KEK-79-4 April 1979 P

PERFORMANCE TESTS OF CROWBAR CIRCUIT FOR PROTECTION

Hisashi KOBAYAKAWA, Koji TAKATA and Isao TOYAMA

NATIONAL LABORATORY FOR HIGH ENERGY PHYSICS OHO-MACHI, TSUKUBA-GUN IBARAKI, JAPAN KEK Reports are available from

Technical Information Office National Laboratory for High Energy Physics Oho-machi, Tsukuba-gun Ibarakl-ken, 300-32 JAPAN

Phone: 02986-4-1171 Telex: 3652-534 (Domestic) (0)3652-534 (International) Cable: KEKOHO PERFORMANCE TESTS OF CROWBAR CIRCUIT FOR KLYSTRON PROTECTION

Hisashi KOBAYAKAWA

Depertment of Physics, Nagoya University,

Nagoya, 464, Japan

Koji TAKATA

National Laboratory for High Energy Physics

Oho-machi, Tsukuba-gun, Ibaraki-ken, 300-32, Japan

Isao TOYAMA

Nichicon Capacitor Ltd.

Kusatsu, Shiga-ken, 525, Japan

Abstract

A crowbar circuit to protect high power klystron for the rf accele­

ration of the PF storage ring has been constructed for the experimental

purposes. Brief description of the test circuit and results of the

performance tests are presented. The circuit is designed for 50 kV

operation and works sufficiently well for our purpose: electrical

energy of 9 kJ switched within 5 ps and thus only 3 J fed to the load. 1. Introduction

The PF storage ring (2.5 GeV) requires more than 500 kW of continuously rated rf power at 500 MHz. The rf generators consist of A each having an output power of 180 kW in the continuous mode.

The klystron will be fed by a high voltagd dc (50 kV,

10 A). The klystron beam voltage is produced by a conventional 12 phase

YA bridge rectifier followed by a single stage LC filter, the design of which achieves a reasonably low level of ripple content of the beam voltage (0.2 %pp) . 2 3) The output of the power supply shall have a high-speed crowbar ' for the protection of the klystron load. The crowbar shall operate automatically, in a few psec, to protect the klystron from excessive beam and/or body current. We have constructed a crowbar circuit for experimental purposes and for the help of actual design of the HV power supply. The circuit contains three (Mitsubishi MI-3200E) in

series connection for the crowbar switch, and also a 7 uF charging

capacitor which will be used as a filter in the actual HV supply and would be the biggest energy source to damage the klystron. The primary

of the HV power supply, 3 (j) 6.6 kV ac, shall be cut off with crowbar

operation within 100 msec, and the crowbar operation will be held on

until the primary is de-energized. The test circuit has an additional

capacitor (0.7 yF) as an equivalent of the primary ciucuit in the actual

HV supply.

Performance tests have been done and the circuit works expectedly

well, suppling only 3 J to the thin copper load at 50 kV charging voltage.

We describe briefly the structure of the test circuit, contents of the

performance test and its results.

- 1 - 2. Brief description of the test circuit

The schematic diagram of the crowbar test circuit is shown in

Fig. 1. The specifications are as follows:

(1) 60 kV high voltage generator;

Output voltage (max. 60 kV) is stepped up from 100 V ac by a variable

transformer and the Cockcroft-Walton rectifier.

(2) Charging capacitor (C_) and a resistor (R_) regarded as a filter

module of the actual power supply;

C = 7 yF, Rf = 20 JJ.

(3) Capacitor (C ) regarded as an additional current source with a

resistor (R ) when the crowbar is operated in the actual HV supply;

C = 0.7 yF, R =1 W2. s * s

(4) Crowbar switch;

(4-a) three ignitrons MI-3200E (Mitsubishi). (4-b) trigger module, containing three high voltage pulse transformers (see Fig. 2 for schematic diagram),

(5) Sensing circuits;

(5-a) 0.1 R Manganin (Mn) resistor: (i) to detect abnormal current

and send a signal to fire the crowbar, (ii) to measure the waveform

of the output current with an oscilloscope.

(5-b) 0.1 0. Mn resistor for waveform measurement of the crowbar

current.

(5-c) RC divider to measure the crowbar voltage (1/1000).

(6) Resistor (R) for the in the output;

R = 4.5 Q.

(7) The high voltage switch and short-circuiting materials (copper wires

or thin metal foils).

- 2 - 3. Crowbar characteristics

We give here briefly the crowbar characteristics for the following

parameters C£ = 7 uF, R = 20 $2, C = 0.7 yF, and E = 4.5 fi.

The stored energy in the charging capacitor (Cf = 7 uF) at 50 kV is approximately 9 fcJ. To see how the crowbar works, we used either thin copper wires 0.18 mmtp, 200 mm long whose critical energy for melting is about 20 J, or aluminum/copper foils with an iron needle for the sparking point. The calculated energy getting into, for example, the thin wire is approximately 55 J without crowbar switch that is enough to melt it away. When crowbar operates this energy is cut down to ~3 J. Expected current waveforms in the load are shown in Fig. 3 when the crowbar is on and off, respectively. Values indicated in the figure are estimated by using the circuit inductance and resistance actually measured.

4. Performance results

Correction of the meterings and RC divider for the voltage measurement

is carefully done. Two shunt circuits (crowbar current, load current)

are also carefully calibrated (time constant is about 1 usee) with both

dc and pulse currents. Three different materials are used as the test

load;

(1) closing with thick copper wire,

(2) melting tests with thin copper wires (0.18mm, 200 mm long),

(3) punching holes on the thin foils (aluminum/copper) by spark currents

from a needle point.

We have studied varying the charging voltage from 10 to 50 kV with

two values of resistors Rf = 20 and 45S2, and with dividing capacitors

C. = 0.2 UF and 2000 pF (see Fig. 2 for the definition of C. ). The ig ig typical waveforms are shown in Fig. 4 for the cases with and without

- 3 - crowbarring. Fig. 4(a) is a picture of the load current, crowbar current

and crowbar voltage at a charging voltage of 45 kV with Rf = 20 fi and

C. = 2000 pF. Fig. 4(b) is the same as (a) except that the load and C is connected with a 9 m, 50 JJ coaxial cable in order to simulate the actual cabling. Rise time (x..) in picture (a) is roughly 0.5 psec, and

3 usee after the high voltage switch is closed, the crowbar ignitrons start firing. Then about 2 sec (x„) later, the load current is almost

cut out. So the current flows into the load during only first 4 psec

and the succeeding current is switched into the crowbar by-pass. Without

crowbarring the decay time (x ) is 190 ysec.

The thin copper wires 0.18 mm, 200 mm long, are used for the

melting study"by which one could imagine the damages on the klystron

electrodes. Without crowbarring, one can see a small spike in the

wavefcrm of the load current (Fig. 5(a)). This is supposed to be a sign

of melt. After this the current still falls with the same time constant.

It may suggest that a conductive region around the vapourized wire

sustains the current. The energy fed before melt, estimated from the

waveform, is about 16 J which is roughly equal to the calculated energy

(20 J) to vapourize the wire. With crowbarring, the main energy in the

capacitor is well by-passed, so the load wire suffers no damages up to

50 kV (Fig. 5(b)).

The holes on the aluminum and copper foils due to the localized

spark current from the needle point are useful to imagine another type

of damage on the klystron electrodes. Sometimes abnormal discharge on

the electrode is localized in a very small area. The electrical energy

and therefore damages also concentrate into this small point and develop

at its vicinity. We can not exactly estimate the relation between the

size of the hole and energy (or voltage). We might say rather qualita-

- 4 - tively that the size (d), diameter of the hole, is approximately linear to the charging voltage (V), as the energy to melt foil is proportional 2 to td with t the foil thickness and the stored energy in condensor is 2 to V , The measured results given in Figs. 6 and 7 indicate a rough relation between d and V. The diameter ratio obtained from the aluminum 1/2 foils of 20 ym and 55 ym is approximately (55/20) , although the extraporated V-d lines seem not to cross the zero point. We cannot, however, explain the difference between the aluminum and copper foils only by taking into account the densities and specific-heats of them.

When the crowbar works, we have still small holes on the foil as given on the right side of the Fig. 7. A small hole, for example, with 1.5 mm on 20 ym foil at 50 kV corresponds to the energy dissipation of 2 ~3 J, roughly.

5. Timings and Waveforms measured

(1) Measurement of breakdown time and jitter.

The breakdown time (T, ) and its jitter (T.) are defined as the

average time and its maximum deviation, taken between the trigger

pulse start and the breakdown of the ignitrons. Fig. 8 is a super­

imposed picture of the waveforms of five discharges, showing T, =

3 ys and T = 0.5 ys at charging voltage of 20 kV (C. = 2000 pF).

(2) Time delay measurements of the trigger circuits

A square wave pulse (V = 25 V, width = 14 ms, rise time = 50 ns)

from a pulse generator is fed into the input circuit, the comparator

(see Fig. 2), of trigger circuit. The pictures of the oscilloscope

at the various positions of circuit triggered externaly by the

square wave pulse are given in Fig. 9. From the waveform pictures

thus taken, we obtained the following delay at each module:

- 5 - (i) T = 0.30 us in the comparator,

(ii) T_ = 0.15 ys in one-shot pulser (waveform shaper) ,

(iii) T„ = 0.35 ys in the pulse amplifier,

(iv) T, = 0.70 ys for the switching,

(v) T_ = 1.5 ys for the ignitor firing with no anode voltage

applied on the ignitrons.

Consequently we have the total delay of 3 ps which is roughly equal to

, measured directly as described in (1). Fig. 9(e) gives a waveform of gate voltage of the thyristor, showing the sign of turning on about 1.5 ys after trigger start. We need more 1.5 ys to turn on ignitor with a jitter of 0.5 ys as one can see in Fig. 9(i).

6. Remarks

Main current from the charging capacitor is sufficiently quickly swithced within 5 ys (x, - 3 ys) after the sensor detects the excessive b current, and surely the load would not be fed more than 5 J. Therefore this circuit would be promising for the klystron protection, because its electrodes are made of heavier materials than the loads used in this experiment. Simultaneous firing of three ignitrons are required for the reliablity. When one of them does not work by any trouble, then others do not. It is not permissible to miss firing of even one stage. Therefore we need warning equipments to notice abnormal signal on the trigger circuit, such as a voltage sensor for the charging capacitor of the 3 thyristor gates. The ignitrons are assured 10 times of firings. Mn resistors (0.1 JJ) are used here as the sensor of the excessive current, but we are planning to replace it by a current transformer in order to eliminate unwanted floating of the load circuit. Finally, we would like to thank Professors K. Fuke and T. Yamakawa

- 6 - for their encouragement and valuable discussions.

References

1) K. Takata, KEK-77-15 ('77)

2) C.R.Durbar, Private communication (Daresbury Laboratory)

3) Narziss, DESY Technical Specification B 2. 447 a,b,c

- 7 - CockrotI TRANS waiton CROWBAR lOOV/ttkV

Crowbar „ trigger ® crowbar voltage crowbar current load current

Fig. 1. Block diagram of the crowbar test circuit Comparator Une- Pulse Thyrislor Pulse- IQ shot amp. 78RF80 uuns- pulser

Fig. 2. Trigger circuits for the three crowbarring ignitrons CI&) (MI3200E, Mitsubishi Electric Co. Ltd.). Ofjz 5/JS 10 }JS L -+-

lT,~o.ofeps / T2 .?rtf i f ino:i T3 — (90 HS fl?f = 2o-^j

iSf: re

Fig. 3. Calculated waveforms of load currents, x is the rise time,

x„ the falling time with crowbarring, and T^ the falling time

without crowbarring. Every time constant is calculated by

using measured circuit impedances.

- 10 - (b)

5^0A/div.

2. s/div.

^ with crowbarrino without crowbarrin

— crowbar current

Fin. (\. Waveforms of load and crowbar current at a charnino voltaoe

9 of 4 5 kV with Rf 0 : and C. = 000 pF. (a) *\ thick cooper

wire was used as a load, (b) Traces are shown when the sap

load is connected by a 9, n lonq 50 coaxial cable. 500A/div

20ns/div

5Q0A/div

lOOus/div

500A/div

2ys/div

Fig. 5. Load currents with thin copper wire (0. "13mpi:, ZOCmrnl),

(a) without crowbarrinq at 45kV, (b) with crowbarrinn at 15,3n,4HV 12 1 i i . , —,

10

: fl/ 20[xm :

! £ "O 6 -

<£ 03"""""^ iu h~ 'JJ L[ _ _ / .-— s: : ft y 2 r- oy

o i i i i i i 0 /0 20 30 &0 ^"0 GO 70 CHARGING- VOLTAGE f/rl/0

Fig. 6. Relation between charging voltage and hole diameter d punched by spark currents without crowbarring. O : aluminum foil • : copper foil

- 13 - CROWBAR UNTRIGGERED CROWBAR TRIGGERED

(a) Al 20 pm thick, needle gap 0 mm

5, 10, 20 kV 5, 10, 20 kV

30, 40, 50 kV 30, 40, 50 kV

(b) Al 55 \m thick

30, 40, 50 kV 30, 40, 50 kV

(c) Cu 35 lim thick

30, 40, 50 kV 30, 40, 50 kV

(d) Al 20 ym thick, 40 kV needle gap

0, 1.6, 3.2, 5, 10 mm

10 mm

Fig. 7

14 r t T; ; - -

l%nLiror> breakdown.

Fig. 8. Load current waveforms. Five discharges are superimposed for

the measurement of the breakdown tine ( ,} and its jitter (-.) SSr-' Input square wove trigger

ISllSafjj width=14ns %} rise time=50ns.

(b) ^| Output signal of comparator, delay time T,=0.30..s.

Output signal of one-shot pulser, delay time T„=0.1 5..S , width=50..s.

(d) Output signal of pulse ampl ifier, delay time T.,=0.35us.

Signal at thyristor gate having turn-on spikes,

delay time T4=0.70..s.

i . 9(A) (f) Primary voltage of the rnls'1 transformer when its secondary is closed, 200V/div, 2..s/div.

(g) Primary current of the puUe transformer when its secondarv is closed. This is measured with 70m Mn wire, shcwino I = 10*2.7/0.07=336A and di/dt= 173A/US.

(h) Secondary current of the oul se transformer, I =77A.

(i) Ignitor anode voltage with chargino vo!tage=0. Five discharges are superimposed. 200V/div, dv/dt=400V/:.s.

B)