amplification with anomalous avalanche diodes

8/12/2019 Amplification With Anomalous Avalanche Diodes

1/8

956 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. MTT- 18, NO. 11, NOVEMBER 1970

generation efficiency (6) nor the generation factor (18) ACKNOWLEDGMENT

can be assumed to be constant. The author wishes to thank D. M. Snider, currentlylhus it becomes necessary to be more speeific. Using

the amplifier in Section IV as an example, we could saywith the hl. I.T. Lincoln Laboratory, who suggested the

that its maximum conversion efficiency with a 5-percentproblem and provided the experimental data.

efficient driver along the trajectory defined by Fig. 4 is

8.5 percent. This information is, at least, specific. [1]

Perhaps the best description would be to give the

maximum de-to-RF conversion efficiency that could be p]obtained on a constant output-power trajectory as a

function of the conversion efficiency of the driver.[3]

REFERENCES

D. M. Snider, .4 one-watt CL\T high-efficiency X-band avalanche-diode amplifier, IEEE Trans. Microwave Theory Tech., this issue,pp. 963-967.M. E. Hines, Negative-resistance diode power amplification, IEEE Trans. Electron Devices, vol. ED-1 7, pp. 1-8, January 1970.F. A. Brand, The numbers game, Microwave J., vol. 13, p. 10,July 1970.

Amplification with AnomalousAvalanche Diodes

H. JOHN PRAGER, b5WBER, IEEE, KERN K. N. CHANG, AND S. WEISBROD

AbstractThe experiment described in this paper is performed

for the purpose of obtaining additional information that might help to

elucidate the operating principle of the anomalous mode.The diodes under investigation are of a silicon p+-n-n+ mesa

structure. The breakdown voltage is 160 volts; the punch-through

voltage is 60 volts.

The test setup is similar to a conventional reflection-type ampli-

fier with the diode mounted in a coaxial-cavity-like circuit. Thereverse-bias puke drives the diode to a low avalanche current levelat which anomalous operation would normally commence. In thiscase, however, the tuning and loading of the cavity and its associatedstubs is intentionally arranged to minimize free-runnin g oscillations.Any residual RF power is reduced to the level of weak instabilitieswithout any strong harmonics or subharmonic. Only in the presence

of an input driving power with a frequency at or near the resonancefrequency of the system, does an amplified power of the same fre-quency appear at the output port.

Starting from the residual level the output power increases pro-

portionally with the input power until at higher drives a saturationlevel is reached. In this particular case, operating at 410 MHz, satura-

tion is reached at 18.5 watts with a maximum power gain of 12 dB anda bandwidth of 26 MHz. This output k approximately 20 to 30 dBhigher in amplitude than any observable peak in the residual broad-band frequency spectrum Operating the diode as a free-running~~hmd~~ o scillator at the same current level and frequency, 18.8 Watts

of RF power is obtained.The tiItial and steady-state power levels are substantially the

same for both amplifier and oscillator. A plausible evaluation of theseresults can be obtained following van der Pols differential equation

for a forced negative-resistance oscillator.

Manuscript received April 12, 1970. This work was supported bythe Air Force Avionics Laboratory, Wright-Patterson AFB, Ohio,LI rider AF Contract F33615-68-C-1688, Project 4460.

The authors are with RCA Laboratories, Princeton, N. J. 08540.

introduction

s

I NCE THE discovery of the anomalous avalanche

diode in early 1967 [1], a great deal of effort has

been expended by several laboratories [2 ] [4] not

only to explain the anomaly but also to harness the effect

by developing better diodes and better associated cir-cuitry. While most of these activities concentrated on

the generation of microwave power, relatively little was

done toward the amplification of microwave power. The

first cursory results of amplifier operation were reported

in August 1967 [5]. This was followed by the paper of

Hoefflinger et al., in May 1969 [6], and that of Dienst

et al. [7] who presented the results of their work on a

self-pumped parametric amplifier with the anomalous

avalanche diode. The applications for microwave

amplification are, however, sufficiently important to

warrant still further investigation.

REVIEW OF OSCILLATING OPERATION OF

ANOMALOUS AVALANCHE DIODES

To better highlight the similarities as well as the

differences between oscillator and amplifier perfor-

mance, we shall briefly review some of the pertinent

features of the anomalous diodes as oscillators.

The first successful oscillation of the anomalous

avalanche diode was accomplished in a coaxial circuit as

shown in Fig. 1. The diode was placed in a General

Radio (GR) Tee fitted with proper bypass capacitors,

while stub tuners were used for tuning to the desired fre-


2/8


3/8

958 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1970

,. .

GOLO WIRE

F

CONTACTp+

N

N+

k MEsA DIAMETER 4Fig. 3. Microphotograph of angle-lapped junction.

S LO PE = 0. 4S 0

uULSEGEN,I

Ei00M,SIGGEN

X-TAi v kIRCPL IODBnSCOPEINPUT -

vULSEGE +STUBI--- -*r--f

m

r POWERBo Lo MET ER M ETER3008400 MHzBANDPASSb

-TAL

oSCOPE

OUTPUT,,

Fig. 5. Test setup for avalanche-diode amplifier.

OFF condition. l~hile considerable diode current is al-

ready flowing, the characteristic increase in current is

still missing. llith the input signal OFF there is as yet no

output power present. The right-hand side of Fig. 6

shows the voltage and current when both the diode pulse

and the input signal are turned ON. The diode current

now shows a small but recognizable increase, and the

voltage shows a corresponding dip typical of the

anomalous mode. To prove that these changes are not

due to a switching phenomenon between two unstable

circuit conditions, we gradually varied the input power

and measured the gradual change in output power,

I I I I I I I I II I I I I I I I I II 10 100

I diode current, and diode voltage. The results have been

VOLTAGE (V)

Fig. 4. Capacitance versus reverse voltage,

oscillator, but although the frequency is the same 400

MHz, the stub tuning and matching conditions as well

as the pulse-operating conditions are changed, to meet

the requirements expected from the amplifier. The input

signal is provided by a pulse-modulated RF generator

operating at the same 400-MHz frequency as the circuit

of the diode. The fact that the frequency of the inputcircuit is exactly the same as the resonance of the diode

circuit is an important feature of this amplifier. It

clearly cliff ers in this respect from a self-pumped para-

metric amplifier where there are three separate fre-

quencies for the signal, the pump, and the idler.

The voltage and current waveshapes of the amplifier

are shown in Fig. 6. The left-hand side of the figure

shows the voltage and current of the diode in the

quiescent amplifier condition; i.e., the diode is ON and at

operating levels, but the input signal is zero or in the

plotted in Fig. 7. The voltage aid current did not change

over as wide a range as they did for the oscillator, and

accordingly the efficiency of the amplifier is as yet well

below the efficiency of the oscillator.

To gain a better insight into the working of the

amplifier, we shall now consider the oscilloscope tracings

of Fig. 8. These are measurements of RF power as seen

by the crystal in the detection circuit following the out-

put of the circulator. In Fig. 8(a) the diode is ON, and

the signal is OFF; this means that while the diode was

drawing current, as is shown in the left-hand side of

Fig. 6, practically no RF power was detectable. Ac-

tually with higher scope sensitivity a small RF power ismeasured, and we shall refer to it as (residual 77 RF

power. The minimizing of this residual RF power during

the tuning and adjustment procedure of the amplifier

setup is a very important feature of its performance. Al-

though this procedure is empirical, it is guided by a

basic plan. Briefly it begins by tuning the diode under

self-oscillating conditions to the desired operating fre-

quency of the amplifier. Stub 4 and, to a lesser degree,

stub 3 are the most effective controls in determining

this desired resonance frequency. with the frequency in


4/8

PRAGER et al.: POWER AMPL1F1CATION WITH ANOMALOUS AVALANCHE DIODES959

VOLTAGE

DIODE ON SIGNAL OFF

CURRENT

Fig. 6. Voltage and current pulse waveshapes of amplifier forsignal ON and OFF conditions. Scales: voltage, .50 V/div; current,1 A/div; time, 1 ps/div.

18

- 16

POWER OUTPUT - 14 *

2.5 xn.

tia=a

; 2.4 u \a \z3 \c1

\x

2.L

182I I

183 184 185VOLTAGE, VOLTS

Fig. 7. Pulse current and pulse voltage acrossdiode as function of power output.

the diode circuit thus established, stubs 1, 2, and again

to a lesser degree, stub 3, are now adjusted until the

oscillatory RF power output is minimized and brought

to the lowest attainable residual level. While the first

adjustment sets the frequency, the second adjustment

changes the dynamic load line until a stable amplifier

condition is reached.

Both of these adjustments are required to meet two

necessary conditions of the amplifier, but they are not

sufficient yet, because a further requirement is that the

power output vary gradually (if not linearly) with input.

Thus the next step in the adjustment procedure is to

verify that this continuity requirement has been met or

to modify the stub settings until adequate control has

been accomplished.

Turning our attention to Fig. 8(b), we note that it is

taken with the diode OFF and the signal ON. Since we are

dealing with a reflection-type amplifier, we have of

course the reflected portion of the input power appearing

(a)

(c)

Fig. 8. RF output voltage of amplifier forthree distinct operating conditions.

at the output port. (We hasten, however, to point out

that for the purpose of measuring the power gain of theamplifier, the input power was measured at the input

port and into a 50-ohm load. The picture shown here is

only for the purpose of illustrating the qualitative be-

havior of the amplifier.) In Fig. 8(c) both the diode and

the input signal are ON and a strong RF lpower outputnow makes its appearance.

Looking at the frequency spectrum under the same

three operating conditions, we observe Fig. 9, where the

center frequency in all three cases is 400 MHz. In Fig.

9(a) with the diode ON and the signal OFF, there is again

very little evidence of any residual oscillation at 400

MHz. Fig. 9(b) shows the frequency spectrum with only

the signal ON, and in Fig. 9(c) both the cliode and the

signal are ON simultaneously. Thus we see again that

only in the presence of the signal is the diode driven to

give the desired power output.

In Fig. 10 we have plotted this power output and the

amplifier gain as a function of input power. I t is not a

strictly linear amplification. To begin with, at zero input

power there is a small amount of residual power output

present. Only after very careful adjustment. of the stub

tuners were we able to reduce this residual output to a

bare minimum. But it is of course an essential feature of


5/8

960 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1970

,POWER OUTPUT

(b)

E ig. 9. Frequency spectrum for amplifier forthree distinct operating conditions.

181 [18

/- ;,6

:~:o .4 .8 1.2 1.6 2.0

POWER INPUT, WATTS

Fig. 10. Power output and gain versus power input.

an ampl if ie r that this residual power be as low as possi-

ble. When we allowed a slightly higher residual level, weachieved a higher power-output level, but also a more

pronounced leveling off at the high end of the curve.

Limitations in the available input-power source pre-

vented further exploration of higher saturation levels.

(This saturation would not be so apparent if the curves

were plotted on loglog paper as is frequently done, and

even the residual power at zero input power would then

conveniently disappear.) Another amplifier character-

istic that is of interest is the bandwidth; this is shown in

Fig. 11. Our amplifier circuit utilizes a narrow-band

FREQUENCY, MHz

Fig. 11. Frequency response curve of amplifier. Power out-put (watts) and power gain (dB) versus frequency (MHz).

1) 400 MHz Z =S.S j5.O 5) 2000 MHz Z=14.5+j16.52) 800 1.0-j2.5 6) 2400 6.5j61.53) 1200 ;: ~~ 7) 2800 8.0 -j754) 1600 . . 8) 3200 1 .0 j 9.0

Fig. 12. Impedence plot for fundamental andharmonic frequencies.

circulator and a triple-stub tuner, which is notoriously

narrow banded. It is therefore not surprising that thetotal bandwidth is only 5.8 MHz or about 1.5 percent.

As part of our experiment we also attempted to mea-

sure the circuit impedance at the point at which the

diode fits into the circuit. For this purpose we removed

the diode from the circuit and replaced it with a 2-inch

length of 50-ohm OSM line which was connected to a

network analyzer. An exact duplicate of this line with a

solid RF short at its end was used to establish and

calibrate the reference point. Fig. 12 gives the results of

these measurements on a Smith chart. The impedance


6/8

PRAGER et al.: POWER AMPLIFICATION WITH ANOMALOUS AVALANCHE DIODSS 961

Fig. 13. Wide-band frequency spectrum of amplifier for signal ON -and OFF conditions.

for the fundamental frequency does not show a very

good match and this might account for the low efficiency

of this particular circuit. The second, third, and fourth

harmonics are typically clustered around a near short

circuit for the real part, and they also show some reac-

tive component. As usual, higher harmonics are much

more reactively terminated.

SPECIAL FREQUENCY STUDY OF THE AMPLIFIER

We had hoped that these amplifier experiments would

not only give us a chance for a feasibility study of the

amplifier per se, but also a chance to collect further data

pertaining to the theoretical model of the anomalousmode. For this purpose we examined the frequency

spectrum of the amplifier, employing a field probe near

the diode and a wide-band spectrum analyzer. This

field probe consisted of a short section of 50-ohm OSM

line and was inserted into the GR Tee through a hole in

its outer shield. The probe was mounted on a microma-

nipulator which allowed precise three-dimensional move-

ment of the probe. Fig. 13 is a composite display of the

wide-band frequency spectrum from O to 6000 MHz, with

each division being 200 MHz. The bottom row shows

the spectrum when the diode is ON and drawing current

and voltage as described before, but with the input

signal turned OFF; consequently, there is no 400-MHztrace present. In the top row both the diode and the

input signal are ON, and a strong trace at 400 MHz as

well as a smaller trace at 800 MHz are now observed.

From our previous results we expected to find a strongfundamental and considerable harmonic content. But

we also found something else: in the signal OFF condition

we observed a small but distinct spectrum line near 1500

MHz and further noticed that this line disappeared in

the signal ON condition. This behavior was unique, since

there was no other frequency just like it. All other fre-

(b)

(c)

Fig. 14. Free-running frequency. (a) For amplifier OFF.(b) For amplifier oN. (c) For oscillator condition.

quency lines show an increase in amplitude when the

input signal is ON. We became very curious about this

frequency and took a closer look at it.Fi~o lA(a) shows this 1.500-MHz line again in the

signal OFF condition; Fig. 14(b) shows the signal ON

condition; and finally in Fig. 14(c) the diode has been

detuned from its amplifier condition until it became an

oscillator at 400 MHz as well as at its harmonics of 800,

1200, and 1600 MHz. And again it can be noted that the


7/8

962

.-

\b)

14~

12

10

8

Ii

0

6

4

4-oo~

POWER OUTPUT AT 400 MHz

(c)

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1970

Fig. 15. Free-running frequency for amplifier at various levelsof power output. (a) ~e = 1490 MHz; signal OFF. (b) .SignalON.

1500-MHz line has disappeared not only in the ON

condition of the amplifier but also in the oscillator case.

Zooming in still closer to this mysterious frequency, in

Fig. 15 we found it to. be at 1490 MHz. The frequency

dispersion in these pictures is now only 10 MHz/cm.Fig. 15 (a) depicts the signal OFF condition, Fig. 15(b)

shows the signal ON condition, and Fig. 15(c) shows how

the amplitude of the 1490-MHz oscillation decreases as

the power output of the amplifier increases continuously.

The observation of this frequency is quite unexplained

because it does not fit into the expected picture. First,

although the diodes under test should have a transit-

time frequency of approximately 5000 MHz, we could

not find any oscillation near that frequency, and 1490

MHz, is too low to agree with any normal transit-time

calculation. Furthermore, 1490 MHz is not harmon-ically related to the 400 MHz of the input and output

signal. Second, there is the unexplained cause of the

decreasing amplitude of this oscillation with increasing

signal input.

DISCUSSION OF FREQUENCY RESULTS

AND CONCLUSION

In trying to find an explanation for these experimental

results, we turned at first to the present theories of the

anomalous avalanche diode. Typically both the

plasma theory [8] and the avalanche pumped

resonance [6] theory rely on the transit-time frequency

(a)

(b)

(c)

(d)

Fig. 16. Locked oscillator.

oscillation for their starting mechanism and view their

operating frequencies in terms of harmonic relation. The

nonharmonic relation between our signal frequency and

the observed system frequency obviously called for still

further explanations. It was for this reason that van der

Pols classical theory of forced oscillations in a circuit

with nonlinear resistance [9] was given serious con-

sideration. Van der Pol treated both the so-called free

oscillation of the system and its forced oscillation

under the influence of an externally applied signal. He

described how the buildup of the forced oscillation

suppresses the free oscillation and also, significantly for

our interpretation, that these two frequencies do net

have to be harmonically related.

Still unanswered is the exact origin of the free oscilla-

tion of the system. It is however entirely possible that a

transit-time oscillation did exist beyond the calculated

value of 5000 MHz and also beyond the range of our


8/8

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1970 963

investigation, i.e., up to 7000 MHz, and that this

transit-time frequency in turn caused a weak anomalous

oscillation at 1490 MHz.

LOCKED OSCILLATOR OPERATION

The type of power amplifier which has been described

here is frequently associated with locked-oscillator

operation. We have therefore endeavored to demon-

strate that type of operation also. This was accomplished

with the same type of diodes used in the amplifier ex-

periments, but the circuit was changed to a simplified

coaxial structure [10 ].

Accordingly, the operating frequency was changed to

1090 MHz. The results are illustrated in Fig. 16, Fig.

16(a) shows the frequency spectrum of the oscillation

before locking, and Fig. 16(b) after locking. The locked-

power output was 25 watts, the locking signal power was

2.5 watts. Fig. 16(c) and (d) show evidence of the fre-

quency locking range with a scale calibration of 1

MHz/cm, thus making the locking range approximately

7 MHz. It should be noted that unlike the amplifier case

there is already a strong oscillation present before the

locking signal is applied and that the change in power

between the unlocked and locked condition is only about

2:1. These two features characterize the two main

differences between a locked oscillator and a power

amplifier.

ACKNOWLEDGMENT

The authors wish to thank Dr. L. S. Nergaard for

suggesting the similarity to van der Pols theory.

[1]

[2]

[3]

REFERENCES

H. J. Prager, K. K. N. Chang, and S. Weisbrod, High-power,high-efficiency silicon avalanche diodes at ultra high fre-quenches, Proc. IEEE (Correspondence), vol. 55, pp. 586-587,April 1967.

R. L. Johnston, D. L. Scharfetter, and D. J. 13artelink, High-efficiency oscillations in germanium avalanche diodes below thetransit-time frequency, Proc. IEEE (Correspondence), vol. 56,pp. 161 11613, September 1968.C. P. Snapp and B. Hoefflinger, An efficient multiresonantavalanche diode oscillator in the 1.5 to 11 G Hz rarwe. Proc.IEEE (Correspondence), vol. 56, pp. 2054-2055, ~ovember1968.

[4] R.- S. Ying, R. G1 Mankarious, and D. L. English, High-effi-ciency anomalous mode oscillation from silicon impatt diodes at6 GHz, 1969 ISSCCDig. Tech. Papers, pp. 86-87.

[5] H. J. Prager, K. K. N. Chang, and S. Weisbrod, Anomaloussilicon avalanche diodes for microwave generation, Proc.Cornell Conf. High F~equency Generation and Amplij cation,pp. 266-280, August 1967.

[6] B. Hoefflinger, C. P. Snapp, and L. A. Stark,, High-efficiencyavalanche resonance Dum~ed amdification. 1969 tit &xo-

[7]

[8]

[9]

[10]

wave Symp. Dig., pp. ~.S5-~60. J. F. Dienst, R. V. DAiello, and E. E. Thomas, Power ampli-fication using an avalanche diode, Electron. Lctt., vol. 5, p. 308,July 1969.A. S. Clorfeine, R. J. Ikola, and L. S. Napoli, A theory for thehigh-efficiency mode of oscillation in avalanche diodes, RCARev., vol. 30, pp. 397-421, September 1969.B. van der Pol, The nonlinear theory of electric oscillations, Proc, IRE+, vol. 22, pp. 1051-1086, September 1934.P. A. Levine and S. G. Liu, Tunable L-band high-power ava-~mc& diode oscdlators, 1969 ISSCC Dig. Tech. Papers, pp.

Correspondence

A One-Watt CW High-Efficiency X-Band Avalanche-

Diode Ampliiier

AbsfracfAn X-band avalanche-diode amplifier circuit is de-scribed which provides accurate and independent in-band impedance

control (R and +Jx) and orthogonal second-harmonic reactance con-

trol. Low-level gains of 13 dB with 500 MHz of bandwidth have been

achieved using thks circuit. Power outputs of 1 watt CW with 5 dB of

gain and 14.7-percent generation efficiency have also been rerdiiedby carefully controlling the in-band and second-harmonic load im-

pedance.Efficient solid-state X-band power generation can be accom-

plished by utilizing the negative-impedance characteristics of ava-lanche diodes in reflection amplifiers. Breadboard 7-GHz circuits have

been constructed which exhibit low-level gains of 13 dB with 500MHz of bandwidth. Power outputs of 1 watt CW, with 5 dB of gainand 14. 7 -percent generation efficiency have been realized in the samecircuit.

The diodes are first characterized by measuring the variation of

small-signal negative impedance with current density and frequency.

Manuscript received March 27, 1970; revised September 7, 1970. This work wassponsored by the Department of the Air Force.

J JpqMe.sucem,ntccuracy * 0 ldSl (Memu ,em,nl Ac, , ,cY i05dS)

4 6 s 10 12 4 6 8 10 12

FREQUENCY (GHz) FREQUENCY (GHz)(a) (b)

Fig. 1. (a) RF +dc port to RF port bias T insertion loss. (b)RF or RF +dc port to dc port bias T isolation.

The diode-pill-package weld flange is used as an impedance reference.A broad-band DC-T is required for this measurement. One was

constructed with the performance shown in Fig. 1 ~1) and (b). A com-puter-controlled network analyzer is used to calibrate out the losses

associated with the DC-T and a So-ohm diode holder. The holder

consists of a 50-ohm coaxial transmission line linearly tapered down

to the dimension of the circuit in which the diode will eventually beplaced. Thermal and electrical ground is provided by the same

amplification with anomalous avalanche diodes

Documents