5

When troubleshooting older electronics, the usual culprits tend to be connections and capacitors. The "solid state" components, if they have failed, usually in the role of victim e.g. bad voltage input.

So I was surprised when I found an IC chip in the middle of an otherwise functional circuit that was not behaving according to its datasheet. (In this case a WD1691V support chip in a TRS-80 Model III floppy interface.) Apparently this is not so unheard of.

For 1970s/1980s era TTL integrated circuits, is there an inherent failure rate due simply to "age" itself? Were some chips — whether through their design or a bad manufacturing run — more prone to random and/or wear-out failures even after an uneventful burn-in?

How likely is it that a typical LSI IC would fail:

  1. simply due to shelf-life, i.e. time since manufacture?
  2. as a gradual side effect of routine usage, i.e. proportional to operating hours?
  3. only because of out-of-spec treatment, i.e. eventually something bad (voltage spike, cosmic ray, diet coke fumes…) comes along and hurts it
Improve this question
4
  • 2
    my cat peed on my Oric Atmos. A few weeks later, the welcome screen background wasn't white anymore but blue...
    – Jean-François Fabre
    Feb 26, 2020 at 22:15
  • @Jean-FrançoisFabre Ha! I think that falls solidly under my last category of failures.
    – natevw
    Feb 26, 2020 at 22:17
  • 3
    The single most reason of failure (beside contact and discharge) is thermal cycling. You may want to look for contemporary studies. Not jsut for hard data, but also to narrow this broad question a bit down.
    – Raffzahn
    Feb 26, 2020 at 22:18
  • 1
    Hard evidence hard to come by. Anecdotally, these IC failures are pretty rare, especially when the containing computer is being treated as a treasured collectible, rather than as a discarded trash.
    – Brian H
    Feb 27, 2020 at 0:29

2 Answers 2

Reset to default
4

All chips will eventually fail due to age if something else doesn't get them first. The bathtub curve reflects this and shows how burn-in improves reliability by weeding out early failures.

Eliminating #3 ("only because of out-of-spec treatment"), the question becomes, what percent of chips die because of age and what percent die because of usage? I think it should be restated as something like, how many hours of shelf life is decreased by an hour of usage? And how many hours of shelf life is decreased by each power on/off cycle (Mean Cycles Between Failure or MCBF)?

Higher chip temperatures (above about 60C) and voltages also shorten a chip's life. And there are probably other metrics that I'm not thinking of.

I'm sorry for the long non-answer. It's possible that the data and even the metrics are secrets which are closely guarded by the chip makers.

Improve this answer
2
  • What are possible causes for such a bathtub curve? Why do higher temperatures alone (without cycling) could decrease chips's life?
    – lvd
    Feb 27, 2020 at 13:42
  • I think it's that obtaining such data through testing is too expensive for anything other than the military or spacecraft use.
    – Brian H
    Feb 27, 2020 at 16:15
2

I'd hypothesize that the most frequent reason for old plastic-cased chips to fail is the plastic package itself. Inevitably there is a strain or pressure from the filling plastic to the crystal itself or to the bonding wires. After a long time, plastic might absorb water and the tension would change, or it might eventually get a small crack inside, or simply due to periodic (caused by thermal cycling, as mentioned by @Raffzahn) or constant stress aging -- crystal or bonding wires get destroyed and fail.

I might be wrong here, but my conclusion is that the ceramic package is more long-lasting. Probably those who have real expertise could (dis)agree now.

Improve this answer
1
  • 2
    Back around 2000, while working in MEMS packaging, I looked through the literature comparing modern plastic packaging to high-rel mil-standard ceramic packaging. In long-term studies of usage in harsh environments, the plastic packaging did substantially better than the ceramic packages (except for the counterfeit chips found in the test!) . Now, current chips and plastic packaging are a well known art, with the early issues with plastics worked out. And, ceramic packaging is now pretty low volume and not broadly used, so some of this is not applicable to 1970s era packaging.
    – Jon Custer
    Feb 27, 2020 at 15:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .