“good'or“bad,”then it is not enough to reduce the test systems are helped greatly in their battle against ob- total number of tests needed from,say,50 to 40 per device. solescence by the fact that many of them are computer- Why bother making 40 tests on a bad device?Why not controlled.Give a computerized system a new software arrange the order of tests so that bad devices are dis- package and it is effectively a new system.(For this qualified as soon as possible-perhaps even after one or reason,if for no other,the computer-operated system two tests?This suggests placing the most critical tests represents more actual value than a hard-wired or tape- first in the sequence (e.g.,testing for catastrophic failures. programmed system.)For all its newly acquired intelli- then functionally testing,then parametrically);but there is gence,however,the older system is not likely to be as more to it than that.If a given test,no matter how productive as newer systems,so its versatility can take critical,is almost always passed,then it is a waste of it only so far. time to put it at the head of the sequence.The possibilities are seen to be endless in theory,but in practice they System compatibility are limited by the flexibility of the test-system hardware Although some bench-top IC testers operate "bare- and software. foot"in incoming inspection facilities,IC testing is by 8.Data logging.Although the concept of productivity and large a systems operation,and the typical IC test in semiconductor testing usually refers to device through- system finds itself surrounded by wafer probers,ovens, put,it is also validly applied to the rate of generation of automatic handlers,scanners,recorders,and other han- test data(summary sheets,distribution analyses,end-of- dling and processing paraphernalia.Thus the designer of life data,etc.).The speed of the data-logging medium an IC test system must take into account the fact that a (teleprinter,line printer,etc.)is important,though less so wafer prober will introduce probe capacitance,that long than one might think:Much data logging is quickly stored leads will be required to connect to automatic handlers, on magnetic tape for later,off-line printout. that the user may want to add auxiliary instruments to the The foregoing considerations apply chiefly to large setup,etc. test systems in production applications and to a lesser It is reasonable to expect the maker,rather than the degree to IC test instruments used in inspection or engi- buyer,of an IC test system to assume most of the inter- neering.A production test system,remember,is as much face worries.The vagaries of automatic handlers and a part of the manufacturing process as a diffusion furnace. probers are well known to any established producer of Since no one is really happy with a manufacturing facility IC test equipment,and it makes little sense for a system operating below capacity,the ideal condition is one in buyer to learn them the hard way. which the test equipment,along with everything else,is being pushed to its limit.In most inspection operations, Easy programming on the other hand,test equipment is purchased not The phrase "easy programming"has been applied because a plant would cease to function without it but to every commercial test system,to the point of meaning- in order to reduce costs of equipment rework and service. An inspection instrument can be idle half the time and still easily earn its keep.This is not to suggest that those who manage inspection facilities are not concerned with productivity;they are,but they define productivity in FIGURE 7.Magnetic-tape cartridge carries test program broader terms,including the total time spent on the end and "device board"houses output loads and other spec- product. ialized test hardware for use with Teradyne J277 MOS Test System,in background. Versatility and obsolescence Any realistic appraisal of IC technology would have to conclude that,far from being a mature field,it has yet to approach full development.The 300 million ICs produced in 1970 include hardly any for consumer and automotive electronics,two areas of major potential.(It has been estimated that the automotive industry alone could con- sume 200 million ICs annually by 1974.)New applica- tions breed basic device changes.So does the current heavy expenditure in semiconductor R&D.Variations on the MOS theme seem to occur weekly.All of this seems to threaten all existing IC test equipment with early obsolescence. Yet most of the several hundred production test systems shipped during 1967 and 1968 are still hard at work for their owners.How can a system designed for 1967 IC technology handle 1971 devices?The answer lies in the fact that the technological ground rules for IC testing have been fairly well established for years.New types of devices require new testing technology,to be sure,but a gate today is tested very much the same way a gate was tested several years ago.The advances have come in the number of gates that can be tested per unit time and in overall system performance parameters.Yesterday's IC Van Veen-An introduction to IC testing 35"good" or "bad," then it is not enough to reduce the test systems are helped greatly in their battle against ob￾total number oftests needed from, say, 50 to 40 per device. solescence by the fact that many of them are computer￾Why bother making 40 tests on a bad device? Why not controlled. Give a computerized system a new software arrange the order of tests so that bad devices are dis- package and it is effectively a new system. (For this qualified as soon as possible-perhaps even after one or reason, if for no other, the computer-operated system two tests? This suggests placing the most critical tests represents more actual value than a hard-wired or tape￾first in the sequence (e.g., testing for catastrophic failures, programmed system.) For all its newly acquired intelli￾then functionally testing, then parametrically); but there is gence, however, the older system is not likely to be as more to it than that. If a given test, no matter how productive as newer systems, so its versatility can take critical, is almost always passed, then it is a waste of it only so far. time to put it at the head of the sequence. The possibilities are seen to be endless in theory, but in practice they System compatibility are limited by the flexibility of the test-system hardware Although some bench-top IC testers operate "bare￾and software. foot" in incoming inspection facilities, IC testing is by 8. Data logging. Although the concept of productivity and large a systems operation, and the typical IC test in semiconductor testing usually refers to device through- system finds itself surrounded by wafer probers, ovens, put, it is also validly applied to the rate of generation of automatic handlers, scanners, recorders, and other han￾test data (summary sheets, distribution analyses, end-of- dling and processing paraphernalia. Thus the designer of life data, etc.). The speed of the data-logging medium an IC test system must take into account the fact that a (teleprinter, line printer, etc.) is important, though less so wafer prober will introduce probe capacitance, that long than one might think: Much data logging is quickly stored leads will be required to connect to automatic handlers, on magnetic tape for later, off-line printout. that the user may want to add auxiliary instruments to the The foregoing considerations apply chiefly to large setup, etc. test systems in production applications and to a lesser It is reasonable to expect the maker, rather than the degree to IC test instruments used in inspection or engi- buyer, of an IC test system to assume most of the inter￾neering. A production test system, remember, is as much face worries. The vagaries of automatic handlers and a part of the manufacturing process as a diffusion furnace. probers are well known to any established producer of Since no one is really happy with a manufacturing facility IC test equipment, and it makes little sense for a system operating below capacity, the ideal condition is one in buyer to learn them the hard way. which the test equipment, along with everything else, is being pushed to its limit. In most inspection operations, Easy programming on the other hand, test equipment is purchased not The phrase "easy programming" has been applied because a plant would cease to function without it but to every commercial test system, to the point of meaning￾in order to reduce costs of equipment rework and service. An inspection instrument can be idle half the time and still easily earn its keep. This is not to suggest that those who manage inspection facilities are not concerned with productivity; they are, but they define productivity in FIGURE 7. Magnetic-tape cartridge carries test program broader terms, including the total time spent on the end and "device board" houses output loads and other spec￾product. ialized test hardware for use with Teradyne J277 KOS Test System, in background. Versatility and obsolescence Any realistic appraisal of IC technology would have to conclude that, far from being a mature field, it has yet to .l. approach full development. The 300 million ICs produced * in 1970 include hardly any for consumer and automotive -i* electronics, two areas of major potential. (It has been estimated that the automotive industry alone could con￾sume 200 munllon ICs annually by 1974.) New applica￾tions breed basic device changes. So does the current heavy expenditure in semiconductor R&D. Variations on the MOS theme seem to occur weekly. All of this seems to threaten all existing IC test equipment with early obsolescence. Yet most ofthe several hundred production test systems shipped during 1967 and 1968 are stiUl hard at work for their owners. How can a system designed for 1967 IC technology handle 1971 devices? The answer lies in the fact that the technological ground rules for IC testing have been fairly well established for years. New types of devices require new testing technology, to be sure, but a gate today is tested very much the same way a gate was tested several years ago. The advances have come in the number of gates that can be tested per unit time and in overall system performance parameters. Yesterday's IC Van Veen-An introduction to IC testing 35