1956/57, No. 12 361

A LOW-NOISE KLYSTRON WITH HIGH POWER OUTPUT

by R. A. LA: PLANTE *) and G. A. ESPERSEN *).

'.. Among the tubes used for gimerating a continuous high power at high frequencies the velocity-modulation valve or klystron occupies an important place. In this article the authors describean investigation on the noise of these tube;. The principle caus~ was shown to be situated in thetuning mechanism, A non-tunable tube, constructed as a result of these experiments, shows avery low ~oise-level and has also some other favourable properties.

Velocity-modulation valves or klystrons havealready been dealt with in two previous articles inthis Review 1). For a description of the working prin-ciples we refer therefore to these papers. In thesecond of them construction details were given ofsome tubes developed for wavelengths between15 and 3 cm (frequencies between 2000 and 10000Mc/s). In this article we will describe some recentinvestigations on the subject of klystrons, moreexpecially with regard to the causes of noise in thesetubes. We will also describe the construction of aklystron that has been developed as a corollary tothese investigations.

Fig. I shows a simplified cross-section of a 3 cmklystron developed in Eindhoven some years ago.The construction is basically the same as that ofthe tubes described in the second article mentionedunder 1), the principle dif-ference being that the out-put cavity resonator is notcoupled to the waveguideby a loop but by a gap inthe resonator wall, the gap.heing connected to a wave-guide flange via a taperedwaveguide. The vacuumseal is obtained with a micawindow of thickness 50

Construction of a 3 .em klystron

621.373.423

Noise of a klystronAs with all oscillators, the current~ and voltages

'generated in a klystron are not purely sinusoidalbut are subject to variations both in amplitude andphase. This also applies, therefore, to the electro-magnetic field in the waveguide coupled to the tube.In many cases-this deviation from a purely sinusoi-dal waveform manifests itself as noise, which res-tricts the usefulness of the tube. For instance, noise'

5cm

9 10 11 2 qOBbb

Fig. 1.Cross-section of a tunable klystron for a wavelength of 3 cm.Various details, includingthe tuning mechanism, have been omitted. 1 modulating cavity, 2 output cavity, 3 modu- .lating gap, 4 inductor gap, 5 cathode, 6 focusing electrode ("grid"), 7 diaphragms, 8collector,9 cooling jacket, 10 feedback hole, 11 drift tube, 12 coupling gap, 13mica window,14 tapered waveguide, 15 glass envelope, 16 waveguide flange.

'. .

. microns. The tube has an Lcathode 2) which, even a~the high emission densityrequired, has a useful lifeof more than 1000 hours. This. tube delivers a poweroutput of 200 W, at an anode voltage of 8800 V,and has an efficiency of 13%.,

*) Philips Laboratorics, Irvington-on-Hudson, N.Y., .U.S.A.1) F. M. Penning, Velocity-modulation valves, Philips tech.

Rev. 8, 214-224, 1946, and B. B. van Iperen, Velocity-modulation valves for 100 to 1000 watts continuous output,Philips tech. Rev. 13, 209-222, 1951/52.

2) H. J. Lemmens, M. J. Jansen and R. Loosjes, A newthermionic cathode for heavy loads, Philips tech. Rev. 11,341-350, 1949/50.

in klystrons used in beam transmitters reduces therange of the transmitter and may ulso limit thenumber of non-interfering channels permissible ina given frequency band 3).

In the ideal case of a completely noise-free signalthe field at every point in the waveguide wouldvary as a function of time according to A cos' wot.

3) See C. Ducot, Beam .transmitters with double frequencymodulation, Philips tech. Rev. 17, 317-327, 1955/56.

362 PHILIPS TECHNICAL REVIEW VOLUME 18

For variations III amplitude A and in phase wotthe formula for the field may be written as:

A ~l+ a(t)~ cos ~wot + p(t)(,

where a(t) and p(t) are random time functions.It has been established that the noise is mainly

due to phase variations and that a(t) is always verymuch smaller than unity.

Experiments, which we shall describe in thispaper, have shown that the noise is predominantlyduc to microphony, to which tubes of this construc-tion are particularly sensitive because of the pre-sence of the two diaphragms (7). The extremely smallvibrations of these diaphragms give rise to slightvariations in the tuning frequency of both cavityresonators, resulting in phase variations and hencefrequency varintions of the output signal.

Another cause of noise, viz. fluctuations in theapplied voltages, especially the anode voltage, canbe kept very small by using adequately stahi-lized voltages.

Noise measurementsFig. 2 shows a block diagram of the system used

for measuring the noise of klystrons. The tube Kunder investigation operates into a waveguide Wterminated by a matched load L. Ignoring theslight losses in the waveguide, the output power ofthe klystron can be ascertained by measuring thepower converted into heat in this load.

Two samples of the klystron output are "taken"from the waveguide Wand directed into separatechannels by two directional couplers 4) DCI and DC2•

The signal supplied by DCI is fed via an adjus-

4) See A. E. Pannenborg, A measuring arrangement for wave-guides, Philips tech. Rev. 12, 15-24, 1950/51.

table attenuator 4), Atl' to a crystal detector Dl'Since the signal voltage appearing on this detectoris very small, the output of the rectifier is propor-tional to the square of the signal voltage and there-fore proportional to the output power of the klys-tron. This D.C. voltage, which contains a small A.C.component due to amplitude variations in the signal,can be read from meter MI'The signal from DC2 is fed via a second adjustable

attenuator, At2' to a discriminator; this consists ofa slightly detuncd transmission cavity, C, to whicha crystal detector, D2' is coupled. The D.C. voltagefrom this detector can be read from meter M2•

As the discriminator detects fluctuations in ampli-tude as well as in frequency, the D.C. voltage outputof this channel contains two A .C. components, oneproportional to amplitude variations and the otherto frequency variations in the klystron signal(there are also some small interaction terms present).When the D.C. outputs of both detectors are equal,the A.C. components in each output are also equal

Q08b7

Fig. 2. System for measuring the noisespectrum of klystrons. J( klystron, Ppower supply, W waveguidc, L load,PIvT power meter, DCI and DC2 direo-tional couplers, Atl and At2 adjustableattenuators, C cavity resonator, Dl andD2 crystal detectors, MI and M2 D.C.voltmeters, TI" balancing transformer,Am amplifier, HA harmonic analyser.

III SO far as they are due to amplitude variations.The A.C. outputs from Dl and D2 are applied to apush-pull-transformer Tr in such a way that thevoltage obtained III the secondary is mainlycaused by frequency variations. This voltage isamplified in an audio amplifier Am and analysedby a harmonic analyser HA. From this voltage wecan calculate the power spectrum of the klystron,the width of this spectrum being a measure for thenorse.

A total of four klystrons of the type shown infig. 1 were analysed in this way, with the resultsshown in fig. 3. Here we plot the power against thefrequency, expressed in W per cis. A measure for

1956/57, No. 12 LOW-NOISE KLYSTRON 363

the width of these spectra is the so-called standarddeviation

-l/J x2Sdx

y - /. -'-=00,----J Sdxo

For the curves shown in fig.. 3 the value of. y liesbetween 3.3 and 12 kc/so

The fact that the noise is mainly due to themicrophony arising from the diaphragms wasplainly demonstrated by the following experiment.·After the spectrum of one of the tubes' had beenrecorded, the diaphragms and tuner were cast inplaster of Paris, mld the spectrum again recorded.The spectra before and after this treatment arerepresented as curves a and b in fig. 4. The rootmean square deviation was reduced in this wayfrom 5.3 to 1.9 kc/so Thè plaster of Paris was thenremoved and the tube was cast in plastic, afterwhich the spectrum was again recorded. The resultis shown as curve c in fig..4. The root mean squaredcviation was now found to be only 200 els.

5

- 1

-30 -20 -10 o 10 2Q JOkeis_X 91162

Fig. 3. Spectrum of four tunable klystrons, recorded with thelayout shown in fig. 2. The power S, expressed in W per cis, isset out along the ordinate; x is the frequency deviation fromthe central frequency of the spectrum.

Construction of a low-noise tube

The main source of the noise having been deter-mined, various methods were considered for reduc-ing microphony. Firstly attempts were made tobuild a more robust tuner to hold the diaphragms.more firmly (see fig. 8), but the results showed verylittle improvement, the r.m.s, deviation ofthe tubesso construeted being of the order of 2 kc/s.· As a10\" noise level was so.important that the tuneabili-ty ,ofthe tube could if necessary: be sacrificed for this

purpose, it was furthermore .considered to cast thetubes in plastic, after first tuning them to a speci-fied frequency, It would also be possible to maketubes with only one diaphragm, i.e. with one of thecavities tuned to a fixed frequency, and finally atube might be designed-with no diaphragms at all.

c

175xl0-5

-10 -5 o 5 TOkels91163

_x

Fig. 4. Spectrum of a tunable klystron, a in normal state, bafter casting. in plaster of Paris, "cafter casting in plastic.

In that case a variable feedback coupling would benecessary to adjust the tube to the conditions fordelivering its rated power: An important objectionagainst the latter method lies in the fact that avariable feedback coupling would 'entail a compli- _cated meehanism, which would very likely alsocause microphony and, consequently, noise.

It was therefore finally decided to develop a tubewithout diaphragms and with a fixed feedbackcoupling. In spite of the absence of a diaphragm thèoutput cavity can nevertheless be tuned. A sim-plified cross-section of this tube is 'shown in fig. 5.The modulating cavity 1 is permanently tuned tothe 'desired resonant frequency. The output cavity2 is adjusted, while the tube is oscillating, by ap-plying to the collector- a large force, denoted by thearrow F. This force is large enough to strain. the'cavity wall beyond its elastic limit. When thisoccurs, t~e cavity can be deformed to the rightamount, at which it remains permanently set. A s inthe tube shown in fig. 1, the feedback coupling iseffected via an opening 'ID in t~e wal} between thetwo cavities.

364 0 PHILIPS TECHNICAL REVIEW VOLUME 18

Characteristics of the low-noise tubeThe cathode of the tube described here may be

either an L cathode 2) or an impregnated tungstentype 5). The life of these cathodes is .better than1000 hours and, compared with tungsten, tungsten-thorium or tantalum cathodes, the heater powerthey require is low; in the present case it is only7 W6). '

The electrode 6 (fig. 5) with which the electronbeam is focused, can be used for amplitude modu-lation and is therefore generally called the "grid".

As described in the articles quoted in 1), a klys-tron can be made .to operate in different modes, i.e.,with different electron transit times between themodulating and output gaps. The tube under dis-cussion can operate in three different modes, denotedby A, Band C, which correspond 0 respectively totransit times of 23/4,21/4 and 13/4periods. The powerdelivered in these modes is respectively 5, 33 and200 watts. Each klystron is adjusted to operate in 'only one of these modes and will not perform prop"erly in other modes.

10 11 2 Q0870

Fig. 5. Cross-section of a low-noise klystron. The figures have the same meaning asin fig. 1. F force required to tune the output cavity 2.

By varying the "grid" voltage it is also possible to_change the frequency betweeri certain limits. Atnormal operating conditions this electrode has thesame potential as the cathode,The diode characteristic of ,the tube, i.e. the

colleètor current plotted as a function of collectorvoltage, is shown in fig; 6. As for every diode, thischaracteristic ca~' be represented by the equation:

i,= AVc3/2•

In this tube the constant A (the pcrveance) IS

approximately,·0.25 X ~0-6 A/V3/2• The tube is

liquid-cooled and the maximum rate of flow re-o quired for water cooling is about 1/2 gallon perminute.

5) R. Levi, New dispenser type thermionie cathode, J. appl.Phys. 24, 233,.1953.This cathode will shortly be discussed.in full detail in this Review.

8) Oxide cathodes, which require an even lower heater power,cannot be usèd in these tubes because their useful life atthe high emission needed would be only a few hours.

2aamAr-----------------------------~

Q0871

Fig. 6. Diode charaèteristic of a klystron as shown in fig. 5.le collector current, Vc collector voltage.

1956/57, No. 12 LOW-NOISE KLYSTRON 365

Tunable klystrons, as illustrated in fig. 1, do notdeliver rated power output over their whole tuningrange. This can be seen from the typical powerversus frequency characteristic of a tunable tube,at two different modes, shown infig. 7. The low-noise

200~r-----------------------'-----,

100

50

o8 11M c]«

------.f 91164

109

Fig. 7. Power output of a tunable klystron as a function offrequency for modes Band C (electron tcansit time 21/4and P/4 periods respectively).

tube, due to its method of manufacture, gives al-ways its rated power. The frequency where thistakes place, can (as indeed the most favourablefrequency of a tunable klystron) be fixed at anydesired frequency with a maximum error of 10 Mc/s.The absence of a tuner makes the construction of

this tube much simpler than that of tunable types.

This is illustrated in fig. 8, in which are shown anolder tunable klystron, a tube with "ruggedized"tuning mechanism and a non-tunable low-noisetube. Because of the remarkable precision withwhich the new tube can be made for any specifiedfrequency, it is possible to switch over or inter-change a series of such tubes for quickly altering atransmitter frequency. This would often be easierthan tuning a normal klystron.

The noise of these tubes is so low that we wereunable to measure the power spectrum with thesystem described. It appeared that the waveguidesin the measuring system were now more micro-phonic than the klystrons under investigation.

The eliminatien of the diaphragms and tuneroffers some advantages in addition to the sub-stantial reduction of noise. In the first place itlowers the cost of the tube. Further, it allows moreaccurate alignment of the electrodes, while theabsence of diaphragms leads to a somewhat betterheat distribution.

Fig. 9 shows the results of some measurements ofthe variation of power output and frequency withcollector voltage. The curves relate to modes BandC (mode A has been omitted because we are interest-ed mainly in higher power outputs). The power isplotted on a normalized scale, obtained by dividingthe power output by the maximum power deliveredin each mode. This maximum power is 33 W formode Band 200 W for mode C, at anode voltagesof 4.35 and 8.85 kV respectively. The collectorcurrent and the efficiency do not differ much fromthose of a tunable tube.

The fact that the frequency variation curve re-verses its direction is not in accordance with thetheory of the electrical phenomena. The assump-tion that thermal effects in the tube are responsible

Fig. 8. From left to right: an early type of tunable klystron, a tunable klystron withruggedized tuning mechanism and a low-noise, non-tunable klystron.

90782

366 PHILIPS TECHNICAL REVIEW VOLUME 18

for-this phenomenon was made and this was veri-fied by two experiments. Temperature variationsinfluencing the static curves in fig. 9 do not 'appearif, the anode voltage is rapidly varied. For thisreason the tube operating in mode B was "swept"in anode voltage at a rate of 60 cis and the frequency

1.0

0.9p

r0.8

0.7

0.6

0.5

0.4.

0.3

0.2

0.1

0

B

generate frequency-modulation in this way, butthis phenomenon can be used for electronic tuningwithin a certain small frequency range. From fig. 11it can be seen that the maximum frequency varia-tion is much higher for mode C 'than for mode B.

The temperature of the coolant has some influence

t 10Mcfs

8 IJf6

r4.

20

-2-4

-6-8

8 9 tOkVVc 91165

65

Fig. 9. Power output P and frequency variation IJf of a low-noisenon-tunable klystron asshownin fig. 5, mode B on the left, mode C on the right. The power is plotted on a normal-ized scale. The maximum power for B is 33Wand for C 200W. The fully-drawn curvesare the results of static measurements. The dot-dash curve is the result of measurementsin mode B with rapidly varying collector voltage.

variations measured through the mode. The resultis represented by the dot-dash curve in fig. 9. Areversal of the frequency curve does not occur now.

In a second experiment a small AC voltage witha higher frequency was superimposed on the anodevoltage, thus producing frequency modulation.From the spectrum of the output signal the fre-quency deviation per volt variation in collectorpotential was calculated. The result is shown infig. 10, together with the power curve shownalready in fig. 9. Whereas it would follow from thestatic frequency curve that, at a collector potenrialof about 3.9 kV, small volta:ge variations would giverise to no variations in frequency, i.e. the modula-'tion sensitivity would be zero, experiment showedthat the modulation sensitivity never becomes zero.In fact the frequency variation per volt follows quiteclosely the slope of the dot-dash curve. in fig. 9.The tube can also be modulated in amplirude by

means of rhé gridvoltage. Fig. 11 shows, for modesBand C, the power in terms of the grid voltage.The latter' voltage has also some influence on the'frequency. The frequency Variation has alsó heendepicted iri. fig. 11. As this variation is principallydue to changes in temperature, it is not possible to

t.O

0.9

P

10.8

0.7

0.6

0.5

0,1.

0.3

02r0'1

I

4.

20

la

~~,IoSkV

-- .... V 91/06

Fig. 10. Frequency variation per V variation in 'collectorvolt-age, dfld Vc, as a function of the collector voltage in mode B.The power output P on a normalized scale is also shown;this curve corresponds to that shown on the left in fig. 9.

, ,

1956/57, No. 12 LOW-NOISE KLYSTRON 367

4-0Mcls 1.0

0.9IJ(

0.8 I 0.8P 30 P .

iD.! J .,0.6 !1 0.6 C

0.5 20 0.5

0.4- OA

0.3 0.310

0.2

Fig. ll. Power output f and frequency variation .dj as functions of grid voltage. Thepower is set out on a normalized scale. The left curve refers to mode B, maximum power33 W; the right curve refers to mode C, maximum power 200W.

90876

Fig. 12. Rieke diagrams for modes B (left)' and C (right). The lines sr constant power(in W) are fully-drawn; the dashed curves are lines of constant frequency deviation (inMc/s). '

on the frequency. We found a frequency variationof 0.2 Mcfs per degree variation in uemperature.Immediately after switching on, the frequencydrift is approximately 2 Mcfs per second for a con-

-100 -200V-Vg

stant coolant temperature. After the warming-upperiod the effect of ambient temperature upon fre-quency is extremely slight.A change in the load impedance affects the output

power and, also causes a frequency shift. A goodimpression of the tube's behaviour at different ~oad

impedances is given by a Rieke diagram, i.e, a polardiagram is which the modulus and 'the argumentof the reflection coefficient of the load coupled tothe waveguide are plotted,"). In this diagram lines

°0~-------~ro~0~------'2~0~V~--- ..... 1'9 91167

of constant power and lines of constant frequencyshift are shown. A typical Rieke diagram for, atube operating in mode B is shown in fig. 12a, and a

7) See e.g. D. R. Hamilton, J. K. Knipp and J. B. HornerKuper, Klystrons and microwave triodes, RadiatiçnLaboratory Series, No. Î, McGraw-Hill, NewYork 194.8,Chapter 15.

368 PHILIPS TECHNICAL REVIEW VOLUME 18

ULTRASONIC MACHINING

Rieke diagram for a tube operating in Ulode C isshown in fig. 12b.If the load impedance is varied such that the

modulus of the reflection coefficient remains 0.2while the argument passes through all angles be-tween 0° and 360°, a certain frequency' variation is'produced which is known as the "pulling figure" ofthe tube. Fig. 12 shows that for modes Band C thepulling figures are about 4 and ç- Mcls respectively.

Some tubes, namely magnetrons and some.klystrons can be damaged when they are used witha load which is not that which gives the maximumpower output. In many cases the window whichensures the vacuum-tight sealing is damaged. In

this ,respect it is of some interest to mention thatthe tube described here has worked in mode -Bwithout difficulties with a load having a voltage-standing-wave ratio of 7. In mode C voltage-stand-

, ing-wave ratios up to 5 ~ave heen used. '

Summary. The flexible diaphragms with which most tunableklystrons are equipped are the most important source of noisein these tubes. Owing to the demand, in. a certain application,for a tube with a'much lower noise level, a non-tunable klystronhas been developed in which these diaphragms are absent.This tube has a very much lowernoise level. It can be used inthree modes, delivering respectively 5 W, 33 W and 200 W.The tube can be frequency and amplitude modulated byvariation of the collector voltage; amplitude modulation canalso be effected by the grid voltage. Variation of the gridvoltage also permits, via thermal effects, a certain limitedtuning of the tube.

11. OPERATING CONDITIONS AND PERFORMANCE OF ULTRASONIC DRILLS

by E. A. NEPPIRAS *) and R. D. FOSKETT *). 534.321.9 :621.95

Machining by means of ultrasonic vibrations is finding widening fields of application.Part I of this article gave an introduetion. to this technique and described some of the ultrasonicdrills developed by the Mullard Research Laboratories. The present (concluding) article con-siders in some detail the factors which are important with regard to cutting speeds, accuracyand surface finish. A brief comparison is made with other special techniques, notably that ofelectro-erosion.

increase in the static load (L), cutting rates (pene-tration in inches per minute) always increase fromzero almost linearly at first to. a maximum at someoptimum load and then decrease, the curve ofcutting rate against L becoming asymptotic to theL axis (figures la and b). Increase of oscillatoryamplitude ç at fixed frequency has the effect ofincreasing' cutting' speeds and at the same timeshifting the .optimum load to higher values (seefig. la):1t seems obvious that at least two separate,effects contribute to' the observed ,results - adamping effect which becomes important- at highvalues of load, superimposed on a linear law ofincrea'se which predominates at low pressures. Thisbecomes more apparent when we study the curvesrelating to tools of varying areas (fig. lb). The'

*) Mullard Research Laboratories, Salfords, Surrey, England. d I1) E.A.NeppirasandR.D.Foskett,Ultrasonicm!Îchining,1. optimum loa increases with the too area, the. Technique and equipment, Philips tech. Rev. 18,325-334" peak becomes' broader, and for tools of very1956/57, hereafter referred to as 1. Erratum to article I: 'large areas the damping effect becomes unimpor-in footnote 15), the subscripts to k should read ).j2 and.not y/2. tant over the range of .loads used. For small-area

Cutting speedThe "machining r,ates attainable with ultrasonic

reciprocating tools are affected by many factors,apart from those quantities fixed by the construe-tional features of the vibrator :_ oscillatory am-plitude, operating frequency. and sta~ic loading -which have been briefly .considered in Part I ofthis article 1). The other factors include: the materialof the tool; its shape; its area; the depth of the cut;the physical properties of the ,work material; andthe abrasive properties including the hardness,grain dimensions, the nature of the suspension me-dium and the concentration ofthe suspension.Drilling tests have been carried out under varying

operating conditions, in an attempt to assess theeffect of all these variables. It was found that with