Emerald Group Publishing Limited
Copyright © 2004, Emerald Group Publishing Limited
Referring to the previous issue that focused on solder joint reliability (Vol. 16 No. 2 Soldering & Surface Mount Technology), as a personal observation, I am continually surprised by the frequent arbitrary use of the temperature range −40 to +125°C (or even to +150°C) for cycling soldered ceramic surface mounted component assemblies.
It must be nearly 20 years ago that Professor David Campbell and his colleagues at Loughborough University in the UK demonstrated clearly that the ranges above were far from the worst-case scenario for 60/40 tin-lead soldered chip resistors on printed boards. The reason put forward was simply that, as the Tg of the printed board was approached, appreciable softening of the substrate material came into the equation – a factor supported by the results published in the above volume by Guo-Quan Lu and colleagues in their paper “Strategies for improving the reliability of solder joints on power semiconductor devices”.
The Loughborough results indicated that an upper temperature limit of 95°C was the “worst case” value for the resistors they tested on FR4 board material. Since then doubt has also been cast on the value of going as low as −40°C, as the difference between results at −20 and −40°C were negligible and probably not worth the extra cost.
Whilst it is appreciated that the range −40 to +125°C (or +150°C) is generally accepted as the harsh environmental limit for military applications. In many applications cycling reliability results over this range should be expected to be considerably better than over the worst case range, i.e. we may be seriously deluding ourselves and some of our customers. This brings us to an argument about reality; what else can we do and what are the most likely temperature-time ranges that operational assemblies will actually experience in use. How can we test in order to approach more realistic worst case combinations?
Normally life experience for electronic assemblies is defined as a collection of likely time-temperature combinations, e.g. x hours at y degrees or in the range −y1 to +y2 degrees, plus z hours at −y3 degrees, or similar. Unfortunately these data do not help in clearly defining a suitable range for solder joint reliability testing, but equally, they do not undermine the case for adopting a move towards “worst case” thinking, rather than simply taking the arbitrary range specified by tradition and habit – especially if the cost of testing is reduced thereby.
Also, there are signs that other reliability parameter tests such as push, shock and vibration may require re-thinking, especially for lead-free solders. It is becoming more essential that standard mechanical strength tests are carried out in closer combination with temperature cycling rather than in isolation.
Industry managers and their customers do not like surprises, especially those appear after medium or longer term product usage. One helpful way of minimising exposure is to carry out hazard and risk analysis on products before they reach the production stage, i.e. to determine what can possibly go wrong, to establish what are the related likelihoods of hazard occurrence, and to assess the consequences of such an occurrence. In this particular case, the products to be assessed are component-substrate solder joints and it would be interesting to know whether any component manufacturer has yet had the wisdom to carry out such analyses on a scientific basis, employing sound mechanical design principles combined with knowledge of the basic physics and chemistry of the structure. As long as the quantified risk results do not fall into the hands of accountants, we will still be able to innovate successfully.
To conclude, it seems that the reliability test protocols adopted by experimenters today, have become tarnished with an institutionalised escapism born of prioritising “comparison of results” rather than of the need to “approach reality”. It is suggested that there is an urgent requirement to completely rethink and redesign our tests. The aim is not to achieve perfection, but to define a set of regimes that are nearer to worst case and possibly cheaper to operate than at present.
One can envisage assemblers throwing up their hands in horror and exclaiming “That’s all very well, but what about the cost ?” and component manufacturers saying “We are not in control of our customer’s soldering processes, so how can we be expected to provide meaningful reliability data ?” True, but if jointly funded initial experiments were carried out independently to determine, for typical applications, processes and material combinations, adjustments in the test temperature ranges and cycling parameters to those most likely to be “nearer worst case than at present”, then certainly in the medium term there could often be savings for all concerned in both subsequent testing costs and also in the time taken to get significant data. Not only that, we could improve our industry’s image as purveyors of meaningful reliability results.
David BoswellE-mail: email@example.com