Ian's Emperical Heatsink Measurements

Or
Why the biggest heatsink in the world can't help you!

While experimenting with MOSFET RF power amps I came up against the age old problem of how much heatsink to I need to get rid of all this heat? The obvious answer is to grab the biggest bit of extrusion you can find from the junk box, and that'll be fine. Unfortunately, life, and heatsinking, isn't that simple.

If you check the data sheets for you favorite power transistor (ie IRF540) you can read some specs such as :

Rating ValueUnit
Thermal Resistance Junction to Case 1 °C/Watt
Total Dissipation at 25° 125 Watts
Derate aboves 25°C 1 Watt/°C

Working back from this, the maximum junction temperature can be 150°C
I.E. There will be a thermal differential between the junction and the case equal to the thermal resistance * the number of watts dissipated. In our case this is 125 Watts / (1°C/Watt) which equals 125°C at full ratings. The case must be at 25°C for these ratings, so the junction is at 25°C + 125°C which equals 150°C

Now assuming we want to derate this a bit, and keep the junction under 100° C and we have perfect thermal transfer to a perfect heatsink and an ambient temperature of 25°C, we can have the device dissipate 75 Watts (ie with a 1°C/watt temperature rise due to thremal resistance WITHIN the TO220 case , 75 Watts will cause an increase in temperature of 75°C above the ambient 25 °C, giving a temperature of 100° C. ). Now this is a bit unrealistic, as we know that heatsinks are reated in °C/Watt as well. If we add a 1°C/Watt heatsink into the equation above we find that we can now only dissipate 37.5 Watts ( 37.5 Watts * (1 °C/watt + 1 °C/watt) = 75 degrees above ambient. Thats not too bad, but what about the transistor case to heatsink? We used a mica washer and a heap of goop there, so that should be OK - shouldn't it ? Well, no. Its probably the weakest point in the whole chain. I couldn't find much information on this, so in true Muppet-Labs style, I built an experiment to try to measure it.

Experiment 1

The test device consisted of a 2SD288 TO-220 transistor pulled out of the junk-box. It appeared to have a fault in that if the C-E voltage rose above 11.24 Volts it started to conduct, not unlike a zener diode. (Thinking about this more I could have had the pin-out wrong and was reverse-breaking down the B-E juntion. It doesn't really matter though. The end result was it was a TO220 device disspating a number known number of of Watts.) Connecting this to my bench supply set to limit at 2.0A gave me a nice constant 22.48 Watt dissipating TO-220 package.

picture of TO220 mounting
I clamped the TO-220 to a largeish heatsink with a large spring clip. The pressure applied was in the vicinity of 2 Kg (~20 N). Between the jaw of the spring clip and the tab of the TO-220 was a LM335 temperature sensor chip. As this was in intimate contact with the copper tab of the TO220, it should give a reasonable good measure of the temperature of the heatsink tab of the transistor.

The LM335 has a voltage drop across it equal to 10mV multiplied by the temperature in °Kelvin. For the purposes of this experiment, I set the delta setting of my multimeter to 3.00V , corresponding to a temperature of 27 degrees. For each test run, I waited untill the temperature dropped (from the prevous run) to 0 on the meter and then started the test. This should give a fairly constant base-line. I ran 22.48 Watts into the TO220 for 60 seconds for each test and measured the temperature rise of the TO220. This will not give an accurate result as the temperature of the heatink will rise somewhat as well, but it at least its an indication. A better test would be to measure the heatsink temperature as well and run each test until the temperature stabilised.

Anyway, here are the results of a number of runs:

Heatsink TreatmentTemp rise in °C after 60 secscalc °C/watt
Silicon Pad (C) 67.03.0
Silicon Pad (C) 75.03.3
Silicon Pad (C) 76.03.4
Silicon Pad (C) 75.23.3
Mica 0.002" no grease76.03.4
Mica 0.002" no grease67.03.0
Mica 0.002" no grease67.03.0
Mica 0.002" no grease66.03.0
Mica 0.002" with grease (A)54.52.4
Mica 0.002" with grease (A)54.42.4
Mica 0.002" with grease (A)53.82.4
bare metal to Metal 40.41.8
bare metal to Metal 43.52.0
bare metal to Metal 40.11.8
Mica 0.002" with grease (B)29.51.3
Mica 0.002" with grease (B)30.31.3
Mica 0.002" with grease (B)30.41.3
Mica 0.002" with grease (B)30.31.3
Bare Metal with grease (A)25.91.1
Bare Metal with grease (A)24.81.1
Bare Metal with grease (A)24.21.1
Bare Metal with grease (A)23.91.1
Bare Metal with grease (B)12.30.6
Bare Metal with grease (B)12.80.6
Bare Metal with grease (B)13.40.6

Note (A): with a VERY thin smear of grease, only just enough to stick washer to heatsink.
Note (B): more grease, but not enough to ooze out under the transistor. Can still feel friction between transistor and washer
Note (C): the Silicon Pad was the type sold by Jaycar as HP-1155

With the Bare metal to Metal with grease, the temperature appeared to have almost stabilised after 60 seconds. With the others the temperature was still rising somewhat.

The reason there are two lots of figures with the heatsink compound is that I wanted to see how much was really needed. - Now I know that a real thin smear isn't quite enough. The (B) runs show a considerably better thermal transfer.
Another observation is that the clamping force (around 2 Kg) is rather light. The Silicon Pad would probably be better with more pressure applied as the pad is fairly resiliant. However given the relatively poor results I think I'll be sticking to mica or no insulator for any serious dissipation in the future.

The calculated °C/Watt is going to be significantly lower than the real value, as the temperature of the seup never reached the point of stabilising, but its an indication at least. Also the heatsink must have heated up somewhat in the 60 seconds, so part of the resistance is due to that as well.

However this was a quite instructive little experiment, and I will now go back to the mica and heatsink goop instead of the silicon pads, except for low dissipation / low mess requirements.

Conclusion

So Ian's simple heatsink advice is....
Use (thin) mica washers and goop.
Don't expect to be able to dissipate more than around 50 Watts from ANY TO220 device if you want the junction to stay under 100°C.


Ian - VK3KRI
Revision Information $Id: heatsink.html,v 1.3 2003/05/17 11:45:58 ic Exp $