Dodgy LEDs!

They may be low-current LEDs (~3mA). LEDs often change colour when overdriven. Perhaps it changes the band gap in the semiconductor, I don't know. But I've seen it many times.

Current in original application was approximately 50mA - driven directly from a CR2032 coin cell, in fact it may have been slightly higher due to the slight drop across the milliammeter I measured it with!

have

very

and

Sounds like a current overload to me. Going orangey usually indicates that. Are you sure you are not 'off by 10x' on the milliamps of drive.

Try one on a bench supply with a current limiting resistor.

-- Graham W

I've seen exactly that effect when deliberately running leds overcurrent as an 'experiment'.

Graham

There are some possibilities. First, how long will the battery maintain this current level?. Second, given the nature of the load, your 'constant current' supply may not be as good as you think it is. It could for instance be oscillating (check with scope, rather than meter), or possibly delivering more current without the extra resistance of the meter present. Though this should also mean that the coil cell current would increase as well, if this is actually being limited by the chemistry rate, rather than the ESR of the circuit, it might remain nearly constant.

Best Wishes

Your post has not made it quite clear: the 25mA that you measured, is that the drive to the current mirror, or is that what you actually measure as going through the LEDs? (In other words, could it be a problem with the current mirror?)

If the 25mA is indeed going through the LED, then it would seem there are only a few possibilities:

If it were me, I think I'd do as Graham suggests, skip the current mirror for now and test them with a known-good bench supply (or 9V battery) and a current limiting resistor. Try to eliminate variables...

Here's a hunch. When fed from a coin cell, the current would be limited by the coin cell's internal series resistance, and can't go above a certain value. Ditto for the voltage.

When fed by any sort of active-drive circuit that's hooked to a low-impedance power supply, the current will be limited by the accuracy, and dynamic characteristics of the active drive circuit. In your case, if I recall correctly, you've got a current reference tied to a current-mirror arrangement.

I wonder whether your drive circuit may have a turn-on or turn-off glitch of some sort, which hits the LED with a momentary overvoltage or overcurrent pulse? Even a very short spike (microseconds or less) which exceeds the LED junction's absolute-maximum pulse ratings could damage the LED.

You might be able to determine if this is the case in a number of ways:

- Use a fast oscilloscope to capture the voltage across each LED during the turn-on and turn-off events

- Use a fast 'scope, plus a small-value current-sense resistor, to capture the current through the LED during the transients

- Do an accurate SPICE model of the circuit, including all of the dynamic characteristics of the LEDs and the current-drive/mirror circuitry You might be able to avoid the problem, pragmatically, by simply slowing down the turn-on and turn-off events (e.g. sticking an inductor in series with your current reference so that it powers up more slowly, and thus delivers a slow rising-edge to the LED, or by placing a bypass capacitor in parallel with each LED to shunt the excessive current pulse to ground), or by adding some independent current limiting to each LED (i.e. a resistor) to ensure that the LED can't be driven to voltages or currents above its abs-max rating no matter what the active current-drive circuit does.

The battery does fade quite quickly - the initial current is somewhere in the region of 80mA, I held the meter on it long enough for it to fall to

I checked 1 of the 4 current mirror nodes - it read slightly higher than the current source JFET at 27mA, which is nowhere near the 50mA it settles at with the CR2032.

My Tek 465 might be fast enough, but what is really needed is a storage scope to capture the turn on instant.

The idea of adding an inductance might be worth a try - just hope it doesn't turn the circuit into an oscillator!

snipped-for-privacy@radagast.org (Dave Platt) wrote in news: snipped-for-privacy@corp.supernews.com:

A SPICE model?!

He grabbed a few unknown LED's out of some kind of keyfob torch.

Forget trying to model the unmodellable to the n'th degree, do it emprically, from a safe starting test, in a way you can test immediately.

Take it back to the most basic possible thing. Grab a few resistors between

Start with the highest, measure the voltage across the diode, and the current through it. Look at the output. Then repeat with each resistor working downwards. Look at the diode to see if you get a broadening or shifting of the colour as well as brightening, and diconnect if you do.

You should be safe with currents up to 50 mA on most LED's. you'll just shorten their lives, but theu won't die in ten seconds this way, so have a go.

If you don't see anything inexplicable, try an LM317 as a current regulator set to 25 mA and run that from the battery. Or use a known good lab bench variable current supply. Don't try anything more complex until you have done this.

You haven't got a fat capacitor involved in this circuit, have you? I smoked a small pointer laser diode because of the cap across the supply lines from a variable supply which applied much more than the set voltage while off load.

No capacitors at all for that very reason, after increasing the current sensing resistor associated with the JFET the LEDs work perfectly. The spread of current for the current mirror's 4 outputs is about 2mA centred around 19mA. It seems there's no inherent problem with the circuit I've devised but the odd behaviour of the LEDs is very puzzling!

Anyhow, now I've proven the circuit with expendable cheap crap LEDs I can go ahead and build the final circuit with decent quality hyper-bright LEDs (probably Toshiba).