9
\$\begingroup\$

I am trying to develop a very high reliability product, with a MTBF exceeding 100,000 hours, hopefully in the 250,000 hours range. Yep, I know - stupidly high reliability, at least 10 years.

Most LED's I've seen have a reliability rating of only 20,000 hours. What happens after this? Do they reduce in output too much? Are they usable as power indicators past this point?

\$\endgroup\$
7
  • 1
    \$\begingroup\$ Make sure your system can withstand all sorts of possible faults and input transients, otherwise being concerned with MTBF for discretes and passives is asinine. \$\endgroup\$
    – Nick T
    Commented Nov 15, 2010 at 3:41
  • \$\begingroup\$ The more expensive version has input transient protection from -40V to +60V, the other version is rated to go from -40V to +24V. \$\endgroup\$
    – Thomas O
    Commented Nov 15, 2010 at 11:14
  • \$\begingroup\$ Shorting outputs together, shorting outputs to the battery/ground, inductive spikes on power lines, also coupled into signal lines, ESD on every pin and more are all things that our finished modules must take without ill effect as per SAE standards and customer requirements. \$\endgroup\$
    – Nick T
    Commented Nov 15, 2010 at 22:10
  • \$\begingroup\$ Shorting outputs together: check - 1 kohm resistors in series. \$\endgroup\$
    – Thomas O
    Commented Nov 15, 2010 at 22:58
  • \$\begingroup\$ Shorting outputs to battery/ground: check - resistors with the clamping diodes in the MCU's. \$\endgroup\$
    – Thomas O
    Commented Nov 15, 2010 at 22:58

4 Answers 4

9
\$\begingroup\$

Must say I've always been highly sceptical of the meaningfulness of MTBF figures. LEDs will gradually reduce brightness over time, but it will be many times their stated life before they cease to be useable as a power indicator. Running at well below the specc'd current will increase lifetime many-fold. And would failure of a power-on LED be counted as a failure of the product?

\$\endgroup\$
3
  • 1
    \$\begingroup\$ We do the same - drive the LED's with less than half the specified current. This increases the lifetime of the LED's significantly by many years, before the LED starts to reduce in noticeable brightness, and many more years before the LED completely dies. \$\endgroup\$ Commented Nov 15, 2010 at 0:32
  • 1
    \$\begingroup\$ Would it be sensible to increase the LED's current once the brightness starts decreasing? Or would that shorten the LED's life faster? Will pulse-width modulation vs. DC make a difference in lifetime? \$\endgroup\$
    – rwong
    Commented Nov 15, 2010 at 7:50
  • \$\begingroup\$ @user1450 If you have a fixed current source driving the LED, the brightness should stay fairly even until total failure (not accounting for things like phosphor degradation in white LEDs). Most LEDs in this world are probably driven from a voltage source through a resistor, however, which is the cheapest way possible and sufficient for almost every application. Adding additional circuitry to account for the aging of the LED would make no sense in the vast majority of these instances. \$\endgroup\$
    – Nick T
    Commented Nov 15, 2010 at 22:12
5
\$\begingroup\$

I think you are utterly doomed trying to get to MTBF of 250000 hours on any usefully complex product, but then you haven't said what it is, so it might be very simple.

LEDs tend to age faster if they're hotter / driven at higher currents, and the failure mechanism is different for white leds (which are fluorescent) and natural-coloured LEDs. But generally they get dimmer as they age.

I would expect that you would get very long life if you significantly under-drove the LED (modern traditional-colour (red, etc)) LEDs are incredibly bright and tend to need very little current to act as a useful indicator, and if you weren't too fussy about the specific output you could play faster-and-looser about the failure condition.

But if I was trying to make a fantastically reliable product, one way I'd do that is to dispense with absolutely every component I possibly could. I expect the power-on LED would fail to make the cut in that regard.

Update: Well, clearly I know even less about reliability than I thought I did - it seems lots of diskdrives claim MTBFs well over 250k hours - so good luck with your product!

\$\endgroup\$
6
  • \$\begingroup\$ Well, I'd like the highest reliability possible because I hate unreliable products, I've had so many and I don't think it will be too difficult to reach such a high reliability. \$\endgroup\$
    – Thomas O
    Commented Nov 14, 2010 at 23:02
  • \$\begingroup\$ It's a fully solid state design and all components are 1.5x to 2x rated where necessary. \$\endgroup\$
    – Thomas O
    Commented Nov 14, 2010 at 23:03
  • 1
    \$\begingroup\$ Does it have software in it? You're totally hosed then. \$\endgroup\$
    – Nick T
    Commented Nov 15, 2010 at 1:16
  • \$\begingroup\$ Yep, but the micros have flash retention of 40 years. \$\endgroup\$
    – Thomas O
    Commented Nov 15, 2010 at 7:57
  • \$\begingroup\$ @Thomas - Flash is hardware. I'm talking about how you would prove that the software will never fail. \$\endgroup\$
    – Nick T
    Commented Nov 16, 2010 at 16:32
3
\$\begingroup\$

MTBF is based off of actual run time. So at a goal of 250,000 hours you are hoping your product can run 24/7/365 for almost 30 years.

A MTBF of 20,000 gives you just over 2.25 years of constant run time at recommended max power. This time is pretty good for most commercial products right now. If your LED is on for 8 hours a day, you are looking at a life of almost 7 years. Now if you decrease the power your putting into the LED you can even extend this life longer.

Now with all of that said, if it is just a power indicator you wont be switching it often so you will put less wear on the device.

I know I haven't directly answered your question, but I am hoping you are starting to realize that MTBF isn't something you should worry about too much. I realize you want your device to last awhile, which is a good thing, but you will probably find that if you always go with the highest life items they will generally be more expensive which will cause your product to be inflated in price, resulting in very few wanting to buy it. What good is a device that lasts for 30 years if no one buys it?

\$\endgroup\$
1
\$\begingroup\$

MTBF is assurance that reasonable derating guidelines are met. Nothing more. An MTBF of 250k hours does not mean any single piece of equipment will last 30 years. It's possible to get that sort of MTBF on equipment that uses electrolytic capacitors, which almost certainly will dry out in much less time than that.

Drive the LED with the minimum amount of current that provides a useful amount of 'signal' (i.e. is viewable under whatever conditions you deem inportant.) The LED manufacturer should have datasheet information about operating current / temperature rise / ambient temperature, so you can judge how 'hard' you're driving it and make your own decision about reliability.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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