Reliability Information Reliability at Vishay Siliconix

1 Reliability InformationVishay SiliconixDocument Number: at Vishay SiliconixThis document provides a general description of Vishay Siliconix s Reliability program. The Reliability data for specific products isavailable from the Information page associated with each part. The general Reliability description provided below and the reliabilityinformation provided for each part may be revised without prior InformationVishay SiliconixDocument Number: :CONTENTSINTRODUCTION TO RELIABILITYD efinition of Reliability1.. Bathtub Failure Rate Curve1.. Calculation of Failure Rate at Test Temperature2.. Observed and Predicted FIT Rates2.. FIT at Operating Temperature3.. PRODUCT Reliability PROGRAMSQ ualification Program5.. Purpose5.. Qualification Categories5.. Qualification Test Plan (QTP)5.. Qualification and Release Process6.. Reliability Monitor Program6.. Purpose/Description6.. Short-Term Monitor6.. Long-Term Monitor7.. Alert Levels7.

2 MOISTURE SENSITIVITY LEVEL (MSL) FOR PLASTIC SURFACE MOUNTED DEVICESTest Method: IPC/JEDEC J-STD-020C9.. Vishay Siliconix Surface Mount Solder Reflow Preconditioning9.. Surface Mount Solder Reflow Preconditioning9.. MOLDING COMPOUND FLAMMABILITY AND ENVIRONMENTAL10.. APPENDIX A11.. Reliability InformationVishay SiliconixDocument Number: to ReliabilityDEFINITION OF RELIABILITYT here are many popular notions of what constitutes Reliability , or what characterizes a reliable product. Asuitable definition of Reliability must be both precise andcapable of expression and measurement in mathematicalterms. One widely quoted and accepted definition of reliabilityis the probability that an item will perform a required functionunder stated conditions for a stated period of time. The required function implies standards or definitions ofsatisfactory performance and acceptable rates of failure. Therequired function is most often defined and measured by anautomatic test procedure.

3 However, the required function mayalso be determined by the customer s expectations in systemuse. Discrepancies between the two sometimes result in theidentification of new failure modes, leading in turn to theinclusion of appropriate tests in future generations of testprograms. Sometimes, especially in complex systems,unsatisfactory performance is not the same as stated conditions are the total physical environment ofthe system or device. They include the mechanical, thermal,and electrical conditions of use or stated period of time is the time during which satisfactoryperformance is desired. This can be relatively short, as in thecase of a missile flight, where the probability that no failure willoccur during the flight is a good measure of Reliability . Otherpossibilities are periods of disuse followed by use, orcontinuous use (such as in maintained systems like telephoneswitches where there is no well-defined mission time).

4 Reliability is the probability of survival as a function of time andstress and can be expressed as follows:F(t) + R(t) = 1; orR(t) = 1 F(t)where F(t) is a cumulative density function, representing theprobability of failure up to time t and R(t) is the reliabilityfunction, a complement of F(t), representing the probability ofsurvival at time that at time zero, the survival is presumed, henceR(0) = 1while at time infinity, the certainty of failure is also to beexpected, henceR(R) = 1 Since F(t) is the cumulative failure up to time t, by definition, thefailure rate at time t can be expressed as dF(t)/dt. Hence, fora constant failure rate ,Since F(t) is the cumulative failure up to time t, by definition, thefailure rate at time t can be expressed as dF(t)/dt. Hence, fora constant failure rate ,dF(t)dt+lFrom the above, it can be proved thatF(t) = 1 exp ( lt)andR(t) = exp ( lt)From these equations, once the failure rate is determined, onecan predict the probability of survival at any given time.

5 Forexample, if the failure rate is 100 FITs (see definition below) thereliability of a component after 100,000 hours of operation canbe calculated asR(100,000 hrs) = that there is a 99 % probability of survival up to100,000 hours of use. Bathtub Failure Rate CurveA constant failure rate assumed in the above example isobviously an oversimplification of the Reliability picture. Atypical failure (or hazard ) rate curve is shown in Figure 1. Thiscurve depicts the general trend of failure (hazard) rate, X(t), forelectronic components along the time of Hazard Rate CurveReliability InformationVishay Number: 7086317-Nov-05 The decreasing portion of the curve represents the earlyfailures and is sometimes called the infant mortality of failures during this period is normally attributedto manufacturing defects. Failures in this period can bescreened effectively if the failure mechanisms are wellunderstood.

6 The screening can be done during manufacturingor production testing, or by using a short accelerated reliabilitystress constant part of the bathtub failure rate curve, called the useful life period, lies in between the infant mortality periodand the wear-out period. Typically, the failure rate during thisperiod is relatively stable and much lower than that of the otherperiods. Failures during the useful life period are generallycalled random failures because they appear to occurunpredictably. Good process control can reduce theprobability of random should be noted that the FIT rate provided for each VishaySiliconix product is meant for this period. The FIT rate iscalculated from high-temperature operating life test wear-out period begins when an electronic componenthas aged, degraded, or outlived its useful operating life. Thematerials used to construct the electronic component start toundergo a fundamental breakdown process and deterioraterapidly in this period.

7 Consequently, the number of failuresduring this period increases with time. Electromigration,hot-carrier induced device degradation, and time-dependentdielectric breakdown are some wear-out mechanismsfrequently discussed in the Reliability good Reliability program should aim toDeliminate, or at least minimize, early failures in the infantmortality period by robust product and process design or aneffective screening methodologyDminimize random failures by good process control, andDpush out the onset of all wear-out mechanisms beyond theexpected lifetime of the components by building-insufficient margins in both the design and manufacturing ofthe electronic example, Vishay Siliconix s Product Reliability Programsare designed to ensure that early and random failures areminimized at time of release (qualification) and continue todecrease as the product matures through a continuousimprovement program.

8 In addition, Vishay Siliconix s ProcessReliability Test Program is designed to ensure that wear-outmechanisms are pushed well beyond the expected lifetime ofthe customer s of Failure Rate at Test TemperatureAs stated earlier, the failure rate during the useful life period isconstant. Testing components operating in this period allowsone to determine the product-specific failure rate. Data from acomponent sample is obtained using an accelerated life test,and the observed failure rate (FR) is expressed asFR+FSS twhere F is the number of failures observed, SS is sample size,and t is time duration of the test in hoursNote that the observed FR is given in the unit of failures perone component-hour. Because this is normally a very smallnumber, FR is usually expressed in % per1000 component-hours, in ppm (parts per millioncomponent-hours), or in FITs ( failure in time or failures perbillion component-hours).

9 Failure rates can be expressed inthese units by multiplying the right-hand side of the aboveequation by 100,000, 1,000,000, or 1,000,000,000,respectively. Hence,FIT+109 FSS tBecause of its widespread use in the Reliability community, wewill use FIT to denote failure rate in the subsequent and Predicted FIT RatesThe observed FIT is obtained from observation of failures in aspecific lot sample tested for a given time. Such a FIT is validonly for the sample that was tested. To obtain a FIT that isapplicable to an entire component population, statisticalconfidence factors should be included, and the FIT becomesFIT+x22 109SS twhere 2/2 is the statistical confidence factor and is a functionof F and confidence level based on a chi-square InformationVishay SiliconixDocument Number: 1. Values of c2/2 Different Numbers of Failures (F) and Confidence Levels (CL)FCL = 60 %CL = 80 %CL = 90 %CL = 95 %CL = 99 % significance of the statistical confidence factor is that itallows prediction of results with a given level of confidence from future sample tests.

10 For example, if the FIT (say, 1,meaning one failure per 109 component-hours) with aspecified CL (say, 60 %) is calculated from the data obtainedduring accelerated tests for a given component population,then 60 % (CL) of all tested lots from the same componentpopulation in the future will have lower at Operating TemperatureSince many Reliability aging processes are thermally activated,testing at high temperatures can help accelerate the aging ofcomponents such that sufficient data can be gathered in arelatively short time as compared to their expected thermally activated process can be described byFITTexp *EakT Statistical confidence factors can be found in Chi square charts or from Poisson charts given in statistics 1 shows values of ( 2/2) for different numbers of failures(F) and for different confidence levels (CL).The FIT at a certain junction temperature may be calculatedfrom the FIT calculated at test temperatures using theArrhenius equation:FITT1 FITT2+exp *EaK 1T1*1T2 where k is Boltzman constant = x 10-5 (eV/_K), T1 and T2are junction temperatures (_K), and Ea is the Arrhenius equation for different activation energies ispresented as a graph in Figure 2.