High-Reliability Designs Reqire Sophisticated High-Reliability Testing

In doing high-temperature/high-reliability designs, one must read and understand much more than just the data sheet. One cannot just look at the sheet and assume the part is good for, say, 125°C in one's application. Many mil-spec parts are qualified for only 1,000 hours in much of their testing over environmental conditions. This may entail a whole lot of surprises if the device must last for 5,000 hours, 10,000 hours, or even longer.

This may leave one with a whole bunch of dead parts if one does not carefully study the test and qualification data for each piece — not just the base parameters.

In equipment used for heating, ventilation, and air conditioning, components in a rooftop unit will often make one large thermal cycle per day, as well as many smaller cycles or ripples of temperature. Design engineers must manage their own temperature tests on each component used in the product.

Typically, such testing must go through many thousands of thermal cycles to look at how aging breaks down the compounds used to make the part. These tests are intended to simulate the desired 200,000 hours or longer of thermal cycle conditions. Note that similar testing would be done for similar industrial applications, such as commercial ovens and petroleum production (well drilling).

IC molding compounds can break down. Varnishes and conformal coatings on discrete components can dry out and crack. Dendrites (metal whiskers) can form in devices. Metal migration can occur. Metal (or semiconductor material) is pushed out of its original position over time at high temperature and in the presence of sufficient electromotive force (voltage) or current flow.

Inductor and motor magnet wire will deteriorate due to fractures in the enamel or polymer coating. Rubber or plastic insulating jackets on cables or individul wires soften, crack, or dry out. In coaxial cables, the dielectric (insulation between the center conductor and the shield) deteriorates in a similar way. But with coax, this causes an additional problem: The characteristic impedance changes. This can cause additional signal attenuation and a degraded standing wave ratio with the attendant signal refections.

Rubber, plastic, cork, and other gasket materials dry out, crack, and let in water or other contaminants. Mold and fungus attack components and coatings. Humidity penetrates coatings and causes corrosion of components and solder joints. If it can go wrong, it usually does.

What does one do? Life testing — boring and grueling life testing — is often the only solution. For less critical designs, put less expensive electronics on a life-limited part, and replace it at service intervals. What are your experiences with long-life parts? How do you keep things running longer than normally seems possible? What deratings for parts do you consider?

Related posts:

17 comments on “High-Reliability Designs Reqire Sophisticated High-Reliability Testing

  1. eafpres
    September 5, 2013

    @William–dead on you are about coaxial cable degradation.  Once had an automotive application that started out using standard RG-174.  Moved to RG-174 with irradiated polyethylene dielectric as it was a little lower loss.  Despite the manufacturer's claims, we had issues in thermal cycling tests.  It was found that the dielectric would undergo a significant shrink in the initital thermal cycling, and this could yank the center wire out of the center pins and other bad results.  Eventually it was found that thermally soaking the cable before cut/strip/assembly made it much more stable.  This requirement was passed down to the manufacturer, and a very unique, low cost cable was born.  Literally hundreds of millions of feet of this stuff are driving around today, without reliability issues.  Without the product level testing, and subsequent process changes, these parts would have failed even though the cable was specified to the required temperature range.

  2. samicksha
    September 6, 2013

    I agree @ eafpres, or we can say Signal Loss, considering skin effect, magnitude of an ac in a conductor decays exponentially, i.e. in a shield of finite thickness, some small amount of current will still be flowing on the opposite surface of the conductor.

  3. BillWM
    September 6, 2013

    COAX — Yep — have killed many of these new scope probes from the various vendors trying to figure out how to do High Rel Testing of High Speed Signals

  4. eafpres
    September 6, 2013

    @William–killing probes sounds pretty interesting.  Can you elaborate more?

    One thing I didn't mention is that the aforementioned RG-174 derivative was OK in short lengths–the cumulative shink was not enough to mess up the connectors etc.  But in long lengths, the cumulative shrink would tear things apart.  That leads to an interesting conclusion.  If you test a design involving coax runs, you could build a test device with, say 10x the length and measure degradation with that and extrapolate.  I've never heard of anyone using a longer coax run to increase the signal vs. time evolution in an accelerated test.  I think it would have merit.

  5. Davidled
    September 6, 2013

    I am wondering how the coax cable is correctly responded to high frequency, after soaking it in the thermal cycles. I think that mainly, coax cable might be degraded even though they meet the requirement for any criteria.

  6. Netcrawl
    September 7, 2013

    @Daej I'm not sure about coax's response on this but we all know that coax cables continue to be improved in response to customer requirements, and modern coax emphasize precision and reliability, I think its still good.     

  7. goafrit2
    September 8, 2013

    The anormaly in the semiconductor industry is the notion that you need this highly paid designers and extremely low-paid constracted staff. That is how it works. You invest in designs and you blow the testing.

  8. goafrit2
    September 8, 2013

    The Apple VP that supervised the design of iPhone antenna that resulted to the antenn-gate “scandal” lost his job. It is high time we understand that great design requires great testing to work.

  9. goafrit2
    September 8, 2013

    >>  have killed many of these new scope probes from the various vendors trying to figure out how to do High Rel Testing of High Speed Signals

    Good point. I continue to ask the day we will have “cordless” scopes. But it seems the EMI and others will not allow that. A product is a complete package – it is all about the weakest link.

  10. BillWM
    September 8, 2013

    Elaboration — Often there is coax that will take 8hrs at say 125 deg C — Try and use it for 80hrs at 125C and the dielectric filler flows and the shield collapses and shorts to the center conductor!

  11. antedeluvian2
    September 10, 2013



     I continue to ask the day we will have “cordless” scopes. But it seems the EMI and others will not allow that.

    It may be closer than you think. Take a gander at this. I actually did a log on the late MCC about it as well, but all that is left is this YouTube video

  12. goafrit2
    September 12, 2013

    @William, the case of coax is a real challenge. The problem is that the degradation happens in ways that one may not even notice it. We may need better dielectric design and stronger shielding to avoid the collapse

  13. goafrit2
    September 12, 2013

    >> I actually did a log on the late MCC about it as well, but all that is left is this YouTube video

    Very cool one – seems like the MCC site is redirected for the content.

  14. yalanand
    November 30, 2013

    @William, thanks for this post. I want to know more about this metal migration problem in IC. How can we solve this problem for higher reliability of IC? Can we do anything during different stages of development of IC to avoid metal migration problems ?

  15. BillWM
    November 30, 2013

    Basically the issues are  temperature and voltage in relation to feature size related — This means if one can afford more NRE for better design software it is able to check for more of these issues — one of my old managers had a patent on reducing electro-migration via special features in  the die as well.   There may be other methods as well.

  16. yalanand
    November 30, 2013

    one of my old managers had a patent on reducing electro-migration via special features in  the die as well.

    @William, thanks for the reply. I never knew we could reduce electro-migration via special features in die. If possible please share more details about this technique. 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.