This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM339: FIT Rating; interpretation with single and multi device parts

Part Number: LM339
Other Parts Discussed in Thread: LMV331, , TL331, LM393

Hi there,

say i have a quadruple comparator with FIT rating 2.2 like the LM399 (A generic single one like the LMV331 has a FIT rating of 11.6).We have now the following discussion

One argumentation goes like: "If a failure occurs 4 Inputs have failed hence it would be better to use single comps so only one fails."

The argument against that: " The shared parts on the quadruple device are the most unlikely to fail when they have reached "the bottom of the tub" because now the inputs themselves are most prone to fail because of external stimulus. And if one of them fails for that reason only the one channel would go down, not all four."

Now my questions:

Is the FIT Rating a good indicator for above and can i use it to close that argument with data?

How do the shared parts in a multi part device influence failure rate? To clarify: Are they most , equally or least prone to fail or in other words is the assumption from above correct?

Can i equally divide the FIT Rating between each channel of the, in case of the LM339, four channels? If not why?

Thanks in Advance

  • On the one hand, the FIT rate applies to the entire device; there is no information about which part of the device caused the failure.

    On the other hand, TI's FIT rate estimation tool (ti.com/quality/docs/estimator.tsp) gives exactly the same values for the LM339‍DR, LM393‍DR, and TL331‍IDBVR. This would imply that only the shared parts (power supply and bias voltage) fail. The only fact I deduce from this is that the estimate is very conservative and that it is not possible to make any quantitative argument based on that.

  • Hi Matthias,

    this is a topic that can discussed on many levels.

    1. Failure rates is a statistical approach and no exact sience. You try to conclude from a observed failure rate gained from a relatively small lot and at very elevated testing temperatures on the general behaviour of other not tested chips. You hope that the fabrication process for any later produced chip is exactly the same as for the tested lot. You hope that the activation energy in the Arrhenius equation is exactly hitting the failure mechanism, which is highly arbitrary because damage can have many different causes with highly different activation energies.

    2. Real failure rates not only depend on the fabrication process but also on the soldering and the handling by the customer. In my experience its far far far more probable that a chip gets damaged because of poor soldering or poor handling by the customer than by mistakes during the fabrication process. Remember the early soldering problems of board computers of Space Shuttle.

    3. What is the purpose of failure rate published by the customer at all? It's important for the manufacturer himself to find out mistakes in the fabrication process. And it can be helpful for claims of damages when a customer assumes to suffer from "too many" damaged chips. But actions for damages are very very complicated. For the customer the published failure rate can be helpful to estimate the behaviour of circuit at elevated temperatures.

    4. Be aware that any published failure rate can only make a statistical statement on how a huge amount of chips is probably behaving at a certain temperature. But it doesn't tell anything about a concrete chip you have in hands. This chip can immediately gets damaged, not only when the chip has reached the end of its life time "calculated" from the failure rate. Yes, even a second chip can be damaged right now, and a third... There's absolutely no guarantee that every chip is doing well up to the end of "calculated" life time. Ok, a simultaneous damage of multiple chips is very improbable but not impossible. There's absolutely no guarantee which the failure rate can make on any individual chip you have in hands.

    5. Now comes the interesting question: What does a damaged chip very concretely mean for you and for your circuit? Does it make a difference whether only one comparator is damaged or four? Is it only a cost factor? Or does it have far reaching safety consequences? If it's only a cost factor because you have to compensate for a financial damage then the damage of only one comparator or of four comparators would not make any difference. But what if you produce an airplane, a missile or a nuclear power station? Can such a product tolerate the damage of a comparator at all? And what if not? What are you doing then? Then you will try to minimize the count of components in your circuit and you will go for redundant systems.

    Kai

  • Thank you, Clemens and Kai, once again for your great inputs.

    Matthias,

    I think the answers to your questions is very subjective and nuanced. A quad channel device would have more transistors and complexity than a single channel device and therefore you could say it's more likely to fail. And if one channel fails, it could affect the operation and behavior of the other channels. But as Kai mentioned, it's more likely you see the failure from a mechanical and board design aspect rather than a fabrication process. I would argue there's a less chance of having mechanical issues and failures of using 1 quad channel device vs. using 4 single channel devices. But there's also a factor in the application/use case of what the comparators are being used for, which is what Kai alluded to at the end. If a quad channel device is damaged, would that cause the entire system to fail? If only 1 out of 4 single channel devices fail, can the system still operate? Perhaps it may be better to have separate single channel devices so the damage is isolated. Sorry we couldn't really you give a definitive answer but it really comes to down to impact of the failure and what it means to your system. 

    -Chi