1
\$\begingroup\$

I am conducting a risk assessment study of a system. To do so, I need to know the failure probabilities of its components. Some manufacturers do include this piece of information in their datasheets, other manufacturers don't and in the worst case some manufacturers don't even have a datasheet.

For example, I have been able to find the failure rate for the ATmega128 on page 8 of the datasheet:

Data Retention

Reliability Qualification results show that the projected data retention failure rate is much less than 1 PPM over 20 years at 85°C or 100 years at 25°C

On the other hand I have been unable to find any useful information for the CYRF6936 in the datasheet.

Finally, I can't find any datasheet for a linear B5K potentiometer.

  • Where/How can I find/estimate failure rates and probabilities when there is no information available?
  • Are there any known generic failure rates/probabilities for electronic components such as sensors or potentiometers?
\$\endgroup\$
  • 2
    \$\begingroup\$ You could look at MIL-HDBK-217, you can find PDF copies online. Better than nothing, but others may have a better reference. \$\endgroup\$ – John D Dec 1 '17 at 1:02
  • 1
    \$\begingroup\$ It never hurts to ask the manufacturers if they have any unpublished reliability data. \$\endgroup\$ – Matt Young Dec 1 '17 at 1:34
  • 1
    \$\begingroup\$ Our practice was to derate components, so for instance resistors would be operated at never more than 50% rated power, so they 'didn't need to be considered' when doing an MTBF, and we only needed to search for real figures for the complicated stuff. Justified? No, not really. Simple, conservative, consistent across projects? Yes. Did it lead to any serious mis-estimates of MTBF? Who knows, nothing glaring surfaced in the decades of subsequent operation. There's such a strong dependence on operating temperature that that will swamp everything except manufacturing hygiene (the biggest factor) \$\endgroup\$ – Neil_UK Dec 1 '17 at 5:27
  • \$\begingroup\$ With potentiometers, one big failure mode is wear, and datasheets tend to have data on that - or is your question limited to trimpots only actuated at production/maintenance/repair times? \$\endgroup\$ – rackandboneman Dec 1 '17 at 19:01
3
\$\begingroup\$

Your best choice is to ask the manufacturer, and ask them to give you a document, not just a mail.

If for some reason the manufacturer can't give you any information, you have to guesstimate how reliable a part is.

There are several standards which can be used to estimate a MTBF for a part.

  • Mil-HDBK-217
  • Telcordia SR-332
  • IEC-TR-62380
  • Siemens SN 29500
  • FIDES 2009
  • 217Plus
  • GJB/Z 299C

I can't really comment on how useful each and everyone of them is, I think in our safety assessments we use different standards.

Siemens and IEC are the ones I already stumbled upon. But in some cases things tend to get very odd. Like failure rates for RAM, where you can get values from almost 0 to so large it dominates the whole system and with a RAM-test you just removed 90 % of the failures and are "done".

On this website I found a few comments on those standards, which suggest that the Mil-HDBK-217 is the most pessimistic guess and if you apply that you'll probably end up on the safe side.

Another source for information is usually the certification authority which is usually involved in the process of getting a SIL certified product for example. But in most cases, they hint you to the same standards, but you get an idea which one they prefer and using that reduces the amount of time spent arguing with them.

\$\endgroup\$
2
\$\begingroup\$

Failure rates are very strongly dependent upon temperature. Thus lots of PCB VDD/GND planes to spread or remove heat to mounting bolts, metal chassis to further extract heat, and fans! will reduce the expected failure rate.

And that is the point of my answer. By significantly dropping the temperatures, the MTBF greatly increases. CONTROL THE HEAT. Design to extract the heat. Copper foil has 70 degree Centigrade Thermal Resistance per square of foil (any size square) per Watt of heat flowing along the sheet of foil; not flowing from face-to-face that 1.4 mils distance, but along the sheet

Can you pass heat from plane to plane, thru the FR-4? YES. But the FR-4 is glass and epoxy, both with HIGH R_thermal_resistance: 200X the R_th of Copper. At 1/16th thickness (0.166 inch), divide that by the copper thickness (default for 1 ounce/foot^2 is 1.4 mils), thusly 0.166/0.0014 == 120x. (So what does this mean?)

For good heat transfer, you need a LOT OF OVERLAP between planes to transfer the heat.

I will revisit this answer later today, with some finite-element modeling (on a napkin).

\$\endgroup\$
  • 2
    \$\begingroup\$ Of course, as soon as you add fans, you can be fairly sure the fans will fail before anything else. Well, at least that is how it seems to me based on products I have purchased (not the ones I have designed, which so far have been low power products with no fans). Actually this may be brilliant advice. Add a fan with a published MTBF. Use that MTBF for the whole system. Saves a lot of work. \$\endgroup\$ – mkeith Dec 1 '17 at 7:35
  • 1
    \$\begingroup\$ And that is the point of my answer. By significantly dropping the temperatures, the MTBF greatly increases. CONTROL THE HEAT. Design to extract the heat. Copper foil has 70 degree Centigrade Thermal Resistance per square of foil (any size square) per Watt of heat flowing along the sheet of foil; not flowing from face-to-face that 1.4 mils distance, but along the sheet. \$\endgroup\$ – analogsystemsrf Dec 1 '17 at 17:26
1
\$\begingroup\$

Where/How can I find/estimate failure rates and probabilities when there is no information available?

You can't. You simply have no way to guess what reliability levels the parts are intended for, let alone actually built to.

Are there any known generic failure rates/probabilities for electronic components such as sensors or potentiometers?

Again, no. If you want to assume the parts were constructed to conform to military standards (and you know which standards were used), then John D's suggestion about MIL_HDBK-217 is perfectly reasonable. But that does not seem to apply the ATMega part.

So, basically, you're on your own.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.