BENSON ELECTRONICS, Inc.
In June o f 1 973 Phil Davidson, Operations Manager f or t he I ntegrated Circuit
Division a t Ben son Electronics, voiced some thoughts concerning Benson's p resent situation. As a p roducer o f components and subassemblies f or t he c omputer i ndustry, Benson's f ortunes are closely t ied t o t he c omputer m anufacturers w ho purcha se much o f Benson's production.
The c omputer i ndustry has continued t o g row d uring the past f ive t o t en years while i t has undergone a significant shake down. The period o f radical product change ha s been replaced by one t hat seeks reliable, c ompact, b ut sophisticated and cost competitive equipment. Although a n umber o f large companies have l eft t he i ndustry, there have also been some basic changes in t he w ay we m u st conduct busines s, so t hat c ompetition is n ow actually much keener and demanding t han b efore.
The increasing cost competition in computer systems is passed down t o t he m arket f or c omputer components even'though t he r ate o f technological advance here has probably continued at a pace t hat is greater t han f or t he c omputer system itself. As a m ajor m anufacturer in t he c omputer industry and as an aggressive b idder f or g overnment subcontracts, Benson has been very successful so f ar, b ut success is b ecoming m ore and more expensive. Reflecting a continuous e ffort t o d evelop and apply t he m ost advanced technology possible t o t he design and manufacture o f c omputer p roducts, Benson spe nt well above the industry average, f or research and development last year and expects t o spend more t hi s year. W ith t he n ew m arket e nvironment w ithin t hi s industry, however, simply spending m ore on research and development isn't enough. Effective product and production pro cess design as w ell as s trict cost controls are becoming more and more i mportant .
As f ar as Benson's thin-film integrated circuit division is concerned, w e c ontinue t o be among t he industry leaders in t hin-film technology, a fact which has enabled us t o m eet i ndustry requ irements f or progressively smaller, more complex and m ore reliable electronic circuitry. However, w e haven 't been able t o make any money. l et m e explain t hat f urther. M ost o f t he production act ua lly shi pped o ut o f t he p lant is in t he f orm o f su bassemblies. M y division manufactures t he t hin-film IC c omponents f or t hese subassemblies and transfers these components t o o ther d epartments a t a standard cost. So f ar we have been unable t o consiste ntly p roduce o ur i ntegrated circuits at o r b elow these standards, so, in effect, I am producing a t a loss. O ur q uality requirements are rigid and are set f or us b y t he customers w e supply, w hether t hey are o ther Benson divisions o r o utside companies.
Consequently, w e have t o l ook at t he p roduction process i tself in o ur e fforts t o c ut costs w ithout sacrificing quality. M ost o f t he areas in o ur o peration have been examined thoroughly; n ew e quipment and techniques are constantly being evaluated f or ways t o increase yields and reduce costs.
I feel o ne area w e should be investigating m ore closely is o ur p resent inspection structure . Ideally, t he costs o f inspection t o weed o ut f aulty c omponents in t he process should be balanced against the cost o f processing these f aulty c omponents f urther, and rejecting t hem later. I'm n ot c onfident t hat o ur c urrent arrangement achieves this balance.

1

BACKGROUND

A t present, t hin-film circuits are manufactured a t Benson by depositing t hin layers o f m etal on a ceramic substrate about 3 inches w ide, 4Yz inches long and onethirtieth o f an inch thick, etching away parts o f the d ifferent layers o f m etal t o f orm t he c onductor and resistor circuits, and attaching "chip" capacitors, small silicon i ntegrated circuits (STCs) and o ther " active" elements t o t he surface o f t he circuit.
The finished circuit is encapsulated in a protective insulation, leaving only t he leads exposed. Exhibit 1 shows t he integrated circuit (IC) a t various stages in t he m anufacturing process.
Benson's principal m anufacturing p lant is m odern and w ell designed. Thinfilm i ntegrated circuit production occupies o nly a small a mount o f t he p lant's f loor space b ut t urns o ut some 1 5 d ifferent integrated circuit designs in large volume.
The l abor f orce is r elatively stable and well t rained. Should additional labor be needed, the supply is ample. A t present t he IC Division is r unning t wo shifts. A t hird s hift is possible w ith o nly a m inimal addition.t o overhead, should an increase in volume w arrant a t hird s hift. Also, i f f or any reason i t is necessary, d ifferent s tations in t he m anufacturing process can go t o t hird s hift operations o r o vertime on an i ndividual basis. This has happened in t he past t o remove t emporary bottlenecks, w hich can s hift f rom o ne station t o a nother when yields a t d ifferent stations change as t he r esult o f n ew t echniques o r t emporary problems w ith a p articular o peration.
A t p resent, all m ajor e quipment is depreciated over a ten-year period. In g enerat such equipment is replaced t ry technically m ore advanced machines and techniques because i t has become obsolete o r u neconomicat n ot because i t has w orn o ut.
In examining t he m anufacturing methods used in t he p roduction o f t hin -film i ntegrated circuits, t he c asewriter f ollowed t he p roduction sequence f or o ne p articular IC, t he H-39, a c ircuit f airly typical o f Benson's IC operations.
THE INSPECTION STRUCTURE

In t he p roduction o f t he H-39, each IC passes t hrough a seri es o f operations on its way t o c ompletion. Each o peration is closely m onitored through sampling, instrumentation and supervision t o ensure t hat t he chemical o r electrical process i tself is f unctioning p roperly. Over t he past year and a h alf Phil Davidson's division has pursued an aggressive and systematic program t o increase t he e ffect iveness o f t he "process" inspections and t o reduce t he lead t ime b etween d etection and subsequent correction o f a m isfunctioning operation. This program has substantially i mproved t he yields a t each stage o f t he p roduction process, b ut i t has by n o means e liminated d efective circuits. There are several reasons f or t his. Several o f t he d elicate o perations are performed manually t hrough t he aid o f a microscope, and t he IC, p rior t o encapsulation, is very vulnerable t o damage f rom h andling. However, p rincipal causes o f t he rejects in many cases are t he technological limitations present in t he p roduction processes themselves. N ot a ll circuits processed through a particular operation w ill c onform t o specifications, despite the fact t hat t he process i tself is functioning " perfectly." In o ther words, even under ideal conditions t he yield w ill n ot be 1 00 p ercent. For this reason, there are also "100 p ercent i nspections/' o r

" part" inspections, in which every single c ircuit is examined, and tested, at various stages in t he process t o isolate and, i f possible, repair defective circuits.
PRODUCTION PROCESS FOR THE H-39

A s tandard size ceramic substrate is purchased f rom an outside source a t a u nit cost o f $1.40. In t he case o f t he H-39, each substrate w ill e ventually be divided up i nto n ine i dentical i ntegrated circuits. A t t he f irst station a m etal called tantalum, chosen f or its resistance and s tability, is evenly deposited on t he " raw" ceramic substrate in a very t hin-film b y an operation called " sputtering." In t his process the substrate is b ombarded b y i ndividual tantalum atoms in a carefully controlled e nvironment. Each t antalum-covered substrate is valued a t $1.90 and t he added cost can be broken d own as f ollows: 10 cents in direct labor, 1 0 cent s in m aterial and 30 cents in allocated overhead. (An overhead rate o f 300 p ercent o f d irect l abor is assigned a t p resent v olume. Actual overhead expense is composed a lmost entirely o f fixed costs consisting o f d epreciation, research and development, and general a dministrative costs.)
Following t he s puttering operation t he electrical resistance o f each substrate is measured t o ensure t hat t he layer o f t antalum is o f t he c orrect thickness. In t he past this inspection has averaged about 90 percent yield (90 p ercent o f t he substrates pass t he inspection)' However, t he engineers i nvolved indicated t hat as t hey c ontinue t o learn m ore about t he s putter ing t echnique i tself, t his yield f igure should i mprove. Each substrate requires about 14 seconds t o test, adding about 3 cents in direct l abor t o t he value o f t he substrate. All rejects are unrepairable, and are discarded.
A fter t esting, t he surviving substrates are transferred t o a c ontrolled access r oom w here a t hin layer o f gold is " evaporated" onto t he substrate (much les s exp ensive t han sputtering, this process is impractical f or tant alum, w hich ha s an extremely high vaporization point, even in a near vacuum). Evaporation adds 10 cents in d irect l abor and 70 cents in materials t o t he value o f t h e substrate. No inspection occurs here o ther t han t o ensure t hat t he chemical process i tself is f unctioning c orrectly.
The substrate then proceeds t o t he p attern generation area, where t he c onductor circuit w ill be fashioned f rom t he layer o f gold and resistors will be shaped f rom t he t antalum. First, t h e substrate is sprayed w ith a chemical called photo-resist, which reacts w hen exposed t o p hotographic light. The substrate is t hen carefully aligned w ith a transparency containing t he c ircuit pattern f or t he H-39, and p hotographically exposed. Following exposure, the substrate is washed, dissolving t he exposed photo-resist, and t hen b aked, leaving a hard, protective covering over those areas t hat w ere underneath t he p attern. These areas w ill f orm t he c onductor circuit' The substrate is then placed in a hydrochloric acid s olution, w hich etches away t he u nprotected gold, leaving t he desired conductor pattern. This entire process is repeated, using a d ifferent p attern, t o f orm t he t antalum resistors. Pattern generation adds an a dditional 6 cents in materials and $1.04 in direct l abor t o t he cost o f t he substrate.
A fter p attern generation t he substrates are baked u nder pressure f or a p redetermined a mount o f t ime, oxidizing a thin laver o f t he t antalum resistors. This

l ayer serves as a p rotective barrier t o u ncontrolled oxidation, which w ould raise t he resistance level o f t he t antalum resistors. The resistors are t hen anodized t o b ring t hem u p t o t heir specified resistance level. This chemical process is regulated by a small c omputer and adds a bout 21 cents in direct l abor and 2 cents in materials t o t he c ost o f each anodized substrate.
The substrate is t hen cleaned, scribed by a laser, and "broken" into nine identical circuits, an operation which adds about 9 cents in d irect l abor cost t o t he value o f t he substrate.
A t t his p oint t he c ircuit undergoes another electrical resistance test and a visual inspection. The purpose o f t he resistance test is t o isolate circuits which contain resistors which are o ut o f tolerance. Resistors whose values are high are n ot repairable, and circuits containing such resistors are discarded. Resistors whose value is l ow are repaired . In t he visual inspection cracked substrates, fa ulty las er scribing, peeling c ircuitry and bad scratches are looked for. None o f these defects are repairable and all circuits containing one o r m ore o f these defects arc di sca rded .
Including t hose circuits successfully repaired, t he final yield f rom t hese t wo inspections is 84 percent. An average cost o f 7 cents in d irect l abor is added t o t he value o f each circuit entering t he inspection and testing sequence. This figure includes t he cost o f all a ttempts a t repair. No material w orth n oting is add ed.
A t t his stage, a frame holding t he required leads is bonded t o ail circuits surviving t est and inspection. The l ea d frames are purchased f or $1.00 a piece, and a bout 3 cents in direct l abor is r equired t o attach t he f rame t o t he IC. Eight s mall
" chip" capacitors are then attached t o t he IC b y t he "solder r eflow" m ethod, in which solder previously placed on t he capacitors is m elted again, f irmly attaching the
"capacitors" t o t he i ntegrated circuit. The chip capacitors cost about 50 cents eac h and a bout 1 4 cents w orth o f d irect l abor is needed t o a ttach all t he capacitors.
Following c apacitor attachment, t hree silicon integrated circuits (SICs) are bonded t o t he t hin-film IC. Each SIC costs about $1.0C, and 16 cents in direct labor is needed t o attach all t hree SICs t o a circuit.
Following SIC b onding, 16 cross-over circuits made o f gold are also bonded t o t he IC t o c omplete t he c onductor circuitry. Each cross-over is purchased f or 20 cents and 27 cents in direct labor is required t o attach all t he cross-overs.
Technological considerations demand t hat t he capacitors be placed on t he circuits before t he SICs are bonded, and handling considerations require t hat t he cross-overs be p ut on a fter t he SICs have been bonded.
A t t his p oint all ICs undergo a rigorous, computer regulated functional t es t which completely checks t he e lectronic circuitry f or each IC.
Seventy percent o f t he ICs i nitially pass t his test. Ten percent o f t he circuits tested (or one-third o f t he rejects) are sound w ith t he exception o f o ne o r m ore cross-overs, which have failed as a result o f e ither f aulty bonding o r handling. An a ttempt is made t o repair these rejects. The cost o f a ttempted repairs averages about $2.70 in materials and d irect l abor f or each faulty IC and about 60 p ercent o f t hese ICs are successfully r epaired . Ten percent o f t he circuits being tested fail solely as a r esult o f a f aulty SIC b ond and are discarded, although a possibility exists t hat in t he f uture these circuits can be repaired. The remaining 10 p ercent fail f or a v ariety o f reasons, usually a c ombination o f reasons, and are n ot repairable. The direct labor cost o f t his functional t est is a bout 6 cents per circuit tested.

A t t his p oint t he lead f rames are t rimmed o n each surviving circuit exposing t he leads themselves. t hen, t he circuits are encapsulated in a protective insulating m aterial. The variable cost o f t hese t wo o perations is 6 c ents in d irect l abor, and 14 cents in material f or each circuit encapsulated. The encapsulated ICs are t hen given a second functional test, identical t o t he f irst except f or t he diagnostics. Ninety percent o f t he ICs pass this test and are readied f or t ransfer o r s hipment. (Ten percent are damaged as a r esult o f t he encapsulation operation.) Once t he ICs are encapsulated n o r epair is possible and t he rejects are discarded. The direct labor cost o f t his second functional t est is also 6 cents per circuit tested.
Seated w ith t he casewriter a fter a t our o f t he H-39 production process, Phil
Davidson continued t o express some o f his thoughts concerning t hin-film p roduction.
FUTURE INVESTMENT

I t's d ifficult t o pigeonhole o ur IC p roduction here a t Benson. O ur o peration is an a nomaly o f sorts. I t has d eveloped almost o ut o f necessity i nto a highly capital intensive process using expensive specialized equipment despite t he f act t hat, as I m entioned earlier, t he t echnology associated with integrated circuitry continues t o change at a rap id rate, often leading t o m ajor process changes as w ell as p roduct changes. In f act, capital investment decisions take up a great deal o f m y t ime. For example, in t he case o f t he H-39, w e are currently considering t he purchase o f a n ew piece o f e quipment t o t est t he c ircuit a fter t he SICs are in place b ut b efore t he crossovers are bonded. A considerable a mount o f value is added at t he cross-over operation and i t m ight be economical t o isolate bad circuits before t he cross-overs are bonded. W ith t his equipment w e feel we could isolate v irtually 100 percent o f t he circuits defective f or o ther reasons . Such a t est would o f course eliminate all faulty, and presently unrepairable, circuits before the cross-overs are installed.
This piece o f t est equipment w ould be similar in nature t o t he e quipment we presently use f or o ur f unctional t est and would cost about $180,000 w ith an a dditional $100,000 involved in i nstallation costs and o ther s tart-up expe ns es. The t est e quipment w ould be used only f or t he H-39 circuit.
Analysis has indicated t hat such a test would cost a bout 4 cents in direct labor p er c ircuit tested and w ould have a yield o f 80 percent. Since we feel t hat c urrently 7 o f 8 circuits successfully go through cross-over bonding, and, as I m entioned earlier, 60 p ercent o f t hose rejected f or bad cross-overs can be successfully repaired, o ur percentage yield f or t he f irst f unctional t est w ould i mprove, although t he actual n umber o f finished circuits suitable f or s hipment would remain unchanged.
AN ALTERNATJVE

On t he o ther hand, o ur f amiliarity w ith the production process f or t he H-39 has developed t o t he p oint w here w e f eel we could isolate 80 p ercent o f t he circuits containing f aulty SIC bonds t hrough a visual inspection immediately a fter t he SIC b onding operation. We k now t hat 7 o f 8 circuits successfully u ndergo SIC bondingi

t herefore, t he visual inspection w ould r eject 10 p ercent o f t he circuits inspected. N ot all circuits surviving this inspection w ould be sound, o f course. In f act, o ur analysis indicates t hat w hile 20 p ercent o f t he circuits entering t he visual inspection w ould be faulty f or o ne reason o r a nother, o nly h alf o f these w ould be caught and rejected a t such an inspection, a llowing t he r emaining 1 0 p ercent t o proceed through cross-over bonding t o t he f irst f unct ional test, where t hey w ould be discovered and rejected.
In summary, i f visual inspection were undertaken, the f irst f unctional test w ould have an initial yield o f 77.8 percent w ith h alf t he rejects being sound except f or o ne o r m ore f aulty cross-overs (again, an a ttempt w ould be made t o repair these rejects). We f eel t hat i t w ould cost about 1 0 cents in direct labor f or each circuit v irtually inspected; however, t he o nly i nvestment w ould be t wo high p ower microscopes, which cost a bout $1,000 a piece and are good f or a bout 5 years o f t woshift operations.
The trick is t o decide w hether e ither o f these process innovations would be w orth w hile, and i f t hey are both w orthwhile which one should w e choose? Is such an expensive piece o f e quipment w orth t he ihvestment? \,!\{hat are t he risks involved? We're n ot y et confident t hat w e have t he answers t o t he se question s.
A t present, Benson Electronics produces about 80,000 H-39 circuits a year.

TAKE-HOME EXAM
1.
2.
3.
4.
5.
6.
7.
8.

D raw t h e f low process chart
W hat is t he t ype o f t he p roduction process?
W hat is t he focus (QCDF) o f t his company? Elaborate.
W hat is t he d irect cost o f p roducing one u nit o f H-39 integrated circuits (IC)?
W hat is t he yield o f t his production process?
How many substrates is r equired t o produce 80,000 H-39 ICs?
Evaluate t he cost and benefit o fthe f uture i nvestment (new test equipment ).
Evaluate t he cost and benefit o f t he a lternative (visual in spectio n).

6