Conference Report
Total Page：16
File Type：pdf, Size：1020Kb
Electrocomponent Science and Technology, 1982, Vol. 9, p. 292-295 (C) 1982 Gordon and Breach Science Publishers, Inc. 0305-3091/82/0904-0292 $6.50/0 Printed in Great Britain Conference Report Conference on Reliability in Electronic and Electrical Components and Systems (Eurocon '82) Held June 14 to 18, 1982, Copenhagen. Continuous growth in the appreciation of the With regard to components themselves sessions importance of reliability of electrical and electronic were devoted to passive components and inter- component systems has been recognised in the connections, discrete semiconductors and integrated corresponding growth in the number of conferences circuits, connectors, memories, opto electronic devoted to this all important subject. The 5th Euro- devices and component testing. con Conference held in Copenhagen June 1982 was The themes and aims of the Conference were well one of these meetings. One hundred and seventy set by many of the invited papers given at the plenum papers were presented in four parallel sessions, with sessions. The very first paper by Sir Herbert Durkin, the papers selected from a total of two hundred and President of the lEE in the United Kingdom on the eighty which were presented to the Technical Pro- "Technical and Economic Implications of Reliability", gramme Committee. Six hundred participants were emphasised the most important point that reliability present. does not happen, it cannot be grafted on after design Eurocon '82 was jointly sponsored by EUREL is complete, and most of all it is not something that (The Convention of National Societies of Electrical requires attention only by one group of people res- Engineers of Western Europe) and by the IEEE ponsible for a particular part of a project. It must be (Institute of Electrical and Electronic Engineers), recognised at all times that quality and reliability with additional support from the Danish Society of must be related to application and that nowadays Chemical, Civil, Electrical and Mechanical Engineers, poor quality and reliability not only has the effect of the Society of Engineers of Denmark, and the Danish increasing costs to the company due to customer Electro Technical Association. returns but also in this age of fierce competition will Not all aspects of the conference were of direct lose the company future orders. It also needed to be relevance to the component and sub system engineer. emphasised these days that the concept of a bath-tub Theoretical aspects of the study of reliability were curve for representing the change of failure rates with well covered by papers on computational methods time in which after an initially high failure rate, the and reliability modelling as well as on the analysis of failure rate is low and relatively constant in middle complex systems, field tree analysis, and studies of life, and then rises again gradually as the end of life redundant systems. Practical aspects of reliability in approaches, is no longer valid unless one is consider- electronic systems were presented with particular ing systems which are mainly mechanical or which reference to telecommunications, and other types of have well defined chemical wearing out processes. systems were examined in the sessions devoted to The evidence is becoming much less well established detection and control, satellites, transportation, that there is a wear out mechanism in electronic consumer products, power distribution and industrial components both passive and semiconductors, and networks, power generation, sub station equipment for these the bathtub curve looks much more like an and dispersed generation sources. The Conference asymptotic fall to a very low or zero failure rate. also recognised the importance of software reliability Kan L. Wong, the Manager of the Design Effective- and data processing systems, with three sessions ness Operations for Hughes Aircraft in the States gave devoted to this area. The costs of achieving reliability a keynote paper on "New Directions for Electronic were recognised by many authors in their papers, but Reliability Engineering in the 80s". In summary Mr. two sessions were specifically devoted to this area Wong felt that now the time had come for all of us covering life cycle costs, maintainability and to direct our attention to the real failures which were availability. Finally, the human and legal aspects of usually called random failures. These failures came reliability were considered in one session of three from interaction of stresses and flaws and it was papers. necessary to eliminate them. He felt that reliability 292 CONFERENCE REPORT 293 engineers were now entering the era of interaction in However, when models are used for system evalua- which reliability engineering and basic engineering tion purposes, statistical differences between predic- must work closely together to create an environment ted and observed failure rates will not be profoundly for achieving even higher reliability failure analysis. important. For ground based telecommunication Neither Sir Herbert Durkin, or Kan L. Wong, equipment the CNET model is probably the best presented any definitive data to support their impor- model presently used. tant conclusions. This was not however the case with (b) The importance of identifying failure mecha- the important invited paper given by Dr. Fred nisms in order to assess the applicability of accelerated Reynolds of British Telecom on "Measuring and life test is emphasised. Modelling Integrated Circuit Failure Rates". This paper, while acknowledging the difficulty of obtain- (c) The collection of field lifetime data is of vital as a source ing field failure data, emphasised the importance to importance of predictive information. the reliability engineer of such an operation. Figures Many of the subsequent contributed papers were presented for certain components which have re-emphasised the themes discussed in the invited been studied in field failure terms since 1967 and later. papers. Dr. Wayne (from Plessey Research, Caswell, Very important summary charts were presented England) discussed the "Thermal Fatigue Failure of giving failure rates, together with other results, Soldered Joints in Printed Circuit Board Assemblies". obtained from a wide number of sources. However, it The failure mechanism associated with the soldered needs to be recognised that most of the reports used joints recrystallising on thermal cycling was recog- for accumulating field failure data gave little help to nised and many excellent electron micrographs were the present day system designer who is now dealing shown demonstrating the micro cracking that results. with different types of devices than those incorporated A best buy with regard to the use of tin cadmium into equipment manufactured in earlier years. The solder joints, when the printed circuit boards had a natural alternative of life test data acquisition was bare copper or tinned finish, was suggested. Such a examined and the various data banks that exist solder showed a fatigue resistance which was approxi- around the world for collecting and processing such mately four times better than that of the traditional information were summarised. tin-lead solder. The problem of accelerated life tests is one that Changes in microstructure were recognised in a paper still requires urgent attention. For many industrial given by Hieber and Pape of the Philips Laboratories applications, the raising of the ambient temperature of in Hamburg on "The Reliability of Thin Film Metalli- the tested device to 125C whilst continually operat- zations from Thermal Ageing Kinetics". Thin films ing it was often used as a suitable stress level. However, of gold and aluminium were studied and changes of even this level requires large samples in order to the microstructure with thermal treatment were ana- generate a significant failure yield, and, furthermore, lysed. It was recognised that these results were of conditions of test used can expose modes of failure interest for films prepared under relatively clean con- not actually seen in real service. The testing time used ditions and that the inclusion of impurities etc. in by most vendors is too short for any postulated life- the films if they have been prepared by sputtering time distribution to be other than exponential so instead of evaporation could possibly lead to other that long term failure rate trends cannot be predicted. failure mechanisms. Credible interpretation of failure behaviour can A most interesting Paper on "Failure Mechanisms only be possible if the failure causes are known, so in MOS Devices, both those in Integrated Circuits that individual activation energies can be associated and Power Devices", was given on behalf of Dr. with them and the effect of changes in testing tempera- Edwards of the Siemens Corporation, Cherry Hill, ture can then be assessed. In this assessment process U.S.A. The failure modes in integrated circuits were mechanisms with low activation energy clearly give recognised as being due to oxide breakdown, electro the greatest cause for concern. migration, hot electron effects and finally soft errors It is impossible in the space available to summarise due to alpha particle penetration. The alpha particles all the important points raised in Dr. Reynold's paper. originated from the substrates being used and could However, his conclusions can be noted as follows:-- have energies as high as 5.5 MeV. Both these last two (a) Reliability models are attractive from a System errors were becoming