ET PROGRESS IN THE PRODUCTION OF INTEGRATED CIRCUITS
CIA HISTORICAL REVIEW PROGRAM RELEASE AS9
This paper, prepared by
Office of Economic Research, analyzes and assesse evidence concerning the output and quality of Soviet integrated circuits nd the apparen need of the USSR for Western IC productionand machinery. This paper has not been formerly coordinated within the US intelligencey.
Soviet Progress in theof" integrated Circuits
USSRery active program toand produce integrated circuits (iCs)." lines have been set up andoperating in at least three plants. Therereports of production lines infacilities and of two additional plantsin producing ICs. Moreover, theexhibited IC devices at internationalyearnd each year the numbercf devices exhibited has grown larger. ome items of IC productionhave also been displayed. Thesebeen interpreted by some observers as anof the growing sophistication andthe USSR in IC production.
the other hand, US laboratory analysis
of samples of Soviet devices that have become indicates that the design of Soviet ICS is relatively primitive and that the quality ispoor and inconsistent and is clearly inferior to counterpart devices produced in the United States. Soviet devices, even those1 factory'markings, appear to be prototypes. laboratory analysisthat the USSR, as recentlytill had not standardized its IC design. In addition, no Soviet series-produced civil electronic equipment is knev/n zo contain ICs, and thare is no evidence tr.at any military equipment being produced is using ICs. If the USSR is producing ICs on an industrial scale, it is unclear where they are going or how they are being used. Also puzzling, if the USSR has developed a large and viable IC industry, is
* Unless otherwise specified, the term integrated circ-it refersono!ithically" structured device in which all components (active and passive) are diffused into the same (silicon) substrate.
che recent Soviet interest in purchasing IC machinery and technologyarge-scale from the West.
3. Because the USSR was very late in acquiring silicon planar technology, and because of persistent difficulties in producing high-quality silicon starting material in quantity, it is believed that ICs began to be produced in manufacturing plants only relatively recently, and that production is stillery snail scale.
4. IC technology is an outgrowth of advances in silicon epitaxial-planar transistor technology, and ICs are produced using essentially the same equipment and technology. It is not clear when the USSR firstapability to manufacture silicon enitaxial-planar transistors. ew silicon diodes and transistors based on diffused_junction technology may have been produced, but almost certainly epitaxial-planar transistors or other devices such as field-effect transistors (FETs) that are based on silicon planar technology were not being produced." In fact, 5 the Soviet semiconductor industry was very small by US standards.** Relatively few types ofwere in production, and most were simple aermanium types based on point-contact, alloyed, and diffused junction technologies. Most of the
to the Soviet press, the first domestic consumer radio to use silicon transistors was theanufactured in It is estimated that the USSR, illion illion semiconductors (mostly germanium and largely diodes) compared to an out-out of moreillion units in the United States. Actually, the level of output for the united States is much higher. Large US firms such as IBM manufacture large quantities ofors for their own use which are net reportedS statistical data.
electronics equipment produced in the USSR at that tine, both civil and military, was based on vacuum tubes.
first evidence that the USSR hadtransistors using planar technologyofficial Soviet brochure published on electronics hardware arewhen only experimental prototypesmade, this evidence cannot be takenplanar devices were in serial productiontima. Probably, the first limitedsilicon transistors based on planarplace* The firsttransistor tovailable forUS experts was rranufacturedndSoviet production line for planarwas seen for the first tinot the Svetlana plant in
Silicon Materials Technology
In contrast,lectronic component production in the United States was based almost entirely onajor shift to the production of silicon semiconductors had taken place, and the basic technology and industrial 'base for the large-scale production of monolithic ICs hod been developed.
** As oflanar and epitaxial-planar transistors accounted for only about ono-ttnth of the total number of types ot transistors carried in Soviet catalogues.
*'* Visitors to Svetlana8 did not report,pparently were not shown,ine.
t aw material, silicon is one of the most abundant substances in the earth's crust, ffowever, in order to transform basic silicon to the stats of very high purity required for the manufacture of advanced types of semiconductors, complexsophisticated controls, and criticalknow-how are required. Silicon material passes through several precision manufacturingi steps toingle or monocrystalline stake.
is believed that planar devicesor ICs) could not have been ina large scale0 at the earliestof the limited availability ofmonocrystalline silicon.t The evidence
of Soviet shortages of -his material abounds. although the USSR had offeredsilicon for sale in the Westin Europe alleged that only veryof electronic grade naterial werefor purchase. Moreover, the qualitythat was purchased was not uniformfrom batch toom-
plained1 that they we're able to acquiresmall quantities of Soviet high-puritystated1 that
both germanium ana silicon crystals producedwe-" quality. Inthat the silicon they
were receiving from the USSR was of very poorFranco-Polish negotiations1 forof semiconductor technology andPoland, the French allegedly made it clearwould not honor their contract commitments used Soviet silicon. At the ThirdConference or. Crystal Growth held in Mar-France, in
Soviet hardware usee in growing single
crystal* was very crude and that the materialdid not measure up to US standards. Soviet officials in attendance indicated that they were dependent on west European sources for crystal pullers for both germanium and silicon and expressed keen interest ir. purchasing US crystal growth equipment.
7. Finally, there is some evidence that the USSR may not be producing enough high-purityto meet its semiconductor needs.' estfirm was recently reported doing "substantial" business supplying electronic-grade silicon to the USSR and to other Communist countries in Eastern Europe. Sales were being made outside of legal channels to circumvent the embargo.
3. Soviet state-of-the-art in integratedlags significantly behind that of the United States. Technical analysis of Soviet ICs by us experts indicates serious deficiencies in design, processing, and fabrication and suggests that, at least ICs were still in developmental prototype production. By and large, the USSR is
manufacturing only small-scale integration (SSI) devices; transistor-transistor logic (TTL) and diode-transistor logic (DTL). uch lesser degree, emitter-coupled logic (ECL) and metal-oxide semiconductors (MOS) are also being made. The production of medium-scale integration (MSI) and large-scale integration (LSI) devices isseveral years away at best, althoughcould be rapidly accelerated if Western know-how and processing technology should become State-of-the-art for individual types of ICs based on laboratory analysis of Soviet devices ia discussed below.
9. Three types of Soviet DTL ICs, manufactured* have been evaluated by two major USof ICs. Both firms assessed the level of technology asy US standards. The most significant finding was the Soviet use of dielectric isolation, which suggests an attempt at radiation hardening. However, radiation protection was only partial, at best, since the Soviet devices had no compensating diodes for protection against photo-current, and theyold-aluminum metal system whichess desirable radiation tolerance than moly-gold or all aluminum. Both evaluationsumber of processing/design weaknesses in the Soviet devices. Theseack of passivation protection over the metallization, oxide defects, very poor bonding and scribing, low device density per chip, poor control of theprocess (indicated by wide variation ofvoltages from device to device), ond useoor photoresist as evidenced by pinholes, tears, and undercutting. The IC packaging (glass-metal flat packs) also contained weaknesses such as war-page induced during the molding process, loads flush with the bottom of the package (increasing the chances ofoor thermal conductivity of the glassigh expansion coefficient o* the Sovietnd glass which implied potential
first year that the USSR ex-
*Tlbitec IC devices.
"* etallic alloy commonly used in makingf the metal parts of metal-glass IC packages.
problems with seal integrity at the Kovar-to-glass interface. The devices appeared to be prototypes.
Hybrid Thin Film"
of two Soviet hybrid devices,to have been manufactured1 anddesigned for computernd evaluatedeading US producer Both were thin-film devices transistors added. The transistorsNPN silicon devices with electricalcomparable to those found in common
US commercial applications. However, these hybrid devices were adjudged extrcnely large for their intended function (each package would take0 square inch of printed circuit boardnd both the workmanship and the quality of part of the thin-film network were poor. The packages were poor also in terms of hermetricity integrity, thermal expansion, and mechanical strength. These devices were believed to be prototypes.
Soviet digital KOS1 have been examined L
' "Hybrid thin film" devices are not strictly speaking integrated circuits. However, they are fabricated by advanced techniques, are commonly used together with ICs in high reliabilityapplications in the United States- and may be considered illustrative of Sovietrt in IC technology taken in the broad sense.
An unmarked dual flip-flop with twoates,uad NOR gate marked* Three Input NAND NCR gates, marked 8PMD16A.
The devices were packaged in glass ceramic (molded borosilicate) flat packs. Such packaging isexpensive and nayilitary One MOS chip contained corrosion products indicating that it was nonhermetic at one time or that contaminants were sealed in. aoLhhewed numerous scratches in the metallization; oxide defects, and much foreign material attached to or in the oxide. Bond wires were in close proximity to the edge of the chips, and there was
no passivation protection over thecribe lines were very rough, mask misalignment was evident, and gate oxide was exposed. The diffusions were very large. Chip size varied because ofscribing: the average size wasils bylis. Given the same size chip, current US MOS capabilities wouldensity about one order cf* magnitude greater. In sum, these MOS ICsvery poor workmanship by US standards and were probably prototypes."*
12. ew TTL devices manufactured0 aod
ave beeii evaluated
Insxperts found tr.at tne auyiB*.for botn years compared favorably inwith counterpart Texas Instruments'and that they are in "intensive" In contrast, US experts in theirthe datareports have
concluded thaT tne Soviet cevices compared very -& ^ovice.H and ti'.at e design oi tho devices suggests an experimental or pilot line, rather than commercial-scaie production. Themajor ce: lc-.encies in design and processing
Soviet "ICs examined to dote exhibit the
use of oassivation protection over the metallization.
The poor quality of Soviet MOS ICs is notsince MOS transistors have themselves- been developed only recently. To date, no samples of MOS transistors have been acquired, and no MOS
transistor production has been observed by Western
visiters.to the USSR,
For example, the Soviet device nput NAND gate'markedas said tospeed ofanoseconds (ns) comparedtypical" speed of s for the TI device The speed of the TI device appears to be incorrect. The TI catalog- Integrated Circuits from Texas Instruments,)ypical values for this devjice. US experts state that if the TI devicetypical" value of I; ns it would he non-competitive in the United States.
(11 Chip size was very largerrore thar. twice that o: -J Tr.e ize reans that less than one-half as many units per wafer car. be produced, and the reject ratio will be nore than twice as high.
(2) input leakage was highabout ten times that of TI circuits indicating poor surface processing and passivation techniques, as well as poor emittertechniques.
S3) The power (current) consumption of the Soviet devices wasigher than in the TI devices, indicatinghigher operating temperatures and
potentially degraded reliability.*
ii) The Soviet chips includedtransistors and resistors, suggesting that the Soviet devices were in aearly stage of design.** Inthe insertion of one such extra resistor in the "totex pole" outputindicates the use of poor gold doping technology.
13. that computer-aided
design (CAD! techniques were used in the-design and lay-out of the Soviet TTLs. For example, asthat masks were generated using CAD, they observe that the metallization masker that was "apparently in computer typescript."
jbelieve that this CAD association is Trr.iy tenuous. The numbers are not symmetrical (blocked) which is typically the case with print-outs end could have been formedariety of hand methods such as overlay hand printers or simply hand-cutting during conventional
corarxc substrate (rather than metallic) is used in the Soviet devices, apparently to lower the chip's high operating temperature. ** For example, TI used redundant components in its early approach (master slice). When processina techniques were perfected, however, the redundant elements were no longer included.
rubylith slice and peel. Moreover, closeof Soviet ICs indicates that maskshand-cut to obtain the metallizationconfiguration of some of the slopes incould not have been generated byto US experts in computer-aided Finally, the rather primitivethe Soviet devices seems inconsistent withcapabilities of
reportedly are being produced atElectrical Engineering Works in Bryansk,unidentified plant in Voronezh, and atplant in Leningrad. In addition,at the scientific research complex (suburb of Moscow) nay be orArfncing ICs.
plants in Borispol',insk, T'.osEbV,and Riga also are in IC development
IC production appears to be poorly organized, inade quately equipped, and grossly inefficient.
output data has been reported to Bryansk, Voronezh, and Leningrad. of ICs (DTLs and TTLs) at the Bryanskbeen estimated byvisited illion to 3pvt wooed onour operation, 6week. Production technology was said to be
ears to IC years" behind that of the Western productionin wide use. ICs
at the final test stage appeared to carryajor US producer of ICs, althoughunable to exaxine the devices closely tothat
IC production line is installedat the Svetlana plant. Output ofiICsthan, and possibly substantially less than,
1 Possiblynd Kiyev are references)sane
nits per month.* Some Western-madeis ir. use. ICs being produced include some ECL devices. The speed claimed for these devicess, well below average US state-of-the-art for ECLsnd even below US state-of-the-art for TTL ICs The relatively slow speed of the Soviet ECLs suggests that these are still under development and probably in experimental
17. InCL C
visited an IC production facility at Voronezh. He estimated total annual output of ICs atillionillion units. The plant produces dielectrically isolated SSI TTL and MOS ICslat pack configuration.t He also estimated the size of the TTL chips atilsils and the MOS chipsils Density (number of elements per chip) of both chips was very low. All of the production machineryappeared to be of Soviet design. The plant had a large number of diffusion furnaces ingle tube models)? however, onlyere in use. In contrast, there were only aboutire bonders.tit It is possible that the plant is scaling up but has not received all of its machinery.
he evidence indicates that the USSR could beillionillion devices in the three plants known to have IC production lines.
Ac"cdrSingwho visited Svetlana
inf he output uiand high-'frequency transistors combined amounted tonits per month.
** Schottky TTLIt may be noted that the speed of ECLs being produced at Svetlana is far slower thans "speed claimed by the Soviets for ECLs no re than three years ago.
r The use of dialectric isolation and flatfor the TTLs strongly suggests
tt Both chip sizes are very large for their functions by US standards.
ttt Thisery low ratio of bonders totubes by us standards. Generally, us firmsery high ratio of bonders to diffusion tubes, anywhere to
Soviet yields are probably very Low, however, and the output of high-quality, useable devices is probably substantially less thanillion. If production is also under way in some of the other facilities mentioned, gross output of ICs rejects) could be greaterillion units per year. In any case, it may be concluded that current levels of output are inadequate to meet domestic needs. olume of output of aboutillion ICs annually would be needed merely to meet the requirements of the Soviet"RYAD" computer program. Probably, the USSR wil* have toreat deal of machinery and technology from the West over the next few years if production of ICs is zo be equal to rising demand.
Soviet Interest in Western Technology
he USSR has cried repeatedlytechnology and xachinery ir. the westproduction of ICs and semiconductors The Soviets have shown interestproduction lines for both bipolar andthin-film substrates, lead frames,boards (double-sided,ransistors, diodes, andrectifiers. In addition, theto try to purchase benders, thin-film deposition photo plotters and patternand repeat cameras, complete CADgrowth furnaces, IC testers (digitalreflow and flat pack soldering testers, in-line auto coaters, for masks, molding compounds, slicers, silicon. It may also be noted that
tne Soviets, reportedly, continue to purchase large quantitiessemiconductor devices.
the past, the USSR has sought toitems of equipment in smallfrom one item to five items. of this type have been successfulfairly extensive. More recently, the USSR has beer, seeking very largeof equipment. ForistCO diffusiontC testers,robe
3indicate that the Soviets may havein acquiring two complete plants bipolar DTL and TTL circuits
and the other an MOS/MSI
In past years, the Soviets have purchased semiconductors and individual items of production machinery through third countries. More recently, however, they have been attempting to deal directly with Western firms. Probably the Soviets have found that purchases of small orders of machinery through third parties is expensive, but seldom (if ever) includes installation and technical training, creates serious maintenance problems, andery inefficient way toighly advanced and viable semiconductor industry.
On the basis of evidence of Soviet interest, the potential sales of semiconductor technology, production machinery, and related items by the West to the USSR over the next several years could be very large. Soviet interest in large-scaleof Western equipment and know-how tends to belie the suggestion that the Soviet semiconductor sectorarge, highly diversified, andadvanced industry.
USSR is engaged in an intensiveproduce monolithic ICs. some successes navebut ouput is still very smallor pilot scale by US standards. It isthat output of useable monolithic ICs infacilities known to be manufacturing ICs)was lessillion, compared to moremillion in the United States."
USSR is believed to be producingdevices, mostly TTLs and DTLs. Some.fCLsalso are being made. Hybrid ICs are under
1 Moreillion units if hybrid, IC devices are included. In addition, this figure does not include large quantities of ICs produced by'firms such as IBM which are for internal use and not
f^rly^ SUt ap?arently not in series Production. MSI and LSI devices apparently are still several years away.
devices are of uncertain andquality end reliability. Soviet ICis significantly behind that ofStates, and Soviet devices areto US counterparts in design and
The USSR does not produce enough ICs to meet its civil requirements. For example, to produce third-generation RYAD computers at the rateer yearapparently the rate originally planned for the current five-year planan annual output of aboutillion ICs would now be needed.evel of output is far beyond present Soviet capabilities.
Little is known about the production and use of ICs for military purposes. However, no Soviet military equipment is known to containcircuits, and it is believed that if sucn equipment exists, it exists in prototype form only.
Even the limited IC manufacturing capability that the USSR now has is the result, in large part, of Soviet success in acquiring crucial items of production equipment from the united states, Western Europe, and Japan. However, failure to acquire at the sarr.e time the know-how needed to
set up, operate, and maintain these equipments has retarded IC production efforts.
USSR is interested in purchasingother semiconductor productionery large scale,urnkey basis. Potential salesvery substantial over the next fewmachinery and technology to produceery large scale, the USSR, willto lag behind the United States, andgenerally, in state-of-the-art and willnot be able to produce quality ICs, atneeded to meet civil and military ne'edsleast several years.Original document.