IRE TRANXACTIONX ON ELECTRON DEVICES 247

The Multiple- lystron*

Summary-The multiple-beam klystron (MBK) is a device for extending klystron power generation capability at a given frequency by a factor of ten or more. The MBK utiEzes a multiplicity of elec- tron beams in conjunction with multiwavelength waveguide cir- cuits. These circuits are periodically loaded and operate in the ?~/Z-mode. The efficiency, bandwidth, gain and stability of the MBK are equal to or better than that oi the single-beam prototype kly- stron. Furthermore, the MBK permits generation of a given power level a t an unusually low voltage, thus minimizing insulation, X-radiation, and modulator problems.

An MBK utilizing ten external-circmit klystrons has been built for 750-Mc operation to demonstrate the principle of the device. A single main magnetic circuit is used to focus all beams simul- taneously. An RF power output ten times that of one klystron was measured, corresponding to an efficiency of 44 per cent. Bandwidth and gain were identical with single-beam prototype operation. Individual beam drop-out tests were made which showed no Cis- ruption of Operation in case of the failure of one or more beams. Limited tests tend to confirm the conclusion that the harmonic content of an MBK can be lower than that of a single-beam klystron.

I. INTRODUCTION HE CONTINUALLY increasing sophistication of radar and communication systems requires ever- increasing transmitter power output. At this time

the most energetic stable power source in the microwave region is the klystron power amplifier. The klystron can be characterized as being a high gain amplifier of moderate bandwidth and efficiency and having a more robust mechanical structure than other types of generators. In the search for devices with higher and higher power capability, the klystron interaction represents a logical starting point. This paper discusses a unique configuration of electron beams and circuits whereby the power genera- tion capability of amplifiers a t a given microwave fre- quency can be increased by ten- to one-hundredfold while retaining other desirable klystron operating charac- teristics such as gain, eaciency, and bandwidth.

Conventional klystrons, either pencil or hollow-beam, have a limiting power capability which depends on fre- quency, cathode emission density, and the thermal dissipation required of various portions of the RF circuits. This limitation results, in part, from the necessity for limiting maximum resonator dimensions to values less than one free space wavelength and also by the require- ment for relatively small drift tube diameters to achieve good coupling between electrons and resonator fields.

performed with the support of the U. S. Army Signa1,Engineering * Received by the PGED, December 1, 1961. This work was

Laboratory under contract DA-36-039-SC-78178. ‘f Mictron, Inc., Loudonville, N. Y. Formerly with General

Electnc Company. $ General Electric Company, Schenectady, N. Y.

For pencil beams, the result is a drift tube diameter of the order

where

X, = free space wavelength u = electron axial velocity c = velocity of light.

Use of a hollow beam increases this limiting diameter by the fact.or c/v. However, at high power levels, where large beam voltages must be used, this factor will in general be less than two. Obviously, no order-of-magnitude increase in power is available by this expendient.

One solution to the problem of very high power genera- tion is to employ circuit structures with extended di- mensions normal to the electron velocity. In such a case at least one dimension of the electron beam must be restricted so that good coupling to circuit fields are maintained. This leads to use of either a sheet electron beam, or a multiplicity of pencil beams. The circuits used with these beams may be either resonant or non- resonant. Devices using a multiplicity of beams in con- junction with resonant circuits have been proposed in recent years,l but no devices representing a reduction to practice have been reported. The adjacent mode inter- ference problem was probably responsible to a great degree for holding back development of this idea. In the work to be described this difficulty has been greatly minimized by utilizing a mode with the maximum fre- quency separation between it and adjacent modes.

Multiple- or sheet-beam devices utilizing nonresonant circuits have also been proposed for obtaining high power output.Zs3 However, while these studies investigated the principles of traveling-wave klystron interaction, neither resulted in a practical demonstration of a significant amount of power output.

At the present time, the most successful practical approach to the problem of increased power output is to combine the outputs from an even number of tubes by the use of hybrid couplers. The MBK constitutes an alter- nate and significantly different approach to ‘(harnessing”

Vitesse, Gauthier-Villars, Paris, France, pp. 721-730; 1951. R. ,Warnecke and P. Guenard, “Tubes A Modulation de

IRE, vol. 40, pp. 308-315; March, 1952. T. G. Mihran, “The duplex traveling wave klystron,” PROC.

Newby, “The periodically-loaded traveling-wave multiple beam V. A. Heathcote, P. A. Lindsay, J. Barraclough, and J. R.

klystron,” Proc. IEE, vol. 105B, pp. 952-965; May, 1958.

248 IRE’ TRANXACTIONX ON ELECTRON DEVICE‘#

beams together which is much less complex and bulky than the hybrid scheme. Furthermore, it is much less sensitive to disruption of operation in the event of beam failure.

In the following sections, the basic principles governing the choice of the operating mode and the design of MBK circuits are outlined. This is followed by a description of experiments in which an order-of-magnitude increase in power output was obtained by the use of MBK principles.

11. CIRCUITS FOR MULTIPLE-BEAM RESONATORS

Essential components of any circuit suitable for an MBK are the interaction gaps. In general, these gaps are small compared with wavelength and can be treated as capacitors. The geometrical configuration of the gaps and tunnels is largely determined once the power level per gap and the frequency are specified. Thus, in a specific MBK circuit design these capacitors are fixed, and the circuit problem reduces to determining the optimum method for coupling power-producing interactions from an array of these gaps. Factors to be considered arc the location of the gaps relative to the field maxima of a given mode in the resonator and the frequency separation of the various possible resonant modes.

A multimode resonator may be derived from a con- ventional klystron cavity by tightly coupling a number of such cavities in an extended array. The resulting structure may be considered as a length of waveguide periodically loaded with capacitors as shown in Fig. 1. The modes of this circuit can be obtained from a knowl- edge of the propagation characteristic of the loaded waveguide as a transmission line. In Fig. 2 a theoreti- cal frequency-phase shift (w-p) curve of a capacitor- loaded guide is shown. Also shown are frequency points corresponding to the resonances of an 8-section shorted length of such a loaded line. Fig. 2 also shows the spatial voltage patterns of the respective modes. The experi- mental data of Fig. 2 were obtained using a circuit de- signed for the 4-beam MBK tests to be described later. The theoretical curve was plotted using the approach outlined in the Appendix.

From the spatial mode plot of Fig. 2, it is evident that only the mode a t p D = T has equal maximum voltages a t all gap locations. This is a desirable character- istic if electron beams passing through all gaps arc to be used most effectively. Reference to the w - p curve, however, shows that in the vicinity of a-mode the modes tend to be crowded together in frequency and therefore adjacent mode interference would be a serious problem. Similar crowding exists for zero-mode (not shown), Le., the frequency a t which the waveguides just cut on. From the w-8 diagram of Fig. 2 it is evident that the mode a t p D = ~ / 2 represents the optimum mode configuration for two reasons. First, maximum frequency separation of adjacent modes exists here. Second, all voltage maxima are of equal amplitude. In this particular mode alter- nate capacitors arc located a t voltage nulls, hence elec-

LCOUPLING IRIS

Fig. I-Distributed beam resonator

I

5 z8001 ?c 8 700

PROPAGATION CHARACTERISTIC

I

800

5 > E w 3

THEORETICAL CURVE

8 700 o EXPERIMENTAL RESONANCE

600;; - - 4 k

PHASE SHIFT PER SECTION

‘ BEAM LOCATIONS MODE) t t

2-Characteristics of a periodically-loaded circuit.

tron beams are not used at these positions, that is, alter- nate gaps would be dummy capacit,ors having the same capacitance as beam gaps.

In the ,8 D = ~/2-mode, each electron beam is situated in precisely the same electrical environment as a beam in the single-mode resonant cavity of a conventional klystron. This situation allows each of the beams of an MBK to deliver energy to the resonator fields fully as efficiently as in the single-beam case. Furthermore, in the MBK the individual interactions or outputs of all beams arc phase-locked because the fields are very tightly coupled throughout the resonator. This advantage of strapping does not exist in a system where outputs of isolated klystrons are combined externally.

The use of a multipoint drive and a particular mode in a multimode resonator has an important advantage with regard to klystron harmonic output. Current bunches arriving at a klystron output gap are extremely rich in harmonic content. In a single resonant cavity, for harmonic power to be developed it is only necessary to have im- pedance present at the harmonic frequencies. In an

1962 Boyd, et al.: The Multiple-Beam Klystron 249

MBK, however, in addition to this impedance requirement, there also is a phase requirement that must be satisfied to obtain harmonic output. In this case harmonic genera- tion requires not only that the harmonic frequency occur in a higher pass band a t a resonance of the loaded circuit, but also that the phase sbift per section be harmonically related t,o that of the fundamental input signal. Minimum harmonic output will be achieved by designing so that harmonic frequencies occur in the stop bands of the circuit. This minimum output should be significantly less than that of a single-beam klystron of the same total fundamental output.

Tuning and coupling to multimode resonators are somewhat more complicated than in single-mode reso- nators. For efficient operation klystron output circuits are usually loaded externally to an impedance level that permits a voltage swing approximately equal to the beam voltage to be developed across the interaction gap. To load the ~/2-mode of an n-section MBK resonator to the same Q as the prototype klystron requires that the load be coupled to the resonator n times as tightly as in the single-cavity case. In practice it was found that iris coupling such as that shown in Fig. 1 could be used to obtain any degree of loading. With iris coupling it was found, incidentally, that as the loading is increased it is necessary to move the position of the iris in the di- rection of the gaps to avoid distortion of the fields a t the gap locations. For the limited number of beams used in our low-frequency test set-ups, loop coupling or probe coupling was used. For instance, in the 10-beam experi- menta1 model MBK, a probe coupler section was used in the center of the resonator. This was done to permit symmetrical trim-tuning by moving two shorts in the end resonators simultaneously.

Mechanical tuning of a multimode resonator can also be achieved by varying other cavity dimensions. End wall tuning is restricted to narrow frequency ranges because the mode pattern tends to become distorted. Side wall tuning of each cavity section is required for wide tuning ranges or large numbers of beams.

111. TEST SETUP AND RESULTS

A. Description of Equipment

For flexibility and convenience the basic principles of the MBK were tested not by designing a single vacuum envelope encompassing a multiple of beams, but by using conventiona1 commercially available external circuit kly- strons inserted into multiwavelength waveguides. This set-up has proven to be very useful in that modifications can be made in the RF circuitry without disturbing the dynamics of the individual electron beams.

The tube chosen for this purpose was the 3K3000LQ klystron. This is a 3-gap klystron, and operates in the frequency range from 610 t o 985 Mc. Although these tubes are rated a t 2-kw RF output, they were operated con- servatively at the reduced beam voltage of 7 kv and a power output of 1.0 kw.

Our ultimate goal was to show that 10 such tubes properly interconnected could operate as an MBK and deliver 10 kw of RF power to a single load, representing an order-of-magnitude increase in power capability. To do this, we first constructed a 4-beam MBK, and when this was found to operate satisfactorily, a 10-beam model was built. A cross-sectional view of the 4-beam model is shown in Fig. 3(b). The prototype cross-section is shown in Fig. 3(a). The tubes are detachable from the circuitry just outside the ceramic insulators, shown in dotted cross section. Midway between tubes, loading capacitors are added in order to obtain %/%-mode opera- tion, as described in Section 11. Tuning is accomplished by movable shorts located a t both ends of the individual waveguides.

Fig. 3-Cross sections of single-tube and four-tube setups.

The 10-beam model was arranged and tuned in much the same manner as the 4-beam model, except that the circuits were horseshoe-shaped rather than linear. A photograph of assembled 10-beam MBK is shown in Fig. 4. The high voltage seals of the tubes can be seen beneath the table. The electron beams travel vertically upward through the three waveguide circuits and are collected by water-cooled collectors. These collectors are insulated from the tube bodies, making it possible to monitor individual body current on each tube. The output coupling cavity utilizes probe coupling and is located directly below the double-slug tuner shown in the upper part of the photograph. Input coupling is obtained by a loop in the corresponding cavity of the input deck. The corresponding cavity of the center deck is not used for coupling but is loaded by a capacitor equal to the average tube gap capacity to preserve n-/2-mode. Tuning is ac- complished by means of double tuning pistons with gear mechanisms placed in the cavities diametrically opposite the coupling cavities. Thus the 12-unit circuit is composed of 10 units enclosing tubes, one coupling unit, and one tuning mechanism unit.

In the 4-tube setup, the tubes were arranged in a straight line, and a single large diameter coil provided the magnetic field necessary for beam focusing. This can be seen in cross section in Fig. 3(b). Two large diameter plates form the top and bottom of the magnetic circuit,

250 IRE TRANSACTIONS ON ELECTRON DEVICES M a y

and the flux is returned by a series of twenty-four vertical iron bars spanning the distance between them. A photo- graph of this arrangement is shown in Fig. 5 for the 4-beam MBK. Exactly the same magnetic circuit was used for the 10-beam MBK, except that the tubes threaded through a series of holes near the periphery of the structure rather than being placed along a diameter as shown in Fig. 5. The 3Q3000LQ klystrons are designed for use with an additional focus coil which is placed between the electron gun and the magnetic flux input plate, as shown in Fig. 3(a). In the multiple-beam set-up, such individual coils were used in each klystron. It should be pointed out

Fig. 4-10-beam MBK amplifier without magnetic field.

Fig. 5-4beam MBK amplifier.

that in an MBK where all beams are located within one vacuum envelope, use of these coils would not be con- venient. By proper beam design they couId be eliminated.

B. Test Results

1) Power Output, Efi iency and Gain The performance of the 10-beam MBK is most readily

evaluated by comparing it with single-tube performance. This comparison is made in Fig. 6 where RF power output is plotted as a function of RF power input. The upper curve represents 10-beam MBK performance; the lower curve was obtained by taking data on a single-beam klystron and subsequentIy scaling the effciency at a drive level of G watts to the average of the individual efficiencies of the 10 tubes used in the 10-beam MBK as determined by data supplied by the manufacturer. In taking these curves, the middle or penultimate circuits were optimized at each drive level. The shapes of the two curves are very similar except that the 10-beam MBK gives 10 times more power output at a drive level 10 times that of the corresponding prototype point. Note that this is done not by raising beam voltage but by increasing current, or in effect, perveance by a factor of 10.

2) Bandwidth The bandwidth of the 10-beam MBK is compared with

single-tube bandwidth in Fig. 7. Two measurements of bandwidth were made. In one case all circuits were optimized a t midband and a curve of power output vs frequency was taken with no retuning. These curves are shown as the dashed lines marked “over-all.” The curves shown as solid lines were taken by keeping drive level constant and retuning both input and middle circuits. The output circuit remained fixed in this test, hence this curve is indicative of output circuit bandwdith. If ad- ditional stagger-tuned bunching cavities were available, presumably the dashed curves could be made to approach the solid curves as in conventional broad-band klystrons. ,v Comparing MBK and single-tube bandwidths, we see the output circuit of the MBK performs as well as the single-klystron output circuit, both bandwidths being about 3.5 Me out of 720 Mc. This represents a bandwidth of 0.5 per cent which is proper for a beam whose dc beam resistance is about 20,000 ohms working into an output circuit with an R/& of about 100. Thus the MBK output circuit operates over the same bandwidth as the proto- type klystron output circuit. The reason for the apparent improvement in the over-all bandwidth of the MBK, shown by the dashed curves, is not clear. It probably can be attributed to different relative placement of the resonant frequencies of the input, middle, and output circ,uits used in the two cases.

3) Miscellaneous Tests

a) Beam failures: The MBK is relatively insensitive to beam failures. If a beam is lost, the power output drops

1962 Boyd, et al.: The Multiple-Beam Klystron 251

/ V, = 7070 VOLTS

/ / r. = 3.40 AMPS f = 718 MC.

q / /

7 / / I / // I / /

- / TEN BEAM MBK

V, = 7070 VOLTS r. = 3.40 AMPS f = 718 MC. / /

I 10 20 30 40 50 60 70 80

.I ' ' 1

' O ! I 2 3 4 5 6 7 8 R F DRIVE POWER IN WATTS

Fig. 6-Multiple-beam and single-beam klystron power output.

I 720 717 718 719

2 0 Y

+- 2 ,Q -I w n

-2

-3

-4 I I I

SINGLE KLYSTRON \ -5

724 725 726 727 728 FREQUENCY I N MEGACYCLES

Fig. 7-Multiple-beam and single-beam klystron bandwidth.

by an amount roughly equal to the power the lost beam was contributing. There is some reaction on the per- formance of the other tubes, but it is not major. For instance, dropping out one beam reduces dc input power by 10 per cent, hence output power would be expected to drop off by a similar factor. When this test was made, output power dropped by 14 per cent. The additional drop-off over that expected is due to a combination of factors, such as reduced effective R/Q, lower circuit efficiency, and improper output impedance. This last factor can be eliminated by reoptimizing the output impedance after a tube is dropped out. A test was made in which three beams of the 10-beam MBK were biased nearly off such that there was a 28 per cent reduction of dc input power. Output power dropped 40 per cent under these circumstances. By reoptimizing the output load impedance, this drop-off was reduced to 35 per cent, which compares favorably with the expected drop-off of some- thing greater than 28 per cent.

b) Harmonic output: Second and third harmonic out- put of the IO-beam MBK relative to the fundamental was measured and compared with harmonic content in a single klystron operating in the conventional manner. Second harmonic output was found to be about 25 db below the fundamental in either case. Third harmonic content of the single klystron was found to be 45 db below the fundamental, while in the 10-beam MBK it was a t least 51 db below the fundamental. This repre- sented the lower limit of the sensitivity of the measuring equipment. While no attempt was made t o design these circuits for minimum harmonic interaction, this limited test tends to confirm the conclusion that the harmonic content of an MBK can be lower than that of a single- beam klystron.

c) Operation at other than n/2-mode: Although n/2- mode represents a very favorable mode of operation as discussed earlier, an MBK is capable of delivering RF energy at any of the resonant frequencies of its circuits. In general there will be a loss of efficiency associated with modes other than .rr/2-mode since the output RF voltages seen by individual tubes are not longer optimum. How- ever, operation is possible a t reduced power level and efficiency. In our particular test set-up, arcing at the loading capacitors further restricted the power output because they were not designed with this type of operation in mind. However, operation was achieved in the ~ / 4 - mode at a beam voltage of 5.5 kv with an efficiency of 27.5 per cent indicating that such operation is possible.

d) Broadbanding of the output circuit: Preliminary cold tests were conducted on the possibility of broadbanding the output circuit of an MBK by critically coupling the .rr/2-mode resonance to an external resonance. It was found that it is possible to achieve a fairly uniform double- humped response at all tube positions under these circum- stances. In one test the 10-unit circuit was driven a t one end by a ridge waveguide coupler which possessed an

252 IRE TRANSACTIONS ON ELECTRON DEVICES M a y

internal resonance. The results are shown in Fig. 8. It is evident in this case that broadbanding action is achieved at all 10 tube positions. No hot tests were carried out with this configuration.

FREQUENCY SWEEP IOMC

TUBE# A t T U B E # Lit

10 5.lmc 5 50mc

4 4.9

3 4.9

7 5.1 2 4 8

6 4 9 I 4.6

Fig. 8-Frequency response at tube locations (resonant coupler).

IV. CONCLUSIONS

A multiple-beam klystron has been demomtrated which promises to extend the power capabilit,y of micro- wave tubes embodying klystron-type interactions by a t least one order of magnitude. As the name implies, this device has the gain, efficiency, bandwidth, and stability characteristic of conventionar klystrons. However, by using a multiplicity of electron beams and a carefully selected mode of resonance, the power output of the MBK can be made to exceed that of the prototype klystron by a factor equal to the number of beams used without sacrifice of any klystron advantages.

The MBX is fundamentally a low-voltage, high-current approach to superpower. Therefore, problems such as high-voltage power supplies, voltage breakdown, and X-radiation protection are simplified. In an MBK each beam operates under the same conditions as the beam of the prototype klystron. Consequently, the electrical and thermal stresses impose no new limits 011 available ma- terials. The basic unit of the MBK is a conventional klystron beam; therefore, an MBK design requires no advance over the present state of the art for the design of the single-beam prototype unit. Furthermore, any future advances in klystron technology may be incorpo- rated into MBK designs.

The ultimate limits of power capability of an MBK are not precisely known. The first limit to be encountered will probably be due to adjacent mode interference. This limit will depend upon circuit design, individual beam impedance, and the ability to control mode excita- tion in the input circuit by selective loading or strapping. The best present estimate for typical existing klystrons is that the limitation will occur somewhere between 40 and 100 beams. A more fundamental limitation will be encountered at power levels where the circuit losses are

comparable to the unit power being developed. At present the latter is not a practical limit.

The multiple-beam klystron offers advantages which are not present in paralleling arrangements of individual klystrons. Such a system is necessarily complex because of external circuitry. Furthermore, individual tube failures are usually accompanied by more than proportion- ate losses because of an unbalance in the coupling system.

APPENDIX

PROPAGATION CHARACTERISTICS OF

LOADED TRANSMISSION SYSTEMS

The method for analyzing loaded transmission lines combines the theories of transmission lines and filter circuits. The type of circuit which is most easily handled is a uniform transmission line or waveguide which is periodically loaded with lumped admittance. The propa- gation constant for such a system is given by4

where

p D is the resultant phase shift per section eo is the phase shift in length D of unloaded guide Y o is the characteristic admittance of unloaded guide b is the loading susceptance.

Assuming the waveguide dimensions are known, it is only necessary to make a single resonance measurement to evaluate the propagation cha’racteristic for all fre- quencies where the loading may be considered lumped. For the particular case of the low-frequency MBK circuit, the propagation constant is given by

where

D = 5.2’‘

ea = - F = 2.77f,,,F W D C

f c = cut-off frequency of unloaded waveguide.

This curve is plotted along with experimental points in Fig. 2.

Inc., New York, N. Y., p. 183; 1950. J. C. Slater, “Microwave Electronics,” D. Van Nostrand Go.,