PROCEEDINGS OF THE IRE

A Half-Watt CW Traveling-Wave Amplifierfor the 5-6 Millimeter Band*

H. L. McDOWELLt, MEMBER, IRE, W. E. DANIELSONt, AND E. D. REED T, MEMBER, IRE

Summary-The traveling-wave tube here described representsthe first practical CW power amplifier with broadband performance inthe millimeter wave band. More than 30 db of gain has been achievedover a bandwidth of 10,000 mc centered at 55,000 mc. A maximumCW output power of I watt has been obtained in this band. This com-bination of high power output and broadband performance representsa significant advance in the millimeter art.

The electrical and mechanical techniques are described whichwere found successful in solving the problems peculiar to the highoperating frequency. These problems are focusing, heat dissipation,intrinsic RF loss and structural precision. Experimental data on anoperating tube are presented which show good agreement with pre-dicted performance. These data suggest that the techniques de-scribed may be extended to allow either an increase in power to sev-eral watts at the present operating frequency or an increase in operat-ing frequency to 150 kmc for output powers of a hundred milliwatts.

INTRODUCTION

FqOR SOME years now the possibility of broadbandcommunications at millimeter waves has beenstudied at Bell Telephone Laboratories.' This

study has centered around the use of circular waveguideas a low-loss transmission medium. Transmission overa 40-kmc wide band, from 35-75 kmc, appears feasiblein a single waveguide, but initial interest has beenfocused on the 50 to 60 kimc region. This interest inmillimeter wave communications has stimulatedvacuum-tube work both on primary signal sources'-5and on amplifiers. In this paper we shall describe an ex-perimental helix-type traveling-wave amplifier with aCW power output of 2 watt in the 50 to 60 kmc band. Thisexceeds output powers previously obtained from CWamplifiers in this frequency range by at least an orderof magnitude.

Results on early work on millimeter wave amplifierswere reported by Little6 in 1951. In his tube a tiny un-supported helix was stretched between two posts andflooded by an electron beam. Cooling was entirely byradiation and output power thereby limited to a few

* Original manuscript received by the IRE, October 1, 1959.f SFD Labs., Inc., Union, N. J. Formerly at Bell Telephone Labs.,

Inc., Murray Hill, N. J.Bell Telephone Labs., Inc., Murray Hill, N. J.

l S. E. Miller, "Waveguide as a communication medium," BellSys. Tech. J., vol. 33, pp. 1209-1265; November, 1954.

2 E. D. Reed, "Tunable low voltage reflex klystron for operationin the 50-60 KMC band," Bell Sys. Tech. J., vol. 34, pp. 563-599;May, 1955.

3 A. Karp, "Traveling-wave tube experiments at millimeter wave-lengths with a new easily built space harmonic circuit," PROC. IRE,vol. 43, pp. 41-46; January, 1955.

4 A. Karp, "Backward wave oscillator experiments at 100-200KMC," PROC. IRE, vol. 45, pp. 496-503; April, 1957.

6 C. F. Hempstead and A. R. Strand, "Versatile source of milli-meter waves," Bell Labs. Rec., vol. 35, pp. 241-245; July, 1957.

6 J. B. Little, "Amplification at 6-millimeter wavelength," BellLabs. Rec., vol. 29, pp. 14-17; January, 1951.

microwatts. The net gain of this tube was about 3 dbat 6 mm. Also in 1951, Millmani7 reported 20 db oflow-level gain and an output power of about 20 mw at7 mm from a tube having an all-metallic filter-typecircuit. This tube had a bandwidth of about 7 per cent.In later millimeter wave work, the helix-type tube wasrevived by Robertson8 in an attemiipt to obtain widerbandwidth. He supported the helix by four knife edgesof quartz, but again cooling was mainly by radiation.He obtained a pulse power output of 5-10 Imlw at a 5per cent duty cycle. The net gaiin was 10-15 db at 6 mm.At the start of this program we thought that heat

dissipation associated with the relatively high poweroutput would force us to use an all-metallic filter-typestructure such as Millman's, with a consequent loss inbandwidth. However, further studies showed that ade-quate power output could be obtained with a helix, ifit was cooled by conduction and interception was keptvery low. These studies led to the helix-type tube de-scribed below.

DESIGN CONSIDERATIONS

The major problems encountered in the developmentof this amplifier were direct consequences of the verysmall helix diameter necessitated by the high operatingfrequency. These problems were:

1) the need for adequate cooling of the helix;2) the production of an electron beam of very small

diameter anid the focusing of this beam throughthe helix without requiring excessive cathode cur-rent density or excessive magnetic field;

3) the minimization of intrinsic helix attenuation;anid

4) the precise alignment of gun, helix and magneticfield.

Our solutions to these problems will be treated sepa-rately in the following sections.

Direct scaling of typical lower-frequency TWT'swould have resulted in a prohibitively small helixdiameter of about 5 mils. To enable us to use a largerdiameter, we chose to operate at the high ka value of0.25. As a result we were able to use a 15-mill diameter-a value which seemed attainable through the techniques

7 S. Millman, "Spacial harmonic traveling-wave amplifier for sixmillimeters wavelength," PROC. IRE, vol. 39, pp. 1035-1043; Sep-tember, 1951.

8 S. D. Robertson, "Broadband helix traveling-wave tube formillimeter wavelengths," IRE TRANS. ON MICROWAVE THEORY ANDTECHNIQUES, VOl. MTT-2, pp. 48-54; September, 1954.

1960 321

PROCEEDINGS OF THE IRE

to be described. At the time, a ka of 0.25 seemed aboutas high as we could safely go without incurring back-ward wave oscillations. Subsequently we found that ourmethod of supporting the helix introduced a strongstop-band around ka= 0.5. This means that we couldprobably increase ka to 0.4 in a future redesign withoutincurring backward wave oscillations. To make the fre-quency response of this tube as flat as possible, wechose ya equal to 1.5. This, together with the choice ofka, led to a comparatively high beam voltage and alarge helix-length-to-diameter ratio. The relevant tubeparameters are as follows.

ka = *25} at 55 kmc-ya = 1.5

Helix inside diameter= 15 milsHelix pitch =110 turns per inch

Helix length =4 inchesSynchronous Voltage = 7000 volts

Beam current= 3 maGain parameter C=0.015

The values of helix length and beam current were calcu-lated assuming an output of 100 mw, an efficiency ofone times C and a low-level gain of 25 db. Some of theseparameters are discussed in more detail later.

HEAT TRANSFER FROM THE HELIX

T'he helix is heated both by RF dissipation and inter-cepted beam current. This heat must be removed whilemaintaining the helix temperature low enough to pre-vent an undue increase in RF attenuation and a conse-quent loss of efficiency. This we have done by providinga direct thermal conduction path from the helix wire tothe outside of the vacuum envelope. As shown in Fig. 1,the helix is glazed to a single wedge of dielectric whichis forced against a massive copper heat sink by strongspring pressure. Dielectric rods of either Bell Labora-tories F-66 steatite or of synthetic sapphire have beenused. Beryllium oxide appears to be a still better ma-terial because of its very high thermal conductivity andit will be tried in the future. In finely divided form,however, it is highly toxic, and, its experimental evalua-tion will have to await the development of special grind-ing and processing techniques. These three dielectricmaterials-steatite, sapphire, and beryllium oxide-have thermal conductivities in the approximate ratio1:10:100.There are three sources of thermal impedance in the

structure of Fig. 1. These are the glazed joint betweenhelix and dielectric rod, the rod itself, and the interfacebetween rod and heat sink. The first of these is rather

I This means of introducing a stop band is similar in principle toan earlier arrangement used by Poulter in which a helix wasmounted in a quartz tube having a groove which intercepted thehelix once each turn. See w. L. Rorden, "A 100 Watt CW TWT atS Band," Electronics Research Lab., Stanford Un versity, Tech.Rept. 351-1; March 9, 1956.

Fig. I-Cross section of helix mounted on heat sink.

small because the glaze, in wetting both the helix andthe rod, provides an intimate thermal contact betweenthem. The other two impedances may, however, be ap-preciable. We obtain a measure of their relative magni-tudes by comparing the power dissipation properties ofhelices with steatite and sapphire rods. This was doneby passing dc current through some sample helicesmounted on a heat sink in a vacuum bell jar. The helixtemperature was calculated from the rise in wire re-sistance and plotted against the dc power dissipated.Fig. 2 shows the results. The curves for helices mountedon a heat sink are approximately straight lines indicat-ing that conduction is the main mechanism of heattransfer and that the thermal impedances are roughlyconstant. Whereas the intrinsic thermal conductivitiesof steatite and sapphire differ by a ratio of ten, thecurves show that the apparent conductivities for thetwo materials differ by a factor of only 3: 200°C/watt/inch-of-helix-length for steatite vs 65°C/watt/inch forsapphire. This means that the thermal impedanceof the interface between dielectric rod and heat sink isnot negligible. Assuming it to be the same for the twomaterials, as a first-order approximation, we calculateits value to be 50°C/watt/inch. The impedance of thedielectric rod is then 150°C/watt/inch for steatite andis 15°C/watt/inch for sapphire. This shows that for thecase of sapphire the interface impedance will largelycontrol the temperature rise of the helix. A rough calcu-lation of the termal impedance of the steatite rod fromits geometry shows good agreement with the aboveresults.

For comparison, Fig. 2 also shows the temperaturerise for a helix glazed to a ceramic rod but suspendedwithout contact to the heat sink so that it can only becooled by radiation. As may be seen, the power that canbe radiated from the helix for a given temperature riseis very much smaller than that which can be carriedaway by conduction.

Most of the helix heating is caused by beam bombard-ment-RF heating occurring in any appreciable mag-nitude only along the last few turns near the output.The beam interception in experimental tubes has rangedfrom 5 to 7 per cent. Assuming that this interception isdistributed over one quarter of the helix length, we findthat the temperature rise is 300°C for a steatite rod helixand 100°C for sapphire. Both of these values are withinpermissible limits.

322 March

lcDowell, et al.: Traveling- Wave A mplifier for the 5-6 Millimeter Band

600w

1-400

200

SAPPHIRE ROD

O 2 34

WATTS/ INCH

Fig. 2-Helix temperature vs power dissipated (per inch) inhelix structure.

At the very output enid, the helix will also be heatedby RF power dissipation. This heating causes an in-crease in helix attenuation as a result both of increasedwire resistivity and increased dielectric loss. The in-creased attenuation, in turn, lowers the power outputthereby giving rise to fading; i.e., if the beam is turnedon suddenly, the power output will go to a maximumvalue and theni decrease gradually over a period ofabout one minute while the helix heats up. The fadingproblem is aggravated by the necessity of stopping theheat sink short of the last few turns of the helix to makeroom for the helix-to-waveguide transducer. This makesthe heat flow path from the end turns to the heat sinklonger than from the remainder of the turns and thusincreases the fraction of the total thermal impedancewhich is due to the dielectric rod. Thus, to minimizefading, it is especially important to obtain a highthermal conductivity dielectric rod. Fading has limitedthe power output of tubes with steatite rods to aboutone-half watt CWV. The use of sapphire with about tentimes the thermal conductivity will reduce fading con-siderably. Beryllium oxide which offers an additionalfactor of ten in thermal conductivity would probablyproduce little further improvement in fading in thepresent tube. However, if means were found to reducethe thermal impedance between dielectric rod and heatsink, this material should make possible a considerableincrease in beam power. Coupled with an increase in kato 0.4, this could conceivably result in a CW output ofseveral watts.

BEAM FOCUSINGThe problem of focusing the tiny electron beam dif-

fers from that commonly encountered in that thetransverse thermal velocities with which electrons areemitted from the cathode become the predominant fac-tor in determining the magnetic field required for ade-quate focusing.To achieve a high-voltage convergent beam of much

smlaller diameter than had been produced in any of ourprevious work on convergent TWT beams, the gun de-signl effort was accompanied by theoretical beam studiesof quite general application. Herrmann,10 who carried outboth the analytical and the experimental phases of theinlitial gun work, extended calculations of thermalvelocity effects in electron beams beyonid the imag-netically shielded gun region"' to include the full regionof imiagnietic focusing between- accelerating anode anidcollector. Two particularly significant conclusions drawnfrom Herrmann's study are as follows:

1) Where thermal velocities play a dominant role,minimum magnetic field is obtained by shieldingthe cathode from the field.

2) Given the desired beam parameters (voltage, cur-rent and radius) and the cathode current denisity,the minimum magnetic focusing field can be pre-dicted without any knowledge of the specific gungeometry used. Comparison of the actual focusingfield required with this theoretical minimum thusaffords a measure of excellenice of the entirefocusing system.

The expression for the minimum field is made up ofthree terms as follows:

B2_ Bb2 + B 12 b

Space ThermalCharge Velocities

CathodeFlux

where

B = minimum total magnetic flux denisity required,Bb = Brillouin field to counteract space charge forces,Bt = field required to counteract spreading due to

thermal velocities,Bk= magnetic flux density at the cathode,Ak= cathode area, andAb= area of the beam.

10 G. F. Herrmann, "Optical theory of thermal velocity effects incylindrical electron beams," J. Appl. Phys., vol. 29, pp. 127-136;February, 1958.

11 The basic approach to calculations in the accelerating region ofa shielded Pierce gun had been developed by C. C. Cutler and M. E.Hines, "Thermal velocity effects in electron guns," PROC. IRE, vol.43, pp. 307-315; March, 1955. More detailed calculations applicableto shielded guns having a larger range of thermal velocities had beengiven by W. E. Danielson, J. L. Rosenfeld, and J. A. Saloom, "De-tailed analysis of beam formation with electron guns of the Piercetype," Bell Sys. Tech. J., vol. 35, pp. 375-420, March, 1956; also byG. F. Herrmann, "Transverse scaling of electron beams," J. Appl.Phys., vol. 28, pp. 474-478, April, 1957.

1960 323

PROCEEDINGS OF THE IRE

It is the space charge ternm which plays a dominant rolein the much discussed case of Brillouin focusinig. How-ever, as the diameter of a beam is made smaller anidsmaller, the motioni associated with transverse emissionvelocities remains essentiallk constant wvhile the trains-verse motion due to space charge decreases with de-creasing beam radius. In our tiny beams it is thereforenot surprising that the mnagnetic field term due totherm-lal velocities is about 2' timiies as large as that dueto space charge. The actual dimlensions of the mlillimiietertube gunl are showni in Fig. 3. The guni is completelyshielded from the magnetic field, making the third termlin the magnietic field expression zero.

For a giveni beamn diamneter and current, the thermalvelocitv terml Bt2 is proportional to cathode area. Con-sequently, to minimize the required field, the cathodecurrent density was made as high as possible, conisistentwith the probability of long cathode life.A relatively lonlg gunl with a small angle of con-

vergenice was used to keep lenis effects small. The meas-ured beatmi conivergenlce oIn scaled-up mnodels of this gunhas been about 20 per cenit greater thani calculated. Toreduce the convergence and optimize focusing in theactual gun, a positive bias of 5 volts is applied to thebeam-forminig electrode. The accelerating aniode voltagemiiust therefore be reduced fronm its design value in orderto mainitain the desired beamii current. This necessitatespost acceleration between the aniode and the helix, thusgivinlg rise to anl additional lenis effect. The magneticfield is introduced rather suddenly by an aperture in amagnietic pole piece brought inside the vacuum envelope.No attempt has been made as yet to optimize the ge-ometry of this region. Experimlenitally, about 95 per centbeam transmissioni from cathode to collector has beenobtainied with 1500 gauss magnietic field. This comparesfavorably with a calculated minimum field of 1200gauss. With some effort at eliminating the post-acceleration and optimizing the entranice conditions ofbeamn into the magnietic field, it should be possible toincrease the transmissioni.

ATTENUATION AND EFFICIENCY

At millimeter wavelengths, helix attenuation becomessufficiently high to have a major effect Oni power output.Fig. 4 shows the expected relationship between thesequantities. \Ve have plotted the ratio of efficiency tothe gain parameter C as a function of the ratio of L/Cwhere L is the loss per wavelength. The intercept of thecurve for zero atteniuation is obtained usinlg results ofan experimental study of TWT efficiency by Cutler,"2and the slope is determinied from another experimentalstudy of the effect of loss on efficiency by Cutler andBrangaccio .13

12 C. C. Cutler, "Nature of power saturation in travreling-wavetubes," Bell Sys. Tech. J., vol. 35, pp. 841-876; July, 1956.

13 C. C. Cutler and D. J. Brangaccio, "Factors affecting travelingwave tube power capacity," IRE TRANS. ON ELECTRON DEVICES,vol. ED-3, pp. 9-23; June, 1953.

ANODE

BFE

.025

0.187

CATHODE F_L ,003

-4 1-.025

Fig. 3 Electroln gtnii. (All (linlensiotis in iinches.)

3.0

2.0

\ 1.0

z 0.8

IUU.U. 0.6

0.4

0.20 10 20

L/C

30 40

Fig. 4-Approximate relationship between traveling-wave tubeefficiency anid L/C.

For a molybdenum helix of the type used in the nmilli-meter tube the L/C ratio as based on actual measuredhelices would be about 25. According to Fig. 4 thiswould result in an efficienicy of about C/3. To reducethe L/C ratio we copper-plate the helix. With reasonablecare in the platinig process, we cani obtaini a plated sur-face with 0.8 the conductivity of solid copper. Thiswould reduce the L/C ratio to about 12 and the resultingefficiency would be about 0.9 C. By platinig the helix wewill gain somewhat more than the factor of about 3 ini-dicated here, since the less lossy helix will in turni dissi-pate less RF power. This means that the power fadecaused by RF heating will be somewhat less. With a Cvalue of 0.015 this gives us an expected electronicefficiency of about 1.4 per cent. This comparatively low

324 March

McDowell, et al.: Traveling- Wave .4 Inplifier for the 5-6 Millimeter Band

value results first from the high-voltage design whichresults in low C, and second, from the effect of attenua-tion which reduces the efficiency by a factor of about 3over that which could be obtained at low frequencies.Presumably the over-all efficiency of the tube could beraised somewhat by low-voltage collection. For the lowQC value (0.07) of this tube, Cutler's experiments pre-dict a small enough velocity spread so that collection atas low as 0.3 of the helix voltage should be feasible.

\'IECHANICAL DESIGN

'I'he very small internal diameter of this helix (15mils) and its very large length-to-diameter ratio giverise to a tube which requires a high degree of precisionin its construction. Helix straightness and the align-menit of gun, helix, and magnietic field axes must beniainitained within a tolerance of about one-half milover the four-inch length (i.e., within an accuracy of 1in 8000). To obtain such precision through the use ofself-aligning piece parts would have required so manytight tolerances as to render the tube impractical. Theproblem of alignmenit may be viewed as two-fold: first,the concentric alignment of parts, and second, the angu-lar alignmlenlt of their axes. We chose to obtain concen-tric alignment by use of optical techniques, and angularalignmient by the use of a number of parallel referencesurfaces. This simplified the design to the point where wecould take full advantage of machininlg operations inwhich high precision is obtained relatively easily. Exam-ples of such operations are surface grinding to produceflatness and parallelism, centerless grinding to produceconstancy of diameter (although not necessarily straight-ness), and machininig with a single lathe setting to pro-duce concentricity of cylindrical surfaces. By extensiveuse of these techniques we have achieved alignment ofgun, helix and magnetic field axes to within the requiredone-half mil without imposing tolerances tighter thanone mil on absolute dimensions. By contrast, the use ofprecise interlocking piece parts would have requiredtolerances tighter than one-tenith mil to give this pre-cision of alignmlent.The helix assembly is shown in Fig. 5. The helix is

wound from 2X4 mil molybdenum tape, glazed to asingle wedge of dielectric, and then copper plated. Thissilgle-rod support has several advantages. First, it givesrise to structural simplicity and avoids the problem ofstress due to mechanical overconstraint. Second, itminimizes the amount of dielectric in the RF fields andthus miniimizes both dielectric loading and dielectricloss. Finally, the perturbation (once each turn) due tothe dielectric rod introduces a stop band into the helixtransmiissioln around the frequency for which ka=0.5.This helps to suppress any tendenicy toward backward-wave oscillation.The method of supporting the helix is shown in Fig. 6.

Thle ceramic support rod is forced into the corner of acopper block by a multiple-finiger spring which contacts

Fig. 5--Photograph of the helix glazed to (lielectric rod (helixpitch is 110 TPI).

Fig. 6-Helix mounited on a heat sink (artist's renderinig).

the rod along the entire length of the helix. This blockserves as a heat sink and has a direct heat conductionpath of low impedance to the outside of the vacuumenvelope. The surfaces of the block against which thehelix is mounted are ground accurately straight, butneed not be accurately located. The spring and the ce-ramic rod are so designed that there is a net componenitof force pushing the rod flat against the bottom surfaceand back into the corner. The dielectric rod is anchoredto the helix block at its midpoint only and is otherwisefree to slide axially with respect to the copper block asrequired by differential expansion during bake-out. Theend surfaces of the block are ground perpendicular tothe two planes which form the corner against which thehelix is mounted, and thus serve as precise referencesurfaces for aligninlg the gun and magnetic field axeswith respect to the helix.The input and output matching sections which con-

tain the waveguide-to-helix transitions, vacuum win-dows, and means for supporting gun or collector aremounted as shown in Fig. 7. These matching sectionsare brazed stack-ups of two copper disks and a steeldisk. The copper disks contain the waveguide, and thesteel disk serves as a magnetic pole piece. After brazing,the flat surfaces are ground accurately parallel, as sUg-gested in Fig. 7, and the inside hole is machined ac-curately concentric with the outside diameter. The

1960( 325

PROCEEDINGS OF THE IRE

Fig. 7-StruCtural features of the iillijineter-wtaxe 1 \\[(artist's renderiig).

waveguide windows are made as separate subassemblieswhich can be individually leak checked and tested fortheir microwave properties. They consist of a one-milthick synthetic mica flake glazed to a kovar cup which,in turn, is brazed across the copper waveguide. Thewindows have a transmission loss of about 1 db andVSWR of about 2 db. Once a window has been foundsatisfactory, it is fastened against its seat in the match-ing block by a heliarc weld. The matching sections arelocated concentric with the helix by optical alignmentand are then screwed to the helix block. (For simplicitythese screws are not shown in Fig. 7.) The outside di-ameters of the pole pieces are thus concentric with thehelix and can be used for aligning the tube in the mag-netic field. Finally, the four-gun positioning studs arelocated on a circle concentric with the helix-again byoptical means.The gun assembly is shown schematically (and not to

scale) on the left in Fig. 7. A subassembly of threeparallel platforms glazed to four ceramic rods is firstprepared. Electrodes are then placed on these plat-forms one at a time, aligned optically, and screwed intoplace. The complete gun is finally mounted on the inputmatching section, and positioned concentric with thehelix by the studs.The tube is completed by adding collector and stem,

and by closing the vacuum envelope by means of heliarcwelds. Throughout the design, care was taken in theselection of materials to insure that during thermalcycling parts did not shift as a result of differential ex-pansion pressures. Fig. 8 shows a photograph of thecompleted tube and Fig. 9 shows the tube mounted ina permanent magnet. This magnet provides 1500-gaussaxial field with the transverse component held to lessthan 1/10 per cent. The magnetic pole pieces inside thetube serve as part of the magnetic circuit and thus shieldgun and collector from magnetic fields. The holes in themagnetic pole pieces determine the magnetic axis of thecircuit; because they are concentric with the helix theymake the magnetic axis coincident with the helix axis.

Fig. 8---Photograph of the millimeter-wave TWlT.

Fig. 9-Photograph of the tube mounited in a permanent magnet-focusing circuit.

Since the gun has been carefully constructed to producea beam which is accurately coaxial with the helix, allthree critical axes-beam, helix and magnetic field-have thus been made coincident, and very good focusinghas been consistently obtained without the need fortedious final alignment procedures.

EXPERIMENTAL RESULTS

A total of seven tubes have given RF output powersof 100 milliwatts or more in the 5-6-mm band. Resultswill be presented for one of the later tubes in whichseveral of the early problems have been eliminated.Fig. 10 shows typical curves of RF power output as afunction of input power at midband for a beam currentof 3 ma. These curves were taken with the helix voltageadjusted for maximum gain at low signal levels. Maxi-mum power output at saturation is obtained at onlyvery slightly higher helix voltages, and the resultantcharacteristics appear almost identical to the onesshown here.

Experiments were performed to determine the maxi-mum output power capability of the tube. In these ex-

326 illarch

McDowell, et al.: Traveling- Wave Amplifier for the 5-6 Millimeter Band

31Ez

3:0C.

I--Q1-

0

500_ _

BEFORE FADING

AFTER FADING100

50 .__

In - -

5 10 50

40

36

32

28

AD

z

z4C,

.4

24

20

16

12100 500 1000

INPUT POWER IN /LW

Fig. 10-Power output as a function of power input.

0 1 2 3 4 5 6

BEAM CURRENT IN MA

Fig. 11-Power output as a function of beamn current.

periments the beam current was changed by varyingboth the anode anid beam forming electrode voltages.The saturation power output, both before and afterfading, was measured; Fig. 11 shows the results. It isseen that a power output of one watt at 5-ma beam cur-

rent was obtained, but that the tube could not hold thislevel and faded back to about one-half watt. As dis-cussed earlier, this fading was caused by heating of theoutput end of the helix by RF dissipation and the con-

8

4

050 52 54 56 58 60

FREQUENCY IN KMC

Fig. 12 Gain as a functioni of frequenicv.

E

z

or

w

0

a.

54 56

FREQUENCY IN KMC

60

Fig. 13-Power output as a ftuniction of frequency.

sequent increase in RF dissipation. Even at the one-watt level there was little increase in helix interception.The power output before fading is approxinmately pro-

portional to the four-thirds power of beaml current as

expected. It should be noted that this tube had F-66ceramic support rods, and that with the better thermalconductivity of sapphire it may be possible to maintaina considerably higher output level.The low-level gain and the saturationi power output

over the band are shown in Figs. 12 and 13 for a beamcurrent of 3 ma. The detailed shapes of these curves arelargely determined by the helix-to-waveguide trans-ducers initially used. A much improved transducer with

1960 327

_V [ 1

Cl)

3-J

z

a:9Id0a.

a.I-0x

II_I

PROCEEDINGS OF THE IRE

a minimum return loss of 15 db from 50 to 60 kmc hasbeen developed but has not yet beeni incorporated intoa tube. With such transducers flatter characteristicswill be obtained.

SCALING TO LOWER AND HIGHER FREQUENCIES

It may be interesting to speculate on the uses of thetechniques developed for this tube and extensions ofthese techniques in high-power tubes of lower frequency.Thus, suppose we scale the present tube down in fre-quency to X-band. 1'Maintaining ya, ka, and the currentdensity constant, and allowing for the decreased attenu-ation at the lower frequency, one watt at 55 kmc wouldcorrespond to 72 watts at 10 kmiic. To cool the helix ofsuch a tube successfully, techniques of improving theheat transfer across the boundary between the dielectricrod and the heat sink would have to be developed be-yond those used in the mm-wave tube. The CW outputat X-band might be further increased to over 200 watts,at the expense of a higher voltage, by inicreasinig ka to0.4. As discussed before, we would be dependinig onl thestop band at a ka of 0.5 to prevent backward-waveoscillations.

As for going to higher frequencies, it appears that anincrease in ka to 0.4 (while maintaining the helix diame-ter the same) might permit operatioin at 80 kmc. Opera-tion at a ka of 0.65 i.e., above the stop band but belowthe first forbidden region of the helix-might pernmit usto go to 150 kmc. Both of these tubes would have CWoutput powver in the range of about 100 milliwatts.

CONCLUSION

The results presented in this paper are of a prelimi-nary niature. However, we have accumiiulated sufficientexperience with this tube to convince us that there willbe no major stumblinig blocks betweeni the poinlt wehave now reached and our ultimate objective of a lonig-life tube with refined and reproducible performance. Bythe use of the mechanical techniques described here itis possible to keep tolerances reasonably liberal eventhough we obtain final alignmenits accurate to withinteniths of a mil. It should therefore be possible to coni-struct this tube in quantity and at reasonable cost. Thuswe feel that we have demonstrated the practicality of abroad-band amplifier capable of delivering power out-puts of up to one watt in the 5- to 6-mmli banid.

ACKNOWLEDGMENT

The millimeter wave amplifier described here hasbenefited importantly from contributions of a niumberof colleagues. L. J. Speck developed the means of con-structinig the tiny helix which is the heart of the tube.He also developed the mica vacuum windows and wasresponsible for the execution of the over-all mechanicaldesign. K. E. Schukraft of the Murray Hill PrecisioniRoom developed the technique of grindinig copperwhich we have used extensively in the construction ofthe tube. G. F. Herrmann designed the electroni guIn anldP. P. Cioffi developed the magnetic circuits. AM. E. Hinesmade valuable suggestionls with regard to the supportrod and to electrical contact to the helix.

CORRECTIONH. A. Wheeler, author of "The Spherical Coil as an

Inductor, Shield, or Antenna," which appeared onpages 1595-1602 of the September, 1958 issue of PRO-CEEDINGS, has requested that the following correctionsbe made to his paper.Formula (29) on page 1600 should read:

R ( 2a\3 1 V 1

wL \ X 1 + 2/k Vr I + 2/k

Formula (31) oni page 1600 should read:

37r2a'62

2X

to agree with formula (3) of the author's earlier paper,reference [18].Formula (46) on page 1602 should read:

( 27L2 1/3

27rM02A /

Alarch328