Enhanced leakage-current pulse-inducedchanges in impatt oscillators

R.J. Gutmann and J.M. Borrego

Indexing terms: Impatt diodes, Leakage current, Microwave oscillators, Solid-state microwave circuits

Abstract: Two types of stables changes in c.w. X-band impatt oscillators have been observed after exposureto pulses of enhanced leakage current. The first change is indicated by a change in r.f. power and frequencywhile maintaining a 'clean' r.f. spectrum, and is attributed to parasitic circuit-induced instabilities (withchanges in oscillator characteristics thought to be influenced by changes in harmonic phasing). The secondchange is indicated by a modulated r.f. spectrum and oscillations in bias voltage after irradiation, and isclearly attributed to bias-circuit oscillations. These conclusions are supported by experimental results withSi and GaAs diodes in a resonant cap waveguide oscillator circuit. Enhanced leakage-current pulses up to5 mA in amplitude were obtained using 10 MeV electron pulses from a linear accelerator. The r.f. impedancewas adjusted by varying the sliding short-circuit position, while the bias-circuit impedance was varied in-dependently by inserting various length cables between an /?Cbias network and the diode.

1 Introduction

Principal effects of enhanced leakage current pulses onimpatt oscillators have been well documented using trans-ient ionising radiation to generate the leakage current.1'2

During the radiation pulse a leakage current is created whichis proportional to the radiation dose rate and the sum ofthe volume of the depletion region and the region fromwhich carriers can reach the depletion region before recom-bining. The effect of this additional leakage current is apremature build up of the avalanche current, resulting inreduce r.f. power and an increase in frequency of oscilla-tion.2'3 These r.f. parameters change gradually with in-creasing leakage current in free-running oscillators, insubstantial quantitative agreement with calculations basedon a large-signal-device model.2 After the enhanced leakagecurrent pulse (i.e. the ionising radiation pulse in this experi-mental technique) the diode and oscillator return to pre-pulse conditions.

However, radiation-induced changes have been reportedin some impatt oscillators in which prepulse oscillator con-ditions are not observed after pulsed irradiation.4"6 Ander-son4 reported a ceasing of r.f. power in low-power (5 mW)silicon impatt oscillators during, or immediately after, a15 ms pulse of a combined neutron and gamma flux of3x 1015n/cm2s and 5x 106 rad/s, respectively. Fullrecovery of r.f. power only occurred after the bias wasreduced to zero and then increased to the preirradiationlevel. At approximately twice these radiation levels irre-versible diode failure occurred.5 Borrego et al.6 reportedsimilar results with medium power (400 mW) Si and GaAsimpatt oscillators and 100 ns pulses of 10 MeV electronshaving a dose rate of 4 x 109 rads/s. Immediately after theirradiation pulse, the r.f. power was reduced and thefrequency of oscillation changed compared to preirradia-tion oscillator characteristics. This new condition wasstable as long as the bias remained fixed, but the pre-irradiation oscillator power and frequency were obtainedif the bias was reduced and then increased to the pre-irradiation level. These effects were dependent oncavity tuning and r.f. loading, although the dependencewas not explored in detail. Diode failure was occasionallyobserved in GaAs Schottky devices.

Paper T28, received 10th August 1976Prof. Gutmann and Dr. Borrego are with the Electrical & SystemsEngineering Department, Rensselaer Polytechnic Institute, Troy,New York 12181, NY, USA

These results indicate that pulses of ionising radiationcan induce changes in r.f. operation of impatt oscillatorsthat persist after the radiation pulse. In many cases, alower-power stable oscillating condition is obtained afterirradiation (without irreversible diode changes), but occa-sionally diode failure is observed. This paper reports similarexperimental results of changes in impatt oscillators in-duced by enhanced leakage-current pulses using transientionising radiation under controlled r.f. and bias-circuitconditions, and relates this data to known impatt insta-bilities.

In the Section 2 of the paper impatt instability mech-anisms are reviewed. In Section 3 the experimental pro-cedure is described, followed by the experimental resultsin Section 4. In the last Section, these results are inter-preted using the impatt instability mechanisms describedin Section 2.

2 Impatt instability mechanisms

Since the aforementioned experimental observations ofionising-radiation induced changes in impatt oscillators,4"6

there has been much progress in understanding instabilitiesin impatt oscillators that could be invoked in a phenomeno-logical explanation. We have delineated the followingpotential instability mechanisms for consideration: biascircuit,7 parametric,8 harmonic related9 and parasiticcircuit induced.10 This Section reviews the most relevantinformation concerning these mechanisms.

Bias-circuit instabilities are caused by an r.f. voltage-induced negative resistance due to the nonlinear depen-dence of the ionisation rate on electric field. In a classicpaper, Brackett7 developed conditions for stabilisationof impatt oscillators against bias-circuit instabilities. Thenecessary condition relates the bias-circuit impedance inthe 1 to 50 MHz frequency range (maximum frequencyof instability dependent on diode parameters and r.f.circuit Q) to the diode impedance at these baseband fre-quencies (the latter is dependent on diode and r.f. cir-cuit parameters). These conditions have been used success-fully by many workers, although GaAs diodes are moredifficult to stabilise as the rectification-induced negativeresistance is larger than in Si devices.7 The presence ofbias-circuit instabilities can be readily observed in ther.f. spectrum of a c.w. oscillator.11

MICROWA VES, OPTICS AND ACOUSTICS, JANUARY 1977, Vol. 1, No. 2 75

Parametric instabilities are caused by the nonlinearinductive behaviour of the avalanche process, a result ofthe exponential dependence of the avalanche current onthe ionisation integral.8 Parametric instabilities can resultin power saturation at unexpectedly low power levels,unconversion of low-frequency noise and spurious r.f.responses. Schroeder12 has demonstrated that the generalstability criterion developed by Hines8 works well forX-band GaAs impatt oscillators. Since the spurious fre-quency is usually not known a priori (except in the de-generate case), the circuit-impedance requirements neces-sary to prevent all possible parametric instabilities can besevere. Parametric instabilities are indicated experimentallyby two responses whose frequencies sum to the funda-mental or by a subharmonic response.

Harmonic-related instabilities are a result of the har-monic sensitivity of large-signal impatt operation. Harmonicgeneration is a result of both the exponential dependenceof the avalanche current on the ionisation integral and thenonlinear dependence of the ionisation rate on electricfield.9 By considering operation of an impatt oscillator attwo harmonically related frequencies, Mouthaan9 hasshown that two modes of operation are possible at a givenfundamental frequency of oscillation (differing in r.f.power and start-oscillation bias current). These two modescorrespond to either of two phases that an oscillation ata harmonic may have with respect to the fundamentalfrequency of oscillation. Brackett13 has shown that thestability criteria for tuned-harmonic operation (high-powermode of Mouthaan) can be quite restrictive and difficultto satisfy in typical impatt oscillators. Since harmonicsare expected in any large-signal oscillator circuit, thepresence of harmonic-related instabilities cannot be de-tected directly by a spectrum analyser (the phase of theharmonic compared to the fundamental must be consideredand the criterion of Brackett13. and Mouthaan9 evaluated).

Parasitic-circuit-induced instabilities are caused by alooping r.f. circuit impedance near the fundamental fre-quency of oscillation. Kurokawa10 has shown that twostable operating points may be possible, depending onthe relationship of the large-signal-device impedance andthe looping-circuit-impedance locus. If two stable oscillat-ing conditions are possible, the actual condition is depen-dent on the oscillator's history; i.e. the oscillator stateis not uniquely defined by diode and circuit parameters.Similar behavior is possible in any electronic oscillatorwith suitable circuit parameters, as described in the classicbook by Slater.14 R.F. cavity and large-signal-device im-

VARIABUE LENGTHCABLE TO DDDE

| BIAS| NETWORK

TO CRO

Fig. 1 Impatt diode bias circuit

76

pedance measurements are needed to confirm the possi-bility of parasitic-circuit-induced instabilities.

Either of these instability mechanisms could be triggeredby the large change in operating conditions caused by theenhanced leakage current pulses.1'2 The purpose of thisresearch was to isolate the instability mechanisms that hadbeen previously observed, and to develop guidelines toprevent the resultant enhanced leakage-current pulse-induced changes in impatt oscillators.4"6

3 Experimental procedures

The microwave cavity used in testing the impatt diodeswas a tunable waveguide cavity consisting mainly of thediode mounted in a resonant cap structure15 with anadjustable micrometer sliding short circuit. A slide-screwtuner was placed after the diode mount for fine tuning,followed by a broadband isolator to complete the oscilla-tor circuitry. The Qext of the cavity varied with differentdiodes, and had a typical value of 100. The cavity was notevacuated during this radiation testing, as the effect ofair ionisation on such impatt oscillators is insignificantat dose rates below 3 x 109 rad/s and well documentedat higher dose rates.1

Following the oscillator circuitry are standard wave-guide couplers to allow r.f. power monitoring by means ofa thermistor, a crystal detector, and a spectrum analyser.The thermistor is used for calibrating the crystal detectorand for measuring the power output of the impatt oscilla-tor between radiation pulses, while the crystal detector isused for monitoring the r.f. power during the ionising-radiation pulse. The spectrum analyser is used for obser-ving the r.f. spectrum before and after pulses of radiation,and for measuring oscillator pushing characteristics (fre-quency shifts as the r.f. power is changed with bias current).A cavity wavemeter is included for precision c.w. frequencymeasurements.

Before irradiation, each diode oscillator was carefullytuned so that the frequency and power changed smoothlywith bias current below power saturation. This was assuredby monitoring the r.f. spectrum with the spectrum analyserand the r.f. power with the power meter as the diodecurrent was varied. All the r.f. circuit tuning described inthis paper was accomplished by varying the position of thesliding short circuit, the slide-screw tuner being unnecessaryfor the diodes reported here.

The circuit used for biasing the impatt diode is shownin Fig. 1. In the 1 to 50 MHz frequency range of typicalbias-circuit oscillations,7 the bias circuit impedance at thediode terminals was varied by inserting different lengthsof 93 SI cable between the RC network shown and thediode (a 60 pF bypass capacitance is located near thediode). The RC network was properly radiation shieldedfor all cable lengths used (0-3 m, 0-10 ft). The voltageacross the diode terminals was monitored with a properlyterminated cable connected as shown in this Figure. Thisbias configuration provided a nearly constant bias currentduring irradiation, with changes less than 4 mA under alltest conditions. The power supply and necessary testequipment (such as c.r.o.s, d.v.m.s and ammeters) werelocated in an instrumentation room adjacent to the radia-tion chamber.

The experimental measurements with enhanced leakagecurrent were performed at the US Air Force CambridgeResearch Laboratories linear accelerator. The acceleratorwas used to generate an electron beam with a 100 ns

MICROWAVES, OPTICS AND ACOUSTICS, JANUARY 1977, Vol. 1, No. 2

pulse duration and an energy of lOMeV. By varying thepeak injector current, dose rates from 2 x 108 to 8 x 109

rad/s were achieved with only a small shift in the beamenergy. The radiation dose level inside the diode cavitywas determined using thermoluminescent dosimetry. Thecalibration between dose rate and leakage current is de-scribed in detail in Reference 2. For X-band impatts,109 rad/s corresponds to between 0-3-0-9 mA of en-hanced leakage current, depending on device character-istics.

Half-watt X-band devices with the characteristics givenin Table 1 were irradiated under these operating proce-dures. To prevent thermal problems, the diodes wereoperated below 350 mW output. Previous experiencewith these and similar devices1 indicates that they arerepresentative of commercially available flat-profile single-drift X-band impatts.

Table 1 . Impatt diode characteristics

Diode number Diode type Cbr

301BZ2A41B10

Si-double epiGaAs-diffusedGaAs-Schottky

V786449

PF0-761-260-83

mA93

100105

The enhanced leakage-current testing focused on theeffect of bias-circuit impedance and r.f. circuitry on en-hanced leakage-current pulse-induced changes in impattoscillators. Bias-circuit oscillations were suspect in thepreviously reported radiation-induced changes4"6 due tothe long cable length that is often used between the diodeand bias network (RC network in Fig. 1) to ensure properradiation shielding. Fig. 2 illustrates the effect of this

BIAS NETWORKA. BIAS EQUIVALENT

— Bias lmpedance(4 ft cable) * —— Bias Impedance

(I ft cable). '

CIRCUIT- Brackett Go As

IMPATT ImpedanceAll frequencies

in MHz

B.Zg WITH FREQUENCY AND CABLE LENGTH

AS PARAMETERS(Brackett GaAs IMPATTsuperimposed)

Fig. 2 Impatt diode bias impedance as a function of frequency

cable length on the bias circuit impedance at the diodeterminals. The RC network of Fig. 1 is considered as a336 ft [(930 ft + 93 ft) in parallel with 500 ft] resistivetermination, connected to the 60 pF bypass capacitorat the diode by the cable. Superimposed is the small-signallow-frequency impedance of a 6 GHz GaAs impatt ascalculated by Brackett.7 Using Brackett's criterion, bias-circuit instabilities are possible for cable lengths greaterthan 1 -2 m (4 ft).

The r.f. circuit is particularly important in parametricharmonic-related and parasitic induced instabilities asdiscussed in Section 2. In this work, the post and capdimensions15 were selected to maximise power and effic-iency of each diode, and were not changed during enhancedleakage-current pulse testing. The slide-screw tuner was notused for fine tuning in this programme to simplify thecavity circuitry: i.e. only the sliding short circuitwas adjusted before and during enhanced leakage testingto maximise r.f. performance (while ensuring smoothtuning as described earlier) and to vary r.f. circuit para-meters.

4 Experimental results

In this Section data obtained with GaAs Schottky impatt1B10 under enhanced leakage-current pulse testing willbe presented in detail, and results obtained with otherimpatts will be briefly summarised.

The GaAs Schottky diode (1B10) was tuned with thesliding short circuit close to the diode while satisfying thesmooth-tuning requirement described in Section 3 (result-and short-circuit position 0-750 cm, with an additional0-950 cm from the waveguide short flange to the centreplane of the resonant cap diode structure). As shown inFig. 3, enhanced leakage-current pulse-induced changesdid not occur at the highest dose rate used (7 x 109 rad/scorresponding to approximately 5 mA enhanced leakagefor 1B10). The r.f. power decreases almost to zero duringthe radiation pulse, owing to the effective leakage currentgenerated by the ionising readiation and cavity ionisationat this high dose rate. The apparent delay between pulsesis merely due to the differences in triggering in dual beam

1I

7 X 10 RADS/SEC/DIV

90 mW/DIV

IOO nSEC/DIV

T

97 X 10 RADS/SEC/DIV

90 mW/DIV

Fig. 3 R.F. power of GaAs-Schottky diode tuned without en-hanced leakage-current pulse-induced changes

MICROWAVES, OPTICS AND ACOUSTICS, JANUARY 1977, Vol. l,No. 2 77

oscilloscopes. In fact, no measurable delay exists betweenthe radiation pulse and onset of the diode response. Thereis a small time delay before the r.f. amplitude builds up tothe steady-state value after the radiation pulse, but ther.f. power clearly returns to its preirradiation value (asdoes the r.f. frequency, bias voltage and other electricalcharacteristics). This lack of enhanced leakage-currentpulse-induced changes was observed for all cable lengthsinserted in the bias circuit at this sliding short-circuitposition, indicating that a variety of bias impedancesin the 1—50 MHz frequency range does not cause anyenhanced leakage-current induced changes.

After extending the sliding short circuit by approxi-mately one-half guide wavelength, almost identical pre-irradiation r.f. tuning was observed. However, after theirradiation pulse the r.f. power reduced from 280 mW to120 mW while the frequency increased from 9-55 to 9-95GHz and the bias voltage increased by 0-7 V. The r.f. powerchange is shown in Fig. 4a, and the bias-voltage changein Fig. 5a. The slight increase in bias voltage after the

7 X I09 RADS/SEC/DIV

T 90mW/DIV

lOOnSEC/DIV

IT XIO9 RADS/SEC/DIV

T9O mW/DIV

Fig. 4 R.F. power of GaAs-Schottky diode tuned with enhancedleakage-current pulse-induced changes

2 V/DIV

lOOnSEC/DIV

2 V/DIV

Fig. 5 Bias voltage of GaAs Schottky tuned with enhancedleakage-current pulse-induced changes

radiation pulse (compared to the preirradiation value)is consistent with the lower amplitude of r.f. oscillation.With a l-5m (5 ft) coaxial cable inserted between thediode and the bias network, 25 MHz bias-circuit oscillationswere observed in addition to the power and frequencychange. The bias oscillations are apparent in the r.f. powerphotograph shown in Fig. 4b, as well as in the diode voltagephotographs in Fig. 5b. In addition, the classic bias-oscilla-tion spectrum was observed after radiation using a spectrumanalyser.11 The spectrum was 'clean' otherwise and was notnoisy with, or without, the bias-circuit oscillations. Inaddition, no spurious responses could be observed with thespectrum analyser after irradiation.

When these enhanced leakage-current pulse-inducedchanges did occur, the original r.f. characteristics couldagain be obtained by lowering the bias current and thenincreasing the current to the prepulse value. In particular,there were no diode failures as sometimes reported.5'6 Thefollowing points should be emphasised:

(i) the r.f. power and frequency tuned smoothly withbias current while maintaining a clean spectrum beforeirradiation at all test conditions used.

(ii) these enhanced leakage-current pulse-induced changescould be reproduced without difficulty. These changes inoscillator characteristics were obtained in diode 1B10with preirradiation r.f. powers greater than 200 mW andenhanced leakage-current pulses greater than 1 -4 mA.

Similar r.f. and bias-circuit tuning and irradiation testingwere performed with the GaAs diffused diode (Z2A4)and the Si diffused diode (301B).16 Enhanced leakage-current pulse-induced changes (r.f. power reduction andfrequency change) could only be observed if the diode biascurrent was set only slightly below the value for powersaturation and/or spectrum deterioration, and higher en-hanced leakage currents (~5mA) were generated. TheGaAs diffused diode was more prone to these enhancedleakage-current pulse-induced changes than the Si diffuseddiode, although the susceptibility was much greater in theGaAs Schottky impatt than either of the diffused diodes.As with diode 1B10, the enhanced leakage-current pulse-induced changes only occurred if the sliding short circuitwas placed an extra half-guide wavelength from the diode-resonant cap structure. Bias-circuit oscillations were notobserved for these diodes, the extra cable length havingno effect on the enhanced leakage-current pulse-inducedchanges with diodes Z2A4 and 301B.

5 Discussion and conclusions

Two different enhanced leakage-current pulse-inducedchanges have been observed which remain stable after theradiation pulse. The first is indicated by a change in r.f.power and frequency while maintaining a 'clean' r.f. spec-trum and is probably caused by parasitic circuit-inducedinstabilities (with changes in oscillator characteristicsthought to be influenced by changes in harmonic phas-ing).10 The second is indicated by a modulated r.f. spec-trum and oscillations in bias voltage after irradiation, andis clearly attributed to bias-circuit oscillations.7 In thisSection these conclusions are discussed and used in ex-plaining previously reported radiation-induced changesin impatt oscillators.4"6

The first enhanced leakage-current pulse-induced changeclearly depends on r.f. circuit parameters (observed withthe sliding short circuit one guide wavelength from diode-resonant cap structure but not with the sliding short

78 MICROWAVES, OPTICS AND ACOUSTICS, JANUARY 1977, Vol. l.No. 2

circuit one-half guide wavelength away), clearly does notdepend on bias-circuit impedance at baseband frequencies(large variations in bias impedance introduced with var-iable length cable between the RC bias network and thediode (without effect), and was most prevalent in theCaAs Schottky diode (occurring at operating power levelsand enhanced leakage currents as low as 200 mW and14 mA, respectively). Since this enhanced leakage-currentpulse-induced change does not depend upon bias-currentimpedance in the 1 -50 MHz frequency range, and there isan absence of bias-current oscillations, bias-circuit insta-bilities are not a contributing factor.

In addition, parametric instabilities are not likely to beinvolved since the only circuit change required to inducethe instability is to increase the sliding short-circuit position.For the lower-frequency oscillations typically observedwith parametric instabilities, the WR-90 waveguide is cutoff (fc = 6-56 GHz). Therefore, the impedance introducedat the diode-resonant cap plane by the sliding short willbe independent of position at frequencies below 6-56 GHz.Furthermore, there were no parasitic oscillations observedwith the spectrum analyser after the enhanced leakage-current pulse-induced change, so the possibility of a para-metrie instability with one frequency high enough to propa-gate in a WR-90 guide (therefore being affected by slidingshort-circuit position) is unlikely.

Harmonic-related instabilities are a potential causeof the first type of enhanced leakage-current pulse-inducedinstability as two stable operating modes are clearly possiblewith a strong second-, or higher-order, harmonic present.9'17

While the increase in sliding short-circuit position to obtainenhanced leakage-current pulse-induced changes wasapproximately one-half guide wavelength at the frequencyof oscillation (and, therefore, the circuit impedance atresonance is nearly unchanged although Qext is slightlylarger), the increase is not a multiple of one-half wave-lengths at the harmonic frequencies due to waveguidedispersion. Therefore, the harmonic impedances at theresonant cap-diode plane from the sliding short are differ-ent in the two positions, potentially causing one operatingpoint to be more prone to change following the enhancedleakage-current pulse. However, the diode power depen-dence on the bias current was identical at both shortpositions, indicating that either

(i) impatt operation has a small dependence on har-monic impedance, or

(ii) the harmonics are trapped by package and mountparasitics (slight frequency-pushing differences with theshort-circuit position attributed to the different qext).In either case, harmonic phasing would not be expectedto depend on the sliding short-circuit position. Thus,while harmonic-induced instabilities are possible in suchimpatt oscillators, they do not appear to have been aprimary cause of the enhanced leakage-current inducedchanges observed in this experiment.

Parasitic-induced instabilities are possible even thoughthe oscillator was adjusted to satisfy the smooth tuningrequirement before the enhanced leakage-current pulse.The device negative resistance decreases in magnitude andthe device capacitance decreases during the pulse to valuesimpossible to simulate without enhanced leakage-currentoperation.2 Therefore, a new stable operating conditioncan exist after the enhanced leakage-current pulse eventhough smooth oscillator tuning is achieved in the absenceof enhanced leakage current.

This explanation is supported by cavity-impedanceMICROWA VES, OPTICS AND ACOUSTICS, JANUAR Y1977, Vol. 1,

measurements taken with a coaxial line replacing thediode holder, the centre conductor of the line extendingacross to the centre of the disc.18 These data are presentedin Fig. 6 in the RX plane for both sliding short-circuitpositions of the diode oscillator 1B10. These data wereobtained using a network analyser without calibration ofcircuit losses and connector mismatch. The looping impe-dance locus as discussed by Kurokawa10 is apparent,supporting the parasitic-induced-instability hypothesis.Moreover, the looping is faster with frequency for the2-880 cm sliding short-circuit position (recall that afterthe enhanced leakage-current pulse the frequency changedfrom 9-55 to 9-95 GHz). From this Figure, it may beexpected that the oscillator would change from 9-55 GHzto perhaps 10-25 GHz after enhanced leakage-currentpulsing with the sliding short circuit at 0-750 cm, butprobably the large-signal diode impedance was not appro-priate.

10

0-9

O-8

0 7

0-6

0 5

0-4

0-3

0-2

01

All frequencies in GHz

Sliding short 0-750 cm

Sliding short 2-880 cm

Sliding short

0-750 cm

2-880 cm

Oscillation

Pre- pulse

9-55

9-55

frequency(GHz)

After pulse

9-55

9-95

0-1 0-2 0-3 0-4 0-5 0-6

Fig. 6 Normalised oscillator circuit impedance as measuredwith packaged diode removed (for diode 1B10)

Although the triggering of the first type of enhancedleakage-current pulse-induced change has been attributedto the looping r.f. circuit impedance, the difference in r.f.power and frequency after enhanced current pulsing andthe lack of change with a slower looping impedance locuscan be influenced by harmonic phase changes. The level ofoscillation remains quite high during the enhanced leakage-current pulse even though the r.f. power is significantlyreduced; most of the power reduction is a result of the re-duced phase angle from the avalanche zone.2-3 Thus,diode harmonic phasing is expected to be most affected

No. 2 79

by the leakage-current pulse, with only relatively smallchanges in harmonic amplitude. As described by Mouth-aan12 lower power can result from such a harmonic phasechange. While we have no firm evidence of the significanceof the harmonic phase, this explanation appears reasonable.

The second enhanced leakage-current pulse-inducedchange is due to bias-circuit oscillations as shown by ther.f. spectrum, the r.f. detected pulse and the bias voltagepulse. Moreover, the bias impedance at the diode terminalsagrees with the Brackett7 model for bias-circuit oscilla-tions, and GaAs diodes are more prone to such oscillations.Although no diode failures were obtained in this pro-gramme, it is well known that bias-circuit oscillations canresult in device failure.

We believe that these two mechanisms, parasitic circuit-induced instabilities causing the first type of radiation-induced change, and bias-circuit oscillations causing thesecond, were probably present in previously reportedexperimental results.4"6 Stable oscillator changes withoutdiode failure4'6 were probably a result of parasitic-circuitinstabilities, principally a result of a looping circuit-impe-dance locus. The power and frequency change are thoughtto be affected by changes in harmonic phasing. In addition,we believe that bias-circuit oscillations are the cause of thepreviously reported diode failures under transient ionisingradiation.s>6 GaAs diodes are more sensitive to this in-stability than Si owing to the larger value of r.f. inducedd.c. negative resistance,7 and failure was observed in onlya GaAs device by Borrego et al~.6 using cable lengths ofl-5m (5 ft) between bias network and diode. With lowpower Si device failure as reported by Anderson,4 the biascircuit was not described, but the radiation chamberdescription indicates that up to 9 m (30 ft) of cable waslikely used between diode and bias network. Such a longbias cable enhances the possibility of bias circuit oscilla-tions with even low power Si impatts.

6 Acknowledgement

This work was supported by the National Science Found-ation under Grant ENG-75-03424.

7 References

1 BORREGO, J.M., GUTMANN, R.J., COTTRELL, P.E., andGHANDHI, S.K.: 'Transient ionizing radiation effects in IMP ATTOscillators', IEEE Trans., 1972, NS-19, pp. 328-334

2 COTTRELL, P.E., BORREGO, J.M., and GUTMANN, R.J.:'IMPATT oscillators with enhanced leakage current', Solid-StateElectron., 1975,18, pp. 1-12

3 MISAWA, T.: 'Satuation current and large-signal operation ofa read diode', ibid. 1970,13, pp. 1369-1374

4 ANDERSON, W.A.: 'A radiation pulse interrupts IMPATTdiode oscillation', Proc. IEEE, 1970, 58, pp. 807

5 ANDERSON, W.A.: 'The performance of IMPATT diodesunder the influence of high-intensity gamma and neutron radia-tion', ibid., 1969,57, pp. 1441 -1442

6 BORREGO, J.M., GUTMANN, R.J., COTTRELL, P.E., andGHANDHI, S.K.: 'Aftereffects in IMPATT oscillators withtransient ionizing radiation', ibid. 1973, 61, pp. 675—676

7 BRACKETT, C.A.: 'The elimination of tuning-induced burnoutand bias circuit oscillations in IMPATT oscillators', Bell Syst.Tech. J. 1972, 52, pp. 271-306

8 HINES, M.E.: 'Large signal noise, frequency conversion, andparametric oscillations in IMPATT diode networks', Proc. IEEE,1972, 60, pp. 1534-1548

9 MOUTHAAN, K.: 'Non-linear characteristics and two-frequencyoperation of the avalanche transit time oscillator', Phillips Res.Rep., 1970, pp. 33-61

10 KUROKAWA, K.: 'Some basic characteristics of broadbandnegative resistance oscillator circuits', Bell Sys. Tech. J., 1969,48,1937-1955

11 Hewlett Packard Application Note 935, 'Microwave powergeneration and amplification using IMPATT diodes', sectionHID

12 SCHROEDER, W.E.: 'Spurious parametric oscillations in IM-PATT diode circuits', Bell Syst. Tech. J., 1974, 53, pp. 1187-1210

13 BRACKETT, C.A.: 'Characterization of second harmonic effectsin IMPATT diodes', ibid, 1970, 49, pp. 1777-1810

14 SLATER, J.C.: 'Microwave electronics', (Van Nostrand, 1950)chap.9

15 GROVES, I.S., and LEWIS, D.E.: 'Resonant -cap structures forIMPATT Diodes', Electron. Lett., 1972, 8, pp. 98-99

16 GUTMANN, R.J., and BORREGO, J.M.: 'Experimental evalua-tion of aftereffects in IMPATT oscillators with transient ionizingradiation', Paper PC-1, present at the IEEE Conference onNuclear and Space Radiation Effects, 1976

17 BRACKETT, C.A.: 'Circuit effects in second-harmonic tuningof IMPATT diodes', IEEE Trans., 1971, ED-18, pp. 147-150

18 EISENHART, R.L., and KHAN, P.J.: 'Theoretical and experi-mental analysis of a waveguide mounting structure', ibid.,1971,MTT-19, pp. 706-719

80 MICROWAVES, OPTICS AND ACOUSTICS, JANUARY 1977, Vol. 1, No. 2