Thermal Compound - Results Posted

But the heat will be more because when badly cooled it will produce more heat and less light.

Only for a short while, until delta T is large enough to ‘force’ heat dissipation to the body. (it’s got nowhere else to go)

So in this case delta T is comparable with voltage. make it high enough and the ‘heat current’ to the body will be high too.

Well, first off, where did NeutralFan take the measurements? I don’t see where in the OP he states the location. If it is the outside body, close to where the shelf contact is, then with better heat transfer you would think the outside body would take on more heat, keeping the MCPCB cooler in the process. I’m guessing it will be slightly hotter but I doubt it will be noticeable.

Cool project though, you sound like me. I always want to do things as efficiently and effectively as possible, the first time.

Makes sense, either way I still think it’s not enough heat to see a measurable difference, unless hes driving that LED at 6A+ or something like that.

@ rdoc613:

The LED produces heat and light.
The heat has nowhere else to go then to the body.
So, the efficiency of the LED basically determines the heat going to the body.
The efficiency suffers when there is a poor thermal path (high delta T) because the LED will then operate at higher temperatures.
At a certain point (saturation) the delta T is high enough to stop the LED from getting even hotter and thus all of the generated heat will be transferred to the body.

I don’t know how big the difference in efficiency between a properly cooled LED and a badly cooled LED can be.

When I put the LED MCPCB back on the shelf over the thermal compound, I first slowly pressed down evenly and gently. Eventually I pressed hard and slightly rotated the MCPCB clockwise and counter-clockwise by approximately 5-10 degrees.

Well unless you push an LED pretty far, it seems like the difference isn’t very significant:

I took the temperature measurements using an infrared thermometer. I pointed it at the body of the flashlight and found the hottest spot. The flashlight was tail standing in turbo mode. The hottest spot was near the top of the head, just below the bezel.

You should use a thermal probe attached directly to the copper MCPCB to get a useful reading without all of the thermal variables discussed above.

I agree with this and that was my rationale for the experiment. Assuming a better heat transfer from the LED MCPCB to the shelf using the dot method versus the spread method, I would expect the temperature to increase faster near where the shelf is located.

The driver was in direct drive turbo mode and the same battery was fully charged before each set of measurements.

In those tests the thermal path seems to turn out well because the LED board is clamped down on the cooling body.
‘Bad cooling’ is not tested there.
But you can see the differences better in tests between DTP and non DTP boards.

In those tests the thermal path seems to turn out well because the LED board is clamped down on the cooling body.
‘Bad cooling’ is not tested there.
But you can see the differences better in tests between DTP and non DTP boards.
[/quote]

Yeah but it’s still bad cooling, you do that to a computer and it will instantly overheat.
It’s just that LEDs don’t make a lot of heat unless you push them hard.

I was going to point to the powder-coated S2+ thread, but… :smiley:

I was a little surprised by the results. It makes sense that the middle dot method is preferred for applying thermal compound. After all, the spread method can lead to air gaps which is not good. I only measured the temperatures up to 4 minutes since I was genuinely concerned that the LED would get too hot if I went longer. I did not want to damage the LED due to not applying the thermal compound correctly.

Was the time I spent to change out the thermal compound worth it? In hindsight I would say no. But since I didn’t know, applying the thermal compound the correct way gave me piece of mind. That in itself is worth the hour or more I spent on this.

Perhaps with a more powerful flashlight the results would have been different, and probably even more critical on a computer CPU, but it appears that the 2 methods I used to apply thermal compound was equally effective.

Here are the results again from the surface spread method:

  • Start – room/flashlight temperature at 67 degrees Fahrenheit
  • 1 minute - 84F
  • 2 minutes - 95F
  • 3 minutes - 104F
  • 4 minutes – 113F

Here are the results after I used the middle dot method, look familiar?

  • Start – room/flashlight temperature at 67 degrees Fahrenheit
  • 1 minute - 84F
  • 2 minutes - 95F
  • 3 minutes - 104F
  • 4 minutes – 113F

Thanks to all that gave their predictions and knowledge about thermal compounds. I took a thermal dynamics class back in college, but that was many years ago and I don’t recall any mention of thermal compounds. Certainly many of you have a lot of knowledge about this and I continue to learn from BLF.

some of the best negative results I’ve seen lately
acutely portended by most
so as long as the surfaces are flat, then it’s superfluous to add compound?

The only reliable way to measure a miniscule thermal difference vs LED junction temp is to monitor LED voltage by at least 3 decimal digits. This is one of the simplest method used by advanced LED driver to regulate temp or to maintain true constant brightness/power.

With low power stuff like most LEDs yeah.
If you’re like me with an XHP70 at 12A making 100W of heat, then no you certainly need thermal paste to keep the LED from melting.
There’s no harm in adding thermal paste though just to be sure.

At the very least it will indicate how well the two surfaces are mated or whether there’s a lip on a wire hole, a screw hole, or the center is high. It’s all to easy to declare something proven yay or nay, it’s another to go deep enough to understand what’s really going on and be able to explain it. It may be the case that the host temp is almost identical but a variation in thermal transfer could mean a still small but more readily detectable difference in mcpcb temp simply because the energy not transferred is left focused in the much smaller mcpcb rather than spread throughout the host. It’s the same effect we see in the difference between DTP and non DTP save that it’s much more crucial at the heat pad solder point because the heat is even more focused there in the die. DTP copper is much more forgiving but it doesn’t mean you can take the next transfer completely for granted. At some point a poor thermal transfer will show up as diminished output, it may take higher current but that’s why we use DTP in the first place and why some go to the extreme of soldering mcpcb’s to heat sinks.

My prediction was based on my experience with thermal compound used mounting heatsinks to cpus. I’ve tried both spreading a thin layer using a plastic card (credit card, id, driver’s license, etc) and the ‘grain of rice’ method and found that there was little difference in the effectiveness even with heavily overclocked multi-core processors. This goes back to the days of good old 400MHz celerons oc’d over 1GHz on a dual processor board to more recent and more modestly oc’d core processors like a core 2 duo rated for 2.4GHz at 3.6Ghz. Using the same compound and heatsink the cpu temp was within a few tenths of a degree under heavy load (running a boinc app on all cores).

It’s likely that in the case of cpu heat sink and thermal pad assembly the manufacturers are a bit more diligent about surface prep than the run of the mill light manufacturer. I daresay that a similarly poor standard of prep would result in more than a few tenths difference, especially if no thermal compound were used or so much that a bad fit was not apparent(both of which are common enough with lights). Granted, the majority of flashlights don’t push the limits nearly as far but it’s just common sense to increase the scrutiny as you increase the power.