power conversion technologies...reprinted from engineering research, development and technology fy...

February 1997

THRUST AREA REPORT • UCRL-ID-125475

This is an informal report intended primarily for internal or limited external distribution. The opinions and conclusions stated are those of the author andmay or may not be those of the Laboratory.

Work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract W-7405-Eng-48.

Power Conversion Technologies

Mark A. Newton, Thrust Area Leader

DisclaimerThis document was prepared as an account of work sponsored by an agency of theUnited States Government. Neither the United States Government nor the University ofCalifornia nor any of their employees, makes any warranty, express or implied, orassumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or representsthat its use would not infringe privately owned rights. Reference herein to any specific commercial products, process, or service by trade name, trademark, manufacturer, orotherwise does not necessarily constitute or imply its endorsement, recommendation,or favoring by the United States Government or the University of California. The viewsand opinions of authors expressed herein do not necessarily state or reflect those of theUnited States Government or the University of California, and shall not be used for advertising or product endorsement purposes.

This report has been reproduceddirectly from the best available copy.

Available to DOE and DOE contractors from theOffice of Scientific and Technical Information

P.O. Box 62, Oak Ridge, TN 37831Prices available from (615) 576-8401, FTS 626-8401

Available to the public from theNational Technical Information Service

U.S. Department of Commerce5285 Port Royal Rd.,Springfield, VA 22161

Reprinted from Engineering Research, Development and Technology FY 96 UCRL 53868-96

February 1997

THRUST AREA REPORT • UCRL-ID-125475





integrating technologies that will significantlyimpact the capability, size, cost, and reliability offuture power electronic systems.

During FY-96, we concentrated our researchefforts on the areas of (1) Micropower ImpulseRadar (MIR); (2) novel solid-state openingswitches; (3) advanced modulator technologyfor accelerators; (4) compact accelerators; and(5) compact pulse generators.

The Power Conversion Technologies thrust areaidentifies and sponsors development activities thatenhance the capabilities of engineering atLawrence Livermore National Laboratory (LLNL)in the area of solid-state power electronics.

Our primary objective is to be a resource toexisting and emerging LLNL programs thatrequire advanced solid-state power electronictechnologies. Our focus is on developing and


6

Contents

6. Power Conversion Technologies

OverviewMark A. Newton, Thrust Area Leader

Advanced Modulator Technology for Heavy Ion RecirculatorsRoy L. Hanks, Hugh C. Kirbie, and Mark A. Newton ...................................................................................6-1

Evaluation of a Compact High-Voltage Power Supply ConceptRobert L. Druce, Randall E. Kamm, and Roy L. Hanks ...............................................................................6-5

Millimeter-wave Microradar DevelopmentStephen G. Azevedo, Thomas E. McEwan, and John P. Warhus ..................................................................6-9

High-Performance Insulator Structures for Accelerator ApplicationsStephen E. Sampayan, David O. Trimble, George J. Caporaso, Yu-Jiuan Chen, Clifford L. Holmes, Robert D. Stoddard, Ted F. Wieskamp, M. L. Krogh, and S. C. Davis...........................6-17

Compact Gas Switch DevelopmentDavid A. Goerz, Michael J. Wilson, Ronnie D. Speer, and Joseph P. Penland ...........................................6-23

Engineering Research Development and Technology


Introduction

Recirculating induction accelerators are beinginvestigated as potential low-cost drivers for inertialfusion energy.1,2 A recirculator is a circular induc-tion accelerator where beams of heavy ions areaccelerated and deflected in a closed path, as

illustrated in Fig. 1. Unlike linear machines, theacceleration sequence reuses each inductionaccelerating cell many times to reduce cost.

The Heavy Ion Fusion (HIF) project at LLNL isconstructing a small recirculator that will demon-strate concept feasibility. This small-scale recircula-tor is a 4.5-m diameter ring that will accelerate

Thrust Area Report FY 96

dvanced Modulator Technology for Heavy Ion Recirculators

6-1

At Lawrence Livermore National Laboratory (LLNL) we are building a small-scale recirculator todemonstrate the feasibility of accelerating ions in a closed circular path. This report describesprogress made in the development of a new solid-state induction modulator cell that is a key elementin this effort.

Roy L. Hanks, Hugh C. Kirbie, and Mark A. NewtonLaser Engineering DivisionElectronics Engineering

Figure 1.Demonstration recirculator underconstruction at LLNL.


singly ionized potassium ions to 320 keV, after15 laps past the 34 induction accelerating cells thatmake up the ring.

Electrical Requirements

A sequence of pulses must be generated by eachof the 34 modulators to accelerate the beam oneach lap around the recirculator. The time between

pulses is equal to the time it takes for the beam tomake one full circuit of the ring. As the beamenergy increases, the lap time decreases from 25 µsto 10 µs, which corresponds to pulse repetitionrates from 40 to 100 kHz.

Another facet of the recirculator experiments isthe demonstration of longitudinal beam compression.As beam velocity increases and compression takesplace, the modulator must generate pulses thatreduce in width from 4 to 1 µs. To achieve compres-sion of the ion beam, additional features are addedto the modulator pulses. These features take theform of a variable risetime on the leading edge ofthe rectangular modulator pulses and a ramp thatvaries in amplitude and risetime appended to thetrailing edge, as shown in Fig. 2. They have theeffect of slowing down the head of the ion beamwhile speeding up the tail.

To meet these requirements, a system wasproposed that uses two modulators within eachinduction cell. Their outputs are summed to achievethe desired voltage waveforms, as shown in Fig. 3.The first modulator produces the main 500-V rectan-gular pulse that features risetimes of 250 ns to 1 µs,pulse widths of 1 µs to 3.5 µs, and pulse top regulationof better than 1%. The second modulator generatesa 30 V to 150 V ramp that modifies the trailing edgeof the main pulse. This ramp varies in risetimes from250 ns to 1 µs during the evolution of the pulse train.

Engineering Research Development and Technology6-2

20 30 210100 220Time (µs)

Cel

l vol

tage

(V)

0

200

400

600

800

Programmable waveform generator

Ear modulator

Accelerator modulatorProgrammable

waveform generator

IBM personal computer

Feedback

Feedback

Accelerator cores

Ear core

Figure 2. Simulation of a typical recirculator pulse scheduleshowing that each lap has a different pulse requirement. Noticea time-scale change after 30 µs, which eliminates 12 pulsesfrom view.

Figure 3. Blockdiagram of inductioncell.


Progress

Initial evaluation of the main-pulse modulatorwas carried out with cores wound with Metglas2605 CO. As a result of testing and research, it wasdetermined that enhanced performance would beobtained with a new core of longitudinal fieldannealed Metglas 2605SC. This material promisesto provide superior performance in terms of greaterpulsed permeability and saturated flux density.

The ramp-generating modulator was firstconstructed using a single piece core of PE11Bferrite material. Early pulse tests indicated anunexpectedly low saturation flux density and perme-ability. We next wanted to evaluate the relativeperformance of CMD5005 ferrite material. Due tothe large diameter of the toroidal core, a singlepiece unit was unavailable. The alternative was topurchase a core constructed of five epoxied

segments. To better understand their magneticproperties, both cores underwent low frequency B-Hloop mapping. The test results, shown in Fig. 4,reveal substantial differences in the performance ofthe two ferrite materials.

From this data it was possible to determine thereason for the poor results encountered in theearlier pulsed PE11B core tests. Due to its “square”B-H characteristics combined with incomplete elec-trical resetting, the core was allowed to operateover only a limited portion of the available B-H loop.

In contrast, the CMD5005 ferrite core behavedmuch like a gapped core inductor. We believe thisis a direct result of its segmented construction,leading to no measurable residual flux density.This has the advantage of requiring only passiveelectrical resetting. On the negative side thepermeability is also 20 times less than that ofPE11B. Soon we will select the appropriate ferritematerial and modify the ramp modulator to accom-modate the choice. This decision will take intoaccount parameters such as core size, transientresponse, electronic circuit complexity, andmodulator cost.

The modulator electronics have undergoneseveral revisions to drive and control circuitry overthe past year. Here are a few examples:

1) Apex WB05 was replaced by the Elantec EL2008.The originally selected high-current buffer wascostly and scheduled to be discontinued in 1997.It was replaced by a higher bandwidth driver forone tenth the cost.

2) Analog Devices AD840N was replaced by theNational LM7171. The video op-amp thatprovides voltage regulation was upgraded in

Thrust Area Report FY 96 6-3

Figure 4. Low frequency B-H core mapping, revealing largedifferences in performance between PE11B ferrite (a) andCMD5005 segmented ferrite (b).

0. 1

35 T

esla

/div

PE 11B ferrite core(a)

0.15 Oersteds/div

0. 1

35 T

esla

/div

CMD5005 ferrite core(b)

1.5 Oersteds/div

Cel

l vo

ltag

e

20 V/div 400 ns/div

Time

Figure 5. Typical ramp voltage waveform.


order to take advantage of the latest generationof high-speed/high-current op-amps withimproved phase margin.

3) Many configurations of the voltage feedbacknetwork have been evaluated to meet therequirements of the induction cell and also toachieve the best balance of transient and DCregulation. Figure 5 shows a typical ramp volt-age waveform using five voltage-regulated rampmodulators and a PE11B ferrite core.

4) The effect of stray inductance and capacitance onmodulator performance was investigated. Theprimary stray circuit elements we considered havethe form of package and interconnect inductanceand stray capacitance that is attributed to themechanical structure of the cell. Based onMicroCap IV simulations, a reduction in MOSFETgate inductance and interconnect lengths, shownin Fig. 6, could yield substantial performanceimprovements in the form of improved rise andfalltime with less overshoot.

Future Work

As a result of the work that was conducted thispast year two issues have come to light.

First, our present active reset circuit is not capableof determining if the core is initially at its positive ornegative remanence state. This information is

important to take advantage of the full core fluxswings. The modulator circuitry will be revised suchthat the core will be initialized to the desiredremanence state.

Second, our modulators are constructed usingdiscrete component on printed circuit boardtechnology. The next step is a collaboration withAlliedSignal on the design and construction of ahybrid microcircuit (HMC) that will increasepackaging density and address the problem ofstray and interconnect inductance. The firstsamples of these HMCs will be evaluated inFY-97. From these tests we hope to demon-strate the anticipated improvements in size andperformance provided by this new technology.

References

1. Friedman, A., J. Barnard, M. Cable, D. Callahan,F. Deadrick, D. Grote, H. Kirbie, D. Longinotti,S. Lund, L. Nattrass, M. Nelson, M. Newton,C. Sangster, W. Sharp, T. Fessenden, D. Judd, andS. Yu (1995), “Progress Towards a PrototypeRecirculating Induction Accelerator for Heavy-IonFusion,” Proceedings of the 1995 ParticleAccelerator Conference, Dallas, Tex., (UCRL-JC-119538).

2. Newton, M., H. Kirbie, and R. Hanks (1995),Development of Advanced Modulator Technology forHeavy Ion Recirculators, Lawrence Livermore NationalLaboratory, Livermore, Calif., (UCRL 53868-95).


Figure 6. MicroCap IV simulations, showing a significant reduction in overshoot if interconnect inductance can be eliminated.

-200

-400

-600

With interconnects

Time (us)2 3 4 50 1

(a)

200

(V) 0

-200

-400

-600

Interconnects eliminated

Time (us)2 3 4 50 1

(b)

200

(V) 0

Cell voltage


Introduction

High-voltage power supplies typically achievevoltage multiplication by means of a trans-former. The size and weight of such powersupplies are ultimately limited by the amount ofwire in the transformer windings and the corematerial. Increased output voltage increases thenumber of turns necessary, while increasedoutput power increases the size of the wire andamount of core material necessary. As a result,high power and high voltage conspire to increasethe size and weight of the transformer and thepower supply. Increasing the operatingfrequency improves conditions somewhat, butwinding capacitance and switch speeds limit thepossible improvement.

The design of high-voltage power supplies usingtransformer technology is a mature discipline,precluding major improvements in efficiency orsize. Charging a capacitive load provides an addi-tional challenge to the overall efficiency, sincetransformer-coupled supplies are constant-voltagesources in series with an impedance. A new,transformerless approach is required to provideadditional significant reduction in size and weightof these supplies.

Progress

Concept

One possible alternative is a charge-pump circuitsuch as that shown in Fig. 1. This circuit uses anopening switch with an inductive store to pumpsmall charge packets into the capacitive load. Sincethis circuit functions primarily as a constant-currentsource, the efficiency can be improved over that of aconstant-voltage source supply for capacitor charg-ing. The main challenge with this circuit is findingan appropriate opening switch to drive the inductivestore. This switch must be capable of operating atthe desired output voltage at a high repetition rateand must have low losses.


valuation of a Compact High-Voltage Power Supply Concept

6-5

We have evaluated a Russian-designed opening switch for use in a compact high-voltage powersupply. The switch is for use in a charge-pump circuit which would eliminate the transformer whileallowing the power supply to charge a capacitor to high voltage. The design parameters of the powersupply were < 1.5 kV input voltage, 100 kV output voltage, 0.15 µF load capacitor, and < 1 s chargingtime. The diode switches tested proved to be too lossy for this application. Additional evaluation isnot planned for this concept unless devices with characteristics better suited to this applicationbecome available.

Robert L. DruceDefense Sciences Engineering DivisionElectronics Engineering

Randall E. Kamm and Roy L. HanksLaser Engineering DivisionElectronics Engineering

IGBT

IGBT

Battery

Openingswitch

Loadcapacitor

Figure 1. Simple charge-pump circuit using an opening switchfor voltage multiplication.

Table 1. List of circuit parameters and range of variation.

Circuit parameter Range

Source Capacitance (C1) 10 –100 nFCircuit Inductance (L1) 1–10 µHCharge voltage 100 – 2,400 V


Opening Switch

The likeliest candidate for an opening switch inthis application is a solid-state device. The switchchosen for evaluation is a diode of Russian design.This device consists of a series/parallel stack ofdiode junctions with each junction tailored to openat a very high rate after being biased with a specificwaveform. The diode stacks are rated at 100 kVopening voltage against 800 A current and 12 nsopening time. The stacks are about 10 cm long. Thediodes must be driven with a ringing waveform togive the proper initial conditions for opening. Thesediodes are designed to be used as a pulse sharpen-ing element. We planned to evaluate them in asomewhat different regime, giving rise to some risk.

Diode Evaluation Circuit

A schematic of the evaluation circuit is shown inFig. 2. The circuit was fabricated using strip linesto give minimal parasitic circuit elements. Thecircuit parameters were varied over the rangeshown in Table 1. The maximum voltage waslimited to a range similar to that attainable bycommercial solid-state closing switches since anypractical power supply would use these devices.Operation of the circuit with no opening switchdiode installed agreed with calculated waveforms towithin a few percent.

Diode Evaluation

Initial DC evaluation showed a forward bias gapvoltage of ≈ 85 V and a forward resistance of about2 Ω. The DC small-signal reverse resistance indi-cated ≈ 10 MΩ. The forward bias gap voltage wasverified with the pulsed experiments but theforward impedance was not consistent and waspossibly masked by other effects. No attempt wasmade to verify the reverse characteristics in thepulsed tests. The diodes showed a definite thresh-old current (and charge) for switching. Figure 3shows a typical inductor voltage and current wave-forms for a switching event. The parameters forthis event were as follows: Vchg = 500 V, L = 7.5 µHand C = 51 nF.

Note that the voltage of the inductor spikes upas the current falls when the diode switch opens.This event had a voltage gain of approximately0.95 into an open circuit. The inductor voltageearly in the waveform is approximately 200 V.This leaves about a 300 V drop across the openingswitch diode. Figure 4 is a plot of the same


Tack switch

Diode opening switch

Circuitresistance Load

inductor L1

Energystore C1

Figure 2. Schematic of the circuit used to evaluate the open-ing switch diodes. The switch is a solid-dielectric tack to givefast turn-on and low impedance over a wide range of voltageand current.

Current

Inductorvoltage

Time (µs)

500

400

300

200

100

0

-100

-200

-300

50

40

30

20

1

0

-1

-20

-30-0.5 0 -0.5 1 1.5 2 2.5 3 3.5 4 4.5

V A

Figure 3. Plot of inductor voltage and total circuit current fora typical switching event. The parameters for this event wereVchg = 500 V, 7.5 µH inductor and 51 nF capacitor.

Figure 4. Plot of inductor voltage and circuit current. Thecharge voltage is 2400 V for this event. The circuit inductanceis approximately 7.5 µH; the source capacitance is 51 nF.

Time (µs)

30002500200015001000

5000

-500-1000-1500-2000

300250200150100500-50-100-150-200

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Current

Inductorvoltage

V I


waveforms with a 2400 V charge and the samecircuit parameters. The voltage gain is approxi-mately 1.2. Note that the voltage gain is increas-ing with increasing voltage and current, indicatingthat the diodes will perform well at much highervoltage and current. Figure 5 is a comparison ofthe voltage across the diode for the same twoswitching events, showing that the diode is


Time (µs)

4000

3000

2000

1000

0

-1000

-2000

-3000

-40000 0.5 1 1.5 2 2.5 3 3.5 4

V

Figure 5. Comparison of the diode voltage for 500 V and2400 V charge switching events. The waveform with thelarger negative spike is the 2400 V charge waveform.

extremely lossy but that the losses are not linearwith voltage or current. We have not been able todetermine definitively the source of this loss.

These waveforms show typical behavior for theswitch. The waveforms with differing capacitanceand inductance in the circuit are similar. The gain istypically lower with less circuit inductance and smallinductance showed a higher switching threshold.

Conclusion

Though the concept of a charge-pump high-volt-age power supply is quite appealing, the new open-ing switch diode technology evaluated is not appro-priate to produce the voltage gains needed in theparameter space required. These devices were eval-uated in a regime that they were not specificallydesigned for, on the possibility that they wouldproduce the necessary voltage gain. In light of thelarge gap between the required parameters andthose demonstrated by the devices tested, wediscontinued the investigations without extensiveresearch into the nature of the losses in the devices.

The results of these investigations do not implythat the devices are not appropriate for pulsesharpening at higher voltage and current.


Introduction

In a previous report,1 we described many of thedifferent types of Micropower Impulse Radar (MIR)systems that have been developed in support ofnumerous government and commercial systems inthe last few years. The main ideas behind MIR,invented by T. McEwan at LLNL, are the generationand detection systems for extremely low-powerultra-wideband pulses in the gigahertz regime usinglow-cost components. These ideas, coupled withnew antenna systems, timing and radio-frequency(RF) circuitry, computer interfaces, and signalprocessing, have been the catalysts for a new gener-ation of compact radar systems. The systems gener-ally fall into four sensor categories: 1) motionsensors; 2) distance sensors; 3) imaging sensors;and 4) communication devices. (For more informa-tion on these systems, and on the overall MIRProject, see past reports2 or our world-wide webpage at http://www-lasers.llnl.gov/lasers/idp/mir/mir.html.)

There are still many new directions that we planto explore, to continue our leadership role in MIR.Rather than repeat the broad list of sensors andsystems within the MIR scope, in this report we willdescribe our efforts in FY-96 to develop higherperformance (yet still small and low-cost) radars

that will expand the technology base and offer moreopportunities. As an entirely new technology area,rather than a single circuit, MIR has the potential toaddress a rich set of applications for which there isnot yet program, government, or commercialsupport. The modular systems and higher-frequencysensors developed in the last year will inspirenumerous novel concepts that have high expectedimpact and return on investment. We plan toembark on areas that will enhance our engineeringexpertise and technology base, while providing newopportunities and capabilities for programs.

Progress

The objective of this project was to expand andadvance our current capabilities in MIR technologyalong several directions. We discuss the key techni-cal MIR developments in terms of the technologyimprovements, with applications for each.

Modular MIR Components

In an effort to produce functional radar modulesthat can have many features to aid system develop-ment, we have designed and developed a family ofMIR boards, antennas, interconnections and soft-ware with standard interfaces that “plug and play”


illimeter-wave Microradar Development

6-9

Stephen G. Azevedo and Thomas E. McEwanLaser Engineering DivisionElectronics Engineering

John P. WarhusDefense Sciences Engineering DivisionElectronics Engineering

The objective of this project was to enhance the low-cost impulse radar systems made famous atthe Lawrence Livermore National Laboratory (LLNL) with new ranges of frequency, resolution, anddirectionality. Moving into these ranges opens up many new areas of application while expanding ourexpertise in small low-power radar systems. Several areas were identified that are leading toprojects in support of LLNL programs, as well as generating new outside funding. Small, low-costradar systems enable many applications that use arrays of transmit/receive elements, and we arepursuing many such imaging system concepts.


together. Figure 1 is a schematic diagram of MIRmodular components that fall into four general cate-gories of software, computer interfaces, timing andbaseband processing circuits, and high-speed frontends. The interconnections between the categoriesare based on industry standard hardware or softwarecomponents (for example, SMA or audio connectors,TTL voltages, and 50-Ω terminating resistances).

The focus of our effort has been on the timing andhigh-speed circuitry, where much of the family hasalready been implemented. Incorporated into thesemodules are many of the anticipated enhancementsneeded to develop future MIR hardware systems.For this reason, the modular MIR boards are gener-ally larger than the first generation of boards, buttheir layout is such that subcircuits can be easily

reconfigured to build custom boards, or potentiallyapplication-specific integrated circuits (ASIC’s), forparticular applications. Both government andcommercial projects are better served by the modu-lar approach, because prototyping of radar systemsbecomes a simpler task and direct characterizationof the individual components is more straightforward than with previous MIR systems.

Significant progress has been made in generatingradar components that produce consistent andrepeatable responses under most conditions. Incontrast to earlier MIR prototypes, the currentsurface-mount board designs are very robust toshock, interference, and temperature changes. Forexample, a constant-fraction discriminator (CFD)has been added to the Dipstick and Rangefinder


The Plug-And-Play Modular MIR Family

Software ComputerInterface

High Speed & RFTiming and BasebandProcessing

Rangefinder/TDR Data

Acquistion &Display

PCMCIA

Serial PortBasic Stamp

Quartz TimeBase

TDR SignalProcessor

RangefinderSignal

Processor

MulticellMotionSensor

ASIC

ASIC

ASIC

30, 60, 200 ps 30, 60, 200 ps

PulsedRF

TX & RF

ImpulsedTX & RF50-OHM

TDR Probes

WideBeam

Antennas

2, 6, 10, 24, 96 GHz

2, 6, 10, 24 GHz

NarrowBeam

Antennas

2, 6, 10, 24 GHz

2, 6, 10, 24 GHz

Circ PolAntennas

High EpsilonLaunchers

0.1 - 10 GHz

30, 60, 200 ps

Pulsed RF

TX & RX50-OHM

2, 6, 10, 24 GHz

MotionSensor

λLaserTR Head

TR Array

2, 6, 10, 24 GHz

5X5, 10X10, Mono& Multistatic,

30, 60 & 200 ps

30, 60, 200 psPrinter PortBasic Stamp

Echelon Bus

Wireless Lan

4 - 20 MACurrent Loop

Micro-power

RX

Micro-power

TX

TriggerMUX

VideoMUX

DataAcquisition

PC Card

Time-of-Flight

Positioner

Subsurface/Biomedical

Imaging

Free Space/Thru-WallImaging

BiomedicalMotion

Analysis &Display

ImpulseTX & RX

PlateMTD TDR

Head

50-OHMTDRHead

PresenceSensor

Figure 1. The MIR modular system.


systems that automatically adjusts a thresholddetector to a temperature-compensated referencevoltage. Then distance or range is measured bypulse-width modulation of the time between thresh-old crossings of MIR impulses.

For a 100-ps-wide pulse, the leading edge rise-time corresponds to about 1.5 cm in range. To holdrange errors below 1 mm, we need a thresholddetection accuracy of better than 6%, regardless ofpulse amplitude. Compensation for voltage andthermal changes (about 1% fluctuation over outdoortemperatures) is automatic and flexible enough tooperate under harsh conditions.

Out of the modular components described inFig. 1, there are several complete radar systemsthat have been developed in the last year.

1. MIR Motion Sensor.3 This is an enhancedversion of the single-board motion sensor thatcan be easily reconfigured to match a specificneed. Like the original, it is range-gated, lowpower (multi-year battery life), low cost, channel-less (multiple MIR units can operate in closeproximity without RF interference), and nearlyimpossible to detect. A photograph of the boardwith simple quarter-wave antennas is shownin Fig. 2.

Only motion-modulated signals or changesfrom a baseline measurement are detected,thereby eliminating false triggers from station-ary room “clutter.” The motion pass-band canbe changed by modifying the on-board filtercomponents to match the application. An inde-pendent laboratory has verified that the MIR

motion sensor can satisfy FCC Part 15 regula-tions. Applications are in security and energycontrol systems, industrial safety, robotics,vibration sensing, and speech processing.

2. MIR Electronic Dipstick.4 This is a two-boardlow-cost time-domain reflectometer (TDR)system that was designed to detect the height offluid in a reservoir or container by measuringthe pulse-echo time of an MIR pulse launchedalong a transmission line—a simple wire. Thetwo modular boards are the quartz time baseand TDR signal processor described in Fig. 1.Measurement of the fluid height is typicallyresolved to 0.1% of maximum range. There aremany applications of the system in measuringfluid and material levels in industrial containers(tanks, vats, silos), hazardous materials, down-hole water levels, automotive tank monitoring,and in providing automatic fill control.

3. MIR Rangefinder.5 This is a five-board completeimpulse radar transceiver system with sweptrange-gate and ultra-wideband antennas. Aphotograph of the full modular Rangefinder isshown in Fig. 3. The five boards used are thequartz time base, Rangefinder signal processor,60-ps impulse receiver, and two transmitterboards (60-ps impulse and 6.5-MHz pulsedoscillator boards). The receiver works equallywell with both impulse and pulsed-oscillatingtransmitters. Waveform outputs of the twotransmitters are shown in Fig. 4. It generatesan equivalent-time A-scan (echo amplitude vsrange, similar to a WW-II radar) with a typical


Transmit antenna

LED alarmindicator

10 mA

Receive antenna

Sensitivity adjustRange adjust

+9-Vbattery

ALARM OUT(switch to ground)

Commonground

Figure 2.Photograph of the MIR MotionSensor with simpleinterconnects.


range sweep of 10 cm to 3 m, and an incremen-tal range resolution, as limited by noise, of0.3 mm. It operates in spectral regions thatreadily penetrate walls, wood panels, and to anacceptable extent, concrete and human tissue.

The MIR Rangefinder is the most sophisti-cated of the dozens of MIR prototypes; it is thebasis of all imaging applications and of manyreimbursable projects. Uses of the Rangefinder

include replacement of ultrasound rangefind-ers for fluid-level sensing (a dipstick withoutthe stick), light-weight altimeters forunmanned airborne vehicles, localizing breath-ing motion behind walls, vehicle height sens-ing, and robotics control. When positionedover a highway lane, it can collect vehiclecount, vehicle profile, and approximate speeddata for traffic control.


RX antenna

SMA coupler

RXmodule

Trig in

Dual phono cable

Analog outputto scope CH1

Range PWMoutput toscope CH2

Ribbon cable - stripe to pin 1

Single phono cable -peel single cableaway from dual

Video out Trig in

TXmodule

SMA coupler

TX antenna

Pin 1Signal

processor PCB

Timing PCB

12-V powermodule

Pin 1

Figure 3.Photograph of theModular MIRRangefinder withsimple interconnects.


Rangefinder Extensions

As mentioned in the last section, the modularMIR architecture makes it possible to extend theRangefinder to new levels of performance and capa-bility. Where the original Rangefinder was a fully-integrated unit with fixed antennas, the modulardesign adds flexibility to our prototyping efforts.During the development of the modular design, wealso added improvements to the standard MIRfeatures (without substantially adding to the cost).For example, the CFD described above is nowdirectly integrated into the Rangefinder design. Alsoincluded are more accurate time-base generation(0.1% over the maximum range using a crystaloscillator), high- and low-frequency cut-off filtercontrols, and wide flexibility in the start location andlength of the scan. In this section, we describesome additional extensions to the Rangefinder madepossible by the plug-and-play concept.

The modular MIR Rangefinder provides theperfect vehicle for ultra-wideband imaging applica-tions, and we are currently working on several suchprojects through outside sponsorship. Multipletransmit/receive modules (yet only one time baseand one Rangefinder signal processor) can beconfigured into arrays of radars and coupled to acomputer to form either synthetic aperture or real-array imaging.6 Radar return signals are digitizedand stored in the computer. Reconstruction ofcross-sectional images from B-scan or waterfall typedata is performed by diffraction tomography soft-ware on the computer.7 Images of the scene aredisplayed directly on the screen within 10 s (in 2-D).

We have demonstrated the use of this radar pack-age for integration into an imaging array that issmall, lightweight, low power, and inexpensive, rela-tive to existing radars. Some of the imaging applica-tions we are exploring are road-bed and bridge-deck

inspection,8 land mine and buried ordnance detec-tion,9,10 detection of underground utility lines,through-wall detection of people (for military, lawenforcement, and search and rescue teams), andnondestructive evaluation (NDE) of concrete (civilstructures, earthquake damage), wood (lumberevaluation, power pole rot), or, to a limited extent,living tissue (hematoma detection, kidney stones).Other materials are also possible candidates formaterial inspection, such as low density foamsand composites.

The current arrays use monostatic imaging, butfuture versions will be capable of multistatic opera-tion. The modular system makes it possible to sepa-rate the antennas to any distance and to performboth reflection and transmission experiments. Inconjunction with the NDE Thrust Area, we are devel-oping systems to explore these and other possibili-ties (see the article by J. Mast, in the NDE section ofthis report).

For the bridge-deck inspection project,8 a high-speed radar (HSR) front-end system was developedthat requires single pulse detection with no averag-ing so that the vehicle can travel at highway veloci-ties and still detect subsurface flaws. While stillsmall, the HSR has higher performance specifica-tions (with associated higher cost and higher power)than its MIR counterpart. However, the HSR front-end hardware was made so that it attaches to themodular MIR antennas and back-end circuitry asanother module to the tool set. This is anotherexample of how all applications can now fit thecommon architecture.

Antennas

It will be necessary to integrate the electronicswith the antennas in future versions of the imag-ing radar systems. This is needed to keep the


Am

plit

ude

(0.5

V/D

iv.)

Am

plit

ude

(0.5

V/D

iv.)

(a) (b)

Time (2.5 ms/Div.) Time (2.5 ms/Div.)

Figure 4. Video output of the MIR Rangefinder using the impulse transmitter (a) and the pulsed-oscillating transmitter (b). Thetransmitting and receiving antennas are pointed directly toward one another with about 20 cm stand-off.


electronics small, robust, and fast. Integratedelectronics of this type will advance our capabili-ties in radar well beyond the current state of theart, and open numerous new areas of programdevelopment. Higher power and other directionalimaging arrays may also be required for specificapplications. A list of antenna constraints for theMIR Rangefinder may include:

1) s11 and s21 characteristics that are flat across avery broad band and exhibit smooth, linearphase s21 roll-off at the band edges;

2) group delay (dΦ/df) substantially less than theimpulse width across the operating band, suchas less than 50 ps across 1 to 10 GHz (equiva-lently, very clean step function response);

3) well-controlled, low sidelobes—no change inpulse shape vs angle;

4) low feedline-coupling into the antenna; and5) low-cost, compact, rugged, and simple

construction.In conjunction with the Computational

Electronics and Electromagnetics Thrust Area, asignificant amount of effort has been directedtoward stable, repeatable, and scaleable ultra-wide-band antenna designs. This work has been instru-mental in improving the beam-width, bandwidth,impedance, launch point, size, shape, and cross-talk characteristics of the complete MIR system.Several antenna designs are currently being usedand are pending patent consideration. The basicimaging antennas have a very broad beam width

and correspondingly low gain. They are suitable forsynthetic aperture imaging where broad illumina-tion is desirable. Narrower beam widths and highergain can be obtained on a broadband basis withhorns, reflectors, or dielectric lenses.

Frequency Extensions

Extending the frequency range of MIR from themaximum of about 4 GHz (microwave) into themillimeter-wave bands has been initiated in the lastyear. We have accomplished this with both impulsesystems (wideband up to 12 GHz) and pulse-drivenoscillators (at 20 GHz and 94 GHz). Figure 5shows the relationships of frequency range to reso-lution and penetration depth. By going to higherfrequencies, wavelengths become shorter, antennasbecome smaller, and resolution improves to themillimeter range. Pulses at these higher frequen-cies can be launched more directionally and withlower sidebands. Penetration into materials is muchless, but the resolution (on the order of 1/4 wave-length) is much greater. Also, for most applicationsthe pulse will be launched in air, where losses aresmall. In all cases, we use the modular MIR timingand video processing on the back end while empha-sizing research in the pulse driver circuitry.

An important advantage of moving into themillimeter-wave bands is that very high-gain anten-nas can be designed, which are small whencompared to antennas with similar gain in the


Res

olu

tio

n c

ell (

λ/2)

in v

ario

us m

edia

(m

)

0.1

0.1

0.01

0.0011.0 10.0 100.0

Frequency (GHz)

1

Proposed high-frequency MIR’s

HSR

Low-frequencyMIR

A

D

C

T

Expectedwideband limits

Police radars(12 & 24 GHz)

= Current= Proposed

Medium:A = AirD = DirtC = ConcreteT = Tissue

StandardMIR

Figure 5. Frequencyranges of current andproposed MIR subsys-tems. Notice that theresolution improves(gets smaller) as thefrequency increases,but the penetrationin most materialsdecreases at thesame time.


current MIR operating frequency band. Reasonably-sized millimeter-wave antennas, with improved gainand directivity can be produced to extend the rangeof MIR, and improve angular resolution and systemportability. In addition, compact high-gain antennaswill enhance the performance in some of the appli-cations for which MIR sensors have already beenproven. Examples of such millimeter-wave MIRapplications include bullet/projectile tracking; high-resolution target acquisition for tanks and otherweapon systems; robotic collision avoidance; airportground traffic surveillance/ tracking; automotivesystems; high-resolution personnel imaging; radaraids for the blind; and NDE of composite materials.

A prototype impulse driven 94-GHz radar hasbeen assembled to gauge the operational capa-bi l i ties of impulse-driven ultra-widebandmillimeter-wave radars. The test system consists ofa standard MIR backend timing and signal process-ing module that drives a high-speed IMPATT diodesampler. This diode oscillates for several cycles at94 GHz and drives a high-frequency transmitantenna. Return signals are measured with a simi-lar sampler on the receiver side, and again attachedto the MIR signal processing module. A plot of theresulting waveform is shown in Fig. 6. Initial testdata indicates a radar bandwidth of >5 GHz andtransmit power >10 mW. Antennas used had a -3 dbbeam width of 10°. Tests indicate that a properlydesigned impulse driven 94-GHz radar should havea transmit power >25 mW and a bandwidth>15 GHz.

With additional development it may be possible toreach full waveguide bandwidth of 35 GHz (75 to110 GHz). Radar repetition rates from kHz to>10 MHz are easily reached. Proper design ofIMPATT diode matching networks for the input-drivenimpulse and RF output matching could allow IMPATTdiodes to be used to create radars in the 1 to 40 GHzband with bandwidths on the order of 10 to 20 GHz.Further work needs to be done to better quantify thefull capability of impulse-driven millimeter-waveIMPATT diodes.

Single-antenna Systems

The early MIR systems all had two antennas, onefor transmit and one for receive. In this last year,we have performed work on single-antenna systemsthat use the same antenna for both functions. Forexample, a type of motion sensor we call the FieldDisturbance Sensor (FDS) has been developed with2- and 4-GHz pulsed-oscillators. The FDS is arange-gated homodyne motion sensor with similar

characteristics to the MIR, yet only a single antennais needed, and that antenna can be a standarddirectional one rather than omni-directional.

The Rangefinder can also be used in a single-antenna configuration by means of another recentlydeveloped “directional sampler” circuit. To receivewhile still transmitting, a method is needed tocancel the transmit pulse at the sampler (receiver)input. In the directional sampler, the transmit pulseis applied to the top of a resistive bridge. The trans-mit pulse is divided equally by the bridge resistorsand applied to a differential sampler comprised of apair of charge-holding capacitors and diodes. Theoutput of the differential sampler is applied to anamplifier where, properly tuned, the sampled trans-mit pulse is differenced to zero.

Future Work

We envision many additional refinements for theMIR systems of the future. Many of those arecentered around insights gained from governmentand commercial interactions. Most applicationsalso require some level of effort, such as a changein range/sensitivity/directionality, or size/-power/penetration, or pulse shape, or signalprocessing, to reduce it to practice. As proprietorsof the MIR technology, we anticipate performingmuch of the “proof-of-principle” development work,while attracting private industry participation formass production.

For FY-97, we plan to continue our progresstoward higher frequencies and more modularsystems. In this way, we can continue our efforts tocharacterize MIR systems and tailor them more


-.6

0

.2

.4

.6

-.4

-.2

-2 .5-.5-1.5 -1 1 1.5 2

Figure 6. Pulsed 94-GHz received waveform from a metalplate at 1.32 m from the transmit/receive horns. The 94-GHzsignal is sampled by the MIR back-end and shifted to the 1 to4 GHz range.


easily to the applications at hand. There is stillsignificant R&D needed to have reliable turn-keyhigh-frequency MIR systems. For example, we havelong considered the possibility of integrating muchof the back-end MIR circuitry into a single siliconASIC. This is still expected soon, but we will mostlikely carry this out in conjunction with industrialpartners. While our licensees and partners arecontinuing to develop commercial applications ofMIR, our internal efforts must stay focused on thelonger-term systems issues that will take us to thenext step.

Acknowledgments

The authors would like to acknowledge the contri-butions of the scientists, engineers, and techniciansinvolved with the MIR Project. These includeJ. Brase, R. Cavitt, G. Dallum, L. Haddad,R. Hugenberger, B. Johnston, H. Jones, J. Mast,D. Mullenhoff, S. Nelson, T. Rosenbury, R. Stever,P. Welsh, and M. Wieting. Administrative help camefrom M. McInnis, C. Bothwell, M. Lynch, F. Reyna,R. Sachau, and S. Turner-Perry.

We also acknowledge the continued support fromthe Laser Programs Directorate (R. Twogood,E. M. Campbell) and the Electronics EngineeringDepartment.

References

1. Azevedo, S. G., T. E. McEwan, and J. P. Warhus(1996), “Microradar Development,” EngineeringResearch, Development and Technology: Thrust AreaReport, Lawrence Livermore National Laboratory,Livermore, Calif., (UCRL-53868-95), pp. 6−17.

2. Azevedo, S. G., and T. E. McEwan (1996),“Micropower Impulse Radar,” Science and TechnologyReview, Lawrence Livermore National Laboratory,Livermore, Calif. (UCRL-52000-96-1/2).

3. McEwan, T. E. (1996), MIR Motion Sensor User’sGuide, (Controlled Distribution), Lawrence LivermoreNational Laboratory, Livermore, Calif., (UCRL-MA-124118).

4. McEwan, T. E. (1996), MIR Electronic Dipstick User’sGuide, (Controlled Distribution), Lawrence LivermoreNational Laboratory, Livermore, Calif., (UCRL-MA-124398).

5. McEwan, T. E. (1996), MIR Rangefinder User’s Guide,(Controlled Distribution), Lawrence LivermoreNational Laboratory, Livermore, Calif., (UCRL-MA-125056).

6. Azevedo, S. G., D. T. Gavel, J. E. Mast,E. T. Rosenbury, and J. P. Warhus (1996), “Arrays ofMicropower Impulse Radar (MIR) Sensors forSubsurface Detection,” Proceedings of the EURELConference on the Detection of Abandoned LandMines, (IEE Conf.), Edinburgh, Scotland, UnitedKingdom, Pub. No. 431.

7. Mast, J. E., and E. M. Johansson (1994), “Three-dimensional ground penetrating radar imaging usingmulti-frequency diffraction tomography,” AdvancedMicrowave and Millimeter Wave Detectors, (SPIE),Vol. 2275, pp. 25−26.

8. Azevedo, S. G., J. E. Mast, S. D. Nelson,E. T. Rosenbury, H. E. Jones,T. E. McEwan,D. J. Mullenhoff, R. E. Hugenberger, R. D. Stever,J. P. Warhus, and M. G. Wieting (1996), “HERMES: Ahigh-speed radar imaging system for inspection ofbridge decks,” Nondestructive Evaluation Techniquesfor Aging Infrastructure and Manufacturing, (SPIE),Vol. 2946,(23) in press.

9. Gavel, D. T., J. E. Mast, J. Warhus, and S. G. Azevedo(1995), “An Impulse Radar Array for DetectingLandmines,” Proceedings of the Autonomous Vehiclesin Mine Countermeasures Symposium, Monterey,Calif., pp. 6−112.

10. Azevedo, S. G., D. T. Gavel, J. E. Mast, and J. P.Warhus (1995), “Landmine Detection and Imagingusing Micropower Impulse Radar (MIR),” Proceedingsof the Workshop on Anti-personnel Mine Detectionand Removal, Lausanne, Switzerland, pp. 48−51.



Introduction

A U.S. patent is pending for this work.We have pursued the development of compact,

high-current (>2 kA), high-gradient acceleratorsystems for various Department of Energy missionsover the past several years. This work has mainly

focused on a new high-gradient, prompt pulse (onthe order of 10 to 50 ns) accelerator conceptcalled the Dielectric Wall Accelerator (DWA).1 Thepulsed electric field in this accelerator is devel-oped by a series of asymmetric Blumleins incorpo-rated into the insulator structure (Fig. 1). When thisstructure is combined with the new high-gradient


igh-Performance Insulator Structures for Accelerator Applications

6-17

Stephen E. Sampayan and David O. TrimbleLaser Engineering DivisionElectronics Engineering

George J. Caporaso and Yu-Jiuan ChenInertial Confinement Fusion ProgramLaser Programs

Clifford L. HolmesApplied Research EngineeringMechanical Engineering

Robert D. Stoddard and Ted F. WieskampDefense Technologies Engineering DivisionMechanical Engineering

M. L. Krogh and S. C. DavisAllied Signal CorporationKansas City, Missouri

It is experimentally observed that insulators composed of finely spaced alternating layers of dielec-tric (<1 mm) and thin metal sheets have substantially greater vacuum surface flashover capabilitythan insulators made from a single uniform substrate. A conclusive theory that fully explains thiseffect has yet to be presented. The increased breakdown electric field that these structures exhibitmay result either separately or in combination from 1) minimized secondary electron emissionavalanche (SEEA) growth; 2) shielding of the insulator from the effects of charging; or 3) a modifica-tion of the statistical nature of the breakdown process by separating the structure into N-1 additionalsub-structures. We have previously performed measurements and reported on small- to moderate-sized insulator structures. In the previous work we showed these structures to sustain electric fields1.5 to 4 times that of a similar conventional single substrate insulator. In addition, we previouslyreported on the capability of these structures under various pulsed conditions, in the presence of acathode and electron beam, and under the influence of intense optical illumination. In this paper wedescribe our on-going studies investigating the degradation of the breakdown electric field resultingfrom alternate fabrication techniques, the effect of gas pressure, and the effect of the insulator-to-electrode interface gap spacing. Additionally, we report on initial testing that subjects the insulatorto the effect of energetic radiation fields.


vacuum insulator technology reported here, short-pulse high gradients of greater than 20 to 30 MV/mmay be possible.2,3

The asymmetric Blumlein functions as follows.Two stacked pulse-forming lines, form the asymmet-ric Blumlein each with a different transit time andideally, equal impedance. In the ideal configuration,these Blumleins consist of alternating layers of twodissimilar dielectric materials with permitivities, εr,which differ by a ratio of 9:1.

When the conductor in common with both lines ischarged to potential, Vo, and shorted on the circum-ference of the accelerator structure, two reversedpolarity wavefronts move at a velocity proportionalto εr

–0.5 toward the beam tube. For a fast pulse line length of time, t, and a slow

pulse line length of time, 3t, an energy gain of 2Vooccurs across a single Blumlein structure into amatched beam load over the interval t to 3t.

The maximum gradient of this accelerator isdefined by the dielectric strength of the walldielectrics and the maximum pulsed surface break-down electric field capability of the interiorvacuum interface in the acceleration region. Mostdielectric materials can support the required gradi-ents; the vacuum insulator structures generally donot. To maximize these gradients, we have under-taken to improve the overall performance ofvacuum insulators.

In addition to this particular accelerator applica-tion, we are pursuing other near-term applications.These include the Advanced Hydrotest Linear

Induction Accelerator (AHF-LIA) proposed byLawrence Livermore National Laboratory, and theDual Axis Radiography Hydrotest (DARHT)Accelerator presently being built at Los AlamosNational Laboratory. In these accelerators, high-performance insulators will be required to optimizeaccelerator gap design for long pulses (on theorder of 2 µs) on the AHF-LIA system and also formulti-pulse options being considered for theDARHT system.

A high-gradient insulator consists of a series ofvery thin (<1 mm) stacked laminations interleavedwith conductive planes. This insulator technologywas originally conceived and disclosed by Gray inthe early 1980’s4 and resulted from experimentalobservations that the threshold electric field forsurface flashover increases with deceased insulatorlength (Fig. 2).5,6 Further investigations showedsubstantial increases in the breakdown threshold ofthese insulator structures over conventional, singlesubstrate insulators.2 More recent data shows anincrease of 1.5 to 4.0 times that over conventionalinsulator technology.3 We have also explored theproperties of these structures in the context ofswitching applications, investigating their behaviorunder high-fluence photon bombardment.7

A certain amount of understanding of theincreased breakdown threshold of these structurescan be realized from the basic model of surfaceflashover. The most simplified vacuum surfacebreakdown model suggests that electrons originat-ing from the cathode-insulator junction are


CLCLCLVacuum wall

++_ _ ++ __

+ _

++_

++ _

_

Charged Switch Acceleration

Bre

akd

ow

n e

lect

ric

fiel

d (

kV/c

m)

101

102

102101100

Insulator length (mm)

Lucite (+45°, 5 µs pulse)

Lucite (-45°, 5 µs pulse)

Plexiglass

Teflon

Polyethylene

Alumina ceramic

Pyrex glass

Figure 1. Principle of the asymmetric Blumlein in the DielectricWall Accelerator.

Figure 2. The effect of length scaling on the surface break-down electric field of vacuum insulators.


responsible for initiating the failure.8 When theseelectrons are intercepted by the insulator, additionalelectrons, based on the secondary emission coeffi-cient of the surface, are liberated.

This effect leaves a net positive charge on the insu-lator surface, attracting more electrons and leading toescalation of the effect or SEEA breakdown.

It has been shown that full evolution of thedischarge occurs within 0.5 mm.9 Thus, placingslightly protruding metallic structures at an equiva-lent interval is believed to interrupt the SEEAprocess and allow the insulator to achieve highergradients before failure. Alternate modifications tothis explanation include the effects of insulatorshielding and equilibration of the induced surfacecharge. As a result, electron impact on thesurface is modified. Or, alternately, by separationof the insulator into N-1 additional decoupled sub-structures, a local breakdown on the insulatorcannot propagate to the remainder of the structure.

In this paper we describe our on-going work inwhich we have performed additional studies on theeffect of various fabrication techniques, the effect ofgas pressure, and the effect of the insulator-to-electrode interface spacing. Additionally, wereport on initial testing which subjects the insulatorto energetic radiation fields.

Progress

Small sample testing (approximately 2.5 cmdiameter by 0.5 cm thick) was performed in a turbo-molecular pumped, stainless-steel chamber atapproximately 10−6 Torr. High voltage was devel-oped with a 10 J “mini-Marx.” The Marx developeda pulsed voltage of approximately 1 to 10 µs (base-to-base) and up to 250 kV amplitude across thesample. Diagnostics consisted of an electric fieldsensor and a current viewing resistor. Failure of theinsulator was determined by a prompt increase inMarx current and a prompt collapse in the voltageacross the sample.

Several small sample insulators were fabricatedby interleaving layers of 0.25-mm fused silica,formed by depositing gold on each planar insulatorsurface by a sputtering technique and then bondingthe stacked layers by heating while applying pres-sure. Bond strength between the gold layer andsubstrate using this technique was measured toexceed 10 kpsi. To perform the breakdown experi-ments, the structure was slightly compressedbetween highly polished bare aluminum electrodesthat establish the electric field for the tests.

To obtain a particular data set, the insulatorswere subjected to several low-voltage conditioningpulses. The voltage was then increased a smallamount incrementally until breakdown occurred.Voltage was reduced for several shots and thenincrementally increased again until a constant valuewas achieved. In these experiments, however, wegenerally observed that the insulators did not condi-tion. Once a breakdown occurred at a particularfield, reducing the voltage slightly and increasing itagain did not cause an increase in breakdown field.

To produce a given data set we applied up to 150to 200 shots to a given structure and attempted todetermine if any damage to the structure occurredthat significantly altered the breakdown characteris-tics. At these applied energies, we generally did notobserve any degradation. These data were thenreduced to reliability plots by determining the totalnumber of successful shots over the total number ofapplied shots. In these data we define the electricfield as the applied voltage divided by the total insu-lator length. We define reliability at a given electricfield as the total number of successful shots over thetotal number of shots.

Using this method, we observed flashover of thesmall samples at approximately 175 kV/cm for thefused silica substrates (Fig. 3). The effect of pulsewidth from 1 to 10 µs on this breakdown thresholdwas well within the statistical nature of our data.The trend in conventional insulator technology(Fig. 4) for 0°-insulators indicates a breakdownthreshold of approximately 50 kV/cm. Thus, with


250

1.2

1.0

0.0

0.2

0.4

0.6

0.8

Rel

iab

ility

20050 100 1500Electric field (kV/cm)

Figure 3. Pulsed surface breakdown reliability of a ground-fused silica high-gradient insulator.


these insulators there was a net increase in theperformance over conventional technology by afactor of approximately 3.5.

To ensure concentricity on these first structures,a finish grinding operation was performed on theoutside diameter. Since this process is a time-consuming second operation, an alternate fabrica-tion means was pursued. To simplify fabrication, weattempted an ultrasonic machining process.Although it was possible to fabricate the part in a

single operation, the surface was left slightlyrougher. Comparison of the breakdown characteris-tics of these samples showed significantly morescatter and on average a slightly decreased break-down threshold of approximately 25% (Fig. 5).

The structures were also subjected to increasedpressures to determine susceptibility to breakdown(Fig. 6). In these data, using the previouslydescribed procedure, a fixed reliability was estab-lished at the various pressures. All data was then


Bre

akd

ow

n f

ield

(kV

/cm

)

500

400

300

200

100

0

Pulse width (ns)

103102101100

Anderson, 1976Ohki, 1982Thompson, 1980Anderson, 1980Glock, 1969Vogtlin, 1989Watson, 1967Milton, 1970

Figure 4. Pulsed surface breakdown electric field as a functionof pulse width for single substrate, straight wall insulators.

Rel

iab

ility

0.0

0.2

0.4

0.6

0.8

1.2

1.0

2001208040 1600Electric field (kV/cm)

Figure 5. Pulsed surface breakdown reliability of a fused silicahigh-gradient insulator fabricated using an ultrasonic fabricationtechnique.

0.0

0.2

0.4

0.6

0.8

1.2

1.0

10110010-6 10-5 10-4 10-3 10-2 10-1

Pressure (Torr)

No

rmal

ized

bre

akd

ow

n e

lect

ric

fiel

d

High gradient structure 0°, 5 µs

Smith (PMMA)45°, 1 µs

Smith (PMMA)25°, 30 ns

Figure 6. Effect of gas pressure on the performance of high-gradient insulators.

No

rmal

ized

bre

akd

ow

n e

lect

ric

fiel

d

0.0

0.8

0.6

0.4

0.2

1.2

1.0

1251007550250Gap spacing (µm)

Figure 7. Effect of an increased electrode/insulator interface gap.


normalized to a mean breakdown electric field.Susceptibility to breakdown stays relatively constantup to about the 10−3 Torr range, at which point, thefield at which breakdown occurs decreases rapidly.Also shown are data from previous work by Smith.10

It appears that the new structures show a lowerbreakdown electric field threshold than that of theprevious data.

Any insulator not in full contact with the elec-trode surface will show a higher susceptibility tobreakdown and lower reliability at a given electricfield. This effect results from the enhanced electricfield that occurs between the insulator/electrodeinterface gap. To investigate this effect with thesenew structures, shims were placed between thecathode electrode and insulator, and the reliabilityat a given electric field were determined. This data,normalized to the configuration where the insulatorwas flush with the electrode, is shown Fig. 7. Inthese tests, we observed the reduction in the capa-bility of the insulator to be strongly reduced fromabout 90% of full capability for a 12-µm interfacegap to less than 60% for a 125-µm interface gap.

We have also begun testing these structures inthe presence of an ion beam and various other radi-ation fields. In this test, we used a high-currentpulsed-ion source. The ions are allowed to impingeon the structures near the cathode triple junctionwhile a high potential is applied across the sample.Our observations to date are somewhat qualitativeand indicate that the ion beam does not induce animmediate and prompt breakdown on impact.Rather, we observe only a somewhat reduced break-down electric field capability resulting from direction impact on the insulator surface.

References

1. Caporaso, G. (1994), “Induction LINACS and PulsedPower,” Frontiers of Accelerator Technology, Proc.1994 Joint Topical Course, Maui, HI.

2. Elizondo, J., and A. Rodriguez (1992), “Novel HighVoltage Vacuum Surface Flashover InsulatorTechnology,” Proceedings of 1992 15th InternationalSymposium on Discharges and Electrical Insulationin Vacuum, Vde-Verlag Gmbh, Berlin, Germany,pp. 198−202.

3. Sampayan, S., G. Caporaso, Y. Chen, C. Holmes, E.Lauer, D. Trimble, B. Carder, J. Elizondo, M. Krogh,B. Rosenblum, C. Eichenberger, and J. Fockler, “HighGradient Insulator Technology for the Dielectric WallAccelerator,” (1995), Proceedings of the 1995Particle Accelerator Conference, (IEEE), New York,N.Y., pp.1269−1271.

4. Gray, E., (1984), private communication.

5. Milton, O. (1972), “Pulsed Flashover of Insulators inVacuum,” IEEE Trans. Electr. Insul., Vol. EI-7,pp. 9−15.

6. Pillai, A. S., and R. Hackam (1982), “SurfaceFlashover of Solid Dielectric in Vacuum,” J. Appl.Phys., Vol. 53(4), pp. 2983−2987.

7. Sampayan, S., G. Caporaso, M. Norton, D. Trimble, B.Carder, and J. Elizondo, “Optically Induced SurfaceFlashover Switching for the Dielectric WallAccelerator,” (1995) Proceedings of the 1995Particle Accelerator Conference, (IEEE), New York,N.Y., pp. 2123−2125.

8. Miller, H. C. (1988), “Surface Flashover ofInsulators,” G. E. Aerospace Report, (GEPP-TIS-1064-UC-13).

9. Glock, W., and S. Linke (1969), “Pulsed High-VoltageFlashover of Vacuum Dielectric Interfaces,” CornellUniversity Report, Laboratory of Plasma Studies,Ithaca, N.Y., (No. LPS 24).

10. Smith, I. D. “Pulsed Breakdown of Insulator Surfacesin Poor Vacuum,” unpublished AWRE report,Aldermaston, United Kingdom.



Introduction

Spark-gap switches are commonly used in Marxgenerators and other high-energy pulsed-powersources. This type of switch consists of two metalelectrodes, separated by a gap filled with pressur-ized gas contained within a dielectric housing. Theswitch operates when the electric field in the gapexceeds the breakdown level of the gas, or whencharge carriers are introduced into the gap by someexternal means.

A Marx generator uses a fundamental circuit thatcharges many capacitors in parallel, then switchesthem into a series configuration to generate a high-voltage output equal to the sum of the voltagesacross each capacitor. In principle, Marx genera-tors are simple, and their capability is largely deter-mined by the basic capacitors and switches thatmake up each stage. In practice, the performanceof Marx generators depends critically on physicallayout and construction details.

In most applications, Marx generators areconstructed with standard, commercially availablecomponents that are arranged in an orderly mannerinto compact assemblies. Since these componentsare made for general use, their packaging and termi-nal styles are usually less than optimal for achievingmaximum performance and the compactness needed

for some applications. Significant improvementscan be realized by modifying standard componentsor by manufacturing special components to fit intointegrated packages.

We have developed a design for a repetitively-switched, Marx-type high-voltage generator basedon custom components that can be closely coupledand integrated into an extremely compact assembly.This ultra-compact Marx (UCM) can be used in avariety of special applications requiring a compacthigh-voltage pulsed-power source. The conceptrelies on a low-profile, low-inductance, high-voltage,spark-gap switch with the following performancelevels:

Hold-off voltage: 100 kVPeak current: 30 kARepetition rate: 10 HzCharge transfer: 0.1 CInductance: 5 nHCapacitance: 130 pF

Progress

We have evaluated many design options andperformed field modeling to evaluate the electricalstresses and quantify field enhancements of differentshaped assemblies. Through an iterative process,we have been able to identify the limiting features


ompact Gas Switch Development

6-23

David A. Goerz, Michael J. Wilson, and Ronnie D. SpeerDefense Sciences Engineering DivisionElectronics Engineering

Joseph P. PenlandDefense Technologies Engineering DivisionMechanical Engineering

We have developed a low-profile, high-voltage, spark-gap switch designed to be closely coupledwith other components into an integrated high-energy pulsed-power source. We performed fieldmodeling to determine the appropriate shape for the highly stressed insulator and electrodes, andemployed special manufacturing techniques to mitigate the usual mechanisms that induce breakdownand failure in solid dielectrics. We have constructed and tested a prototype switch unit and achievedsatisfactory operation at 100 kV levels. Preliminary tests to evaluate repetitive operation and lifetimehave been encouraging.


and devise suitable design and construction meth-ods to satisfy the basic requirements. Figure 1a isa sketch of the compact gas switch.

The shape of the metallic and dielectric parts iscrucial to properly manage the electric fields andkeep the stresses below the threshold for flashoveror breakdown of a material. Figures 1b and 1cshow field modeling results where contours ofequipotential lines are plotted. The metal electrodesand insulator surfaces are appropriately shaped toreduce electric field stresses in the weakest regionswhere dissimilar materials meet, and to spread thefields evenly throughout the dielectric materials,allowing them to operate closer to their intrinsicbreakdown levels.

Figure 1b shows that for the chosen dimensionswith an applied voltage of 100 kV, the electrical fieldis less than 30 kV/cm at the triple-point regionwhere the metal, plastic, and gas all meet, and thefield is less than 130 kV/cm along the envelope ofplastic material containing the pressurized gas.These stress levels are below the thresholdsreported by others as troublesome.1,2

The average field across the thinnest annularregion of plastic is 500 kV/cm, whereas the highestfield at the outermost enhanced region is 575 kV/cm.

The gap between the cathode and anode electrodesis sized according to the desired operating level andgas pressure for a particular gas species or mixture. Anominal gap of 2.0 mm will normally breakdown at100 kV with SF6 pressurized to 100 psia. Fieldmodeling allowed us to correct the empirical rela-tionship for gas breakdown of planar gaps by includ-ing the effect of field enhancement at the edges. Asevident in Fig. 1b, the fields are enhanced along thesurface of the electrodes, reaching 560 kV/cm. Thisis described as an enhancement factor (f*) of 1.120,given by the ratio of 560 kV/cm to 500 kV/cm.3

We constructed a special test fixture to evaluatevarious switches and demonstrate their performanceover the desired range of operating parameters.Figure 2a is a photograph of the assembled unit nextto a high-voltage probe used to monitor the chargevoltage. The slotted metal cylinder is the outer currentreturn path for performing current ringdown tests.The finned cylindrical structure below the test fixtureis a high-energy current-viewing resistor used as adiagnostic. Figure 2b is a sketch of the cross-sectionof the test fixture showing a 1-cm-thick switch bodynestled between a ring of ceramic capacitors on topand a metal spacer below, with channels for flowinggas through the switch electrodes.


500 kVcm

130 kVcm 500 kV

cm

Metal

∈ℜ=6.0

∈ℜ=4.5

∈ℜ=4.5

∈ℜ=3.2

∈ℜ=6.0

500 kVcm

275 kVcm

575 kVcm

2mm

Soliddielectric Metal

O-ring(a)

(b)

Electrode

(1b) (1c)

Gas

560 kVcm

(c)

30 kVcm

Figure 1. (a) Schematic of compact gas switch; (b) and (c) show field modeling results.



yyyy

yyyyy

yyyyyyy

yyyyyyyyy

yyyyyyyyyyy

yyyyyyyyyyyyy

yyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyyyyyyy










yyyyyyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyyyy

yyyyyyyyyyyyyyy

yyyyyyyyyyyyy

yyyyyyyyyyy

yyyyyyyyy

yyyyyyy

yyyyy

yyy

CapacitorCapacitorCapacitorCapacitor CapacitorCapacitor

HVcable

Gasline

(a)

(b)

Metal spacerMetal spacer

Switch body

MetalMetalelectrodeselectrodes

Metalelectrodes

AcrAcrylicyliccylindercylinder

Acryliccylinder

GasGasdielectricdielectric

Gasdielectric

Metal spacerMetal spacer Gas channelGas channel

CVR bodyCVR body

InsulatorInsulator

CVRterminal

Spark gapSpark gap

Metal spacer

Metal spacer Gas channel

CVR body

Insulator

Spark gap

Slotted metal cylinder

Figure 2. (a) Photograph of apparatus; (b) sketch of cross-section of test fixture.


Three different sets of electrodes were used tocharacterize the switch with nominal gap spacingsof 2.0, 2.5, and 3.0 mm. Figure 3 shows resultsfrom testing the static breakdown performance of thecompact gas switch. The solid markers representthe breakdown measurements that were made ateach pressure increment of 5 psi. The open markersrepresent the well-known relationship for breakdownin SF6 gas3 corrected for the field enhancements forthese particular shaped electrodes. While there issome statistical variation in the breakdown levels,the general agreement with the empirical modelindicates that the switch does indeed function asintended. Considerable testing confirmed that theelectric field stress has been properly managed atthe triple points and along the insulator surface, andthat the switch performs as expected.

Further testing was done to exercise the compactgas switch at expected peak currents and charge-transfer levels to determine whether the insulatorwould survive repetitive high-energy pulses, orwhether electrode erosion would become trouble-some. Figure 4 shows a typical voltage trace froma 4-s burst-mode operation of a 3-mm gap switchwith SF6 at 65 psia and the 100-kV power supplycurrent limited by a 20-MΩ series resistor.Figure 4 shows the current trace from the ringdownevent. In this case the oscilloscope was set up toaverage the first 30 ringdown current waveforms.

This mode of operation was also used to deter-mine the statistical variation of voltage breakdownlevels to evaluate how well the switch would function

in a Marx-type high-voltage generator. Figure 5shows a histogram of the number of switch break-downs versus voltage level. For this 4-s burst themean operating voltage was 88.8 kV and the stan-dard deviation was 4.5 kV or 5.0%. This statisticalvariation is adequate for a Marx generator to func-tion reliably with a reasonable amount of voltagecoupling between stages to ensure successiveswitch operation.

To evaluate the expected lifetime of such acompact switch assembly, multiple 4-s bursts weretaken, and the voltage and ringdown current wave-forms were recorded for each shot. The switch stilloperated satisfactorily after more than 175 bursts,totaling more than 7000 individual shots. This seriesof tests was concluded when one of the capacitors


Pressure (psia)20 30 40 50 60 70 80 90 100

Stat

ic b

reak

do

wn

(kV

)

100

80

60

40

20

0

Model SF6 gap d/f=0.30/1.170Model SF6 gap d/f=0.25/1.175Model SF6 gap d/f=0.20/1.120SW#2 with 3.0-mm gapSW#2 with 2.5-mm gapSW#2 with 2.0-mm gap

SW#2 Vbd vs P vs gap

Figure 3. Test results from static breakdown performance oncompact gas switch.

-20

-15

-10

-5

0

5

10

15

20

Cur

ren

t (k

A)

0 0.5 1 1.5Time (µs)

Average of first 30 ringdown waveforms

SW#2c, 3-mm, 65psia, 091696

(b)

100SW#2c, 3-mm, 65psia, 091696(a)

80

60

40

20

00 4321

Time (s)

Self

-bre

ak v

olt

age

(kV

)

Figure 4. (a) Typical voltage trace; (b) current trace from ringdown event.


failed. The switch was inspected and later reinstalledfor further tests. Figure 6a shows a close-upphotograph of the Lexan switch housing. No trackingor deterioration could be seen.

The anode and cathode button electrodes wereweighed before and after the lifetime tests describedabove. The mass loss was 5.4 mg for each elec-trode. The total charge transferred through theswitch during the period in which these electrodeswere installed was 30.4 C. The specific mass lossamounted to 0.178 mg/C.

Several types of insulator materials were used toevaluate machining methods. Most of the electricaltesting was done with a polycarbonate material(Lexan). Figure 6b shows a close-up photograph ofone switch housing made from a large Lexan cylin-der. Stress fractures were apparent after machiningthe inner switch cavity. This particular switch stilloperated satisfactorily for more than 1000 shots;however, the crazing became progressively worse.Switch housings were also manufactured usingalumina-trihydrate (ATH) loaded epoxy andEPON-825 thermoset resin. ATH machined veryeasily, whereas some difficulty was experiencedmachining the EPON material.

Future Work

We plan to continue the development of thecompact gas switch and demonstrate its perfor-mance in an ultra-compact, high-voltage Marxgenerator. More effort will go into evaluating higher-strength materials and developing advanced manufac-turing methods for making intricately shaped parts.

We will conduct further tests on the differentmaterials and assemblies to fully characterizetheir performance levels. We anticipate a need toperform 3-D analysis of electric field enhance-ments to determine optimal shapes of parts andinterconnections for an integrated Marx package.


20

15

10

5

0

Co

unt

30 40 50 60 70 80Range (kV)

90 100

@60 psia, Avg=88.8 kV, StdDev=5.0%

091396 Vbd(P) Stats

Figure 5. Histogram of switch breakdowns as a function ofvoltage level.

(a)

(b)

Figure 6. (a) Photograph of tested Lexan switch housing;(b) photograph of machined Lexan switch housing with stressfractures.


Acknowledgments

The authors gratefully acknowledge the expertiseand support of the Computational Electronics andElectromagnetics Thrust Area and the valuablecontribution of W. Ng who performed thefield modeling.

References

1. Laghari, J. R. (1981), “Surface Flashover of SpacersIn Compressed Gas Insulated Systems,” IEEETransactions on Electrical Insulation, Vol. EI-16 (5).

2. Laghari, J. R. (1985), “Spacer Flashover inCompressed Gases,” IEEE Transactions on ElectricalInsulation, Vol. EI-20 (1).

3. Alston, L. L. (1968), “Breakdown Characteristics inGases,” High Voltage Technology, Oxford UniversityPress, pp. 45–58.


EnginReseaDeveloand Te

EnginReseaDeveloand Te

THRUST AREA REPORT

Engineering R

esearch, Developm

ent and Technology • FY 96

Thrust Area R

eport • UC

RL 5

38

68

-96

Technical Information DepartmentLawrence Livermore National LaboratoryUniversity of CaliforniaLivermore, California 94551

power conversion technologies...reprinted from engineering research, development and technology fy...

Documents