Series connection of 10 LEDs?

Well, let's not get bogged down in this, but you're going to have to overstress them a lot to get a significant failure rate with a one-off simple test. Arrhenius equation is already out of its depth when applied to failure rates (AFAIK it is basically about chemical reactions) and becomes quite meaningless once you exceed the manufacturer's ratings to force a failure. New failure mechanisms spring into existence at higher temperatures. Once the die melts, for example, the MTBF is zero :)

I'm always intrigued by manufacturers' current ratings. If anything is hurt by current as such, rather than heat, it should show up if you pulse the LED at low duty cycle keeping the same mean current. AFAIK, this effect does not take place - at least, not until the RMS current gets to silly values that damage the chip bonds. I suppose there could be a kind of second breakdown effect but I don't think this can generally be produced in diodes. Other than that it's hard to imagine a mechanism for current as such to cause failure - carrier flow does not disrupt the device structure.

Really there's not much else apart from temperature that you *can* push. Voltage is determined by the diode and the current, you'd need enormous pulses to increase the filed strength sufficiently to cause migration of doping agents in the chip. Hardly in the same league as a string of LEDs running off a DC PSU.

Hmm. Did I say "let's not get bogged down in this"? i seem to have rambled on a bit.

Not if you do it right.

Nobody is suggesting that you connect them "in parallel". Connecting LEDs in parallel is bad practice: one will hog _all_ of the current, and when it fails, the next higher Vf unit hogs all of the current, and so on, like popcorn.

What's being recommended is running them simultaneously, each with a reasonably well-calibrated, stable current source. (or a resistor ;-)) All of them set to the test current, of course. An LM317 can achieve arbitrary precision and arbitrary voltage compliance, with a little design work and maybe trimming.

Especially if you're doing research on brand new technology! This is no time to scrimp! :-)

Oh. You didn't say that at first.

Never mind. ;-)

Cheers! Rich [original context follows]

See Mil handbook 217. Most failure modes do follow the Arrhenius relationship, this includes metal and glass crack propagation, all kinds of ion / contaminant diffusion, most kinds of fatigue, etc., Just the same, there are often boundary limits where it breaks down. Test design relies on knowing where these limits are and leaving convincing clearance from them.

Too bad that LED's usually fail open.

Cheers,

Phil Hobbs

Well, it still works after 20+ years. Must not have been too bad.

It is a talking alarm clock which has a selectable gender voice. When it is time to wake up a voice says "Good Morning it's 6:30 am," and then the radio comes on.

If you set the clock for evening time, then the voice says "good afternoon," instead of "good morning."

--Mac

No, on both counts.

There is a two-position switch to change the brightness of the display. When the clock was new I used to keep it on dim all the time. Now the bright setting is much dimmer than the dim setting used to be.

The clock also has a large announce button on it. When you press that, a voice tells you what time it is. Sometimes in the daytime I use that button to figure out the time, since I can't read the display.

Mostly I just use it as an alarm clock, so it serves its purpose fine. I guess I should get a new one, anyway, but I'm not sure if you can still find talking alarm clocks in the US.

--Mac