Series connection of 10 LEDs?

Well, blue diode maybe.. protector Vf > LED Vf. Otherwise you're testing the protection diode instead of the LED ;)

Tim

If it's driven with a constant current source why not just a diode in parallel with each LED?

Sure, wrongly written in haste. A couple of diodes works, depending on the LED though. I'm not a huge fan of zeners for such things.

Isn't this what antifuses are for?

I'd go with a zener too. Remember that the zener will dissipate E*I so go with the smallest voltage that's still above the LED's working voltage.

Compare the cost of a zener with the cost of a current limiting resistor though. Wiring a pile of LEDs in parallel, each with a resistor, to a standard 5Vdc power supply, might be less expensive.

That would work. You just need to select a zener that draws essentially zero current at the normal LED drop but comes on without using up all the voltage compliance of your current source. Of course, it has to handle the power of the current times its voltage drop, but if these are ordinary LEDs that should be no problem. I think a 4.7 volt, .25 or .5 watt zener would work with most LEDs.

Hmmm, If you accelerate the human to near C... ;-)

Or accelerate the failure modes in a T&H oven, BIAS, etc. We used to do 30:1 sort of acceleration factors[*] for silicon devices (microprocessors and memory) that needed rating for 100kPOH.

If you're short on constant-current power supplies, you can give each diode its own LM317 regulator, run as a current regulator.

A normal diode would absorb most of the current, leaving the LED almost off. A zener diode is better, as the OP proposed.

Or even three :)

Agreed, I suppose you could go down to 2.7V but generally zeners add so much voltage to the chain that if more than one LED goes o/c you start to run out of drive voltage.

I have to admit I'm curious about the point of this, though, as LEDs are very reliable and you are unlikely to learn anything about MTBF in a normal human lifetime :)

If he pushes the limits (current, temperature, etc), he can easily see degradation in short time periods, but it is not trivial how to then extrapolate those results to nominal conditions. There's a lot of theory behind that.

Google for a LED switching supply.Those can convert up from the powersupply, and sense the resulting current at ground connection. If a Led fails "shorted",the others just keep on working. They work by switching a coil current into the stack of LEDS,and adjust switching freq. to get the wanted current. Note that it works without series resistor, so you use your energy very efficiently.

Put two more leds across each of the ones on test - or use a single one of a different colour which operates at a higher voltage.

Thus if a red leds are on test, wire a yellow one across each.

Just curious. What do you mean by "sense the resulting current at ground connection" and "without series resistor"?

Also, varying the switching frequency usually does not modify the average of any current or voltage, but only the amplitude of the ripple. This assumes that the converter stays in continuous conduction mode. An easy way of changing average voltages and currents is by modifying the duty ratio.

Best,

On Tue, 17 Jan 2006 18:21:19 +0000, Derek Potter wrote: [snip]

I have a 20+ year-old LED clock. It is so dim I can only read it at night. But your point is taken. ;-)

--Mac

There is a small resistor from the bottom LED to ground.This is filtered,and connected to the current sense input of the switch IC. The switch IC connects a coil to the power supply, until a set current is flowing. The IC is then swithes the coil accross the pile of LEDS,the coil will dump its stored power into the LEDS,just like it does into any parallel protection diode. Thats one currentpulse. If the sense input senses average LED current as to low yet, another dump cycle is performed. This results in a high enough dump rate to get the required average LED current. To decrease the peak LED current, you can also put a filtercap across the LED stack,so only the filtercap sees the current pulses.

Al this does not require a current limiting series resistor,but only a sense resistor. Googling for LED powersupply,one of the first entry's:

Was it always that dim and have you measured the LED current?

Thank you all very much for your answers!

This newsgroup is really a gold mine!

Just to add a precision: it is true that LED are reliable, but this is a research project, and the LED and not standard products, the technology is still not stabilised. That is why they need a life test in a thermal chamber.

I prefer to connect them is series, because I need to be sure that they all take the same current. Connecting them in parallel is difficult, as there are some technological variations if their Vf voltagen and for this reason their current will be different.

I will first try to connect the zeners, and if this doesn't work well, I'll then use a LM317 as a current regulator for each ;)!

Thank all!

Adrian Tuddenham wrote:

It was standard practice to use subspec components in radio alarms, so I'm not sure one can conclude anything certain fom your observation.

NT