Help! Rebel getting brighter over time!

[gazmono]

Hi, I am a technician working on a stop-frame animated feature film. Over the last 6 weeks I have started using Luxeon Rebels in 'Practical' props. A 'Practica'l is any light which appears on screen and has to actually work, eg Table lamps, candles, torches, etc.
I incorporated a Rebel LED into a model flashlight (only 10 x 40 mm in size). It's been going well, giving a great little beam which looks good on camera. On a recent shot however, one of them (I made 3 ) has started to increase dramatically in intensity. the cameraman had to reduce the voltage by two volts over the course of a day.
Any ideas what could be causing this?
Stop-frame animation takes place over such a long time scale everything has to stay at the same level. A small light change over the course of a day will translate as a flash on screen. (An animator is only expected to do two seconds a day!).
Is it known for High power Leds to increase in output over time?
Do they need a burn-in time?
Maybe the heat-sinking is inadequate?
I got around the problem of needing a large heat-sink by only delivering full power to the led when an exposure is being taken.
I am powering the leds using variable-voltage electronic lab power supplies, with a dropper resistor to limit the current to 750 mA max, and to enable it to be used over the 12 volt range of the power supply (and to standardise the voltage of all the practicals).
This particular' torch' has been run at a higher voltage than the others this time.
Any help from you boffins would be much appreciated.

 

[mudman cj]

I suspect that running this particular torch at a higher voltage resulted in greater current and therefore higher die temperature. This tends to burn in the LED and reduce the Vf so that less voltage will be needed to achieve the same drive current. I recommend switching to a current limiting supply instead of using limit resistors since you have minimal heat-sinking.

 

[gazmono]

Thanks very much. That makes sense. I have got some LM317's here, so I will look into making a variable current limiting circuit.( I still need dimming ability, hope that doesn't mean getting into PWM). Ta very much.

 

[mudman cj]

This is not your usual solution, but since this is not a situation where efficiency really matters you could dim using a pot. Add the potentiometer in parallel with the LED so that an adjustable fraction of the current bypasses the LED through the pot. The circuit doesn't have to work any harder and the LED will still never get too much current. You will want to use a small range pot with logarithmic taper for best control of LED brightness.

 

[CKOD]

The problem is as LEDs heat up, the forward voltage drops, on a fixed voltage power supply, current increases. The resistor limits this but its not constant, the current will still increase some. You need to drive the LED with a constant current, and since you have a 317 on hand, youre good to go.
http://en.wikipedia.org/wiki/LM317
They have a schematic there, the regulator tries to keep the V-out pin 1.25V above the ADJ pin. So a pot/rheostat instead of the fixed resistor pictured will let you have a variable current, so your current will be 1.25(v)/R=I. Figure out what the lowest current you'll want and pick an appropriate sized pot. 250 ohms gets you down to 5mA, so find a log-taper pot in that range, just mind your power dissipation in the pot. 1/2 watt pot would limit you to 400mA without overheating it. Ideally, you could get another regulator with a lower reference voltage, but thats not super critical since youre running off a power supply.

 

[gazmono]

This is great, but I need to be able to drive the leds up to 750 mA. This means using a logarythmic 1 Watt potentiometer, (50 or 100 ohms would do). I cant seem to find them anywhere, only linear taper.
Should I struggle with a linear pot, or is there an alternative circuit that will do the same job?

 

[CKOD]

This is great, but I need to be able to drive the leds up to 750 mA. This means using a logarythmic 1 Watt potentiometer, (50 or 100 ohms would do). I cant seem to find them anywhere, only linear taper.
Should I struggle with a linear pot, or is there an alternative circuit that will do the same job?

You can give it a shot with linear, it might just be a tad twitchy at the higer end of current ranges, figure out what your maximum current will be and put a resistor in series to keep you from turning it to 0 and just giving it the full supply voltage. Since you want 750mA that'd be 1.66 ohms, Plus this would reduce your power dissipation some in the pot. If you dont have a resistor, you could probably use another pot put out of the way, so one is to set the maximum current, one is to adjust the output in use.

 

[Th232]

Not sure on your budget and all, but a good number of TaskLED's drivers offer the option of using a 50k logarithmic pot to alter the brightness. Since you aren't dissipating power with them you don't need to get a 1W pot.

Admittedly, it's rather expensive if all you're driving is a single Rebel, but it's still another option, and it wouldn't surprise me if there were other drivers out there with similar functionality.

 

[gazmono]

Thanks for all the help guys. Very helpful and concise. After the shot finished I took the 'torch' apart and discovered that there was a gap between the aluminium 'slug' that the led was mounted on and the metal casing of the torch, not only that but pvc tape was holding the wires in place, acting as insulation and there was no heat transfer paste. It must be a prototype that got used without having proper heat-sinking work done on it. Actually I am pleased because I have learned about constant current and built a variable constant current cct, which seems to work well, I am making more to use with all future LEDS. Thanks once again, this site is great.

 

[mudman cj]

You're welcome. We'll keep our eyes peeled for "CPF boffins" in movie credits. :grin2: