Reliability of components that have been produced for many years

[matthuszagh]

(Disclosure: I don't know much about semiconductor manufacturing, and I expect that will come across in this post.)

I'm troubleshooting a device that uses 2N1711 transistors. Someone else, who experienced the same instrument symptom, mentioned that the 2N1711 can become leaky, causing the fault. I have yet to measure whether mine are indeed leaky, but this got me thinking about what I will replace it with if I find that it is indeed faulty. The 2N1711 are still produced and so it's likely I'll replace them it with new 2N1711. Apparently, the 2N1711 have been manufactured for at least 50 years (I have a 1972 data book that lists it). This leads to my question: for components that have been produced for so long, does the reliability improve over time? That is, if I took a NOS 1980 2N1711 and compared it to one produced in 2015, is it likely that the modern part would be more reliable/last longer than the old part? It's worth mentioning that while this comparison probably applies to semiconductors (which I think exhibit minimal shelf aging), it wouldn't apply to some other components such as electrolytic capacitors, which do have somewhat of a shelf life.

This question is based on the idea that manufacturing capabilities have improved over the preceding decades. It seems that this might also translate into improved manufacturing quality of old processes (though it also seems possible there would be no such improvement). The final step in this thought is that companies would adopt improved methods for old parts as they become available. A final possibility is that even if manufacturing quality of old processes improves over time, components might have been produced well within manufacturing capabilities such that process improvements cause a negligible effect on reliability.

I expect a lot of people have better insight into component manufacturing than myself. Any thoughts on this?

[ebastler]

I don't think there is a general answer to this. Yes, manufacturing capabilities should have improved over time, and any early teething issues should have been identified and ironed out. But on the other hand, production may have moved to a low-cost factory and a brand selling generic components at low margins at some point, and quality may have taken a hit.

Also, I would not rule out shelf-life issues for semiconductors. Oxidation of the exposed contacts is the obvious effect, but I have read repeatedly about micro-cracks forming in the packages and environmental degradation reaching the innards too.

My personal conclusion would be to
- buy modern parts where available,
- buy NOS components if they are the only game in town or needed for "museum level" authenticity,
- don't replace all components of a certain type in a shotgun approach, only the ones that have failed -- unless you have observed a pronounced increase in failure rate in the actual product.

[Kleinstein]

With rather old transistors it is possible that the new ones that are sold today can be NOS ones, so from wafers pocessed decades ago or they could be produced in a different more modern process and just behave similar in the main parameters. The later is especially the case with parts from Chinese or similar produces that now also make old standard parts and never did the original design.

The improvements in the processes are more toward allowing smaller dies, better purity that may lead to less 1/f noise and better reproducibitly. Much of this does not effect reliabilty but more the performance. As an exaple new TL082 usually all meet the stricter noise specs of the TL072.
A change in the internals can improve things but could also just mean cheaper parts.

If old parts had reliablity issues, like the purple plague things are likely better with modern parts, or those old parts are no longer available.
On the other side they can no longer use lead and this could in a few cases cause a higher likelyhood for failure from tin wiskers.

[David Hess]

The 2N1711 are still produced and so it's likely I'll replace them it with new 2N1711. Apparently, the 2N1711 have been manufactured for at least 50 years (I have a 1972 data book that lists it). This leads to my question: for components that have been produced for so long, does the reliability improve over time?

Parts which have a long production history usually move between processes as those processes are discontinued in favor of newer processes, so any benefit of long production on the same semiconductor process is lost.

Reliability however does generally improve over time with manufacturing experience.  For instance we now know about the problems with tin whiskers and packaging, encapsulation materials have gotten better, etc.

[dobsonr741]

Given it’s a hermetic case, I expect no environmental degradation at all. If it was sitting on the shelf for 50 years electrical stress is excluded. IMO it is as good as it was when manufactuingle epitaxial transistors are simple, the advances are for efficiency of manufacturing, not necessarily improving reliability, that is already high.  They usually die not b/c degradation rather over dissipation/high current.