December, 2014

Plasma Science and Fusion Center Massachusetts Institute of Technology

Cambridge MA 02139 USA

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.

PSFC/JA-14-45

An Overmoded W-band Coupled-Cavity TWT

Kowalski, E.J., Shapiro, M.A., Temkin, R.J.

1

An Overmoded W-Band Coupled-Cavity TWTElizabeth J. Kowalski, Student Member, IEEE, Michael A. Shapiro, Member, IEEE,

and Richard J. Temkin, Fellow, IEEE

Abstract—A 94 GHz overmoded TWT has been designed,fabricated, and successfully tested. The TWT operates in therectangular TM31 mode of the cavity, while lower order modesare suppressed using selectively placed strips of a lossy dielectric.The 87-cavity TWT circuit was directly machined from Glidcop,a dispersion-hardened copper. The TWT was tested in a 0.25T solenoidal magnetic field in 3 microsecond pulses. Operatingat a voltage of 30.6 kV with 250 mA of collector current, theTWT was zero-drive stable and achieved 21±2 dB linear devicegain with 27 W peak output power. Taking into account 3 dBof loss in both the input and output coupling circuits at thewindows, the gain of the TWT circuit itself is estimated to be27±2 dB linear circuit gain with 55 W of saturated circuit outputpower. Using the 3D PIC code CST Particle Studio, the linearcircuit gain was estimated to be 28 dB and the saturated outputpower 100 W, in good agreement with the experimental results.The measured bandwidth of 30 MHz was significantly smallerthan the predicted value of 250 MHz. The overmoded TWT is apromising approach to high power TWT operation at W-Bandand to the extension of the TWT to terahertz frequencies.

Index Terms—TWT, overmoded, W-Band, vacuum electronics

I. INTRODUCTION

Coupled-cavity Traveling Wave Tube (TWT) amplifiers arereliable and compact vacuum devices at high frequencieswith a variety of applications. At conventional microwavefrequencies, TWTs have great operation parameters, with widebandwidth, high average power, high efficiency and large gain.However, bandwidth, power, and efficiency of TWTs waneabove 30 GHz. There is a need for TWTs in the W-band (75–110 GHz), and many research initiatives aim to expand theoperation of TWTs to higher frequencies.

There have been some successful TWTs built in the W-band.Most notably, a coupled cavity TWT developed at Varian,Inc. achieved 8 kW of peak power at 10% duty factor near95 GHz [1], [2]. However, advances in W-band TWT designhave been limited since that achievement. More recently, acomplex integral-pole-piece-ferruleless-coupled-cavity folded-waveguide circuit built by L-3 Communications, Inc. achieved100 W average power, 250 W peak power, 30 dB of gain,and 1% bandwidth [3]. In addition, a larger bandwidth froma similar design at L-3 achieved 4 GHz of bandwidth and 75W average output power for pulsed operation (150 W peakpower) at 94 GHz [4].

There are many new ideas for high power W-band TWT de-signs. These investigated structures include a micro-fabricatedhelical structure with two electron beamlets [5], a suspendedladder structure [6], a wood-pile electromagnetic-bandgap

E. Kowalski, M. Shapiro, and R. Temkin are with the Plasma Science andFusion Center, Massachusetts Institute of Technology, Cambridge, MA, 02139USA e-mail: ejk@mit.edu, temkin@mit.edu

waveguide [7], ridge-loaded folded waveguide [8] [9], orselectively metalized micro-fabricated folded waveguide [10].These designs are promising, but have not yet been exper-imentally validated. Experimentally, a cylindrical dielectricOmniguide photonic bandgap TWT has demonstrated up to4.5 dB of gain over a 10 GHz bandwidth in W-band[11].In addition, a wideband W-band TWT with a serpentinewaveguide has been designed and is under development [12].

An alternative approach to achieving high power TWToperation at high frequency is the use of sheet beam ormultiple beam devices. A high power TWT in Ka-Band hasbeen recently demonstrated using a sheet beam [13]. The sheetbeam approach has also been successfully extended to W-Bandin the form of a sheet-beam EIK klystron [14]. The sheet beamis a very promising approach but it differs from the presentapproach in that it is overmoded in only one dimension.

There are many efforts at higher frequencies using ad-vanced fabrication techniques. Microfabricated TWT designsaccept the limitations of the small wavelength devices anduse advanced manufacturing techniques to achieve precisionstructures. A 220 GHz TWT amplifier with five electronbeams has successfully demonstrated 28 dB of gain, 56 Wof output power and a 5 GHz bandwidth at 214 GHz [15].A 220 GHz sheet-beam folded-waveguide TWT made withnano-CNC fabrication techniques is in development whichwas simulated to have 35 dB of gain and up to 300 W peakpower [16] [17]. Using UV-LIGA fabrication, a 60 W, 220GHz folded waveguide TWT was successfully built at theNaval Research Lab [18]. The TWT has a beam tunnel ofabout 100 microns and a 50 GHz bandwidth. Though verysuccessful, the small size of the beam tunnel in such devicesis a difficult experimental constraint to overcome and requiresa large magnetic field for electron beam confinement.

In this paper, we present the design and test of a novel,promising design for a TWT, namely an overmoded, or over-sized design. Typically, coupled cavity TWTs are designedto operate in the fundamental mode, but the power handlingcapabilities of these devices drops rapidly with frequency pri-marily due to the decreased size of the device which is neededfor small wavelengths. The small cavities at high frequencieslead to manufacturing difficulties and current limitations dueto the size of the beam tunnel. By expanding TWT operationto a higher order mode of the cavity, thereby utilizing largercavity sizes, the difficulties associated with higher frequenciescan be minimized.

The overmoded, or oversized, cavity allows for a largebeam tunnel in the TWT which provides the space for ahigh interaction current while operating with a relatively smallmagnetic field (2.5 kG). In addition, the oversized cavitycan be direct machined with CNC milling instead of relying

2

Dielectric Loading

RF in

RF out

Electron

beam

19 cavities

ch

sh

cw

sw

dh

dw

p

2r

Fig. 1. A diagram of the overmoded TWT with dimensions labeled. A 19-cavity structure is shown, which was used in cold test. The final TWT had87 cavities. Vacuum components are shown in purple, dielectric is shown inpink.

1

0

0.5

TM11

TM31

TM21

(c)

(a) (b)

Fig. 2. The electric field magnitude for the (a) TM11 (b) TM21 and (c) TM31

modes in the TWT cavity as calculated by HFSS.

on nano-fabrication techniques. The design and experimentaldemonstration of an overmoded TWT cavity offers a newapproach to generating power in W-band and the possibilityto extend TWTs to higher frequencies and powers.

II. DESIGN

A. TM31 Mode Cavity

The TWT was designed to operate in a higher order modeof a rectangular cavity. In doing so, the cavity is larger than anequivalent fundamental mode cavity for the same frequency.This leads to several advantages, such as easier machining,higher beam current, lower magnetic field, and lower heatloading. The specifics of the cavity are shown in Fig. 1. Atwo-cavity diagram of the TWT highlights the staggered slotdesign, where the vacuum components are shown in purple.Dielectric components are shown in pink, which load thecavities at certain locations.

For this design, the cavity was chosen to operate in theTM31 mode. This mode allowed for a larger cavity struc-ture and sufficiently high coupling coefficient for operation.Without dielectric loading, the lower order modes would besusceptible to oscillation in the TWT. A high loss dielectric(aluminum nitride – silicon carbide composite, STL-100,

60 70 80 90 100 110 120−12

−10

−8

−6

−4

−2

0

Frequency (GHz)

Transmission per cavity (dB)

without dielectric

with dielectric

Fig. 3. The transmission through one cavity vs. frequency. Each pass-bandcorresponds to a different mode in the cavity: TM11, TM21, TM31, and TM41

centered at 65, 75, 94, and 112 GHz, respectively.

Sienna Technologies, Inc.) was selectively placed above andbelow each cavity to increase the losses in these modes, as isdemonstrated in Fig. 2. Fig. 2(a) and (b) show that the electricfields of the TM11 and TM21 modes are peaked in the locationsof the dielectric. Conversely, Fig. 2(c) shows that the TM31

mode has little field or interaction with the dielectric loads.This is shown quantitatively in Fig. 3, where the transmission,S21 per cavity is shown as a function of frequency withand without the dielectric loading. These simulations werecalculated in Ansoft HFSS. The simulations shows large lossesin the two lower frequency modes (TM11 and TM21) and themode at 112 GHz (TM41). The dielectric has very little effecton the TM31 mode, centered at 94 GHz.

The dimensions of the cavity were selected through iterationin order to meet design goals of the TWT. It was desired tohave a TWT with at least 30 dB of gain and above 100 W peakpower. In this design, bandwidth was not optimized. Throughsimulations, the final parameters were determined as shown inTable I. The beam tunnel is 0.8 mm in diameter, one of thelargest beam tunnels for a W-band TWT. For comparison, arecent design of a high power, serpentine waveguide TWT hasa beam tunnel diameter of only 0.23 mm, which is less than30% of the size of the present TWT beam tunnel [12]. Theseparameters for the overmoded TWT led to ideal operation at93.9 GHz with 3.2 Ω coupling impedance and 0.3 dB/cavityloss. The loss is quite high in the TWT circuit, which allowsfor the circuit to be implemented without a sever, since theround trip losses will be less than the gain of the TWT.The design considered ideal machining conditions (marked“simulated” in the table), while the final machined parameters(marked “machined”) were adjusted to account for fillets dueto the radius of CNC tools used.

B. Simulations

The TWT was initially designed using Pierce theory, andthe design was later refined using computer simulations. Theexperiment operated with a 2.5 kG solenoid magnet, whichprovided an engineering restriction on the electron beamparameters. The specifics of simulation parameters are listed

3

TABLE ISIMULATED AND MACHINED DIMENSIONS OF FINAL TWT STRUCTURE

Property Label Simulated (mm) Machined (mm)

Cavity Width cw 5.60 5.79Cavity Height ch 2.54 2.54Cavity Length cl 0.40 0.40

Slot Width sw 0.60 0.60Slot Height sh 1.30 1.32

Beam Tunnel Radius r 0.40 0.40Dielectric Height dh 0.30 0.53Dielectric Width dw 0.70 0.58

Period p 0.80 0.80Fillet Radius rf 0 0.13

TABLE IISIMULATED TWT OPERATION PARAMETERS

Property Variable Value

Voltage V0 31.1 kVCurrent I 310 mA

Beam Radius rb 0.3 mmMagnetic Field B0 2.5 kG

Number of Cavities n 87Synchronous Operation Frequency f 94 GHz

Coupling Impedance at 94 GHz K 3.2 ΩLoss per cavity l 0.3 dB

in Table II. The final TWT had 87 cavities, taking advantageof the length of the magnetic field flat top available.

Analytical calculations for the gain in the TWT wereperformed following Pierce theory. In general, the gain followsthe relationship

G = −9.54 +BCN [dB] (1)

where B is a weighting factor dependent on frequency, space-charge, and loss, C is the Pierce parameter, and N is the num-ber of wavelengths [19]. For the overmoded TWT parametersand operating conditions, B = 30.16, C =0.0203 and N =65.24 wavelengths. The analytical formula results in 30.4 dBof linear gain for the TWT at 94 GHz.

Latte simulations were performed by importing the dis-persion relation, loss characteristics, and coupling impedancewhich were calculated via HFSS. LMSuite Latte is a 1-Dcode developed for studying helical TWTs [20]. In addition,the full 87 cavity structure was simulated in CST ParticleStudio along with the expected magnetic field and expectedelectron beam parameters. These parameters were provided viathe design of the 31 kV, 310 mA electron gun using LeidosMichelle software [21] to study the beam optics. Simulationresults showing the bandwidth, gain, and saturation of theTWT are shown in Figure 4. Latte and CST results showgood agreement for the linear gain and saturated power. Forthe operation parameters, the TWT is expected to have 32 dBof gain and 300 W peak power at 93.9 GHz. The calculatedinstantaneous bandwidth is 250 MHz, although optimizingbandwidth was not a design consideration for this experiment.

93.4 93.6 93.8 94 94.20

5

10

15

20

25

30

35

Frequency (GHz)

Gain

(dB

)

30.5kV

31.1kV

32kV

100

102

104

100

101

102

103

Input Power (mW)

Ou

tpu

t P

ow

er

(W)

Latte

CST

(a)

(b)

Fig. 4. Simulation results for the TWT, showing (a) the bandwidth of theTWT as calculated in Latte and (b) the gain relationship as calculated in bothCST and Latte.

5.8 mm

WR10

Beam Tunnel

9 cavities, glidcop(a)

(c)

(b)

6.8 cm

Fig. 5. (a) Assembly of the dielectric loading in the cavity structure alongwith coupling waveguides. (b) Details of a 9-cavity test structure. (c) Thefinal 87-cavity structure that was installed in the TWT.

C. Cold Test

Low power measurements were performed in order to verifymachining of the TWT cavities and material selection. Themeasurements were performed with a 2-port Agilent VNAwith WR-08 (90-140 GHz) millimeter waveguide extensions.

4

90 95 100 105 110 115

−60

−40

−20

0

Frequency (GHz)

S21 (dB)

S21 (dB)

−60

−40

−20

0No Dielectric

HFSS

19 cell Glidcop

With Dielectric

19 cell Copper

HFSS

19 cell Glidcop

19 cell Copper

Fig. 6. Low-power transmission measurements of the TM31 and TM41 modethrough the 19-cavity glidcop and copper structures (top) without and (bottom)with dielectric loading. HFSS simulations are shown for comparison.

92 92.5 93 93.5 94 94.5 95 95.5 96 96.5 97 97.5−70

−60

−50

−40

−30

−20

Frequency (GHz)

S21 (dB)

Measured

Simulation

Fig. 7. Low-power transmission measurements of the TM31 mode in the final87-cavity structure. The simulated transmission is shown for comparison.

Test structures were made with 9- and 19-cavity lengths out ofOFHC copper and AL60 LOX glidcop. They were made suchthat they could be tested with or without dielectric loadingin place. For the testing without dielectric loading in place,copper inserts filled the loading slots on the cavities.

Details of the cold test and final structures are shown in Fig.5. A breakout CAD drawing of the assembly is shown in Fig.5(a). The cavities are machined in either side of a split block.Since the gap between the blocks is at a null in magnetic field,there are minimal losses due to RF leakage [22]. The dielectricloading is placed tangent to the cavities in four locations (twoon the top, and two on the bottom of the cavities) and is held inplace with copper inserts (shown in blue). WR-10 waveguidescouple power into and out of the coupled-cavity circuit. Theentire structure is clamped together. There is no need for abraze due to the demountable design of the TWT experiment,which will be discussed in the next section. Fig. 5(b) showsthe 9-cavity cold test structure details, and Fig. 5(c) shows thefinal 87-cavity structure.

Test structures were made out of both OFHC copper andAL60 LOX glidcop in order to determine if there was a more

suitable material for machining. Glidcop is an alloy of copperthat has been impregnated with a small amount of alumina,leading to a harder material without much loss in conductivity.Low-Oxygen (LOX) glidcop also has a small quantity ofboron. Glidcop has successfully been used in many vacuumelectronics and accelerator applications. AL60 LOX glidcophas 0.6 wt. % alumina, 250 ppm boron, and a resistivity of2.21 µΩ-cm.

The transmission, S21, through 19-cavity cold test structuresmade from both glidcop and copper is shown in Fig. 6. Thetransmission through the glidcop structure was an improve-ment over the copper structure. In addition, the copper sawmore reflection. This result is due to the fact that copper ismore malleable and is more susceptible to machining errors,leading to difficulties when creating small, consistent, andprecise cavities. The hardness of glidcop leads to more preciseand consistent cavities. Following the success of the cold test,the final TWT 87-cavity structures were made out of glidcop.

The effectiveness of the dielectric loading was also mea-sured with the smaller cold test structures. The dielectric usedwas an aluminum nitride – silicon carbide (AlN–SiC) com-posite from Sienna Technologies. The dielectric was chosenbecause it has a high loss tangent, good thermal properties,and will perform well in vacuum conditions. Fig. 6 shows the19-cell structures without (top) and with (bottom) dielectricloading in place. The TM31 and TM41 modes were observedduring measurement centered around 95 GHz and 112 GHz,respectively. It can be seen that the dielectric loading reducedthe transmission through the cavities for the TM41 mode whilehaving little effect on the transmission of the TM31, in goodagreement with the HFSS simulations. Fig. 6 also showsthat the glidcop structures have lower loss than the copperstructures. Consequently, the final 87-cavity structure for theTWT experiment was made with AL60 LOX glidcop.

Prior to installation, the transmission properties of the 87-cavity structure were also measured in cold test. With thedielectric loading in place, the TM41 mode was not visibleabove the noise floor of the VNA (about -60 dB) for eitherstructure. Fig. 7 shows the measured S21 for the TM31 modeof the structure along with the simulated transmission. Themeasured and simulated transmission of the TM31 mode agreewell in magnitude and bandwidth; there is a frequency shiftin the measured transmission that is due to small machiningerrors.

III. EXPERIMENT

A. Set-Up

The overmoded TWT was designed and built at MIT. Thedesign is made to be modular for quick experimental turn-around times and has a wide range of operation conditions.The experiment is operated pulsed at 1 Hz with a pulseforming network that provides a 3 microsecond flat-top voltagepulse variable from 10-100 kV. Due to breakdown limitations,the TWT was limited to operation below 35 kV. The fullexperimental assembly and laboratory set-up are shown inFig. 8. A 4-coil solenoid magnet, shown in outline in Fig.9, provides a 2.5 kG field for the electron beam. Iron pole

5

94 GHz EIK

(driving source)

2.5 kG

Solenoid Magnet

30 kV

Electron Gun

Fig. 8. The TWT experimental set-up which was used for testing. The pulseforming network, power supplies, and controls are not shown.

Solenoid Magnet, 26 cm

Electron Gun

Collector

RF Windows

to pump

to pump

Fig. 9. A cross-sectional CAD view of the TWT experiment. The 87-cavityTWT is highlighted in the cutout, showing the alignment surface to the anodeand WR-10 coupling waveguide. The cavities are highlighted in blue and thesimulated electron beam is shown in pink.

pieces on the front and back of the magnet were designedsimultaneously with the electron gun to provide the nominalfocusing field. In addition, a gun-coil provides up to 260 Gto fine-tune the focusing field. The magnet is 26 cm in lengthwith a 5 cm diameter bore and a 10 cm magnetic field flat-top.

The input power for the TWT was provided by two sources.A low power solid-state Amplifier Multiplier Chain (AMC)from Millitech provided up to 32 mW of power from 90-100 GHz. A high-power Extended Interaction Oscillator (EIO)from CPI, shown in Fig. 8, provided up to 300 W of powerfrom 93.5–95.7 GHz. In combination, these sources providedthe power necessary to characterize the linear gain, saturation,and bandwidth of the TWT under test.

Details of the vacuum components, including the electrongun and TWT circuit, are shown in Fig. 9. The 31 kV, 310mA electron gun was designed in Michelle, a beam opticsmodeling software, to fit the parameters necessary for theexperiment [21]. A 3.2 mm diameter cathode manufactured byHeatwave Labs provides up to 5 A/cm2 on the surface, whichis focused to a 0.6 mm diameter beam. The TWT circuit isabout 7 cm in length, and is precision aligned to the anode

0 5 10 15 20 25 300

50

100

150

200

250

300

350

Voltage (kV)

Curr

ent (m

A)

Expected

Current at Collector

Fig. 10. Results from the beam test performed prior to TWT installation. Themeasured current at the collector vs. voltage is shown along with the totalexpected current, as calculated via Child-Langmuir.

surface. WR-10 waveguide couples power into and out of theTWT circuit with very low loss in the waveguide run fromthe window to the circuit. The windows are fused silica, witha thickness that transmits from 92–97 GHz with less than0.1 dB losses, as measured with a VNA. To transmit throughthe window, the WR-10 waveguide is uptapered to WR-28waveguide before the window and then downtapered backto WR-10 waveguide after transmission through the window.These tapers are used at both the input window and the outputwindow. We measured that 3±0.5 dB of loss occurs at eachwindow due to these tapers. This loss number is needed toanalyze the measured gain to differentiate between device gain,measured from the input to the output port of the device, vs.gain on the circuit itself. A simple OFHC copper collector isused, and there are 2 L/s pumps on both the gun and collectorsides of the experiment.

The tube is free-standing within the magnetic bore. It isheld in place and aligned with 3-axis translation stages oneither side of the magnet. The tube must be assembled andinstalled in the magnet simultaneously. Since the tube cannotbe easily accessed after installation, the components underwenta bakeout at 150–200 C prior to final assembly. This issufficient processing for the pulsed experiment.

B. Beam Test

Prior to installation of the TWT circuit, a beam test wasimplemented to verify electron gun operation. For this test, thetube design shown in Fig. 9 was modified. The TWT circuitand waveguides were replaced by a 1.1 cm diameter, 30 cmlong beam tunnel.

The beam tunnel was oversized so that it could be elec-trically isolated from the body without having to rely on theanode’s alignment surface for beam transmission. In addition,the anode reduces to a diameter of 0.8 mm for 0.25 inchesprior to the beam tunnel. This test ensures that the beam iscoupling through the anode which is the desired radius of thefinal beam and that the beam can be directed to the collector.The test involves the measurement of three currents in thetube: body, collector, and beam tunnel.

The results from the beam test are shown in Fig. 10. Thisfigure shows the measured collector current vs. operationvoltage as well as the total expected current. The total expected

6

−30 −20 −10 0 10 20 30 40 50−10

0

10

20

30

40

50

Input Power (dBm)

Outp

ut P

ow

er

(dB

m)

AMC

EIO

Linear Fit

Fig. 11. Output power vs. input power for the TWT at 94.26 GHz, showinglinear gain and saturation. The linear fit of lower power data corresponds to21 dB of gain.

0 1 2 3 4 5 6 70

5

10

15

20

25

30

35

time (microsec)

Vo

lta

ge

(k

V),

Dio

de

(a

rb.) Voltage

Diode

Fig. 12. Data of a sample pulse taken at 94.26 GHz during operation withthe EIO input. The 2.8 microsecond voltage pulse is shown along with thediode trace used to measure the output power of the TWT.

current should follow the Child-Langmuir equation for aspace-charge operated electron gun,

I0 = PV3/20 (2)

where P is the perveance of the gun. Michelle code simulationsindicated the expected perveance to be 0.059 micropervs forideal operation. At 31 kV, the measured collector currentwas 306±6 mA, in good agreement with the expected totalcurrent of 310 mA. The currents measured on the body andbeam tunnel at 31 kV were 20 mA and 23 mA, respectively.Therefore, 88 % transmission of the electron beam to thecollector was measured. Although there is a relatively highbeam interception, the overall total current was more thananticipated and more than 300 mA of current was transportedto the collector.

C. TWT Results

After the beam test was completed, the overmoded TWTcircuit was installed in the tube and tested at 30.6 kV. After aconsiderable effort at alignment, the highest current measuredat the collector was 250 mA, significantly reduced from the306 mA observed in the beam test. For these conditions, theTWT device linear gain was measured to be 21±2 dB. Themagnetic field was about 2.0 kG for this operation. At thesesettings, the TWT was zero-drive stable. This result is shownin Fig. 11, which shows the measured output power vs. inputpower to the TWT device. Here, we define the TWT device

−30 −20 −10 0 10 20 30 405

10

15

20

25

30

35

Circuit Input Power (dBm)

Cir

cuit

Ga

in (

dB

)

AMC

EIO

CST simulation

Fig. 13. Circuit gain vs. input power to the circuit for the TWT at 94.26GHz, showing 27 dB circuit gain. Simulation results from CST which havebeen adjusted for experiment current conditions are shown.

94.22 94.24 94.26 94.28 94.3 94.320

5

10

15

20

25

Frequency (GHz)

Dev

ice

Gai

n (d

B)

Fig. 14. Device gain vs. frequency for the overmoded TWT during low-power,linear gain operation at 30.6 kV.

gain as the gain measured between input and output ports ofthe TWT. The input port is the WR-10 waveguide just priorto the uptaper to WR-28 guide on the air side of the inputfused silica window. The output port is defined as the WR-10waveguide located after the downtaper on the air side of theoutput window. The low power data points used the solid-stateAMC as the input driver, while the higher power points usedthe EIO. The low power points were fit to a linear curve whichcorresponds to 21±2 dB of gain. At high power, saturation isobserved at 27 W of device output power.

A sample set of data from the 94.26 GHz high gain pointoperation is shown in Fig. 12. For this trace, the TWT wasdriven with a pulsed input provided by the EIO. Fig. 12 showsthe voltage pulse along with the output diode signal whichmeasured the power. The diode is calibrated along with avariable attenuator at the output to calculate the output powerof the TWT, and noise from the measurement cables has beenfiltered out of the data. The trace shows that the early part ofthe voltage pulse had sufficient ripple, ±2% peak-to-peak, tostrongly affect the gain of the TWT. The values of TWT powerand gain were taken as an average over the flatter portion ofthe voltage waveform, from about 3.5 to 4.5 microseconds onthe waveform shown in Fig. 12. A flatter voltage waveformmight have slightly improved the present results.

To compare to theory, we must take into account couplinglosses in the device. Previously, we noted that there was 3dB of loss at both the input and output coupler. Taking into

7

account the 6 dB of cumulative coupling losses, there was27±2 dB of circuit gain measured in the TWT. Fig. 13, showsthe data from Fig. 11 where the device gain has been calculatedand adjusted by 6 dB to represent circuit gain, and error barshave been added. Adjusting the saturated output power by 3dB, there was 55 W of peak circuit output power at saturation.

A theoretical analysis was performed for the TWT using 250mA interaction current (the observed current on the collectorduring linear gain operation), while keeping all other factorsthe same as previous analysis. With this current, the lineargain simulated in CST Particle Studio was 28 dB, and thesaturated output power was 114 W. These results are shownnext to the measured points in Fig. 13 and show the goodagreement between simulation and theory.

A bandwidth measurement was done by maintaining con-stant operation parameters and varying the input frequencywith the solid-state AMC drive input. The bandwidth of theTWT is shown in Fig. 14 for the high gain operation point.For the 94.26 GHz high gain point, the voltage was kept at30.6 kV and the bandwidth was measured to be 30 MHz.This bandwidth is smaller than the predicted bandwidth of250 MHz. We believe that this may arise from fabricationerrors, beam quality, and voltage ripple during the flat top ofthe pulse.

IV. DISCUSSION AND CONCLUSIONS

An overmoded W-band TWT has been built and tested,showing successful operation in the TM31 cavity mode. TheTWT is zero-drive stable, and no evidence of oscillations inthe lower-order TM11 or TM21 modes was observed. Theovermoded TWT has one of the largest beam tunnels for aW-band TWT and is able to operate with a 2.5 kG solenoidmagnet and an electron gun that was designed to produce 310mA at 31 kV. Simulations in both 1-D Latte and 3-D CSTParticle Studio for ideal operation conditions predicted 32 dBof linear gain and 300 W peak output power.

Testing of the electron gun showed transmission of 306±6mA to the collector, corresponding to 88 % transmission of thebeam. During operation of the TWT, under the best alignmentconditions, 250 mA was transmitted to the collector. Thelarge beam interception limited the gain and power output ofthe device. The origin of the large beam interception is notunderstood at this time. However, in device operation with 3microsecond pulse lengths at a repetition rate of 1 Hz, novisibly damage to the structure or other deterioration wasobserved. Adjusting simulations to correspond to operatingconditions of 250 mA, CST Particle Studio predicted 28 dBof linear gain and 100 W of peak output power.

The overmoded TWT was measured to have 21±2 dB lineardevice gain at 94.26 GHz, and a peak output power of 27 W.Adjusting for 6±1 dB of input and output coupling losses,there was 27±2 dB of linear gain in the TWT circuit and 55W of peak output circuit power. This is in good agreementwith theory.

This overmoded TWT has shown successful operation in theTM31 cavity mode, and the research can be used to expandTWTs to operate at high powers in oversized cavities at higherfrequencies.

ACKNOWLEDGMENT

The authors are grateful to William Guss and Ivan Mas-tovsky for their help with the experiment.

This research was supported by the Air Force Office ofScientific Research Program on Plasma and Electroenergetics,grant FA9550-09-1-0363.

REFERENCES

[1] B. G. James and P. Kolda, “A ladder circuit coupled-cavity TWT at 80-100 GHz,” in 1986 Intl. Electron Devices Meeting, vol. 32, pp. 494–497,1986.

[2] B. G. James, “Coupled-Cavity TWT designed for future Mm-wavesystems,” in Microwave Systems News and Communications Technology,vol. 16, pp. 105–116, 1986.

[3] A. J. Theiss, C. J. Meadows, R. Freeman, R. B. True, J. M. Martin, andK. L. Montgomery, “High-Average-Power W-band TWT Development,”IEEE Trans. on Plasma Science, vol. 38, pp. 1239 –1243, June 2010.

[4] R. Kowalczyk, A. Zubyk, C. Meadows, M. Martin, M. Kirshner, R. True,A. Theiss, J. Rominger, and C. Armstrong, “A 100 watt W-band MPMTWT,” in IEEE Intl. Vacuum Electronics Conf., May 2013.

[5] C. Kory, L. Ives, M. Read, J. Booske, H. Jiang, D. van der Weide, andP. Phillips, “Microfabricated W-band traveling wave tubes,” in 13th Intl.Conf. on IR mm Waves and THz Tech., vol. 1, pp. 85–86, Sept 2005.

[6] C. L. Kory, M. E. Read, R. L. Ives, J. H. Booske, and P. Borchard,“Design of Overmoded Interaction Circuit for 1-kW 95-GHz TWT,”IEEE Trans. on Electron Devices, vol. 56, pp. 713–720, May 2009.

[7] I. Ederra, I. Khromova, R. Gonzalo, N. Delhote, D. Baillargeat, A. Murk,B. E. J. Alderman, and P. M. de Maagt, “Electromagnetic-bandgapwaveguide for the millimeter range,” IEEE Trans. on Microwave Theoryand Techniques, vol. 58, pp. 1734–1741, July 2010.

[8] J. He, Y. Wei, Y. Gong, and W. Wang, “Linear analysis of a Wband groove-loaded folded waveguide traveling wave tube,” Physics ofPlasmas, vol. 17, no. 11, article 113305, 2010.

[9] J. He, Y. Wei, Y. Gong, W. Wang, and G.-S. Park, “Investigation on aW band ridge-loaded folded waveguide TWT,” IEEE Transactions onPlasma Science, vol. 39, pp. 1660–1664, Aug 2011.

[10] S. Sengele, H. Jiang, J. H. Booske, C. L. Kory, D. W. van der Weide, andR. L. Ives, “Microfabrication and Characterization of a Selectively Met-allized W-Band Meander-Line TWT Circuit,” IEEE Trans. on ElectronDevices, vol. 56, pp. 730–737, May 2009.

[11] D. Shchegolkov, W. Haynes, R. Renneke, and E. Simakov, “Millimeter-wave gain experiments with a wide-band omniguide traveling-wavetube,” in IEEE Intl. Vacuum Electronics Conf. (IVEC), pp. 297–298,April 2012.

[12] A. M. Cook, C. Joye, J. Calame, K. Nguyen, D. Chernin, A. Vlasov,D. Abe, and B. Levush, “Development of a wideband w-band serpentinewaveguide twt,” in IEEE Intl. Vacuum Electronics Conf. (IVEC), pp. 1–2,May 2013.

[13] D. E. Pershing, K. T. Nguyen, D. K. Abe, E. Wright, P. B. Larsen,J. Pasour, S. J. Cooke, A. Balkcum, F. N. Wood, R. E. Myers, andB. Levush, “Demonstration of a wideband 10 kW sheet beam TWTamplifier,” IEEE Transactions on Electron Devices, vol. 61, no. 6,pp. 1637–1642, 2014.

[14] J. Pasour, E. Wright, K. T. Nguyen, A. Balkcum, F. N. Wood, R. E.Myers, and B. Levush, “Demonstration of a multikilowatt, solenoidallyfocused sheet beam amplifier at 94 GHz,” IEEE Transactions on ElectronDevices, vol. 61, pp. 1630–1636, June 2014.

[15] K. E. Kreischer, J. C. Tucek, M. A. Basten, and D. A. Gallagher, “220ghz power amplifier testing at northrop grumman,” in 14th Intl. IEEEIntl. Vacuum Electronics Conf. (IVEC) Paris, France, May 2013.

[16] A. Baig, D. Gamzina, R. Barchfeld, C. Domier, L. R. Barnett, andN. C. Luhmann Jr, “0.22 THz wideband sheet electron beam travelingwave tube amplifier: cold test measurements and beam wave interactionanalysis,” Physics of Plasmas, vol. 19, no. 9, pp. 93–110, 2012.

[17] Y.-M. Shin, A. Baig, L. Barnett, N. Luhmann, J. Pasour, and P. Larsen,“Modeling investigation of an ultrawideband terahertz sheet beamtraveling-wave tube amplifier circuit,” Electron Devices, IEEE Trans-actions on, vol. 58, pp. 3213–3218, Sept 2011.

[18] C. D. Joye, A. M. Cook, J. P. Calame, D. K. Abe, A. N. Vlasov, I. A.Chernyavskiy, K. T. Nguyen, and E. L. Wright, “Microfabrication andcold testing of copper circuits for a 50-watt 220-GHz traveling wavetube,” in Proc. SPIE, vol. 8624, 2013.

8

[19] S. E. Tsimring, Electron Beams and Microwave Vacuum Electronics.Wiley Interscience, 2007.

[20] J. Wohlbier, M. Converse, J. Plouint, A. Rawal, A. Singh, and J. Booske,“LATTE/MUSE numerical suite: an open source teaching and researchcode for traveling wave tube amplifiers,” in 30th IEEE Intl. Conf onPlasma Science, p. 177, June 2003.

[21] J. Petillo, P. Blanchard, A. Mondelli, K. Eppley, W. Krueger, T. Mc-Clure, D. Panagos, B. Levush, J. Burdette, M. Cattelino, et al., “TheMICHELLE electron gun and collector modeling tool,” in Proc. of the2001 Particle Accelerator Conference (PAC), vol. 4, pp. 3054–3056,IEEE, 2001.

[22] S. M. Lewis, E. A. Nanni, and R. J. Temkin, “Direct machining of low-loss THz waveguide components with an RF choke,” IEEE Microwaveand Wireless Components Letters, vol. 24, no. 12, pp. 842–844, 2014.