0
\$\begingroup\$

I have heard it from multiple people over the last 5 years that Optoisolators are failure prone and should be avoided when able. I'm working on a project (for personal use) that I was going to put an optoisolator on to isolate a logic system from a 170V rail. I do plan to use this device for teaching and would hate for it to stop working because an isolator failed. I know there are magnetic coupling ones now that may be better?

The real question is that I have not been able to find any real data on the internet that confirms optoisolators fail easily or with respect to other solutions.

The only thing I found was a report by Sharp Found Here.

I'll also add all I need to do is move a square wave pulse of 100-500kHz from a logic supply that is floating to a -170V rail. So it can tolerate slow edges and such that a digital com system could not.

\$\endgroup\$
  • \$\begingroup\$ Can you provide more information about the multiple sources? Were these posts in forums, or verbal communications, or if not, then what were they? Basically, if you are able to provide links to them, it would help us evaluate them. \$\endgroup\$ – mkeith Oct 24 '15 at 17:20
  • \$\begingroup\$ Verbal from Engineers and Professors. \$\endgroup\$ – MadHatter Oct 24 '15 at 17:41
  • \$\begingroup\$ It is worth noting that LED's have a failure mode where the light output drops over time (opto's LED's or normal LED's) so unless you have made sure the drive current is suitable for the degraded performance after 5 years of 24/7 the use circuit can fail. ... also failure is defined as output light dropping by half. \$\endgroup\$ – Spoon Oct 24 '15 at 23:15
5
\$\begingroup\$

The best guide to component reliability I've ever come across is MIL HDBK 217 F. It details the reliability of components from the analysis of thousands of pieces of faulty/failed equipment gathered over many, many years.

For a simple, fixed value (< 1 Mohm) film resistor (section 9.2), the basic failure rate \$\lambda_b\$ is 0.001 failures per million hours at stress rate 0.5 and operation at 20 degC.

For an opto isolator (section 6.11) of the type with a photo diode the failure rate is 0.0025 times per million hours and, for a phototransistor it is 0.013 times per million hours.

So MIL HDBK 217F is suggesting that an opto might be about ten times more likely to fail than a simple resistor but it's still in the vicinity of once every 11,000 years!

Compare this with a diode (section 6.1) used for power rectification - it has a failure rate of 0.069 times per million hours - several times worse than an opto-isolator.

How many of these are you using in your design?

\$\endgroup\$
  • \$\begingroup\$ One, I figured since they are used in a lot of power supplies they can not just be failing all the time... Thanks for the Doc, I'll take a look. \$\endgroup\$ – MadHatter Oct 24 '15 at 17:42
  • \$\begingroup\$ I meant how many power rectifiers are you using. Be aware that the handbook takes about an hour to grasp the concepts from scratch but if you are wanting this sort of data it's well-worth the time investment. I still use it today and all our customers (who want an overall figure for a piece of equipment) still request we use this document despite its age. \$\endgroup\$ – Andy aka Oct 24 '15 at 17:53
  • \$\begingroup\$ Also note for multiple opto' where one failing the FMEA will add the per failings (still a small chance) equally failing safe and failing ... "bad" is systematically different \$\endgroup\$ – JonRB Oct 24 '15 at 18:06
  • \$\begingroup\$ @JonRB I'm not sure what you are saying. \$\endgroup\$ – Andy aka Oct 24 '15 at 18:20
  • \$\begingroup\$ Not refuting your reference (I use that, plus a few others). Just adding when the FMEA of a system is calculated other aspects come into play. Take an OPTO, take 6off OPTO's to interface to an inverter... one failing has a rate of 0.069 but any one of those could fail, thus the failure of the system due to the opto's in the inverter section would be 0.069+0.069+... ie 0.069*6 \$\endgroup\$ – JonRB Oct 24 '15 at 18:23

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.