31-12-2015, 08:34 AM
Further to the question of reliability issues with semiconductors. What has to be bourn in mind is that once released to the market, and included into some technical marvel outside the manufacturer's control, the reliability of that part is then at the mercy of the equipment designer, and the end user, my own experience that I recounted in the first post on this topic was proof positive of this. Take a transistor, that has the correct electrical specifications, and include it in your brainchild, then use that product in a hostile environment, outside the manufacturer's specs, and outside its manufacturing deficiency's, (sorry, but there is no such thing as perfect, only different levels of imperfection) and it will fail. Add further protection to the equipment at assembly (in my example tropicalization) and that part will possibly then be more reliable than including a better specified part in the first place (in my example TO-18 package transistors with the same specifications). The forgoing are amongst the reasons why semiconductor manufacturer's won't consider guarantee claims on devices that have been soldered in to equipment. As one sales guy from Fairchild Semiconductor said to me once, " Most of the data downloads that are done for our products are front page only." To keep things in perspective, (and to pick figures out of the air), if in 40 years the BC107 had an in service failure rate of 1/1000, and the BC147 had one of say, 20/1000, the BC147 has twenty times worse reliability than the BC107. But it also means that 980 BC147's are still working. Pessimism is a good personality trait in a design technician. Regards. Andy.
