lessness.In fact,writing the program (once one decides example,the programming of a pulse parametric test what he wants to test)is a fairly simple matter with any on Teradyne's J277 MOS Test System;see Fig.7. system.A language and some syntax must be learned, To set up the timing of clocks and output strobe the but the learning generally requires only a day or two,and operator might write: sometimes much less.One is ill-advised to select a test CLOCK 1 system because it is easier to write programs for it than 100 NS FROM TO for another system.In fact,the easier language may re- GAP1 70 NS FROM CLOCK 1 flect a more limited testing capability.The most effective CLOCK 2 210 NS FROM GAP 1 utilization of a test system normally results when a good STROBE 1 300 NS FROM TO technician,having learned all the nuances of an extensice (In this example a gap is defined to indicate that the system language,calls into play all the testing power at clocks do not overlap.)The operator commands the his command system to measure the leakage current on pin 2,rejecting The concept of easy programming is more valid,how- anything more negative than -500 mA at -18 volts, ever,when applied to the type(rather than the size)of as follows: a programming language.Languages are sometimes characterized as "high level"or "low level."A high- MCON PIN 2 level language is designed exclusively for the testing job MEASURE -500 MA AT -18V;RNEG at hand,whereas a low-level language is for general- Not all test-system languages are as close to plain English purpose use.A high-level language geared to IC-testing as the foregoing,but many use mnemonics that are easily terminology obviously will be easier to program in that mastered. application than a low-level language.Consider.for Sizing the equipment to the job Most of the points discussed so far relate chiefly to the use of IC test systems.The world of IC testing,however, embraces bench-top IC testers as well as large,produc- FIGURE 8.Two computer-operated test systems.(Top) Teradyne J283/Si57 system for functional,dc,and pulse tion-oriented systems.In some ways the smaller instru- parametric testing.(Bottom)Datatron 4400 for functional ments are the equal of their massive counterparts.A and dc parametric testing. bench-top tester may well be able to perform functional tests on a given IC as fast as any large system can.Small testers are in fact sometimes interfaced to systems,in order to screen out functional failures as fast as possible, before proceeding to parametric testing. At the low end of the price spectrum of IC test equip- ment is the manually programmed tester costing a few hundred dollars.It can functionally check a simple IC and that is about all.It is useless for the testing of ICs in any volume. A fairly large selection of equipment is available in the $3000-$8000 range.Instruments in this class are typically comparison testers,programmed by plug-in printed- circuit cards,with a binary generator handling the func- tional-testing requirements at very respectable speeds. Instruments of this type are not to be taken lightly. Interfaced to an automatic handler,a $5000 IC tester can throw several thousand ICs an hour into "good" and "bad"bins,on the basis of both functional and parametric test results. In determining whether or not a bench-top instrument will satisfy one's testing needs,the best approach is prob- ably to examine the things that such an instrument cannot do,then determine whether any of these are crucial in a given application.Specifically,and with very few excep- tions,the bench-top tester cannor generate test data; match the multiplexing ability of a large system;provide for dynamic testing;handle as many different types of IC as a computerized system can;perform thorough func- tional testing of devices having sequential logic;make highly accurate and precise parametric tests;or make burn-in or life tests. These limitations rule out the bench-top tester for production and QA testing and incoming inspection where test data are required.Once it is determined that the bench-top tester is inadequate for the job at hand, the next step up the performance scale is the S50000- and-up computer-operated test system.(An exception is IEEE spectrum DECEMBER 1971lessness. In fact, writing the program (once one decides example, the programming of a pulse parametric test what he wants to test) is a fairly simple matter with any on Teradyne's J277 MOS Test System; see Fig. 7. system. A language and some syntax must be learned, To set up the timing of clocks and output strobe the but the learning generally requires only a day or two, and operator might write: sometimes much less. One is ill-advised to select a test CLOCK I 100 NS FROM TO system because it is easier to write programs for it than GAP 1 70 NS FROM CLOCK I for another system. In fact, the easier language may re- CLOCK 2 210 NS FROM GAP I flect a more limited testing capability. The most effective STROBEI 300 NS FROM TO utilization of a test system normally results when a good technician, having learned all the nuances of an extensive (In this example a gap is defined to indicate that the system language, calls into play all the testing power at clocks do not overlap.) The operator commands the his command. system to measure the leakage current on pin 2, rejecting The concept of easy programming is more valid, how- anything more negative than -500 mA at -18 volts, ever, when applied to the type (rather than the size) of as follows: a programming language. Languages are sometimes characterized as "high level" or "low level." A high- MCON PIN 2 level language is designed exclusively for the testing job MEASURE-500 MA AT-18V; RNEG at hand, whereas a low-level language is for general- Not all test-system languages are as close to plain English purpose use. A high-level language geared to IC-testing as the foregoing, but many use mnemonics that are easily terminology obviously will be easier to program in that mastered. application than a low-level language. Consider, for Sizing the equipment to the job Most of the points discussed so far relate chiefly to the use of IC test systems. The world of IC testing, however, embraces bench-top IC testers as well as large, produc- FIGURE 8. Two computer-operated test systems. (Top) tion-oriented systems. In some ways the smaller instru￾Teradyne J283/Si57 system for functional, dc, and pulse parametrictesting. (Bottom)Datatron4400forfunctional ments are the equal of their massive counterparts. A and dc parametric testing. bench-top tester may well be able to perform functional tests on a given IC as fast as any large system can. Small testers are in fact sometimes interfaced to systems, in order to screen out functional failures as fast as possible, before proceeding to parametric testing. At the low end of the price spectrum of IC test equip￾ment is the manually programmed tester costing a few hundred dollars. It can functionally check a simple IC, and that is about all. It is useless for the testing of ICs in any volume. A fairly large selection of equipment is available in the $3000-$8000 range. Instruments in this clas are typically comparison testers, programmed by plug-in printed￾circuit cards, with a binary generator handling the func￾tional-testing requirements at very respectable speeds. Instruments of this type are not to be taken lightly. Interfaced to an automatic handler, a $5000 IC tester can throw several thousand ICs an hour into "good" and "bad" bins, on the basis of both functional and parametric test results. .36 In determining whether or not a bench-top instrument . .,, ~~~~will satisfy one's testing needs, the best approach is prob￾ably to examine the things that such an instrument cannot do, then determine whether any of these are crucial in a given application. Specifically, and with very few excep￾tions, the bench-top tester cannot generate test data; match the multiplexing ability of a large system; provide for dynamic testing; handle as many different types of IC as a computerized system can; perform thorough func￾tional testing of devices having sequential logic; make highly accurate and precise parametric tests; or make burn-in or life tests. These limitations rule out the bench-top tester for production and QA testing and incoming inspection ~~~~~~~~ ~~~~~~~where test data are required. Once it is determiined that the bench-top tester is inadequate for the job at hand, the next step up the performiance scale is the $50000- and-up computer-operated test system. (An excepition is 36 MEEE spectrum DEcEmBER 1971