6.This instrument is programmed by a plug-in card The number of stations and the level of independence containing a bank of 40 resettable slide switches. of each directly affect productivity.A four-station system, for example,may or may not be able to distribute its General considerations in services among three probe stations and one final test the evaluation of Ic test equipment station,or among four stations testing four different The first requirement of any testing,productivity con- devices,or among three classification stations and one siderations notwithstanding,is that the tests be valid. data-logging station. In IC testing this means (1)identifying those parameters 3.Handling speed.If devices are manually inserted into that adequately define the“goodness'"or“badness'”ofa a test socket of a single-station system,the operator's device,(2)ensuring that the conditions of measurement handling rate will almost certainly be the major factor relate meaningfully to the conditions of use,and (3) limiting throughput.It is the great imbalance between achieving the desired accuracy of measurement. testing and handling speeds,in fact,that provides the Specifying the accuracy of IC measurements is an rationale for multiplexing. especially precarious business because the test environ- The use of automatic handling equipment speeds up ment is unpredictable-and often hostile.Extremely high the mechanical end of the process to the point where accuracy and precision are rarely primary considerations, several thousand devices an hour can be tested at a single since the required levels are usually well within the state station,assuming that the test sequence per device is fairly of the art.Instead,the emphasis is on repeatability,with simple. periodic verification of accuracy.The word"calibration" The fastest mechanical handler cannot begin to match is anathema to most test-system users,because it implies the test system's speed for simple devices (e.g.,well under downtime.When a system is found to be outside toler- 100 ms to test a gate).Where automatic handlers are ance one approach is isolation and quick replacement of to be multiplexed,however,test time and handling the guilty component. converge,and the relation between the two must be Although accuracy and precision-in the traditional noted in setting up the test installation and program. instrument sense-are not primary issues,the "analog" 4.Programming and debugging.A computer-operated performance of an IC test system most certainly is.A test system is normally supplied with an executive pro- "clean"transition,without overshoot or ringing,is es- gram that contains the formatting and subroutines needed sential if fast devices are to be exercised without ambi- for a given type of application (end-of-life testing,classi- guity.The signal should be clean at the device under fication,etc.).The user must write and enter his own test,which in production testing usually means at a test specifications (bias conditions,limits,bin criterions, prober or handler separated from the signal source by a etc.),and then he must verify and correct ("debug")his length of cable. program as necessary.The debugging procedure can easily consume great amounts of time (an hour a day is Productivity not at all unusual),so a test system must be evaluated in Since economics are such a dominant factor in semicon- the light of its ability to time-share programming and ductor production,one tends to think not of the cost of a debugging with normal testing and the availability of tester but of the cost of testing.Thus productivity is a debugging aids (panel indicators,check-sum verifica- much-used term and a much-sought-after characteristic tion of software,etc.). of test equipment.Productivity in testing is a function of 5.Downtime.Downtime is the total time during which many variables,including the following: a test system is not available for normal testing.In any 1.Test speed.Although speed is the most obvious calculation of productivity,downtime obviously comes factor,it is rarely the most important.Even the slowest right "off the top,"all other factors coming into play test system is so fast that the real productivity limitations only when the system is operative.It is a difficult param- arise elsewhere. eter to quantify when making a purchasing decision,but 2.Multiplex capability.Many test systems provide for it is so important that an effort must be made.Consider some degree of multiplexing,or the simultaneous use of the difference in productivity between a test system regu- two or more test stations with a single test system larly experiencing 20 percent downtime (a not unusual figure)and one with 1 percent downtime (also a not unusual figure). 6.Retest load.If a system is not testing devices properly, FIGURE 6.An incoming-inspection instrument for linear the lots rejected by quality assurance (QA)or incoming ICs (the General Radio 1730 Linear Circuit Tester). inspection will usually find their way back to be tested again.A system that is out of calibration will therefore directly penalize system productivity. 7.Test-plan efficiency.In most cases,the number of tests required to test an IC adequately is not subject to rigid definition.One can determine easily enough the number of truth-table combinations for a simple IC,but where this number becomes overwhelmingly large,atten- tion shifts to the number required for adequate (rather than exhaustive)testing,leaving room for wide system-to- system variations.Judgment dictates the proper extent of dc parametric or linear testing per device,but system flexibility converts this judgment into productivity.If the testing objective is usually to characterize an IC as IEEE spectrum DECEMBER 19716. This instrument is programmed by a plug-in card The number of stations and the level of independence containing a bank of 40 resettable slide switches. of each directly affect productivity. A four-station system, for example, may or may not be able to distribute its General considerations in services among three probe stations and one final test the evaluation of IC test equipment station, or among four stations testing four different The first requirement of any testing, productivity con- devices, or among three classification stations and one siderations notwithstanding, is that the tests be valid. data-logging station. In IC testing this means (1) identifying those parameters 3. Handling speed. If devices are manually inserted into that adequately define the "goodness" or "badness" of a a test socket of a single-station system, the operator's device, (2) ensuring that the conditions of measurement handling rate will almost certainly be the major factor relate meaningfully to the conditions of use, and (3) limiting throughput. It is the great imbalance between achieving the desired accuracy ofmeasurement. testing and handling speeds, in fact, that provides the Specifying the accuracy of IC measurements is an rationale for multiplexing. especially precarious business because the test environ- The use of automatic handling equipment speeds up ment is unpredictable-and often hostile. Extremely high the mechanical end of the process to the point where accuracy and precision are rarely primary considerations, several thousand devices an hour can be tested at a single since the required levels are usually well within the state station, assuming that the test sequence per device is fairly of the art. Instead, the emphasis is on repeatability, with simple. periodic verification of accuracy. The word "calibration" The fastest mechanical handler cannot begin to match is anathema to most test-system users, because it implies the test system's speed for simple devices (e.g., well under downtime. When a system is found to be outside toler- 100 ms to test a gate). Where automatic handlers are ance one approach is isolation and quick replacement of to be multiplexed, however, test time and handling the guilty component. converge, and the relation between the two must be Although accuracy and precision-in the traditional noted in setting up the test installation and program. instrument sense-are not primary issues, the "analog" 4. Programming and debugging. A computer-operated performance of an IC test system most certainly is. A test system is normally supplied with an executive pro- "clean" transition, without overshoot or ringing, is es- gram that contains the formatting and subroutines needed sential if fast devices are to be exercised without ambi- for a given type of application (end-of-life testing, classi￾guity. The signal should be clean at the device under fication, etc.). The user must write and enter his own test, which in production testing usually means at a test specifications (bias conditions, limits, bin criterions, prober or handler separated from the signal source by a etc.), and then he must verify and correct ("debug") his length of cable. program as necessary. The debugging procedure can easily consume great amounts of time (an hour a day is Productivity not at all unusual), so a test system must be evaluated in Since economics are such a dominant factor in semicon- the light of its ability to time-share programming and ductor production, one tends to think not of the cost of a debugging with normal testing and the availability of tester but of the cost of testing. Thus productivity is a debugging aids (panel indicators, check-sum verifica￾much-used term and a much-sought-after characteristic tion of software, etc.). of test equipment. Productivity in testing is a function of 5. Downtime. Downtime is the total time during which many variables, including the following: a test system is not available for normal testing. In any 1. Test speed. Although speed is the most obvious calculation of productivity, downtime obviously comes factor, it is rarely the most important. Even the slowest right "off the top," all other factors coming into play test system is so fast that the real productivity limitations only when the system is operative. It is a difficult param￾arise elsewhere. eter to quantify when making a purchasing decision, but 2. Multiplex capability. Many test systems provide for it is so important that an effort must be made. Consider some degree of multiplexing, or the simultaneous use of the difference in productivity between a test system regu￾two or more test stations with a single test system. larly experiencing 20 percent downtime (a not unusual figure) and one with 1 percent downtime (also a not unusual figure). 6. Retest load. If a system is not testing devices properly, the lots rejected by quality assurance (QA) or incoming FIGURE 6. An incoming-inspection instrument for linear inspection will usually find their way back to be tested teGeea aio13 iea ii Testeagain. A system that is out of calibration will therefore ~~~~~~~ ~~~~~~directly penalize system productivity. 7. Test-plan efficiency. In most cases, the number of tests required to test an IC adequately is not subject to rigid definition. One can determine easily enough the number of truth-table combinations for a simple IC, but where this number becomes overwhelmingly large, atten￾tion shifts to the number required for adequate (rather than exhaustive) testing, leaving room for wide system-to￾system variations. Judgment dictates the proper extent of dc parametric or linear testing per device, but system fiexibility converts this judgmnent into productivity. If the testing objective is usually to characterize an IC as 34 ~~~~~~~~~~~~~~~~~~~~~~~~~~IEEE spectrUM DECEMBER 1971