, minor scalloping will occur in the beam due to these mismatches
which cannot be entirely avoided experimentally. In addition, an axis offset or tilt
52
between the electron gun and magnetic field can cause oscillations in the field. In
general, any deviation from ideal operation leads to decreased performance of the
TWT from theory. It is best to match the magnetic field in order to get an ideal flow
for the electron beam, with minimal scalloping and velocity spread.
2.2.3 Collector
After the beam is formed and used in the circuit, the beam exits the magnetic field
and expands so that the collector can gather the beam. Typically composed of copper,
the collectors purpose is to ensure that the beam is terminated in a way that is good
for the vacuum system. For high power vacuum systems, this can be a challenge. One
must ensure that the beam disperses enough that is does not completely damage the
copper of the collector and that outgassing in the system is kept to a minimum.
Depressed collectors can be used to create a more efficient experiment. Only
a fraction of the power in the electron beam is transferred to the electromagnetic
beam. To combat lost energy, the collector can be biased with a voltage, and the
beam can be collected in stages. This is a way of recycling the power in the system
and designing a more efficient device. Multi-stage depressed collector designs can
increase experimental efficiencies by up to 65 % [17]. This consideration can be key
to a practical application of a TWT. However, for the experiment considered in this
thesis, a simple copper collector was used, and efficiency was not a priority in the
circuit.
53
it reduces the group and longitudinal phase velocities of an electromagnetic wave in
order to match the phase of the wave with the electron beam that passes through
a beam tunnel in the structure. The electron beam and electromagnetic wave are
traveling at a speed much less than the speed of light. By matching the velocities of
the beam and wave, an interaction takes place where the wave causes velocity shifts
in the electron beam. Figure 2-2 shows how part of the beam consistently sees an
accelerating field and part of the beam consistently sees a decelerating field. This
effect causes longitudinal bunching of the electron beam. After bunching occurs, the
electromagnetic wave is able to extract energy from the beam. The extraction of
energy causes the wave to be amplified and the beam to de-bunch. At this point
of de-bunching, maximum amplification has occurred and saturation is reached in
the device. If the interaction does not end at this point, the beam and wave will
continue to exchange energy and the amplitude of the wave will oscillate. As such,
the interaction should be ended before the saturation point of the structure.
The dispersion relation between the electromagnetic field and electron beam is
easy to understand by considering small perturbation theory. The electric field of
z -traveling wave goes as E exp[j(t z)], where is the frequency of
the +
the wave, and is the wavenumber. If is imaginary, the electric field will have
an exponentially growing (or decaying) real component, exp[Im()z]. To determine
the growth of the wave, we need to establish a relationship between the and by
using continuity equations for the beam and electromagnetic field relationships. To
do this, the electron beam and electric field will be considered separately to establish
a relationship between the disturbed current and the electric field in the circuit.
First, parameters with small perturbations are introduced to understand the mo-
tion of the electron beam with the electric field. Assume that the electron beam has
small disturbances in the velocity, current, and density that correspond to the electric
field disturbance, such that
vtot = v0 + v, (2.20)
itot = I0 + i, (2.21)
54
tot = 0 + , (2.22)
the initial quantities are indicated with a subscript of zero, and the small pertur-
bations are indicated with the bar above the variable. The perturbations are all
proportional to exp[j(t z)], and small in comparison to initial quantities. First,
the electron equation for current is introduced,
I0 + i
= (0 + )(v0 + v) (2.24)
A
where A is the cross-sectional area of the beam. This can be reduced to second order
(ignoring very small terms),
i = (
v0 0 v)A (2.25)
Next, consider the current equation, which establishes the relationship between cur-
rent and density,
J = (2.26)
t
Since the perturbations are of the form exp[j(tz)], the derivatives can be reduced,
so that
ji
= j (2.27)
A
Combining equations (2.25) and (2.27), establishes an equation for the perturbed
velocity and current,
v0i
i = 0 v A (2.28)
A
which can be rearranged,
i
v = ( v0 ) (2.29)
0 A
Next, for the electron beam, consider that the electrons are accelerating in the
electric field, which we will call EzT , such that EzT = Ezn + Esc . In this case Ezn is
55
the electric field in the circuit, and Esc is the space charge field that is due to the
beam itself and its interaction with the beam tunnel in which it is propagating. The
electrons accelerate as
dv e
= EzT (2.30)
dt m
The velocity derivative must be split up into its partial derivatives,
dv v v z
= + (2.31)
dt t z t
z
with t
= v0 , the partial derivatives can be analyzed with the perturbed functions,
e
v (j jv0 ) = (Ezn + Ezsc ) (2.32)
m
i
Ezsc = j (2.34)
A
The above equations assume an infinite beam, but since the beam is finite, a space-
charge reduction factor, R, is introduced to account for the finite beam radius,
i
Ezsc = jR2 (2.35)
A
The reduction factor is, essentially, a reduction in the plasma current that the electron
beam sees due to the conductive wall of the beam tunnel; a plot of the space charge
reduction factor is shown in Figure 2-6. In this plot, you can see that R is dependent
on the relative size between the electron beam radius, b, and beam tunnel radius,
a. For simplicity, this can be re-written in terms of the reduced plasma current,
56
eb
R a b
eb
Figure 2-6: A plot of the space charge reduction factor, R vs. e b for different fill-
factors, a/b. To the left an illustration of an electron beam with radius b inside a
beam tunnel with radius a. The case for a beam in free space is a/b = and for a
beam that is the same diameter of the beam tunnel is a/b = 1 [20].
q = Rp such that
mq2
Esc = j i (2.36)
e0 A
i = 0 eA jEzn
(2.37)
2
mv0 (j j)2 + q2
v0 v02
i = je I0 Ezn
q2
(2.38)
2V0 (e )2 + v02
A simplified version of this equation does not consider space charge. If Ezsc = 0,
i = je I0 Ezn
(2.39)
2V0 (e )2
57
where I0 and V0 are the current and voltage operation conditions of the electron gun.
Now, we will consider circuit equations to explain the electromagnetic field, and
find a relationship for i and Ezn due to the SWS. Quite simply, the impedance of the
circuit is described as the coupling impedance, K, such that
|Ezn |2
Kn = 2 (2.40)
2cn P
where P is the power in the field and cn = /vpn , where vpn is the phase velocity
of the wave, and the subscript n indicates the mode of operation (typically n = 1,
and only c is indicated). Since the beam has a finite cross-sectional area, K can
be averaged over the area which the beam would occupy. The electric field in the
coupling impedance equation refers to the unperturbed field present in the structure
without the beam. This electric field has been integrated along the beam tunnel
axis, so we are dealing with the average electric field that the electron beam sees per
period, Z
1
Ezn = Ez (z)ejcn z dz (2.41)
p
where p is the period of the circuit. The coupling impedance is purely dependent on
the geometry of the circuit, assuming ideal beam interaction within the circuit at the
frequency of interest [70].
The coupling impedance is a measure of how well the electromagnetic wave can
interact with the electron beam in a given circuit. A higher coupling impedance
indicates a better transfer of energy between the beam and wave. Typically, TWTs
have a coupling impedance anywhere between 150 . Using the coupling impedance
along with circuit equations, the electric field can be described as, [70]
j 2 c Kn
Ezn = i (2.42)
( 2 c2 )
The two equations from the electronics and circuits methods of analysis, (2.38)
58
and (2.42), can be combined to arrive at the dispersion relation, without space-charge
c 2 e 2C 3
1= (2.43)
(e )2 (c2 2 )
or with space-charge,
c 2 e 2C 3
1= h 2
i (2.44)
(e )2 + v2q (c2 2 )
0
KI0
C3 = (2.45)
4V0
Using the dispersion relation, we can solve for the gain that occurs in the circuit.
First, consider synchronous operation with the beam and wave, such that vp = v0 .
For now, we will ignore space charge and loss in the SWS. In this case, c = e . The
solution for , where Ez ejz , is of the form
= e + (2.46)
From equation (2.43), = (1)1/3 e C. This leads to three solutions for , corre-
sponding to three different waves. Two slow forward waves are described with
!
C 3
= e 1+ j C (2.47)
2 2
59
results in a fast backward wave, with = e + e C 3 /4. The gain, G, in dB that is
seen in the circuit is a result of the slow, forward growing wave,
2
1 3 C2N
G = 10 log10 e2 [dB] (2.48)
3
For non-synchronous motion, where the wave and electron beam are not traveling
at the same speed, a non-synchronous parameter is introduced, b, such that
ve vp
b= (2.50)
Cvp
therefore,
c = e (1 + Cb) (2.51)
= e + e C, (2.52)
with = jxy. Then, the exponentially growing field is of the form E exp[e Cxz].
Assuming small b, the dispersion relation reduces to
2 ( b) = 1 (2.53)
60
where can be numerically solved and has three distinct solutions. The gain equation,
(2.49), remains the same, but B is now defined as
Note that the solution for synchronous motion occurs when x = 3/2.
Next, circuit loss is added into the solution. Assuming a loss per unit length in
the SWS, , such that E exp[z] in the circuit, we can define a parameter d,
d = /(e C) (2.55)
c = e (1 + Cb jCd) (2.56)
Taking to be of the same form as equation (2.52), the dispersion relation reduces
to
2 ( b + jd) = 1 (2.57)
Once again, can be numerically solved and has three distinct solutions. The gain
equation remains the same with B as defined in (2.54). However the solution for x
(and calculated gain) will be less than case without loss considered.
Finally, consider the effect of space charge in the interaction. Taking the case with
loss and non-synchronous motion, we can solve equation (2.44) for the conditions of
equations (2.56) and (2.52).
1
= + 4QC (2.58)
b + jd
2
q
4QC = (2.59)
C
61
60
Ideal
w/ loss
50
30
20
10
0
5 4 3 2 1 0 1 2 3
nonsynchronous parameter, b
The analytical theory relies on the assumption of synchronous motion (or near-
synchronous motion), meaning that the phase velocity of the wave along z must
be close to the velocity of the electron beam. Since the electron beam is typically
non-relativistic, it is useful to have a SWS, which causes the electromagnetic wave to
travel slower than the speed of light in the axial direction.
There are several different ways to create a SWS, which is characterized as a
62
Electron Beam
Figure 2-8: A diagram of a helical TWT slow-wave structure with the electron beam
traveling through the middle of the helix. Figure from [70].
structure where the group and phase velocities of the wave are less than the speed
of light. In this section we will discuss, in detail, helical TWTs and coupled-cavity
TWTs.
A helical SWS was the design of the first invented TWT. A diagram of a helical struc-
ture with an electron beam is shown in Figure 2-8. In this device, the electromagnetic
wave travels along the wire conductor that has been wrapped into a helix. The wave
travels at the speed of light along the wire. However, in the frame of the electron
beam, the wave travels much slower than c along the z-axis. The field travels along
the helix, with a certain pitch factor, , which slows down the wave such that
vp = c sin (2.60)
where the pitch factor can be described in terms of the length between helical rota-
tions, or period, p, and the radius of the helix, R,
p
tan = (2.61)
2R
The phase velocity calculation, (2.60), indicates that the phase velocity does not
depend on frequency. Though some generalities have been taken into account, there
63
is still very little dependence on frequency in the phase velocity, therefore helical
TWTs are able to support very wide bandwidths; up to 3 dB bandwidths (50 %) are
common in low frequency TWTs. However, the field near the center of the electron
beam is small, and the coupling factor from a helical TWT is typically low (about 1
).
The frequency dependence of the helix shows up when you consider the number
of wavelengths per period, Np .
Np = (2.62)
2R
This parameter establishes a relationship between wavelength, , and the radius of
the helix, R. Consider a helical structure of radius R at various frequencies. For
high frequencies, Np is small. In the extreme case (Np < 1) the electric fields remain
very close to the helical wire; this will severely decrease the interaction that can take
place (K 1), and the structure will not function as intended. At extremely low
frequencies, Np is large. As Np increases, the electric field will be farther away from
the wire and the interaction will decrease due to wave interference with itself [70].
The ideal situation is about Np = 4. In this case, Np is appropriately sized such
that the electric field is contained near the wires where magnetic flux exactly cancels
between the turns of the wire which enhances the field at the location of the electron
beam.
The trouble with helical TWTs is that, as the frequency of the device increases,
the size of the helix becomes incredibly small. Considering this relationship between
Np and TWT performance for a helical TWT at high frequencies, a very small radius
needs to be constructed for ideal performance. For example, a 94 GHz TWT with
Np = 4 must have a radius of R = 127 m, which is incredibly small. Outside
of construction considerations, this radius leads to a very small beam-tunnel which
could not support a large current. Even to get a small amount of current through the
beam tunnel would require a large current density and a strong magnetic field must
be used. In addition, the delicate nature of small structures leads to the inability for
high frequency helical TWTs to support high average powers. As such, it is necessary
64
Figure 2-9: A folded waveguide coupled-cavity design. (Picture from [37].)
to explore other types of TWTs for the W-band if we would like to find high power
devices.
A generic folded waveguide TWT is shown in Figure 2-9. The waveguide has been
folded to a certain period length to create a SWS and allow for an electron beam to
travel through the waveguide and interact with an RF wave. The phase velocity of
the wave along the electron beam tunnel, vp , is nearly equivalent to the velocity of
the electron beam, ve , allowing a transfer of energy from the beam to the wave. A
staggered coupled-cavity TWT is formed when the upper and lower portions of this
waveguide are made into coupling slots instead of loaded waveguides, as shown in
Figure 2-10, with useful parameters labeled. The staggered slots allow for a specific
phase advance per cavity for a forward wave interaction [37]. The cavity can be
any shape that resonates with the operation frequency; often circular or rectangular
cavities are used.
For a coupled cavity TWT, a generic dispersion relation is shown in Figure 2-11(a).
As seen, the dispersion follows a near-sinusoidal relationship between phase advance
and frequency. Assuming interaction near the center of the cold circuit bandwidth,
65
(a) (b)
1.4
Cavity mode 75 GHz
Electron Beam 1.2 95 GHz
110 GHz
0.8
0.6
0.4
0.2
0 0
0 90 180 270 360 0 10 20 30 40 50
Phase per period (Degrees) Voltage (kV)
Figure 2-11: (a) A generic dispersion relation for a coupled-cavity TWT shown along
an intercepting electron beam-line. (b) The period requirement for certain opera-
tion voltages along various operation frequencies in the W-band as calculated with
equation (2.63).
p = (3/4)ve /f (2.63)
assuming that operation is wanted near the center of the bandwidth. Figure 2-11(b)
shows the relationship between operation voltage and period of a SWS. One can see
that operation points with a larger voltage are desirable due to the larger period
size; this would make a structure at high frequency easier to fabricate. However,
there are negative consequences of operating at high voltage. First, in experimental
implementation higher voltages are harder to implement. Second, the voltage plays
an effect in the gain of the TWT, recall that the gain is proportional to the Pierce
1/3
parameter, C, which is, in turn, proportional to 1/V0 . Therefore, a balance must
be struck between these trade-offs to determine operation voltage.
In general, many of the parameters for the coupled-cavity TWT are determined by
the desired operation of the device. In general, the frequency of operation determines
the size of the cavities. The period between the cavities sets the operation voltage.
The beam tunnel size must be below cut-off for the operation wavelength, and must
be chosen such that it allows minimal coupling between the cavities; this allows for
66
a larger bandwidth of operation. In design, the size of the coupling slots between
the cavity will determine both the coupling impedance and the operation bandwidth.
These parameters will be discussed in detail in the next chapter. In general, the
bandwidth of a coupled-cavity TWT will be much smaller than the helical TWT, but
it will have a much larger coupling impedance (on the order of 10 ). The robust
nature of the cavities also allows for larger power loading.
67
68
Chapter 3
Design of a 94 GHz Overmoded
Coupled-Cavity TWT
Having established the theory of a TWT, the specific design of an overmoded coupled-
cavity TWT will be discussed along with the advantages that can be achieved with
an overmoded device. The design specifics for the 94 GHz overmoded TWT will be
presented along with simulations of the experiment.
The goal of this thesis is to design and build a W-band TWT with a center-frequency
of 94 GHz that demonstrates more than 30 dB of gain and a high saturated output
power, greater than 100 W. In order to meet these design constraints, achieving a large
bandwidth or high efficiency in the device is not a primary concern. However, the
tunable bandwidth should be greater than 1 GHz, and the instantaneous bandwidth
should be more than 100 MHz. An overmoded TWT is an ideal design to meet these
specifications since it will allow for a large amount of beam power to be coupled into
the device.
The design will have an overmoded cavity with a large beam tunnel. This will be
used to operate at a low magnetic field, which will support an electron beam with
a low current density. The design will be limited by a 2.5 kG solenoid magnet that
69
was available for the experiment which limited the current available for the design.
Alternatively, a large magnetic field could have been used to increase the current
available in the system.
The experiment will also be pulsed operation, not continuous wave (CW). As
such, heating considerations have not been taken into account in the design. Water
cooling channels could be added to the design to scale the operation to longer pulse
widths, but that is outside of the scope of this thesis and would complicate the initial
proof-of-concept for the overmoded TWT design. The lab is equipped with a 2.8
microsecond pulse generator that can provide a well-shaped voltage pulse between
10100 kV.
The specifics of the magnet and pulse modulator will be discussed in the next
chapter, but the parameters are worth noting since they add engineering constraints
to the cavity design and must be considered for TWT simulations.
The coupled-cavity structure allows for many elements to be changed to alter the
electromagnetic fields in the TWT. For example, ridges can be placed along the beam
tunnel in the cavity to increase the coupling impedance of the TWT without reducing
the bandwidth of the structure. These ridges cause an increase in the electric field
along the axis of the TWT [28]. Unfortunately, the ridges are rather small, difficult to
manufacture, and subject to damage at high powers. Another possibility for change
in the structure is the shape of the TWT cavity. Oftentimes, cylindrical cavities are
used instead of the rectangular cavities depicted in Figure 2-9 to take advantage of
matching the shape of the electromagnetic fields to the beam tunnel. In addition,
there are many possibilities for placing the coupling slots between the cavities and
in manipulating the phase advance seen between the cavities. Many other cavity
designs are possible, but are typically unrealistic to build. Since this thesis will have
an experimental validation of design, an emphasis was put on the practicality of
manufacturing the cavities. In this thesis, a coupled-cavity TWT has been designed
70
which takes advantage of the robust nature offered by operating in a higher order
mode of the rectangular cavity for the frequency of interest; this operation mode is
what designates the device as an overmoded TWT.
jkx kz
Ex = H0 cos kx x sin ky y sin kz z (3.1)
k02
jky kz
Ey = H0 sin kx x cos ky y sin kz z (3.2)
k02
j(kx2 + ky2 )
Ez = H0 sin kx x sin ky y cos kz z (3.3)
k02
ky
Hx = H0 sin kx x cos ky y cos kz z (3.4)
k0
kx
Hy = H0 cos kx x sin ky y cos kz z (3.5)
k0
Hz = 0 (3.6)
71
z y
x
d b
a
Figure 3-1: An illustration of the electric and magnetic fields in the TM110 cavity
mode. (Figure modified from [12].)
where
kx = m/a (3.8)
ky = n/b (3.9)
kz = s/d (3.10)
where m, n, and s are integers. The cases that lead to the best interaction with the
electron beam require that the electric field exist only in the z-direction, or Ex =
Ey = 0. This condition is easily satisfied for the TMmn0 modes where
kz = 0. (3.11)
The cavity frequency for the TMmns mode is determined via the wave equation,
where
2 = k 2 (3.12)
72
This is the resonant frequency for a perfect rectangular cavity. In reality, the fields
and frequency of the cavity are affected by both the beam tunnel and the coupling
slots in the circuit, which will shift the frequency higher or lower, depending on where
they are placed in the circuit.
In general, a cavity has a strict operation frequency with a very small bandwidth,
since the bandwidth is proportional to the resonant Q of the cavity [40]. However,
by adding coupling slots in the cavity, the Q decreases and the bandwidth becomes
larger than just the single operation point. In this way, the size of the coupling slots
determines the bandwidth of the TWT as well as having the ability to enhance or
decrease the resonance of the cavity and, therefore, the coupling impedance of the
TWT.
Typically, the cavity of a TWT is chosen such that the operation mode is the
fundamental cavity mode, TM110 . By choosing a fundamental mode, a majority of
the electric fields power is in the center of the cavity and able to interact well with
the electron beam passing through. However, the size of the TWT is predetermined
due to the nature of the modes in the waveguide. For a 94 GHz TWT, the sides
of a square cavity, a and b, would be 2.26 mm (slightly less than 0.1 inch). This
is small enough to cause manufacturing difficulties when using standard machining
tools and materials for vacuum, especially when tight manufacturing tolerances are
taken into account. In addition, the size of the coupling slots and beam tunnel must
be smaller than the cavity size, which causes an even larger burden on manufacturing.
A small beam tunnel also leads to either small currents or large magnetic fields for
confinement. The fundamental mode may be the obvious choice for operation, but it
is not the only possibility.
An overmoded coupled cavity TWT operates in a higher order mode of the cavity.
By choosing to operate in a higher order mode, the size of the cavities increases for
the same frequency of operation. Figure 3-2 shows the first four modes of a cavity
of the form TMmn0 which has been tuned for operation at 94 GHz. In this case,
73
TM 11 TM 21
1
b=2.54 mm
0.5
0.5
a=2.05 mm a=4.10 mm
TM 31 TM 41
a=6.15 mm a=8.20 mm
Figure 3-2: Contour plots of the electric field, E = Ez , for some of the lowest order
cavity modes. For each mode, b has been set to 2.54 mm, and a is determined such
that the mode will oscillate at 94 GHz.
b has been taken to be consistent among the cavities, and a has been changed to
select the frequency of operation. A larger cavity has several advantages: easier to
manufacture, a larger beam tunnel, and lower heat loads are just a few. These points
can lead to many more advantages in the design of the TWT, as well. For example,
a larger beam tunnel allows for an increase in the operation current, or it can lead
to a lower current density and operation with a lower magnetic field. The design of
the overmoded TWT will be determined in detail in the next section. Of the cavities
depicted in Figure 3-2, only the fundamental mode and the TM310 (or TM31 mode,
for simple designation) have a peak electric field at the center of the cavity. For
simplicity in design, the TM31 mode was chosen as the high order mode of operation
for this design in order to keep the electron beam at the center of the cavity.
Without considering coupling slots or the beam tunnel, a rectangular cavity which
operates at 94 GHz in the TM31 mode is simple to design. The frequency of the TM31
74
cw
sh
p
ch
sw 2r
Figure 3-3: Two cavities of the TWT structure linked via staggered coupling slots,
as seen with an isometric view and from the top of the structure. The purple areas
show the negative vacuum space, which is surrounded by copper metal to make the
cavity. The dimensions for the slot width, sw, slot height, sh, cavity width, cw, and
cavity height, ch are indicated.
mode is s
2 2
1 3 1
f310 = + (3.14)
2 a b
Since there is no electromagnetic field variation in the z-direction, d, is independent
of the frequency of operation. Since d is in the same direction as the electron beam,
the simple relationship was chosen,
d = p/2 (3.15)
to allow for uniform beam bunching. The period length, p, is set by the desired
voltage of operation. Therefore, only two free variables are left to adjust to set the
frequency of the mode, a and b.
For simplicity that will be apparent when designing input and output couplers, one
of the sides of the cavity should be the same height as standard WR-10 waveguide.
WR-10 waveguide is the standard rectangular W-band waveguide, which is 0.10 x
0.05 inches in size (2.54 x 1.27 mm). Therefore, b = 2.54 mm, and equation (3.14)
can be solved such that a = 6.15 mm. These dimensions are solved for various modes
in Figure 3-2.
In reality, coupling slots and a beam tunnel must be added to the design, as
75
shown in Figure 3-3. The variables a, b, and d have been changed to ch, cw, and
p/2, respectively, since the cavity equations are no longer completely valid. The full
structure will be a series of these cavities repeated over a certain length. The addition
of the coupling slots and beam tunnel will change the frequency of operation for the
cavity slightly. In addition, these features will change the Q and the bandwidth of
the cavity.
In any given cavity, it is important to note that an infinite number of higher order
modes exist at different frequencies. When building a TWT, there is a possibility of
oscillations happening in the mode of interest as well as in any of the other modes
in the waveguide. These oscillations would disrupt the operation of the TWT and
prevent high powers from being reached. For operation in the fundamental mode of a
cavity, oscillations in other modes are generally not a problem since all other modes
will exist at higher frequencies and have higher start currents for oscillation than
the fundamental mode. For the overmoded cavity, other modes must be considered,
particularly lower order modes with low start currents for oscillations.
The three lowest order modes are shown in the dimensions for the 94 GHz TM31
mode cavity in Figure 3-4. These modes were calculated in ANSYS HFSS, a high
frequency 3-D finite element electromagnetic field solver. Unless otherwise noted,
all field solutions in this thesis were obtained with HFSS. The fundamental mode
of this cavity is at about 67 GHz and is a large threat to operation. Since it is the
fundamental mode, it has a low start current for oscillations, high electric field on axis,
and a large predilection for the electron beam to interact with the mode. The other
mode shown, TM21 , is of only minor interest. Under the right operating conditions,
there is a possibility for oscillations, but there is no peak electric field in the center
of the cavity (where the electron beam is located). Therefore, an offset in the beam
tunnel would be needed to excite the mode, and it would be a small interaction even
under that condition.
With the addition of coupling slots and a beam tunnel, the bandwidth of the modes
76
TM 11 TM 21
TM 31
1
0.5
Figure 3-4: The electric field patterns of the three lowest order TM modes in a
rectangular cavity tuned to 94 GHz for the TM31 mode.
is increased. These features are outlined in Figure 3-4. The dispersion relation for
the three lowest order modes is shown in Figure 3-5, where the frequency dependence
of each mode is shown as a function of the phase advance per cavity. This plot
also shows the electron beam line, defined as = kz ve , which has been tuned for
operation with the 94 GHz mode. In the locations where the electron beam crosses
with a dispersion relation, the phase velocities of the two are matched and they are
in synchronous motion. This intersection means that there can be a strong coupling
interaction between the beam and the mode. The interaction which occurs with the
fundamental mode could, potentially, cause an oscillation at about 67 GHz.
In order to eliminate the threat of the fundamental mode from interacting with
the electron beam, additional losses should be added to reduce the start current for
the TM11 mode. If the walls of the cavity were made out of a lossy material, all
of the modes in the waveguide would see a high loss in the circuit. For example, if
the walls of the cavities were made of a dielectric material instead of copper, there
would be a significant amount of loss in the system. Ignoring the fact that this would
be an impractical cavity design, the lossy dielectric would reduce the potential for
77
100
95
90
Frequency (GHz)
85 TM11
TM21
80
TM31
75 30 kV Beam
70
65
60
0 50 100 150 200 250 300 350
Phase per Cavity (deg)
Figure 3-5: The dispersion relation for the lowest order TM modes of the cavity along
with an electron beam line optimized for interaction with the TM31 mode.
oscillation of all of the modes in the system - including the TM31 mode.
78
cw
dh
sh p
ch
sw dw 2r
Dielectric Loading
19 cavities
RF out
RF in Electron
beam
Figure 3-6: Placement of dielectric on the top and bottom of the cavity. Purple is
the negative vacuum space, pink is dielectric, and the vacuum space is surrounded by
copper. Two cavities are illustrated which are repeated indefinitely for an infinitely
long TWT with a specific phase advance specified between the two-cavity sections.
A sample circuit with 19 total cavities is shown with RF input and output marked
on the first and last cavities, respectively.
see high losses due to the dielectric loading, while the TM31 mode has a minimal loss
due to the dielectric loading. In fact, the small loss that the TM31 mode sees is useful
for TWT operation. Over the entire length of the circuit (10100 cavities), the TM11
mode will see significant losses and will not be a threat to oscillations. The TM31
mode will have slightly more loss than in the pure copper structure, but the relatively
high-loss circuit eliminates the possibility of round-trip oscillations in the TM31 mode
due to reflections from the output, as will be shown in simulation. In the final design,
a sever is not needed due to these losses, as will be discussed later.
The material used for the lossy dielectric loading must have several key properties.
First, and foremost, it must be safe to use in vacuum, otherwise it cannot be used
in the TWT. Secondly, in order for the dielectric to perform as we have described,
79
TM 11 TM 21
TM 31
1
0.5
Figure 3-7: The electric field patterns of the three lowest order TM modes in a
rectangular cavity tuned to 94 GHz for the TM31 mode with dielectric loading added.
it must have a large loss tangent at high frequencies. Many dielectrics will not pro-
vide sufficient losses for frequencies above 10 GHz, and this material must suppress
oscillations at 67 GHz and above. In addition, it would be ideal to have a dielectric
with good thermal properties so that it will be able to deal with the possible stresses
of the TWT. Several possible dielectrics fit these categories, but aluminum nitride
composites are particularly good at providing high relative permittivities and large
loss tangents at high frequencies. The aluminum nitride (AlN) composite from Sienna
Technologies that will be used in experiment has a loss tangent of tan = 0.26 and
a relative permittivity of r = 34 at 12 GHz [60]. Previously, this material was used
to show the suppression of 10 GHz modes in a proof-of-concept TM31 coupled-cavity
design on which this work is based [10] These values were used in HFSS simulations,
though they may be lower at higher frequencies. Cold tests, as reported in Chapter
4, show that the AlN composite properties are sufficient for the overmoded TWT
performance.
80
0
10
without dielectric
with dielectric
12
60 70 80 90 100 110 120
Frequency (GHz)
Figure 3-8: The transmission through a single cavity with and without dielectric
loading for the lowest four TM modes. The dielectric used in simulation had a high
loss tangent, tan = 0.25, and r =24 for all frequencies.
The goal of the TWT design is to create a cavity structure, with coupling slots and
a beam tunnel that will allow for high gain and high peak power. In order to achieve
this, a large Pierce parameter is wanted in the circuit since the gain is proportional
to C. To reiterate, the Pierce parameter is defined as,
KI0
C3 = (3.16)
4V0
In essence, this amounts to two parameters that can be manipulated in the circuit:
the coupling impedance, K, and the ratio of current to voltage, I0 /V0 .
A large coupling impedance primarily results from having a peak electric field
region on-axis and with alignment to the beam tunnel. As a reminder, K is a result
of the integral of the magnitude of electric field along the beam tunnel. In addition
to the peak field on-axis, the design of the coupling slots should ensure that the
wave does not couple to the next circuit via the beam tunnel; this behavior would
81
(a) (d)
1400 6
1200
Current (mA) for 30kV, 2.5kG
5
1000
K at mid bandwidth
4
800
600
3
400
2
200
0 1
0.1 0.2 0.3 0.4 0.5 0.6 1 1.2 1.4 1.6 1.8
Beam Radius (mm) Slot Height (mm)
(b) (e)
12 12
10
8 8
6 6
4 4
2 2
0 0
0.2 0.3 0.4 0.5 0.6 0.7 1 1.2 1.4 1.6 1.8
Beam Tunnel Radius (mm) Slot Height (mm)
(c) (f )
0.028 4
0.026 3.5
C at mid bandwidth
K at mid bandwidth
0.024 3
0.022 2.5
0.02 2
0.018 1.5
0.016 1
0.2 0.3 0.4 0.5 0.6 0.7 0.4 0.5 0.6 0.7 0.8 0.9 1
Beam Tunnel Radius (mm) Slot Width (mm)
Figure 3-9: The variation of the (a) current, (b) K and (c) C, as dependent on the
beam tunnel radius with constant coupling slot parameters and assuming a 2.5 kG
limiting field. Alongside the variation of (d) K and (e) cold-circuit bandwidth vs.
slot height and (f) K vs. slot width (Graphs taken with all other parameters kept
constant.)
82
15
r=0.4mm
r=0.5mm
Coupling Impedance, K, ( )
r=0.6mm
r=0.7mm
10
0
92 94 96 98 100
Frequency (GHz)
Figure 3-10: The effect of beam tunnel size on calculated coupling impedance through
the middle of the circuit. For each simulation in HFSS, all other parameters were
kept consistent.
reduce the coupling impedance since the oscillations between the cavities would be
less pronounced.
Therefore, the relative size between the coupling slots and beam tunnel greatly
affects the coupling impedance observed in the circuit. The relationships between
these circuit dimensions, the coupling impedance, and Pierce parameter are shown
in Figure 3-9. In essence, it is wanted for both the coupling slots and beam tunnel
to be of sizes that are below cut-off for the 94 GHz mode (this equates roughly to
having dimensions less than 1.5 mm). To couple through the slots and not the beam
tunnel, the slots should always be kept slightly larger than the beam tunnel; this
parameter ensures a larger coupling impedance. Figure 3-10 shows how the coupling
impedance is inversely proportional to the beam tunnel radius. However, more current
83
can be pushed through the circuit with a larger beam tunnel. As it is, this allows
for an optimum beam tunnel and slot dimensions to be found. Figure 3-9(c) shows
a peak in the Pierce parameter at about 0.3 mm radius, though this data was taken
while keeping the coupling slots at a consistent size. With further optimizing of
the circuit, the beam tunnel became slightly larger allowing for more current in the
system without severely limiting the coupling impedance.
The limitations on the current in the design are also a large factor in the gain that
the TWT will see and play a large part in the design of the circuit. One of the goals
of using an overmoded circuit to allow for a larger beam tunnel in the design. We see
with the limitation due to the coupling slots, that the beam tunnel can be large for
the TM31 circuit design, but it is useful to know the amount of current that can be
supported in the design for particular beam radii. In that limitation, the magnetic
field also sets certain limitations in the design that are important to discuss. Due
to the Brillouin limitation (described in Chapter 2), only a certain current can be
supported in the TWT at a particular voltage. Figure 3-11 shows the limit for the
84
2*Brillouin = 2.5kG
90
2rb = 0.6mm
80 =0.8mm
=1.0mm
70 =1.2mm
60
I0 / V0 (mA/kV)
50
40
30
20
10
0
10 20 30 40 50 60 70
V0 (kV)
Figure 3-11: I0 /V0 ratio vs. operation voltage, V0 for different electron beam radii
(and subsequent electron beam tunnel radii). The I0 /V0 limitation was calculated
considering a magnetic field of 2.5 kG, and operation at 2 times the Brillouin field
condition.
current and voltage that can be supported by a 2.5 kG field with various beam radii,
assuming an operation condition of two times the Brillouin field (generally recognized
as a safe operation condition), Bz = 2BB . From this graph, its clear that the I/V
ratio scales significantly with the size of the beam radius. In this regard, a larger beam
tunnel will allow the TWT to reach high gain while still operating in a relatively small
magnetic field.
Another component of gain is the interaction length and period of the TWT.
Arbitrarily, we could set the length at any value to obtain the gain that we desire.
(Though a long TWT will limit the peak power and bandwidth that can be obtained.)
In reality, several factors limit the achievable gain in the device. These factors can
be split into mechanical factors and interaction factors. The main mechanical factor
that limits the length of the device is the solenoid magnet that will be used in this
experiment, which has a limit on the flat-top field region that can be used in the
85
fillets
Figure 3-12: The final cavity design showing the dielectric loading and manufacturing
fillets which create a small internal radius on part of the cavity structure.
interaction length, about 10 cm. The interaction factors that limit the length of the
circuit deal with saturation, bandwidth, and velocity spread. As the circuit becomes
longer, the velocity spread of the electron beam limits the gain per unit length in the
device and can be detrimental. In addition, the circuit must end prior to saturation
occurring; this is a design consideration which limits the peak power achievable in
the device. Also, longer circuits have more gain but less bandwidth. Therefore, the
trade-off in factors must be considered. With iteration in design, it was found that the
entire length of the circuit should be kept to less than 7 cm, which allows for a high
gain and high peak power though it limits the bandwidth and operation efficiency.
For the circuit, the period length is simply determined by the operation voltage. At
high voltage, less current can be pushed through the system, and the gain decreases.
However, at low voltages, the cavities become difficult to manufacture. For this
reason, a modest voltage of about 31 kV was chosen, leading to a period length of
0.8 mm. Combined with the 7 cm interaction length, this leaves room for about 90
cavities in the full circuit. In the end, an 87 cavity structure was chosen.
One thing to consider in the design is the manufacturing process. A picture of the
vacuum space for the final TWT design is shown in Figure 3-12, where fillets have
been added to some of the corners of the cavities. This design takes into account man-
ufacturing practicalities; the cavities will not be perfectly rectangular. This rounding
is due to the manufacturing via CNC machining that will be discussed in Chapter 4.
The rounding, if not accounted for, would lead to an offset in the cavity frequency
from the design. However, the rounded corners can lead to long simulation times.
86
Table 3.1: Dimensions of Final Structure
100
95
90
Frequency (GHz)
85 TM11
TM21
80
TM31
75 30 kV Beam
70
65
60
0 50 100 150 200 250 300 350
Phase per Cavity (deg)
Figure 3-13: The dispersion relation for the first three cavity modes of Design A, as
calculated by HFSS.
For the majority of simulations, fillets were not included; the design with no fillets is
referred to as Design A. Meanwhile, Design B included fillets and simulations with
this design were only used to check that the coupling impedance and frequency of
the manufactured components would be similar to Design A. The final dimensions
for both Design A and Design B of the TWT circuit are depicted in Table 3.1.
Using the final dimensions from Table 3.1, the parameters of the circuit were
calculated in HFSS. Figure 3-13 shows the dispersion relation for the first three modes
of the structure specified as Design A along with a 31 kV electron beam. The 31 kV
87
(a) Design A (simulated)
20 97
TM 31
96 31 kV Beam
Frequency (GHz)
15
95
K ()
10 94
93
5
92
0 91
92 93 94 95 96 97 98 0 90 180 270 360
Frequency (GHz) Phase per Cavity (deg.)
10
94
5 93
0 92
91 92 93 94 95 96 97 0 90 180 270 360
Frequency (GHz) Phase per cavity (deg.)
Figure 3-14: The coupling impedance and dispersion relation as calculated by HFSS
for the cavity design (a) for Design A without fillets and (b) for Design B with fillets.
electron beam interacts with the mode at 93.86 GHz. Figure 3-14(a) shows the
dispersion and coupling impedance calculations for both designs discussed operating
in the TM31 mode. In experiment, an electron beam between 26 kV to 36 kV could
interact with the TWT at different frequencies. The calculations for the coupling
impedance of the TM31 mode assume an ideal electron beam. At 94 GHz, the coupling
impedance is 3.2 for Design A. For completeness, Figure 3-14(b) shows the coupling
impedance and dispersion relation as calculated for Design B, which accounts for fillets
in the cavities due to the manufacturing process. The frequency has shifted higher
by 1 GHz, however the impedance is still comparable to Design A. Since Design
88
A is easier to simulate, those parameters will be used in the next section. These
calculations show that Design B is close enough for reasonable expectations in circuit
parameters.
Note that this coupling impedance is small in comparison to the coupling impedance
for a fundamental-mode coupled-cavity TWT. This is due to the fact that the elec-
tromagnetic field is spread throughout the cavity more than the fundamental mode
would be. However, the gain lost due to a low coupling impedance will be compen-
sated by a larger beam tunnel allowing for a large current to interact with the wave.
The beam tunnel of the design is 0.8 mm in diameter and this beam tunnel is larger
than all other W-band TWT experiments. In addition, the magnetic field of 2.5 kG
is lower than all other W-band TWT experiments. These decisions lead to a design
with a modest current.
For the mode of operation, the loss in the circuit is due to a combination of ohmic
and dielectric losses in the cavity. These losses are due to both the copper walls of
the cavity, which were simulated with a finite conductivity, and the dielectric loading
in the structure. HFSS predicted that the loss would be about 0.3 dB per cavity in
the TM31 mode; about half of the loss is due to the copper conductivity and half is
due to the AlN composite dielectric loading.
It should be noted that the design discussed in the previous section was kept simple
due to manufacturing considerations. Many other components can be taken into
account with the design, but were rejected because they impeded the practicality of
implementation in the circuit.
For example, the coupling slot in the final design was kept along the edge of the
cavity. This detail simplified the manufacturing. Yet, the slot could be designed to
be anywhere inside of the cavity. Placing the coupling slot at a location with a larger
electric field would lead to a higher coupling impedance but less bandwidth in the
circuit. Though the design could be optimized for 94 GHz operation, the difficulties
it presented in manufacturing were too large and designing an offset in the coupling
89
(a)
(b)
Figure 3-15: (a)The input coupler from standard waveguide into the first cavity of
the TWT. (b) Reflection from the cavity structure back to the standard waveguide
port for different sizes of the coupling aperture. The optimized aperture design is
shown.
90
electromagnetic wave is completely absorbed. The sever is typically a matched load
or an area of large diffraction losses placed about midway through the circuit. This
causes a high loss point in the circuit where the RF wave is dissipated, but the beam
remains bunched. The bunched beam re-enters a second length of interaction circuit
after the sever and continues to amplify the RF wave. A sever has little effect, less
than 3 dB, on the overall gain of the circuit while preventing oscillations. If the circuit
were not lossy, reflections from the output could result in oscillations in the circuit.
In general, the round trip transmission of the circuit must be below zero for output
reflections to not be problematic. That is, the gain that is seen in the forward wave
must be less than the loss that is seen in a reflected reverse wave. Simulations will
show in the next section that there is enough loss in the system (0.3 dB/cavity) due
to the dielectric loading to prevent oscillations. Therefore, a sever is not necessary
for stable operation.
91
Table 3.2: TWT Operation Parameters
the known parameters, such that C = 0.0203. The number of wavelengths in the
synchronous system is
N = Lf /vp = pnf /ve (3.17)
where L is the entire length of the circuit and n is the number of cavities in the
circuit. With 87 cavities in the interaction structure, N = 65.24 wavelengths in the
structure at 94 GHz. In this case, the ideal gain in the circuit is calculated to be 52.8
dB.
Simulations from 1-D LATTE were used to collaborate with the analytical the-
ory and 3-D simulations. LMSuite LATTE is a code which solves the non-linear
gain equation for TWTs [74]. Since it is a quick 1-D code, it is easy to run sim-
ulations while designing the TWT with parameters such as the dispersion relation
and coupling impedance calculated via HFSS. A comparison between LATTE and
Pierce analytical theory for the design parameters shows good agreement. It should
be noted that LATTE calculates the velocity of electrons assuming non-relativistic
92
60
Ideal
50 w/ loss
w/ SC
40 w/ loss and SC
Gain (dB)
30
20
10
0
91.5 92 92.5 93 93.5 94 94.5
Frequency (GHz)
Figure 3-16: Linear gain vs. operation voltage for 94 GHz operation and design
parameters discussed in Table 3.2. Calculations with and without space charge and/or
loss were considered.
35
30
25
Gain (dB)
20
15
10 30.5kV
31.1kV
5
32kV
0
93.4 93.6 93.8 94 94.2
Frequency (GHz)
Figure 3-17: Linear gain vs. frequency for LATTE simulations at various operation
voltages, V0 , with all other operation conditions as described in Table 3.2.
93
operation. For the device designed in this thesis, relativistic effects cannot be ignored;
therefore, this graph and all LATTE results reported in this thesis correct for rela-
tivistic velocity from the LATTE results. This is done by calculating the relativistic
operation voltage during post-processing from the electron velocity, ve , that was used
in LATTE simulation,
!
1
Vrel = 511 p 1 [kV] (3.18)
1 (ve /c)2
Note that this calculated voltage does not agree with the voltage set in the LATTE
simulation. Figure 3-17 shows the gain vs. frequency output from LATTE for various
operation voltages. Via these simulations, the operation parameters shown in Table
3.2 were optimized along with cavity parameters which were determined via HFSS
calculations for the coupling impedance and dispersion relations as shown in the
previous section.
3-D CST Particle Studio is a PIC code which was used to simulate the entire operation
circuit with a realistic electron beam and magnetic field. These calculations refined
the design from the 1-D simulations and agreed well with previous results. CST
simulations also helped direct the design to prevent oscillations in the system: both
oscillations in the fundamental mode of the cavity as well as round-trip oscillations
in the operation mode were eliminated.
Figure 3-18 shows the entire 87 cavity structure as it was simulated in CST. The
cavity parameters and operation parameters used have already been discussed. The
RF input and output ports are at faces in the cavity which face away from the circuit,
as shown in Figure 3-18. These ports are the same height as WR-10 waveguide. In
the final design, a simple taper along the width of the WR-10 waveguide is utilized
to couple from fundamental waveguide into the overmoded cavity.
CST 3-D particle studio launches an electron beam with an angular velocity and
a percentage of velocity spread. For these calculations, 1 % velocity spread was
94
RF in
electron
beam
RF out
Figure 3-18: The full 87 cavity structure with dielectric loading as simulated in CST.
The vacuum sections of the circuit are shown in purple, and the dielectric is shown
in pink. An electron beam source transmits electrons through the beam tunnel. RF
input and output ports are outward-facing sides of the the first and last cavities.
(a) (b)
4
150 10
Input
Output
2
100 10
Magnitude
Power (W)
0
50 10
2
0 10
0 2 4 6 8 10 0 50 100 150
Time (ns) Frequency (GHz)
Figure 3-19: Output power vs. time for a 90 mW input signal along with an FFT of
the output signal, showing oscillation at the TM31 input frequency of 93.86 GHz.
95
RF out
Energy of Particles at the end of the interaction circuit:
5p
Figure 3-20: The energy of the particles in the beam tunnel at the end of the circuit,
and the particle density along the length of the cavity.
considered in the circuit (by simulating the circuit with less dielectric loading), round-
trip transmission greater than zero is observed along with oscillations that develop
at the operation frequency. The ringing that is seen later in the pulse is due to low
sampling of the data (which was recorded as electric field magnitude). Figure 3-19(b)
shows the frequency components of the output and input signals in the device. The
simulation indicates that the output signal is oscillating at the frequency of interest.
No evidence of the TM11 mode at 65 GHz is seen in the signal, indicating that the
dielectric successfully suppresses the fundamental mode from self-oscillating as well.
There is a slight dip at 64 GHz and at 130 GHz which are orders of magnitude less
than the amplified signal, but the signals never develop throughout the duration of
the simulation, indicating that they will not be a problem.
The bunching of the circuit is shown in Figure 3-20 for a 10 mW input signal.
It is seen that by the end of the circuit, electrons have both gained and lost energy
(indicative of correct bunching in the TWT). An analysis of the particles in the
simulation shows that bunches have formed at the end of the circuit. Since the
bunches have not yet dispersed, this indicates that the simulation for 10 mW of
96
3
10
Latte
CST
1
10
0
10 0 2 4
10 10 10
Input Power (mW)
Figure 3-21: Output power vs. input power for the simulated TM31 coupled-cavity
TWT as calculated in 1-D LATTE and 3-D CST Particle Studio. Linear gain and
saturated output are shown.
97
98
Chapter 4
Cold Tests of the Overmoded TWT Circuit
One of the main difficulties in making W-band TWTs is machining the detailed
structures necessary for interaction at high frequencies. By making an overmoded
design, machining is easier than other TWTs due to its oversized nature. However,
machining still requires careful consideration. To deal with small features, many W-
band TWTs utilize nano-fabrication techniques [19] [3]. Development has also gone
into UV-Liga fabrication for high-frequency structures [35]. The work in this area is
important to the development of TWTs, especially at high frequency. However, with
the overmoded design, the structure is able to be built with less complex machining.
One goal of this thesis is to demonstrate the experimental operation of a W-band
TWT using traditional machining techniques.
Two machining possibilities were considered for this structure: Wire EDM and
99
direct CNC machining. First, Wire Electrical Discharge Machining (EDM) was used
along with a 4-plate design. The 4-plate design, shown in Figure 4-1(a), has the
cavities machined through the inner two plates; wire EDM was used to cut the beam
tunnel into the middle of block and to cut the cavities and input and output ports
through the entire width of the inner two plates. The coupling slots between the
cavities were machined separately. Second, the entire structure was redesign as a 2-
plate structure and constructed with a Computer Numerical Control (CNC) milling
machine (and an appropriately experienced machinist), as depicted in Figure 4-1(b)
and (c). With both techniques, precise alignments and sub-mil (1 mil = 0.001 inches
= 0.0254 mm) tolerances were achievable. With CNC milling, it was necessary for
the design to have a sufficiently small width-to-depth ratio for the end mill that was
used during machining. That is, the depth of the cavity is limited by the width of
the cavity due to the fact that an end mill must be long and thin enough to create
those dimensions within the metal block without breaking.
Both the 4-plate and 2-plate structures could be brazed together in order to finish
the structures. Commercial TWTs are typically finished with this process since the
cavities themselves must hold vacuum and brazing the cavities together would achieve
this requirement. The ability to hold vacuum is not necessary for the cavities tested in
this thesis. Chapter 5 will discuss the tube design for the experimental implementa-
tion of the TWT, which holds the overmoded TWT inside of a vacuum test chamber.
Since the cavities in this design do not have to hold vacuum themselves, brazing the
final structure was not a consideration in the machining. Since brazing often leads to
complications in the design, the decision was made to opt-out of brazing which led to
a quicker turn-around and analysis time for test structures. This decision does not
invalidate the design for commercial use.
The 4-plate structure has an easier construction than the 2-plate structure, but
is inherently flawed in design. During cold-test, it became evident that extreme
pressure was needed to sufficiently clamp the 4-plate structure together and achieve
transmission through the structure. As can be seen in Figure 4-1(a), when the plates
are mated together, there is the possibility of three gaps which would cut through
100
(a) (b) (c)
Figure 4-1: The assembly of the (a) 4-plate design, and (b) A CAD model of the
2-plate design for the 9-cavity cold test structure. (c) The assembly of the 2-plate
design with dielectric loading added.
the design of the structure. Indium foil was placed between the cavities and outer
plates (to fill the top and bottom gaps) in order to achieve measurable S-parameters
through the 9-cavity structure. Though indium foil has a low melting temperature,
it could be used in vacuum and was malleable enough to fill the gaps left between the
plates. However, the transmission achieved was not sufficient. There was significant
coupling at 94 GHz into the structure, but very high losses and minimal transmission.
The poor transmission in the 4-plate structure was due to the coupling of the
electromagnetic wave into the gaps between the plates. Figure 4-2 shows how a wave
will travel through WR-08 waveguide with gaps at the top, middle, and bottom of the
guide. The gaps make parallel plate waveguides that intersect the WR-08 rectangular
waveguide. The fundamental waveguide mode, TE01 couples into the TEM mode of
the parallel plate waveguides formed by the top and bottom gaps, and power is lost.
Fortunately, there is little coupling of power into the middle gap, as shown in Figure
4-2. These results are universal for any fundamental waveguide arranged in this
manner; the electrical field is aligned correctly to couple well into the top and bottom
gaps and does not couple into the middle gap [50]. A slight misalignment of the gap
101
Gaps
Waveguide input
Figure 4-2: HFSS simulations showing a WR-08 waveguide with gaps at the top,
mid-plane, and bottom of the waveguide. A 95 GHz wave was excited at the input.
The simulated electric field magnitude in the middle gap and top/bottom (identical)
gap is shown, indicating that the electric field couples well from the waveguide into
the top and bottom gaps.
from center would lead to some losses into the middle gap, but this loss could be
eliminated by introducing a quarter-wavelength choke flange in the gap, which would
reflect any energy that couples into the gap [45]. Unfortunately, the TWT structure
is too complex for the choke flange to be added, but the loss associated with the
middle gap is small enough that it should not be problematic.
Though the 4-plate design was simpler than the two plate design, the transmission
was inadequate. On the other hand, the next section will show that the 2-plate
design had adequate transmission since a mid-plane gap is not susceptible to very
high losses. Though the dielectric placement is complex in the 2-plate design, it is
not insurmountable. For placement, slots must be machined from above the cavity
and require care in machining, as shown in Figure 4-1(c).
The tolerances of the design were determined via HFSS simulations. For most
features, 1 mil tolerances were sufficient for frequency and coupling consistency.
The placement of the plates was held to the highest tolerance achievable and aligned
with precision pins. Rounding on the corners with CNC milling was kept to 5 mils;
the corners that must be rounded during fabrication were highlighted in Chapter 3
102
with the discussion of fillets in simulations.
4.2.1 Materials
103
Assembly without
dielectric
dielectric
placement WR10
19 cavi es; OFHC 9 cavi es; Glidcop
Dielectric placement
5.8 mm
Beam Tunnel
WR10
Figure 4-3: CAD drawings and machined 9 and 19 cavity cold test structures which
were made out of OFHC copper and LOX glidcop. Slots were machined into the top
of the cavities which could be filled with a copper insert or with dielectric loading.
is less likely to deform under stress and easier to machine with clean and consistent
edges and surfaces.
Glidcop is another material that was considered in this thesis. Glidcop is a trade-
mark copper alloy of Hoganas, Inc, a company that specializes in alloys and manu-
facturing techniques. It consists of copper that has been impregnated with a small
amount of alumina (Al2 O3 ), which leads to a small reduction in the electrical and
thermal conductivity, but enhances the high temperature strength of the alloy. This
leads to a material that is less susceptible to deformation during the stress and tem-
perature rise of machining, but it has all of the positive aspects of copper for a vacuum
application. Various types of glidcop exist with different percentages of alumina con-
tent, where higher percentages have a correspondingly higher threshold for thermal
stresses and lower electrical conductivity (or higher resistivity). For example, AL25
glidcop has 0.25 wt.% alumina and 1.98 -cm resistivity, while AL60 glidcop has
104
(a) 9-cavity copper without AlN
0
HFSS
10 9 cell copper
S 21 (dB) 20
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
Figure 4-4: The measured transmission through (a) 9- and (b) 19-cavity OFHC copper
structures compared to HFSS simulations.
0.6 wt.% alumina and 2.21 -cm resistivity [65] [66]. (For reference, OFHC copper
has an electrical resistivity of 1.72 -m.) Though glidcop has a non-insignificant
oxygen content, low oxygen (LOX) glidcop is made by impregnating a small amount
of boron into the cavity (nominally 250 ppm). LOX glidcop is safe for vacuum and
for cathodes and can be brazed, if necessary.
Recently, glidcop has been used in several vacuum devices. Most notable, it has
been explored for high gradient accelerators due to its ability to withstand large
temperatures [71] [2]. Glidcop has also been used for high temperature joints in
the ITER fusion reactor [73] [24]. In addition, glidcop has been successfully used
to construct electron gun anodes [14] and high power gyrotrons cavities [69], though
105
(a) 9-cavity glidcop without AlN
0
HFSS
10 9 cell glidcop
S 21 (dB) 20
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
Figure 4-5: The measured transmission through (a) 9-cavity AL60 LOX glidcop and
(b) 19-cavity AL25 LOX glidcop structures compared to HFSS simulations.
it has never been used in a TWT amplifier circuit. These experiments have shown
that glidcop is able to perform well under high vacuum testing, is safe for operation,
and will perform well under high temperature. It has also shown that it will perform
well in the possible case of beam interception on the glidcop structure with limited
out-gassing and damage due to melting.
For testing, short structures were constructed which consisted of 9-cavities and
19-cavities. These structures were used to test machining techniques, tolerances,
coupling into the circuit, losses in the circuit, and the effectiveness of the dielectric
loading. Transmission measurements for the copper and glidcop structures are shown
in Figures 4-4 and 4-5, respectively. In the figures, the TM31 and TM41 pass-bands are
106
visible from 9298 GHz and 110115 GHz, respectively. Dielectric loading was omitted
from this test and copper placeholders were used to fill in the slots that were machined
to hold dielectric. In the copper measurements, its clear that there is leakage of the
TM41 mode into lower frequencies; this is evidence of poorly formed coupling slots
and unmatched surfaces in the 2-plate structure. In addition, the TM31 agrees well
with theory, but shows some additional losses in the 9-cavity structure. Figure 4-
4(b) shows that the 19-cavity copper structure intensifies the losses observed in the
9-cavity structure. Figure 4-5(a) shows the measurement for a 9-cavity structure
made from AL60 glidcop, which agrees very well with HFSS simulation. In addition,
Figure 4-5(b) shows the 19-cavity results for AL25 glidcop, and indicates that the
longer structure (though made of a softer glidcop alloy) is still in good agreement
in theory. This indicates that there are minimal compiled manufacturing errors that
would become evident in longer structures.
The resonances observed in both the copper and glidcop structures (also evident
in the HFSS simulations) are due to reflections between the cavities. They are visible
in these structures due to the low number of cavities that were machined. As more
cavities are added to the structure, these resonances will become less resolved across
the bandwidth of the mode and there will be a more consistent transmission across
the band.
Though both copper and glidcop show good transmission through the cavity in
both the 9- and 19-cavity structures, it is clear that the glidcop structures show
better agreement with HFSS simulations. The glidcop structures showed increased
transmission over the copper structures, even though they have a higher electrical
resistivity. This is due to the fact that glidcop is a harder material than copper and
was easier to machine to specifications without being susceptible to deformation. Due
to its nature as a good vacuum material and the minimal losses measured in the cold
test structures, AL60 LOX glidcop will be used in the final cavity design.
107
(a) 9-cavity copper with AlN
0
HFSS
10 9 cell Copper
20
S 21 (dB)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
Figure 4-6: For copper structures, the measured transmission through (a) 9-cavity
structure with dielectric loading as compared to HFSS simulation along with (b)
comparison to results without dielectric loading, and (c) the measured transmission
for a 19-cavity structure with and without dielectric loading.
108
(a) 9-cavity glidcop with AlN
0
HFSS
10 9 cell glidcop
20
S 21 (dB)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
30
40
50
60
90 95 100 105 110 115
Frequency (GHz)
Figure 4-7: For glidcop structures, the measured transmission through (a) 9-cavity
structure (AL60) with dielectric loading as compared to HFSS simulation along with
(b) comparison to results without dielectric loading, and (c) the measured transmis-
sion for a 19-cavity structure (AL20) with and without dielectric loading.
109
4.2.2 Dielectric Loading
The dielectric used for loading the cavities must have several key properties. The
most necessary property is that it must have high loss properties at high frequencies.
It also needs to be compatible with vacuum. Though the dielectric loading wont
see much stress, it would also be useful to have low heat-loading characteristics in
the material. Common dielectrics with low heat-loading and minimal outgassing that
are commonly used in vacuum include macor, fused silica, beryllium oxide and alu-
minum nitride. Macor and fused silica have low loss tangents and will not sufficiently
suppress the unwanted modes. Some Beryllium oxide, BeO, composites are able to
provide the necessary loss tangents, but BeO is a dangerous material to machine
due to health risks and is not commercially used in vacuum systems at the present.
Aluminum nitride composites have high loss tangents, are easy to machine, and are
easily available.
Previously, a TM31 mode cavity was designed at 17 GHz and successfully tested to
show lower order mode suppression with Aluminum Nitride (AlN) dielectric loading
from Sienna Technologies, a ceramic supplier that deals with specialized composites
[10]. The AlN composite suppressed other modes in the circuit, while leaving the
TM31 mode with little additional losses. There are many types of AlN composite
dielectrics, but the most promising one to use is the STL-100c AlN-SiC which has a
dielectric loss tangent of tan =0.32 at 12 GHz and a relative permittivity r =25.
The AlN-SiC 100c composite from Sienna Technologies has good thermal and vac-
uum properties and was already shown to work in the Ka-band for mode suppression,
so it was used as the dielectric for the overmoded TWT in this thesis. The dielec-
tric pieces were machined with high tolerances to the width and height dimensions
discussed in Chapter 3. The pieces are 2.54 cm in length, so multiple pieces must be
used along the length of the final structure to cover all of the cavities.
The AlN dielectric was placed on top of the cold test cavities, and a copper cover
kept the dielectric in place while the structure was under test, as shown in Figure 4-3.
Results for the copper structures are shown in Figures 4-6. Figure 4-6(a) shows good
110
Table 4.1: Coupled-Cavity Loss with Dielectric Loading (dB/cm)
agreement of the 9-cavity structure with theory. The dielectric loading very effective
suppressed the TM41 mode. Figures 4-6(b) and (c) explicitly show the effect that the
dielectric has on the transmission of the TM31 and TM41 modes in the structure by
comparing the same structure with and without the dielectric loading in place. Its
clear that the TM31 mode is affected very little by the dielectric loading. However,
the TM41 mode has additional losses when the dielectric is added to the structure.
In fact, its clear that the 19-cavity structure has about twice as much additional
loss due to the dielectric as the 9-cavity structure. This is due to the fact that the
dielectric is adding a distributed loss into the circuit; so more cavities will induce a
proportionally larger loss into the system.
The dielectric results for the glidcop structures are shown in Figure 4-7. Figure 4-
7(a) compares the measured transmission of a 9-cavity glidcop cold test with dielectric
loading against HFSS simulations. Once again, good agreement is shown for the TM31
mode between measurement and theory, an improvement over the copper results.
Figures 4-7(a) and (b) show the effect that the dielectric has when added to the
glidcop structures by comparing measurements made with and without dielectric for
the 9- and 19-cavity structures, respectively. It is clear that the dielectric causes
decreased transmission in the TM41 mode in the same manner as it had with the
copper structure.
From the S-parameter measurement, the loss per unit length of the structure can
be calculated. The losses as they were calculated are displayed in Table 4.1. It was
found that about a 6 dB coupling loss was present in the glidcop structure, which
agrees with HFSS simulation (slightly more coupling loss was present in the copper
structure). For both materials, about 0.3 dB/cm loss was present with dielectric
111
loading; also good agreement with theory.
These cold tests have shown that the AlN composite dielectric is selectively sup-
pressing the unwanted modes in the cavities, while having little effect on the wanted
TM31 mode of operation. Therefore, the AlN composite will be used in the final struc-
ture. In addition, the successful performance of the glidcop structures also reinforces
the decision to use glidcop rather than OFHC copper in the final structure.
After construction of smaller cold test structures, the final full-length structure was
constructed with 87 cavities. Figure 4-8(a) shows the 2-plate cavity structure that has
been designed for installation in the TWT experiment. The details of the experiment
will be discussed in Chapter 5. The assembly of the structure and placement of the
dielectric loading on the cavities is highlighted in Figure 4-8(b). This figure also
highlights the conical tip around the input of the beam tunnel, which is an alignment
surface to allow for the beam tunnel to be aligned with the electron gun and magnetic
field axis, as will be discussed in Chapter 4. Figure 4-8(c) shows the fully assembled
of the cavity along with input and output WR-10 waveguide components; a standard
USB flash drive provides a scale.
Two 87-cavity structures were machined for experimental testing, which will be
referred to as Structure A and B for simplicity. Structure A was manufactured first
and underwent experimental testing while Structure B was being finalized. The two
structures are identical, except for random machining errors, and Structure B was
cleaned excessively to remove any machining anomalies, burrs, or errors, in the cavi-
ties.
The S-parameter measurements for Structure A are shown in Figure 4-9 with and
without dielectric loading. With the dielectric loading in place, the last two cavities
of the structure are left empty of dielectric loading and the dielectric is held in place
by the copper inserts. For the measurements without dielectric loading, the slots are
kept empty, simulations have shown that this set-up is nearly equivalent to the results
112
(a)
(b)
(c)
Figure 4-8: (a) The details of the 87-cavity structure with a U.S. dime for comparison.
(b) Assembly of the structure, showing the front of the beam tunnel and alignment
surface. (c) Assembly of the structure with input and output WR-10 Waveguide.
113
0
No AlN
10 With AlN
Noise
20
30
S21 (dB)
40
50
60
70
80
90 95 100 105 110 115
Frequency (GHz)
Figure 4-9: Transmission, S21 , measurements for the first 87-cavity structure with
and without the dielectric loading installed. The transmission in the TM31 and TM41
modes are shown. The noise floor of the measurement is shown for reference.
114
20
Measured A
Measured B
30 Simulation
S21 (dB)
40
50
60
70
92 92.5 93 93.5 94 94.5 95 95.5 96 96.5 97 97.5
Frequency (GHz)
Figure 4-10: Transmission, S21 , measurements for the 87-cavity structures with di-
electric loading as compared to calculated transmission from HFSS simulations for
the TM31 mode.
well with theory. In addition, the cold circuit bandwidth for Structure A is 2 GHz
less than the predicted bandwidth of the coupled-cavity circuit, and it is about 1.5
GHz less for Structure B. Structure A has more loss than predicted due to machining
burrs incurring additional losses and reflections. Though the gain seen from Structure
A will be less than predicted in Chapter 2, the structure has sufficient transmission
to test the overmoded TWT.
Since Structure A had less transmission than anticipated, a second structure (B)
was made to test the robustness of the machining process. The same specifications
were used in machining Structures A and B, but Structure B was extensively cleaned
and removed of machining anomalies (primarily consistent of small burrs on the di-
electric slot edges). A direct comparison between the two structures is shown in Figure
4-11 without the dielectric loading installed. With the extensive cleaning, some im-
provements were noted: a significant increase in transmission (especially with the
dielectric loading in place), more consistent transmission over the bandwidth, and
about 500 MHz of bandwidth was recovered. In hot test, the gain seen in Structure
B may agree better with theory, but it may still have a smaller bandwidth.
Overall, three TWT hot tests were performed. The S21 transmission measurements
115
No dielectric
40
50
60
70
92 92.5 93 93.5 94 94.5 95 95.5 96 96.5 97
Frequency (GHz)
Figure 4-11: Transmission measurements for first and second 87-cavity structures
with no dielectric loading installed. The second structure was extensively cleaned of
machining burrs.
25
1: A, full
30 2: A, half
Transmission, S21 (dB)
35 3: B, full
40
45
50
55
60
65
70
92 92.5 93 93.5 94 94.5 95 95.5 96 96.5 97
Frequency (GHz)
Figure 4-12: Transmission measurements for the assembled structures that underwent
hot test: Structure A (full loading), Structure A (half loading), and Structure B (full
loading).
for the TM31 mode of the three combinations tested is shown in Figure 4-12. First,
Structure A was tested with full dielectric loading installed. Full loading means
that the first 85 cavities of the structure were filled with dielectric, and the last
two cavities were left without loading. This combination allows for slightly higher
116
gain without the possibility of inducing oscillations. Second, Structure A was tested
with half of the dielectric loading installed; the first 40 cavities of the structure had
dielectric loading, and the last 47 cavities were left with a vacuum gap where the
loading would be. Third, Structure B was tested will full dielectric loading. The full
experimental set-up will be described in Chapter 5, and the results from these three
combinations under hot test will be described in Chapter 6.
117
118
Chapter 5
Experiment Design and Set-up
The overmoded TWT experiment was designed and built at MIT. Unless otherwise
noted, all custom components were designed in-house and fabricated at the MIT
Central Machine Shop. The design is made to be modular. It has a wide range of
operation parameters, the test stand can be disassembled, and the TWT structure
and RF components are replaceable. Due to its nature as a modular experimental
design and the need for quick experimental turn-around times, it is not comparable
to a commercial W-band TWT design in size or portability. However, the interaction
structure within the experiment is of the same quality as its commercial counterparts.
An overview of the full laboratory set-up is shown in Figure 5-1. Here, the EIO
driving source is shown; not shown is a low-power solid-state W-band source that will
also be discussed and can easily replace the EIO. The solenoid magnet is displayed,
and the electron gun is highlighted. Also shown is the transformer tank for the pulse
modulator, which provides a high voltage pulse for the electron gun. All of these
components, along with the RF windows and the safety interlock system, will be
described in detail in further sections of this chapter.
The main component of the experiment is the vacuum tube which is placed inside
of the solenoid magnet. Figure 5-2 is a CAD rendering which displays a cross-sectional
119
120
94 GHz EIK 2.5 kG 30 kV 2 microsecond
(driving source) Solenoid Magnet Electron Gun pulse modulator
Figure 5-1: The full experimental set up, highlighting the EIO driving source, solenoid magnet, electron gun, and pulse
modulator.
to pump Solenoid Magnet, 26 cm
RF Windows
Collector
to pump
Electron
Gun
Figure 5-2: A rendering of the TWT tube assembly inside of the solenoid magnet.
The cut-away highlights the TWT structure, where the coupled cavities have been
colored cyan. The simulated electron beam trajectory is pink.
view of the TWT tube inside of the magnet. All assembly drawings and diagrams were
made in Autodesk Inventor. In this figure, one can see how the electron gun, TWT
structure, RF coupling, and collector are arranged inside of the tube. This tube is
designed such that it must be assembled around the solenoid magnet. The simulated
worst-case electron beam is shown in pink in the drawing. There are two 3-way
translation stages on either side of the magnet (not shown) in order to finely adjust
the position of the tube. The x- and y-adjustments on the electron gun (high voltage)
side of the magnet are remotely controlled so that alignment can be performed while
under operation.
Of particular note with the tube design is the TWT structure placement, which
is highlighted in Figure 5-2. Four screws hold the alignment surface of the structure
onto the anode of the electron gun. This alignment ensures a solid mating between
121
Figure 5-3: The 2.5 kG solenoid magnet installed on the experimental optical table for
testing. The collector-side iron pole piece can be seen. The magnet is water-cooled.
the center-lines of the electron gun and structure, which is critical for operation. The
structure was made to be modular. Any other structure with the correct mating
surface can replace the TWT structure that is currently in place. This easy switching
of components allows for a quick turn-around time between experimental tests.
The solenoid electromagnet used in this experiment is a 2.5 kG, four-coil electromag-
net. Shown in Figure 5-3, the magnet is a wire-wound magnet from Arnold Magnet
Technologies (Ogallala Electronic Division) and purchased from Varian Associates; it
was previously used for a Pierce wiggler gun experiment [8]. The magnet is water-
cooled with copper tubing installed around the solenoid windings. The four coils
are each individually controllable, allowing for variable magnetic field shaping. The
solenoid magnet can provide up to 3 kG peak field and is nominally operated at
2.5 kG peak field. With appropriate drive currents on the four coils, the flat-top of
the magnet is 10 cm long. For convenience, the two inner coils of the magnet are
122
connected in series to a power supply, and the two outer coils are also connected in
series to a power supply. TDK Lambda GEN40-85-3P230 3 kW power supplies are
used, offering precision control for the current and voltage applied to the magnet. If
necessary, the four coils can be controlled individually by separate power supplies.
A small solenoid electromagnet is installed on the electron gun side of the magnet,
to be used as a gun coil. The gun coil is able to provide a magnetic field up to 260 G to
help shape the field at the location of the electron gun. The gun coil is controlled with
a TDK Lambda GENH20-38-U 750 W power supply which offers precision control.
An iron pole piece was designed for the magnet in tandem with the electron gun
design. The iron pole piece provides a sharp rise in the magnetic field which works
with the electrostatic focusing of the anode-cathode configuration in order to provide
a high compression electron gun. A back pole piece was also implemented on the
collector side of the magnet in order to control the electron beam expansion as it
exits the magnetic field.
The code Poisson was used for calculations and design of pole pieces to shape the
magnetic field. Poisson/Superfish is a 2-D Poissons equation solver distributed by
the Los Alamos Accelerator Code Group at Los Alamos National Lab. Superfish is
an electromagnetic field solver often used for designing accelerator cavity structures
[4]. The solenoid magnet with the gun coil and iron pole piece was calculated as a
magnetostatic problem with ferrous materials. The pole piece was designed in tandem
with the electron gun in order to provide the appropriate beam shaping. The final
geometry of the pole piece from the Poisson simulation is shown in Figure 5-4, along
with the calculated magnetic field lines. The main coils of the solenoid are outlined
in green, and the gun coil is also shown in the geometry.
In the geometry for the Poisson simulation, the iron pole piece appears to be
complicated. In reality, this complicated design was modified from the original, much
simpler, design. Unfortunately, when manufacturing considerations had to be taken
into account, the simpler design was not physically realizable since it needed to exist
in the same physical location as the anode. That is, the pole piece had to be shaped
around the electron gun geometry, while also allowing room for adjustments and
123
35
30
25
r(cm)
20
Gun Coil
15 Pole Piece 4 coils Pole Piece
10
0
10 15 20 25 30 35 40 45 50 55
z(cm)
Figure 5-4: Poisson calculation of the solenoid electromagnet with the gun coil and
pole pieces. Magnetic field lines are shown in pink and magnetic field vectors are
marked in red.
alignments that would be necessary in the final implementation of the system. These
requirements led to the tapered design of the iron pole piece.
The magnetic field was measured prior to installation of the TWT test stand
with a 1-D Hall probe. Experimental measurements of the magnetic field are shown
in Figure 5-5. The measurements for 1.8 kG operation are shown, and agree very
well with Poisson calculations. The electron gun was designed to have optimized
performance at this magnetic field. This conservative point of ideal operation will
allow for adjustments to be made in the magnetic field. That is, a higher field can
be obtained if it is found necessary under non-ideal experimental conditions. These
adjustments will allow of optimizing the performance of the TWT.
Between two stages of experimentation, the magnetic field was re-measured with
a 3-D Hall probe. Azimuthal measurements around the center of the magnetic bore
indicated a small, 2.5 0.5 mm, misalignment between the magnetic field axis and the
cylindrical bore center at 55 degrees from the top of the magnet. This misalignment
was fixed by re-machining the iron pole piece to have a larger inner hole that could
124
Figure 5-5: Measurement of the z-directed solenoid magnetic field with the iron pole
pieces and gun coils installed and operating at 1.8 kG (30 A driving current on each
coil), with comparison to Poisson simulations.
be centered on the magnetic axis (instead of being centered on the bore). This
adjustment allowed for the tube to be aligned to the magnetic axis, allowing for better
beam formation in the electron gun. The on-axis axial magnetic field changed by less
than 10 G with the new pole piece. The effect of the iron pole piece adjustments and
tube alignment to the magnetic axis will be discussed further in Chapter 6.
The Pierce electron gun for the TWT was designed and built at MIT. The TWT
was designed to operate with a 31 kV, 310 mA electron beam. The electron beam
diameter must be less than the beam tunnel diameter, 0.8 mm, and was designed to
be no more than 0.6 mm in the ideal case. Minimal axial velocity and beam scalloping
was required for ideal operation of the TWT.
125
Figure 5-6: The 2-D cylindrical electron gun geometry (units in mm). The cath-
ode consists of the emitter surface (pink) and the focus electrode (red). The anode
and beam tunnel (green) are kept at ground potential. The black line indicates the
rotational axis.
5.3.1 Design
The design was completed in Leidos (formerly SAIC) Michelle, an electron gun mod-
eling software [54]. The geometry of the focus electrode, emitter cathode, and anode
were modeled in 2-D for simplicity. Though, 3-D simulations were possible to detect
alignment and manufacturing tolerances. Michelle calculates the emitted electron
beam and transportation of the beam through the anode and beam tunnel with an
external magnetic field applied. The axial magnetic field was calculated in Poisson
and was used in the design of the electron gun.
Through an iterative design process, a Pierce electron gun was designed to meet
the parameters of the experiment. The final parameters of the cathode and anode
are shown in Figure 5-6, which displays the design in cylindrical coordinates. The
emitter surface is a 3.2 mm diameter spherical cap, and the focus electrode is 54 from
126
Figure 5-7: The beam envelope calculated via Michelle for 1.8 kG flat-top magnetic
field. The axial position z = 0 is marked as the coordinate axis in Figure 5-6 and
physically corresponds to the front surface of the magnet.
the center-line. The distance between the anode and cathode is nearly 12 cm. This
distance allows for a high compression beam with minimal scalloping to be formed.
Not shown in the picture is the anode extending to a 30 kV ceramic standoff, which
connects to the cathode and provides vacuum.
During space-charge limited operation at 31 kV, 320 mA was emitted from the
cathode. The cathode emitter has less than 5 A/cm2 current density on its surface
during space-charge limited operation. The electron beam profile which results when
a magnetic field of 1.8 kG is applied is shown in Figure 5-7. In the beam tunnel,
the electron beam is less than 0.5 mm in diameter under ideal conditions. When
the magnetic field was adjusted to a slightly higher value for safe operation, minimal
scalloping kept the beam diameter to be less than 0.6 mm in diameter. The beam is
confined for the duration of the flat-top field (from 110 cm). After exiting the beam
tunnel, the beam expands, which allows energy to dissipate when it deposits onto the
127
128
Figure 5-8: The equipotential lines and particle trajectories as calculated by Michelle for the 30 kV electron gun. The cathode
emitter surface is on the left, where the particles originate. The figure is truncated at the right, where the beam enters the
beam tunnel.
129
Figure 5-9: The electric field in the electron gun for operation at 31 kV as calculated in Michelle. The peak electric field occurs
on the anode surface. The beam tunnel is truncated on the right to emphasize the electric fields within the region of the A-K
gap.
collector. Adjusting for experimental operation and machining tolerances, 310 mA of
current was expected from the cathode.
Figure 5-8 shows the equipotential field lines between the anode and cathode for
the given geometry. In addition, the electron particles simulated are shown coming
off of the emitter surface, gaining energy in the A-K gap, and entering the beam
tunnel; minimal crossover is seen between the particles, showing that minimal axial
velocity is present in the simulation. Keep in mind that this simulation is for the ideal
case of alignment between the cathode, anode, beam tunnel, and magnetic axis; any
misalignment in those components will cause an axial velocity to form. For simplicity,
we have estimated there to be 1 % axial velocity spread for the TWT simulations
that were referenced in Chapter 3.
The magnitude of the electric field for the electron gun operating at 31 kV is
shown in Figure 5-9. The peak electric field occurs on the anode surface, 62 kV/cm,
which is low enough for safe operation during vacuum. The design goal was to keep
the electric field below 80 kV/cm. At all other surfaces, the electric field is kept
reasonably low.
5.3.2 Manufacturing
After design, dimensions were scaled to account for thermal expansion during hot
cathode operation. A CAD cross-section of the final design of the electron gun is
shown in Figure 5-10. A 1 kV feed-through leads to the heating element on the
cathode emitter, which heats up to 1200 C. The cold cathode has a resistance of
1 , and the fully heated cathode has a resistance less than 5 . The heater power
should not exceed 8.5 W for safe operation. A standard 30 kV ceramic standoff was
used, and the design is able to withstand up to 35 kV operation safely.
Figure 5-10 highlights the pole piece and anode configuration of the electron gun.
The pole piece is machined in two parts and is installed around the electron gun; the
inner radius of the pole piece limits the size of the tube in that area. For vacuum,
the anode creates a bad pumping conductance between the electron gun and collector
130
~ 15 cm
sides of the tube. For reference, conductance, C, can be calculated through a pipe as
where D is diameter and L is length of the pipe (in inches) [15]. So, the small diameter
of the anode and the beam tunnel leads to a small conductance between the electron
gun and collector sides of the tube. The mechanical design could be modified to
increase this conductance in this region, however both sides of the tube are connected
to 2 L/s ion pumps, which provide adequate pumping speed for the entire device.
The dual pump approach prevents the need for additional conductance to be added
around the iron pole piece.
The electron gun was made with 110 mil tolerances, as necessary, and precise
alignment between the anode and cathode was achieved. The machined cathode and
anode are shown in Figure 5-11. Prior to final assembly, the final welding joint was
machined for accurate alignment and separation between the anode and cathode.
The electron gun components were machined at the MIT Central Machine Shop,
with exception of the the heating element and cathode emitter, which was manufac-
131
Cathode
Anode
Figure 5-11: The cathode and anode prior to final welding of the electron gun. Also
shown is the beam test assembly, where the assembled electron gun is visible on the
left, with an ion pump attached.
132
tured by Heatwave Labs. The cathode used was modified from a standard dispenser
cathode, model STD 134 [29]. A precision spot weld secures the cathode stalk to
the base assembly. The copper anode was brazed onto the stainless steel surrounding
structure, while all other joints were welded. Figure 5-11 also shows the electron gun
after final welding when it was installed onto the entire tube assembly, which shows
the electron gun ceramic stand-off and ion pumps. The tube assembly shown was
baked at 150 C prior to being installed in the solenoid magnet. This assembly was
used to test the quality of the electron beam produced by the gun. Results from the
beam test will be shown in Chapter 6.
Of final note on the machining of the electron gun is cooling considerations. The
electron gun is air-cooled since the cathode is small and the power required to heat it
is, at maximum, 8 W. The cathode stalk was machined for minimal expansion and for
heat dissipation. A moderate air-flow is maintained around the electron gun during
heating and operation.
For the high power source, an EIO manufactured by CPI Canada was used. It is
shown in Figure 5-12. The output of the EIO is in W-band waveguide, WR-10.
In the picture it is shown connected to an isolator, two variable attenuators, and a
detector diode, since power measurements of the EIO were being performed at the
133
Figure 5-12: A picture of the 94 GHz EIO in the laboratory with an isolator and
attenuation on the output.
time. The EIO is capable of producing more than 300 W of power between 93.5
and 95.7 GHz. The frequency is manually tunable via a motor controller. It is able
to provide 15 microsecond pulses at a rate of 1 Hz, which matches well with our
experimental requirements.
The EIO offers the advantage of an extremely high input power. This will allow the
circuit to reach saturation while under testing. It is also useful in finding operation
points when under initial testing of the TWT, since more power can be used in
detection of small amplification parameters. Unfortunately, the EIO has a small
bandwidth, and it not easily or consistently tunable to a certain frequency; leaving it
difficult to measure gain-bandwidth characteristics.
134
20
18
Power Out (dBm) 16
14
12
10
8
6
4
2
0
90 91 92 93 94 95 96 97 98 99 100
Frequency (GHz)
The broad bandwidth, solid-state Active Multiplier Chain (AMC) used for this ex-
periment was purchased from Millitech. The AMC-10-RNHBO is a 6-times multiplier
used in tandem with a 3 dBm, 15-17 GHz CW frequency generator. The AMC can
provide at least 15 dBm of power (32 mW) between 90-100 GHz, as shown in Figure
5-13.
The AMC is useful in providing a source that is highly tunable in frequency over
a broad bandwidth. Unfortunately, it provides a relatively small amount of power to
the amplifier circuit when compared to the EIO. The AMC is not powerful enough
to drive the circuit into saturation under the expected operating conditions.
For these reasons, a combination of the two sources was used in the amplifier
experiment. The low power AMC was used to find low power linear gain and gain-
bandwidth characteristics of the TWT. The high power EIO was used to initially
optimize the circuit (via tuning of positions and operation parameters of the TWT)
and to find high power gain and saturation characteristics of the TWT amplifier
circuit.
135
5.5 RF Vacuum Windows
RF windows are necessary for input and output coupling into the TWT circuit, under
vacuum. There are two possible ways to go about the window design. The simplest,
though most expensive route, would be to use WR-10 waveguide RF windows that
have been specifically designed for 94 GHz. This is the standard approach used in
TWT industry, and improvements on W-band windows are still under development
[32] [72]. Unfortunately, engineering to industry standards while maintaining a TWT
tube that has an interchangeable circuit is prohibitive. In addition, transmission of the
fundamental mode in the circuit at 62 GHz must also be considered; though the mode
should not be present, this TWT is being built to verify that prediction, so a very
narrow-bandwidth window could be prohibitive to our testing. Therefore, windows
fitted in WR-10 waveguide and tuned to 94 GHz were not used. The alternative
option for windows is to use over-sized windows that are fit to a standard conflat
flange; these windows work well with the interchangeable circuit and modular TWT
design and can be manufactured to a specified width such that they will not have high
reflection at the 62 GHz fundamental mode while maintaining good transmission at
94 GHz.
The windows used in this experiment are standard windows from MPF Products,
a company that specializes in ceramic-metal brazing and vacuum components. The
windows are fused silica (SiO2 ) that has been brazed onto a conflat flange. The width
of the silica has been chosen to have a transmission resonance at 94 GHz. This is
calculated via the simple 2-boundary dielectric problem, where a dielectric slab is
considered, the transmission, T can be calculated as
(1 R)2
T = (5.2)
(1 R)2 + 4R sin(/2)2
with
2n
= cos w (5.3)
c
[40] [26] where is the incident angle (in this case, we assume normal incidence,
136
0
Reflection, S 11 (dB)
10
20
30
40
50
Transmission, S 21 (dB)
0.5
1
Window 1
1.5 Window 2
Calculation
2
85 90 95 100 105
Frequency (GHz)
Figure 5-14: Measured and calculated transmission through the fused silica RF win-
dows.
= 0), R is the reflection of the material on a single interface, and the relative index
of refraction for silica is nr = n/n0 = 1.955 in the W-band [13]. The width of the
window, w is calculated such that the T = 1 at 94 GHz ( = 294109 rad/s). There
are multiple solutions to the equation. A window width of w = 3.26 mm (precisely
measured as 0.12840.0005 inches) was chosen because a standard part at that width
was available from MPF Products and fit our purposes. The calculated and measured
S-parameters are shown in Figure 5-14, and show good agreement, assuming R = 0.1.
The windows have a resonance at the design frequency of 94 GHz, and show good
transmission at that point. A reflection is evident in the measurement due to the
resonances in the coupling waveguides. There is more than -0.01 dB transmission
from 9297 GHz.
For the internal waveguide that couples into the TWT circuit, the WR-10 waveg-
uide was tapered to WR-28 waveguide prior to transmission through the window, as
can be seen in Figure 5-15. This allows for cleaner transmission through the fused
137
Figure 5-15: One of the fused silica RF windows in a 2-3/4 con flat flange, installed
on the TWT. The input WR-28 waveguide can be seen
silica window due to easier alignment of the waveguides, and less overall power loss
in the device. Outside of the window, a matched WR-28 to WR-10 waveguide taper
is used to couple in power from a driving source at the input. At the output, the
transition couples power to a diode to measure the output power from the TWT.
These tapers have the potential to set up a resonance in the device, leading to 12 dB
of trapped power losses. However, the transmission improvements seen when using
this tapered section were greater than the detriments of trapped power losses.
138
Gyrotron PFN TWT
Heater Variac Heater Variac
Active
Experiment
Switch
Figure 5-16: A block-diagram of the PFN and high voltage pulse transformer set-up
with dual-experiment access.
that share the same laboratory space: either the TWT discussed in this thesis or a
megawatt-class 110 GHz gyrotron [67]. The active experiment switch, thyratron, and
capacitor bank are located in the control rack (see next section), and are shown in
Figure 5-17. The thyratron triggers a 2.8 microsecond pulse from the capacitor bank
at a rate set by the experiment controls. These components are all shared between
the two experiments, but switch-over of operation between experiments can be done
extremely quickly.
139
PFN active experiment
switch. The inactive
leads are grounded.
capacitor bank
thyratron and
thyratron heater
Figure 5-17: The modulator set-up, showing the high-voltage active experiment
switch between the megawatt-class gyrotron and TWT experiments, the PFN ca-
pacitor bank, and the thyratron.
140
High Voltage Pulse Transformer
230:30
0.94
F
6V
2A
Safety 1:12 Heater
switch Cathode
from RM
PFN 950 Body
80
Collector
0.94 TWT
1:3 F
from Heater
Variac
Figure 5-18: An electronic schematic of the high voltage pulse transformer and con-
nections to the TWT experiment. The safety switch, a high voltage crowbar switch,
is only energized and open when all safety interlocks are satisfied.
The controls for the experiment along with the Safety Interlock System are shown in
Figure 5-19. In the left rack is the cathode heater, along with a calibrated output
of the heater voltage and current. Also included are interlocks to trip the heater in
case of power loss or high vacuum in the tube. Below the cathode heater control are
the trigger control, high voltage supply, vacuum pump control and power supplies
for the solenoid magnet coils and gun coil. The middle rack holds the PFN for
high voltage modulation, highlighted in Figure 5-17. The right rack holds the safety
interlock system and active experiment switch-over control board for either the TWT
experiment or the alternative experiment in the lab which uses the same PFN. The
safety interlock system contains sensors for vacuum pressure, magnetic field, water
flow, and PFN settings. In addition, there is a light curtain to prevent personnel
from accessing the high voltage area while the experiment is operating. If access is
required, the light curtain interlock has a 10 minute override switch.
All interlocks must be satisfied to turn on the high voltage power supply and
operate the experiment. If any interlock is switched off while operating, the high
141
Figure 5-19: The control system for the experiment and safety interlock system.
voltage will automatically turn off and the safety grounding crowbar switch will be
unarmed in the high voltage pulse transformer. Other experiment elements will turn
off, as needed. For example, if the cooling water to the magnet is interrupted, the
high voltage power will turn off along with the magnet power supplies in order to
prevent the magnets from over-heating. The selective experiment control allows for
safe operation of the sensitive components of the system.
142
Figure 5-20: Installation and alignment of the TWT structure onto the anode. The
alignment surface is held in place with 4 screws on the back-side of the copper anode
piece.
5.8 Installation
The installation of the TWT experiment is shown in Figures 5-20 to 5-23. The tube
must be assembled and installed in the solenoid magnet simultaneously. Since the
tube cannot be easily accessed after installation, components underwent bake-out at
150200 C prior to assembly.
To begin installation, the electron gun is aligned to the TWT structure. Figure
5-20 shows the TWT structure being connected to the back of the anode on the
electron gun. Four screws secure the alignment surfaces and ensure proper mating
between the parts.
After the structure and outer vacuum tube are connected, the tube is inserted
into the magnet, and the rest of the components on the collector side of the tube
are put into place. Figure 5-21 shows the assembly of the WR-10 waveguide bends
with the insertion of the support structure which holds the bends in place inside of
the tube. This support aligns the waveguide bends to the straight waveguide sections
from the input and output of the TWT structure. Figure 5-22 shows the completion
of installation with the connection of the RF windows on 2-3/4 inch conflat flanges,
collector, and ion pump for the collector side of the vacuum tube. Also visible in this
143
Figure 5-21: Installation of the WR-10 90 degree waveguide bends and support struc-
tures into the tube.
figure is the gun coil and electron gun on the right side of the magnet.
Figure 5-23 shows the final laboratory set-up, where the EIO is connected to the
input port (right) and an attenuator and detector diode are connected to the output
port (left). An acrylic box isolates the high voltage electron gun for safety.
The ion pump controls and grounding wires are routed through Rogowski coils
which monitor the pulsed currents deposited on the collector and body of the tube
from the electron beam. An oscilloscope near the control panel reads in these two
current signals and the voltage pulse. As needed, the RF output power, input power,
or power reflected at the input of the TWT can also be monitored via WR-10 detector
diodes. Experiment measurements and results will be discussed in the next chapter.
144
Figure 5-22: Completion of installation for the TWT experiment. Standard 1-1/3
inch and 2-3/4 inch conflat flanges were used in design.
145
Figure 5-23: The fully assembled TWT experiment, showing the EIO input and diode
detector at the output of the circuit.
146
Chapter 6
Experiment Results
The TWT experiment underwent four distinct builds in the process of testing. First, a
beam test was done without a TWT structure in place in order to verify the operation
of the test stand and electron gun. Second, the first 87-cavity TWT was installed,
Structure A, with full dielectric loading in place. Third, Structure A was re-installed
with half of the dielectric loading in place. Finally, the second 87-cavity structure,
Structure B, which was extensively cleaned, was installed and tested. In the test of
Structure B, the pole pieces on the magnet were adjusted to compensate for the axis
offset which was found to be present between the magnetic field and bore center.
The results of these experiments will be discussed in this chapter.
A test of the electron gun was performed in order to verify operation of the TWT
experiment test stand and confirm operation of the electron gun to specifications.
The goal of the beam test was to verify the I-V relationship (Child-Langmuir limit)
of the gun and measure the transmission of current to the collector. In addition, this
test verified that all equipment and measurements were functioning properly.
A cross-sectional rendering of the beam test experiment is shown in Figure 6-1.
The electron beam travels from the cathode and through the anode, which has a
0.8 mm diameter (the diameter of the beam tunnel for the TWT design). For the
147
Solenoid Magnet, 26 cm
Collector
to pump
Electron
Gun
1.1 cm
Electron beam 1.4 cm 0.8 mm macor dummy beam
beam tunnel tunnel
Figure 6-1: A cross-sectional rendering of the beam test experiment. This experiment
uses the same components as the TWT experiment, except the TWT structure has
been replaced by a dummy beam tunnel. The cut-away shows the beam tunnel which
is electrically isolated from the anode with macor.
beam test, a copper pipe with a diameter of 2.5 cm was used as the dummy beam
tunnel between the anode and the collector. Instead of using the alignment surface
on the anode (which the TWT structure will use), the beam tester was chosen with a
large diameter that did not require precise alignment. Due to this design, the dummy
beam tunnel could be electrically isolated from the body of the tube. Therefore, the
current emitted from the cathode was separated into three measurable components:
body current, beam tunnel current and collector current. Due to the nature of the
beam test, the body current primarily consists of current that is deposited onto the
anode.
The current output of the electron gun was tested over a range of voltages. Align-
148
(a)
35
30
25
Voltage (kV)
20
15
10
5
0 1 2 3 4 5 6 7
Time (microseconds)
(b)
350
300
250
Current (mA)
Total
200 Collector
Beam Tunnel
150 Anode
100
50
0
0 1 2 3 4 5 6 7
Time (microseconds)
Figure 6-2: Sample traces from the beam test experiment, showing measurements for
(a) (negative) Voltage pulse applied to the cathode and (b) the currents measured on
the collector, beam tunnel, and anode (body) during the 3 s pulse. These data were
taken with a voltage pulse of 31 kV, and the collector current was measured as 301
mA, with cumulatively 358 mA total current measured.
149
350
Expected
300 Current at Collector
250
Current (mA)
200
150
100
50
0
0 5 10 15 20 25 30
Voltage (kV)
Figure 6-3: The Child-Langmuir curve for the electron gun, comparing the theoretical
I-V relationship of the designed electron gun to the measured current at the collector.
The gun ideally operates at 31 kV with 310 mA of current.
ment between the tube and magnetic field was performed at low voltages and minimal
heater settings (about 1 kV with 10 mA) in order to prevent damage to the anode
surface. Operation was found to be best in a wide range of magnetic coil current
settings which resulted in a peak on-axis field of about 1.52.5 kG (as expected when
compared to theory). A sample trace showing the voltage operation at 31 kV, and
collector current measurements is shown in Figure 6-2. These curves indicate that
the flat-top of the pulse is 2.8 s, with 2 % variation in amplitude. In addition, the
current emission directly follows the rise and fall of the voltage pulse. The currents on
the solenoid magnets were finely tuned to achieve maximum transmission of current
to the collector. Several optimum operation currents which corresponded to peak
magnetic field values ranging between 1.52.5 kG were observed.
The electron gun performed well and meets specifications. Measured data is shown
in Figure 6-3, which shows the calculated Child-Langmuir I-V relationship for the
electron gun as compared to the measured current at the collector of the gun. The gun
was designed to operate at 31 kV, 310 mA. It should be noted that at this operation
voltage, the collector current was 3066 mA, the body current was approximately
150
20 mA, and the beam tunnel current was 23 mA. Therefore, there is nearly 88 %
transmission of the beam to the collector at the operation point.
It is of note that the electron gun emits more current than anticipated in de-
sign. The current deposited on the collector is within range of what was expected
for the total current of the design. Therefore, the electron gun is operating within
specifications for our experiment. Since the experiment is pulsed, the expected beam
interception will not be a problem for TWT operation.
151
(a) 8
6
Device Gain (dB)
0
AMC
EIO
2
10 5 0 5 10 15 20 25 30 35 40
Input Power (dBm)
(b)
40
AMC
EIO
30 Linear Fit
Output Power (dBm)
20
10
10
20 10 0 10 20 30 40 50 60
Input Power (dBm)
Figure 6-4: (a) Device gain and (b) output power vs. input power observed at 95.5
GHz with V0 =27 kV. A linear gain of 6 dB (device gain) is shown with 7.5 W
saturated output power measured.
152
the linear gain regime which was measured with the solid state AMC input and
the saturation of the circuit which was measured with the EIO input. The highest
saturated power observed was 8 W at 95.5 GHz.
It is pertinent to note the difference between device gain and circuit gain. All
parameters discussed have been for measurements of device gain. That is, device
gain, Gd , is equal to the power measured at output, Pout , divided by power at the
input window, Pin . On the other hand, circuit gain, Gc is the gain which occurs only
in the TWT circuit. Circuit gain takes into account the losses due to the rest of the
TWT device, Ld , primarily losses at the windows and waveguide outside of the inner
TWT circuit. In other words,
Gc = Gd Ld (6.1)
In this set up of the 94 GHz TWT, there are significant losses due to the coupling
through the windows. Misalignment between the WR-28 waveguides on either side of
the windows contribute a large amount of losses, and are difficult to align correctly.
This is a fault in the design. It is estimated that at least a total of 36 dB of loss
occurred in this experiment due to the window set-up. Therefore, the 8 dB measured
device gain can be estimated as 1114 dB circuit gain observed for the TWT at 95.5
GHz.
It should be noted that gain was observed at other frequencies with higher opera-
tion voltages, but not nearly as strong as the 95.5 GHz point. Since the transmission
through Structure A has resonances, the frequency is not consistently tunable with
voltage adjustments. Therefore, the secondary point of operation was 94.55 GHz with
32 kV operation, where up to 4 dB of device gain was observed.
The magnetic field could be adjusted and fine tuned in the 4-coil solenoid. These
adjustments led to different operation regimes in the circuit. Figure 6-5 shows the
output power detector diode signal and collector current traces that were observed
in the system at the 94.5 GHz operation point with V0 =32 kV. Similar conditions
were observed for the 95.5 GHz operation point. Figure 6-5(a) shows an operation
point that was taken with an inconsistent magnetic field. The field was adjusted
153
10 300
(a)
250
8
150
4
100
2
50
0 0
0 1 2 3 4 5 6 7 8
time (microsec)
50
(b) 200
40
150
30
100
20
50
10
0 0
0 1 2 3 4 5 6 7 8
time (microsec)
Figure 6-5: Sample traces of the collector current and output power measurements
for Structure A under test in the TWT experiment. (a) was observed during low gain
operation, where the system was aligned for high current transmission and consistent
signal, while (b) was observed during high gain operation.
154
8
6
Device Gain (dB)
4
4
95.5 95.52 95.54 95.56 95.58 95.6
Frequency (GHz)
Figure 6-6: Device gain vs. bandwidth for operation at V0 =27 kV (340 mA total
measured current).
to allow for high current transmission, 280 mA, to the collector. This resulted in
a field that was about 2.5 kG at the outer edges of the magnet and less than the
Brillouin field in the center of the magnet. A consistent RF input was used, and the
output power detector diode signal is flat and appears smooth in the middle of the
pulse (ignoring signal noise). However, only 1 dB of gain was observed under this
operation condition. Conversely, Figure 6-5(b) shows the high gain operation point.
This point was taken with a very low, but consistently flat, magnetic field, of about
0.9 kG, which is at the edge of acceptable operation for the magnetic field. This
operation condition for the magnetic fields led to less than 200 mA of current at the
collector and a very inconsistent current trace throughout the time of the pulse. In
addition, the RF pulse, which showed 8 dB of gain, was only amplified during the
peak in the collector current. Though less current was transmitted to the collector at
this operation condition, more interaction occurred between the electron beam and
RF signal.
Many factors could have led to the odd operation conditions needed for high gain
in the circuit. One possibility is that these operation conditions were due to the
magnetic field axis being shifted from the center of the magnetic bore and pole piece
center. By operating at a low current, a consistent interaction occurred between the
155
electron beam and RF wave, but the current was unable to be transmitted to the
collector. The shift of the magnetic field axis will be adjusted in Structure B testing.
For the 95.5 GHz high gain operation point, the frequency dependence on gain is
shown in Figure 6-6. These data points were taken by measuring the linear gain due to
the low power AMC input at various frequency inputs with V0 =27 kV. This voltage
optimized the output for 95.56 GHz operation (the highest gain point observed). A
3-dB bandwidth of 60 MHz was observed for these operation points.
The limited bandwidth and small gain can be traced back to the cold test trans-
mission measurements, where a significant frequency dependence was observed. The
best transmission point occurs around 95 GHz (and it follows logically that this fre-
quency may have shifted to a slightly higher frequency under the final installation
and vacuum operation). Therefore, the discrepancies measured between the cold test
circuit performance and simulation accurately reflect the frequency characteristics
observed in experiment.
Due to the high loss in the structure when full dielectric was in place, which was 20 dB
greater than the anticipated loss (as shown in Chapter 5), it was determined that less
dielectric could enhance the gain results seen in the structure. CST simulations
indicate that less dielectric would result in TM31 round-trip oscillations. That is, the
loss associated with the TM31 mode would be less than the gain in the circuit and
would create an unstable condition in the TWT. However, these simulations predicted
much less loss in the circuit than was measured and may not be valid for Structure
A with more loss present.
Structure A was installed with dielectric in place for the first 42 cavities. For the
remaining cavities, a void was left where the dielectric slots were machined. During
operation, consistent TM31 mode oscillations were present in the device at 95.75 GHz
for a 27 kV voltage pulse. No fundamental mode, TM11 , or backward wave oscillations
were seen in the device. At 94.5 GHz operation with V0 =30.6 kV, less than 1 dB of
gain was observed during zero-drive stable operation. The electron beam was likely
156
modulated by oscillations during the voltage rise, preventing the previously observed
operation. The oscillations in the circuit prevented the high gain point in the previous
test from being reached.
Structure A showed substantial results, but it was possible to improve the gain of the
TWT. Two sources of non-ideal behavior were discovered in the system: misalignment
of the magnetic field and burrs in the structure. These errors could be adjusted with
a second build of the TWT.
The misalignment of the magnetic field was already discussed in Chapter 5. To
fix this in experiment, the front iron pole piece was modified to have a center hole
that was positioned off of the magnetic bore axis by 2.5 mm, and positioned in the
correct orientation during installation of the tube. The back iron pole piece was also
adjusted to be aligned with the magnetic center.
Upon inspection of both Structures A and B, it was evident that many burrs were
present in the structure after machining and standard cleaning for vacuum. Structure
A was installed without further cleaning, but Structure B underwent a much more
extensive cleaning with 0.2 mm acupuncture needles, and the machining burrs were
successfully removed from the structure. Upon cold test measurements, shown in
Chapter 4, it is evident that the extensive cleaning enhanced the transmission in
Stricture B, decreasing ohmic losses and reflections.
6.3.1 Results
After cleaning, Structure B was installed in the TWT with full AlN composite dielec-
tric loading. As was the case with the full dielectric test for Structure A, the area
on top of the last two cavities was not filled with dielectric and was left as a void.
Since the tube could not be aligned to the center of the magnet bore, alignment of
the device was more complicated, but successfully achieved at low voltages.
157
(a) 30
25
Device Gain (dB)
20
15
10
5 AMC
EIO
0
30 20 10 0 10 20 30 40 50
Input Power (dBm)
(b) 50
AMC
EIO
40 Linear Fit
Output Power (dBm)
30
20
10
10
30 20 10 0 10 20 30 40 50
Input Power (dBm)
Figure 6-7: (a) Device gain vs. input power and (b) output power vs. input power
measured at 94.27 GHz. The input power was provided by a low power solid-state
AMC or high power EIO, as indicated. The TWT has 20.9 dB of linear gain and 27
W peak power.
158
Zero-drive stable operation was achieved, and gain was observed at several frequen-
cies. The largest low power linear gain measured was 20.91.7 dB, which occurred
at 94.27 GHz with a voltage of 30.6 kV. The results for the gain measured in these
conditions is seen in Figure 6-7. Two data sets are shown; the low power set used the
solid-state amplifier multiplier chain (AMC) for the input signal, and the high power
set used the W-band EIO. The magnetic fields were kept with a consistent flat-top
at about 1.6 kG, which allowed 250 mA of current on the collector. The linear fit in
Figure 6-7(b) for the AMC data corresponds to 20.9 dB of device gain.
As was the case with Structure A, the circuit gain can be calculated by eliminating
other system losses from the device gain measurement. Primarily, these losses are
due to the input and output coupling through WR-28 waveguide and the fused silica
windows. This gain is split between ohmic and reflective losses that occur in the
WR-28 waveguide and tapers.
The coupling set-up was tested with a PNA. The transmitting millimeter wave
head had WR-10 waveguide which was connected to a WR-10 to -28 taper which was
placed against a fused silica window, and the receiving head had the same set up on
the other side of the fused silica window. These cold tests showed that about 12 dB
of power would be lost due to trapped power in the WR-28 waveguide (with a large
frequency dependence), and about 14 dB of power would be lost due to reflections
between the WR-10 to -28 waveguide tapers and misalignment between the two WR-
28 cut-off waveguides on either end of the window. Therefore, a total of 26 dB is
expected as the coupling loss at the input or output of the TWT circuit.
In the TWT experiment, the EIK was used to measure the losses through the
entire TWT device. The losses in the TWT device, including the input and output
coupling through two fused silica windows and the TWT circuit, were measured to
be 36.30.6 dB at 94.3 GHz. At this frequency, the measured transmission through
the TWT circuit, Structure B, in cold test (as shown in Figure 4-12) was 30.10.5
dB. Therefore, the losses due to coupling at the input and output of the circuit can be
extracted from these measurements. The input and output coupling losses, combined,
were measured in the device to be 6.20.6 dB at 94.26 GHz. For simplicity in analysis,
159
35
30
Circuit Gain (dB)
25
20
15
10 AMC
EIO
CST simulation
5
30 20 10 0 10 20 30 40
Circuit Input Power (dBm)
Figure 6-8: Circuit gain in the TWT as compared to simulated gain with CST for
250 mA beam current. The input power has been adjusted to indicate the power
input into the TWT circuit.
its assumed that half of these losses, 3.1 dB, occur at the input and half occur at the
output. This measurement is in good agreement with the independently measured
coupling losses expected in cold test, 26 dB. With this measured coupling loss taken
into account, the circuit gain of the TWT is 27.11.8 dB. This is rounded to be a
reported circuit gain of 272 dB.
The solid-state W-band AMC was used for low power operation, but in order to
achieve saturation, the EIO was used. The input device for each measured point is
indicated in Figure 6-7. The input power was regulated with a variable attenuator at
the input. The frequency of the EIO is tuned manually in the device, and, therefore,
cannot be set as accurately as the AMC. The maximum output power observed was
275 W at 94.26 GHz. This point can be seen in Figure 6-7(b) where the measured
output power levels off at 27 W. Assuming that the measured coupling loss is split
evenly between the input and output circuits, the saturated output power at the end
of the TWT circuit would be 55 W.
A comparison of the measured circuit gain and the theoretical calculation of the
gain in the TWT is shown in Figure 6-8. In the theory, the current through the TWT
160
25
30.6 kV
28.7 kV
20
Device Gain (dB)
15
10
5
94.24 94.26 94.28 94.3 94.32 94.34 94.36 94.38 94.4 94.42 94.44
Frequency (GHz)
Figure 6-9: Device gain vs. frequency in the TWT for two different operation points;
30.6 kV operation shows the high gain point, and 28.7 kV operation shows a wider
bandwidth operation point.
has been adjusted to be 250 mA to match the experimental current and all other
parameters are kept as stated in Chapter 3. With this adjustment, the theoretical
circuit gain for the TWT is 28.3 dB and the measured circuit gain is in good agreement
with theory. In addition, the theoretical saturated output power is 113 W. The
measured saturated power is likely lower due to increased beam interception during
high power operation.
The bandwidth of the TWT for two different operation points is shown in Figure
6-9. The high gain operation point, as discussed above, is shown as 30.6 kV operation.
The bandwidth of this operation is 30 MHz, centered at 94.26 GHz. Another operation
point with a larger bandwidth is shown for 28.7 kV operation. The 3-dB bandwidth
for this point is 40 MHz, centered at 94.45 GHz, but the 6-dB bandwidth is 120 MHz,
quite larger. Theoretically, there should be 200 MHz of bandwidth at the high gain
30.6 kV operation point. The limited measured bandwidth could be due to several
factors that will be discussed in detail in the next section: limited current, transverse
magnetic field, and inconsistent flat-top voltage.
161
6.3.2 Analysis
Samples of the diagnostic measurements for a typical high gain pulse are shown in
Figure 6-10. In this pulse, a low power signal at 94.26 GHz is provided by the AMC at
the input. A 30.6 kV input pulse of 2.8 s is shown in Figure 6-10(a) along with the
diode trace. The detector diode is placed at the output of the circuit, after a variable
attenuator, and has been calibrated so that it directly measures the output power
of the circuit. It should be noted that the measured output power presented in this
chapter was the average power detected during the last microsecond of flat-top of the
pulse, from 3.64.6 s. Figure 6-10(b) shows the current measured at the collector
and body of the TWT. For comparison, the current traces where no RF power is
present in the TWT for the same operating conditions are also shown. The presence
of low power linear amplification has little effect on the current in the system.
From these diagnostics, there are two curious measurements that must be ad-
dressed: there is a very high percentage of body current and the diode measurement
is not consistent for the duration of the pulse.
In a standard TWT, the body current should be less than 5 % of the total current,
even less for a commercial, CW TWT. Since this is an experimental pulsed device, the
TWT can operate with high body current, though it can introduce some problems and
should be addressed. The large quantity of body current in the system is likely due to
the magnetic fields in the TWT. Any inconsistencies in the magnetic field will result
in beam interception throughout the TWT. The magnetic field has been changed from
Structure A to fix some misalignment, but it is still not ideal. Comparison between
the body current measurements for the two experiments during high gain operation
shows better current transmission in the testing for Structure B. However, there is
still between 4050 % beam interception through the duration of the voltage pulse.
In addition, the current transmission shows a dependence on time during the flat-top
of the voltage pulse. During ideal operation, this time-dependent behavior would
not be present. Several factors could contribute to time-dependency, like a rising
pressure during the pulse due to beam interception or beam rotation due to transverse
162
(a) 94.26 GHz, 20.9 dB gain
35 35
30 30
25 25
Voltage (kV)
Diode (mV)
20 20
15 15
10 10
5 5
0 0
0 1 2 3 4 5 6 7
time (microsec)
(b)
300
250
200
Current (mA)
150
100
Body
50 Collector
Body, no RF
Collector, no RF
0
0 1 2 3 4 5 6 7
time(microsec)
Figure 6-10: The diagnostic measurements for a high-gain operation point at 94.27
GHz with a low power AMC input, showing (a) voltage and output detector diode
signals vs. time and (b) current signals vs. time for the duration of the 3 s pulse.
163
magnetic fields. It should be noted that up to 280 mA of collector current was possible
(approximately 70 % current transmission), but these operation conditions did not
correspond to a high gain point.
There are several reasons as to why the diode measurement is not consistent across
the pulse. The most notable reason is that the output power is related to the beam
current in the TWT. Since the current on the collector rises during the duration of
the pulse, it follows that the interaction current in the TWT also rises during the
pulse, leading to more gain in the TWT at the end of the pulse than at the beginning.
However, the effect in the output pulse is too large to be accounted for by this change
1/3
in interaction current, since the gain is only proportional to I0 . Another reason for
this behavior could be the beam shape and quality due to the magnetic field. If the
beam is rotating early in the pulse, it would not have a strong interaction with the
RF field.
Another factor to consider for both the inconsistent collector current and amplifi-
cation is the flatness of the voltage pulse. Typical operation of the TWT would allow
for small variations in the voltage pulse, less than 1 %. The voltage pulse for the
TWT, as measured, has 2 % variation, with a stronger variation evident during the
first microsecond of the flat-top. Gain at a certain frequency is dependent strongly
on voltage, and large voltage variations will result in atypical bunching of the elec-
tron beam. Since the bandwidth is small in the device, the TWT is less tolerable
to voltage variation, intensifying the effect of the flat-top variation on the perfor-
mance of the TWT. The variation is larger at the beginning of the pulse, leading to
a time-dependent variation in the field.
During saturation, the diagnostics look very similar to high gain operation, with
some other distinct characteristics. A sample pulse for operation during the 27 W
saturation point is shown in Figure 6-11. The EIO provided a 10 W input signal
from 2-6 s to overlap with the flat-top of the voltage pulse. The same conditions of
operation for the linear gain points were kept. The current traces are shown alongside
current traces where no RF signal was present in the TWT.
As was seen in the high gain case, amplification is only present during the later
164
(a) 94.27 GHz, 27 W Saturation
35 35
30 30
25 25
Voltage (kV)
Diode (mV)
20 20
15 15
10 10
5 5
0 0
0 1 2 3 4 5 6 7
time (microsec)
(b)
300
250
200
Current (mA)
150
100
Body
50 Collector
Body, no RF
Collector, no RF
0
0 1 2 3 4 5 6 7
time(microsec)
Figure 6-11: The diagnostic measurements for the saturated power operation point
at 94.26 GHz with a high power EIO input, showing (a) voltage and output detector
diode signals vs. time and (b) current signals vs. time for the duration of the 3 s
pulse.
165
part of the pulse, seen in the output diode detector output from 35 s. The diode
pulse appears wider since the high gain points later in the pulse (45 s) are severely
saturated, while the lower gain points (from 34 s) have only just reached saturation;
this effect causes the pulse to appear more level. Of note is that the rf pulse is not
amplified at the start of the flat-top (from 23 s). This behavior in the saturated
power indicates that poor interaction is happening at the start of the pulse, possibly
due to the voltage variation during this time frame.
Also seen during saturation is an increase in the beam interception. This effect
can be seen by a comparison between the collector current with and without RF
in the TWT. During saturated output power, the collector current dips (conversely,
the body current rises) by over 50 mA. This is due to the fact that the beam is de-
bunching at the end of the TWT, after energy has been transferred from the beam
into the wave. At that point, the electron beam parameters are no longer matched
to the magnetic field of the system; the beam spreads radially, leading to current
interception in the beam tunnel. In order to increase gain in the device, the magnetic
field was adjusted to 1.6 kG, instead of the 2.5 kG simulated. This magnetic field
adjustment could prohibit the de-bunching electrons from reaching the collector.
Due to beam interception during saturation, there is less current at the end of the
TWT at high powers than at low powers. Even though there is a small amount of
power output compared to the beam power, the collector current is seen to decrease
during saturation. By losing current in the TWT, the effects of saturation are seen
at lower input power levels than observed in simulation, leading to the difference seen
between simulation and experiment in the curve of the data at high power levels.
In addition, the maximum output power will be less than simulated due to there
being less overall current in the TWT at the saturation point, leading to only 55 W
maximum output power instead of the predicted 114 W. Despite these circumstances,
theory and experiment match well.
The data discussed so far was all collected during zero-drive stable, pulsed oper-
ation. However, the TWT did see some oscillations evident in the device under very
specific and controllable circumstances. A sample oscillation along with the diagnos-
166
(a) Oscillation
35 25
30
20
25
Voltage (kV)
Diode (mV)
15
20
15
10
10
5
5
0 0
0 1 2 3 4 5 6 7
time (microsec)
(b)
300
250
200
Current (mA)
150
100
Body
50 Collector
Body, no RF
Collector, no RF
0
0 1 2 3 4 5 6 7
time (microsec)
Figure 6-12: The diagnostic measurements for an oscillation observed during the start
of operation, showing (a) voltage and output detector diode signals vs. time and (b)
current signals vs. time for the duration of the 3 s pulse.
167
tics for the pulse is shown in Figure 6-12. This oscillation occurred during the start
of operation, when the pressure in the TWT was quite low. After initial pulsing,
the oscillation was not present in the device while operating at 1 Hz. Figure 6-12(a)
shows the 30.6 kV voltage input pulse and the output diode detector signal, which
was calculated to be 2030 W in power. The current traces along with the standard
operation current traces are shown in Figure 6-12(b)
The frequency of the oscillation could not be measured due to the limited time
of operation in conditions that could induce the oscillation. However, the signal
was detected through WR-08 waveguide, indicating that it is has a frequency above
74 GHz (the cut-off frequency of WR-08 waveguide) and likely oscillates in the TM31
mode of the TWT cavity.
During this initial operation condition with very low pressure, the collector current
was significantly higher than during normal 1 Hz pulsed operation. Assuming that
the collector current functions the same way during oscillation as during saturation,
we can speculate that the current in the circuit without an induced oscillation would
be nearly 300 mA. This estimation is judged due to the 290 mA peak of collector
current at the end of the pulse. The oscillation causes the TWT to reach a saturation
condition and for the de-bunching electrons to intercept the beam tunnel, in the same
way that the saturated RF pulse, shown in Figure 6-11, caused the collector current
to be decreased. Since there is nearly 400 mA of total current in the system but
the TWT was only designed for 310 mA of current, it is likely that there is enough
current at the start of the TWT to induce an oscillation. However, during typical
operation, the current in the TWT is reduced and oscillations do not occur.
Overall, the current in the system is higher than anticipated, as was shown during
the beam test prior to TWT installation. The electron gun must be run at a high
heater value in order to operate in the space-charge limited regime of the electron
gun. The electron gun heater power can be reduced slightly (from 8.2 W to 8.0 W
in experiment), which still allows operation in the space-charge limited regime while
emitting 20 mA less total current. In this condition, the start-up oscillations are
avoided when pulsed operation is begun. However, this reduces the current in the
168
system during stable operation, too, and the overall gain seen is lower.
If the operation conditions of the TWT were to be adjusted in some way to allow
for less beam interception, the oscillations could be avoided in the device all together
by allowing the electron gun to operate at a lower temperature while maintaining an
ideal beam current for amplification. Regardless, these adjustments would only be
to fine-tune the operation of the TWT and possibly match conditions closer to ideal
with a higher interaction current in the TWT.
169
170
Chapter 7
Conclusions
This thesis has covered the design and testing of an overmoded coupled-cavity TWT
which has been implemented at MIT. The TWT involves an overmoded, or oversized,
coupled-cavity structure which was designed to operate at 94 GHz. The experiment
has shown successful operation of the TWT.
The overmoded TWT operates in the TM31 rectangular cavity mode. In doing so,
the size of the cavity is larger than a comparable fundamental-mode cavity, allowing
for many size related benefits. Most notably, the cavity size allows for a large beam
tunnel; the 0.8 mm diameter beam tunnel is the largest of any W-band TWT. The
large beam tunnel allows for a large current in the system with a small magnetic field,
and, subsequently, more gain in the device. In addition, the oversized cavity allows
for easier manufacturing of the structure and the ability for the TWT to be scaled to
even higher frequencies.
The lower order modes in the TWT are suppressed using dielectric loading. The
dielectric is selectively placed along the edges of the cavities to interact with the
TM11 and TM21 modes, causing high losses; at the same location, the dielectric does
not effect the TM31 mode. A lossy aluminum nitride composite dielectric was used
in implementation which offered the desired high loss tangent characteristics while
being compatible with vacuum conditions. The effect of the dielectric loading was
171
tested extensively in cold tests of the cavity structures.
The overmoded cavities were built and tested in several different cold tests in
order to verify manufacturing methods and HFSS simulation results. The cavities
were directly machined with a standard CNC mill, and the designs were built in 9-
and 19-cavity cold tests out of both OFHC copper and glidcop. The tests showed
good agreement with theory and frequency-matching to simulation. In addition, the
aluminum nitride composite dielectric loading worked as simulated, demonstrating
the suppression of unwanted modes and minimal additional losses to the TM31 mode.
The glidcop structures performed with significantly less loses than the copper struc-
tures, agreeing much better to theoretical results and having less machining errors and
anomalies. Due to these tests, it was determined to machine the final 87-cavity struc-
tures out of glidcop and with the same machining method and tolerances. Though
the ohmic losses measured in the 87-cavity structures were large, they agreed with
theory and were large enough to reduce round-trip oscillations in the TWT device.
Therefore, no sever was needed in the TWT.
The entire TWT experiment was built and tested at MIT. The electron gun was
designed in Michelle for the parameters necessary. A beam test prior to implemen-
tation showed 85 % beam transmission to the collector through an oversized beam
tunnel, with 3066 mA of current on the collector. Though the electron gun provided
more total current than anticipated, the Child-Langmuir limit of the in-house built
electron gun agreed well with theory.
172
The experiment was designed to have an interchangeable TWT structure and the
TWT was tested in three stages: Structure A with full dielectric, Structure A with
half dielectric, and Structure B with full dielectric. The experiment was operated
with 2.8 microsecond high voltage pulses.
Initial testing of Structure A with the full dielectric loading in place showed 8 dB
of gain with 10 W peak power at 95.5 GHz. Structure A was also tested with half
dielectric loading, where oscillations were observed.
In the second stage of testing the TWT, the magnetic field was adjusted and
Structure B was precision cleaned during finishing. The precision cleaning led to a
higher transmission through the 87-cavity TWT. The TWT with Structure B achieved
212 dB device gain and 27 W output power. Adjusting for device losses, the TWT
achieved 272 dB circuit gain with 55 W peak circuit output power. In testing, only
250 mA beam current was achieved, limiting the theoretical gain to 28.3 dB and peak
power to 114 W, in good agreement with measurements.
173
operation of an overmoded TWT.
In building this overmoded design, as with any first-time device, the success of the
experiment was not guaranteed. The experiment demonstrated an overmoded TWT
of comparable gain and power to present-day W-band TWTs. In addition, the oper-
ation of this design has shown that an overmoded design is capable of handling high
gain and nominal output powers at high frequency. The selective dielectric loading
successfully suppressed oscillations from unwanted modes, and the lossy design pre-
vented round-trip oscillations from the operating mode. The overmoded TWT does
not have the same frequency limitations as the traditional fundamental mode TWT,
and the design has the ability to be extended to higher frequencies. The fabrication of
the overmoded TWT in the W-band was performed using direct machining, another
advantage of this design over fundamental-mode TWTs. The implementation and
research for this overmoded TWT can be used to design and build other overmoded
TWTs at higher frequencies.
174
Bibliography
[1] U. S. Army. Army Science and Technology Master Plan. 21 March 1997.
[3] Anisullah Baig, Diana Gamzina, Robert Barchfeld, Calvin Domier, Larry R Bar-
nett, and Neville C Luhmann Jr. 0.22 THz wideband sheet electron beam trav-
eling wave tube amplifier: cold test measurements and beam wave interaction
analysis. Physics of Plasmas, 19(9):93110, 2012.
[6] Bruce Carlsten and Steve Russell. Microwave sources course notes. United States
Particle Accelerator School (USPAS), June 2012.
175
[10] Edward N. Comfoltey. Design of an overmoded W-Band coupled cavity TWT.
Masters thesis, Massachusetts Institute of Technology, Department of Electrical
Engineering and Computer Science, Cambridge, MA, 2009.
[11] MIT Lincoln Laboratory Communications and Community Outreach Office. An-
nual Report. MIT Lincoln Laboratory, 2011.
[12] David H. Staelin and Ann W. Morgenthaler and Jin Au Kong. Electromagnetic
Waves. Prentice-Hall, Inc., 1994.
[14] David E. Dean, Thomas D. Schaefer, Gregory A. Steinlage, and Liqin Wang.
Bearing temperature and focal spot position controlled anode for a CT system,
March 13 2007. US Patent 7,190,765.
[19] D. Gamzina, Robert Barchfeld, L.R. Barnett, N.C. Luhmann, and Young-Min
Shin. Nano CNC milling technology for terahertz vacuum electronic devices. In
IEEE Intl. Vacuum Electronics Conf. (IVEC), pages 345346, Feb 2011.
[22] Yubin Gong, Hairong Yin, Lingna Yue, Zhigang Lu, Yanyu Wei, Jinjun Feng,
Zhaoyun Duan, and Xiong Xu. A 140-GHz Two-Beam Overmoded Folded-
Waveguide Traveling-Wave Tube. IEEE Trans. on Plasma Science, 39(3):847
851, Mar 2011.
176
[23] V. L. Granatstein, R. K. Parker, and C. M. Armstrong. Vacuum electronics at
the dawn of the twenty-first century. Proceedings of the IEEE, 87(5):7027161,
May 1999.
[27] J. He, Y. Wei, Y. Gong, and W. Wang. Linear analysis of a W band groove-loaded
folded waveguide traveling wave tube. Physics of Plasmas, 17(11), November
2010.
[28] Jun He, Yanyu Wei, Yubin Gong, Wenxiang Wang, and Gun-Sik Park. Investi-
gation on a W band ridge-loaded folded waveguide TWT. IEEE Transactions
on Plasma Science, 39(8):16601664, Aug 2011.
[31] Peter Horoyski, Dave Berry, and Brian Steer. A 2 GHz bandwidth, high power
W-band extended interaction klystron. In IEEE Intl. Vacuum Electronics Conf.
(IVEC). IEEE, 2007.
[32] Yinfu Hu, Jinjun Feng, Jun Cai, Xianping Wu, Shaoyun Ma, Bo Qu, Juxian
Zhang, and Tongjiang Chen. A broadband microwave window for W-band TWT.
In IEEE Intl. Vacuum Electronics Conf. (IVEC), pages 376377, April 2008.
[34] B. G. James and P. Kolda. A ladder circuit coupled-cavity TWT at 80-100 GHz.
In 1986 Intl. Electron Devices Meeting, volume 32, pages 494497, 1986.
177
[36] Colin D. Joye, Alan M. Cook, Jeffrey P. Calame, David K. Abe, Alexander N.
Vlasov, Igor A. Chernyavskiy, Khanh T. Nguyen, and Edward L. Wright. Micro-
fabrication and cold testing of copper circuits for a 50-watt 220-GHz traveling
wave tube. In Proc. SPIE, volume 8624, 2013.
[37] A. S. Gilmour Jr. Principles of Traveling Wave Tubes. Artech House, 1994.
[39] Rudolf Kompfner. The Invention of the Traveling-Wave Tube. San Francisco
Press, 1964.
[41] C. Kory, L. Ives, M. Read, J. Booske, H. Jiang, D. van der Weide, and P. Phillips.
Microfabricated W-band traveling wave tubes. In 13th Intl. Conf. on IR mm
Waves and THz Tech., volume 1, pages 8586, Sept 2005.
[48] Millitech, Smiths Microwave. Active Multiplier Chain Data Sheet, 2014.
http://www.millitech.com/.
178
[49] Matthew Morgan, Sander Weinreb, Niklas Wadefalk, and Lorene Samoska. A
MMIC-based 75-110 GHz signal source. In IEEE MTT-S Intl. Microwave Sym-
posium Digest, volume 3, pages 18591862. IEEE, 2002.
[50] Emilio A. Nanni. A 250 GHz photonic band gap gyrotron amplifier . PhD thesis,
Massachusetts Institute of Technology, Department of Electrical Engineering and
Computer Science, Cambridge, MA, June 2013.
[56] David M. Pozar. Microwave Engineering. John Wiley, second edition, 1998.
[57] Herbert J. Reich, Philip F. Ordnung, Herbert L. Krauss, and John G. Skalnik.
Microwave Theory and Techniques. D. Van Nostrand Company, Inc., 1953.
[58] Albert Roitman, Peter Horoyski, Mark Hyttinen, Dave Berry, and Brian Steer.
Extended interaction klystrons for submillimeter applications. In IEEE Intl.
Vacuum Electronics Conf. (IVEC), pages 191191. IEEE, 2006.
[60] E. Savrun, V. Nguyen, and D. K. Abe. High thermal conductivity aluminum ni-
tride ceramics for high power microwave tubes. In IEEE Intl. Vacuum Electronics
Conf. (IVEC), pages 3435, 2002.
179
[61] James Schellenberg, E. Watkins, M. Micovic, Bumjin Kim, and Kyu Han. W-
band, 5 W solid-state power amplifier/combiner. In IEEE MTT-S Intl. Mi-
crowave Symposium Digest, pages 240243, May 2010.
[62] S. Sengele, Hongrui Jiang, J. H. Booske, C. L. Kory, D. W. van der Weide, and
R. L. Ives. Microfabrication and Characterization of a Selectively Metallized W-
Band Meander-Line TWT Circuit. IEEE Trans. on Electron Devices, 56(5):730
737, May 2009.
[63] P. H. Siegel. Terahertz Technology. IEEE Trans. on Microwave Theory and
Techniques, 50(3):910928, March 2002.
[64] E. Simakov, D. Dalmas, L. Earley, W. Haynes, R. Renneke, and D. Shchegolkov.
Progress on the Omniguide Traveling-Wave Tube Experiment. 52nd Meeting of
APS Division of Plasma Physics, November 2010.
[65] Spotwelding Consultants, Inc. Technical Data Sheet: Glidcop AL-25, 2014.
[66] Spotwelding Consultants, Inc. Technical Data Sheet: Glidcop AL-60, 2014.
[67] David S. Tax. Experimental study of a high efficiency step-tunable MW gyrotron
oscillator. PhD thesis, Massachusetts Institute of Technology, Department of
Electrical Engineering and Computer Science, Cambridge, MA, September 2013.
[68] A. J. Theiss, C. J. Meadows, R. Freeman, R. B. True, J. M. Martin, and K. L.
Montgomery. High-Average-Power W-band TWT Development. IEEE Trans.
on Plasma Science, 38(6):1239 1243, June 2010.
[69] M. Thumm. Progress on gyrotrons for ITER and future thermonuclear fusion
reactors. IEEE Trans. on Plasma Science, 39(4):971979, April 2011.
[70] Shulim E. Tsimring. Electron Beams and Microwave Vacuum Electronics. Wiley
Interscience, 2007.
[71] Robert Valdiviez, D. Schrage, F. Martinez, W. Clark, et al. The use of dispersion-
strengthened copper in accelerator designs. XX International Linac Coference,
2000.
[72] Shuzhong Wang, Cunjun Ruan, Xiudong Yang, Ding Zhao, and Changqing
Zhang. The design considerations of W-band broad band output window. In
IEEE 14th Intl. Vacuum Electronics Conf. (IVEC). IEEE, 2013.
[73] R. D. Watson, F. M. Hosking, M. F. Smith, and C. D. Croessmann. Development
and testing of the ITER divertor monoblock braze design. Fusion Technology,
19:17941798, 1991.
[74] J.G. Wohlbier, J.H. Booske, and I. Dobson. On the physics of harmonic injection
in a traveling wave tube. IEEE Trans. on Plasma Science, 32(3):1073 1085,
June 2004.
180