A Test for Inadequate Heatsinking

Ever since reading one of Beretta1526's posts in which he mentioned a loss of intensity (measured by light meter) when he hadn't glued one of his emitters to the heatsink well enough, I've been tossing that bit of information around in my mind. Later Doug S confirmed that this is indeed a consequence of heat buildup in the emitter.

Based on that raw information, I propose a very simple test to determine if a given emitter has been heatsinked well enough.

Using a light meter, aim the light at the sensor and turn it on as would be done when taking a standard lux measurement. Don't use the "peak" setting on your light meter, but instead allow the readings to fluctuate freely. Try to find the most intense portion of the beam, then keep the light as still as possible.

Now wait, allowing heat buildup and eventual stabilization (hopefully) in the heatsinking system. Compare the intensity reading you got when first turning on the light with the later reading after thermal transfer has stabilized. The greater the drop from first to second reading, the less adequate the heatsinking system is.

I'm not sure if it's reasonable to expect no decline whatsoever, but that would certainly be the goal. Tentatively speaking, I'd be concerned if there were a big drop, such as 10% or more, between the two readings.

As one example, a recent light I built out of a Brinkmann 2AA body and BB750 LuxIII sandwich showed initial intensity of 974 lux, but after a few minutes of runtime, when the body of the light had gotten noticeably warm, the intensity had fallen to 830 lux. It didn't seem to fall any further from there. I interpreted the results as indicating that my heatsinking was inadequate.

This to me seems like a great and simple way to test the thermal adequacy of your mod in an objective manner.

You need to use a regulated power supply of some sort for this, because, you may not be able to differentiate the drop in light output from the normal drop in battery voltage because of the power drain. In flashlights with larger thermal mass and smaller power sources, such as a 2 D-Cell Maglite with a 5W LS and 6AAs, the light output may drop off more from the battery drain than it will from excessive heat.

The light output will drop off slightly anyways during normal operation, since your flashlight body will heat up. Given unchanging power input, if your flashlight body heats up by 20C, then the junction temperature also rises by that much. A 20C increase in junction temperature accounts for a 5% loss in light output. So a 5% loss can easily be attributed to acceptable heatup when there is very good thermal contact between the emitter and flashlight body.

What Darin says above. I have done some testing of the type you describe and it is a viable approach. There are some things you need to understand. Your results are dependent on how quickly you take that first measurement. I have done this with some high speed instrumentation. The highest *rate* of output drop occurs in the first 20 milliseconds, the decline then continues at lower rate that changes again at about 3 seconds if there is a high resistance slug to heatsink joint or just declines smoothly through the 3 second point if this joint is good.
An entirely different approach that is easier [assuming that you have the equipment] is to apply a constant current to the LED and capture its Vf with a long duration sweep of a digital storage oscilloscope. With this method, you can actually measure the LED temperature rise *if* you trust the datasheet value of -2mV/C tempco for Vf. I am a bit suspicious of this value, I have tried to measure it repeatedly and keep getting a somewhat higher value.

evan9162, that's a good idea about the regulated power supply. I do think batteries would work fine in many cases. One way to do a sanity check as far as seeing if battery dropoff might have caused the loss of intensity would be to let the light cool and try again. If intensity doesn't return to something close to the original level, then the batteries might have lost some oomph. (Or worse, you might have cooked your emitter. /ubbthreads/images/graemlins/icon15.gif)

Regarding the cutoff point for loss of intensity indicating a problem, there's nothing special about the 10% figure I offered. The point is, the more of a decline, the less effective the heatsinking.

Doug S, when I measured the 974 lux in my example, I got the reading within the first couple seconds, but there's no way I got it in 20 milliseconds! /ubbthreads/images/graemlins/icon15.gif

The interesting thing is that the lux readings declined for a few minutes, then seemed to stabilize at 830 lux, give or take a couple lux. I took that to mean the system had reached thermal equilibrium. Was I mistaken about that?

I'm admittedly not an expert on thermodynamics, but I was operating on the concept that the better the thermal path from slug to heatsink, the lower the temperature in the LED when a steady state was reached. That's what I was going for with all this. In Beretta's original example, he noted the intensity decrease, then reglued the emitter to the heatsink and reported no intensity decrease despite an extended run. Sounds like he was measuring intensity maybe a few seconds after startup vs. another reading after maybe five minutes of continuous runtime.

[ QUOTE ]
milkyspit said:
Doug S, when I measured the 974 lux in my example, I got the reading within the first couple seconds, but there's no way I got it in 20 milliseconds! /ubbthreads/images/graemlins/icon15.gif



[/ QUOTE ]
If you are not that quick, then consistancy is the next best thing if making comparisons. For example, you might try for consistantly getting the reading at 2 seconds.

[ QUOTE ]
milkyspit said:
[
The interesting thing is that the lux readings declined for a few minutes, then seemed to stabilize at 830 lux, give or take a couple lux. I took that to mean the system had reached thermal equilibrium. Was I mistaken about that?



[/ QUOTE ]

Probably so. Most lights should take much longer to reach thermal equilibrium.

[ QUOTE ]
milkyspit said:
I'm admittedly not an expert on thermodynamics, but I was operating on the concept that the better the thermal path from slug to heatsink, the lower the temperature in the LED when a steady state was reached. That's what I was going for with all this.

[/ QUOTE ]
You are correct.

DougS, can you elaborate on any theory about the 20ms and three sec thing?

I can only imagine that the die might reach saturation at 20ms, and the slug could reach saturation at 3 sec, or something? Am I out in the farmland here?

[ QUOTE ]
AilSnail said:
DougS, can you elaborate on any theory about the 20ms and three sec thing?

I can only imagine that the die might reach saturation at 20ms, and the slug could reach saturation at 3 sec, or something? Am I out in the farmland here?

[/ QUOTE ]
Sigurd, while saturated is probably not quite the right word, your thinking is correct. The 20ms is associated with the mass of the die and the 3sec with the mass of the slug. When you apply heat to one end of a path that has several "thermal capacities" connected with joints with thermal resistance, the curve of the rate of temperature rise of the heated end will have points where the slope changes [decreases] more rapidly. Each of these points is associated with a thermal junction where heating must occur on the upstream side before significant heat will flow across the joint. I hope this makes sense. It is a bit of a complicated concept.

It makes a lot more sense now, thanks.