Thermal impact of different insulators

[smithy666]

I've been looking at a few different transistor insulation products, which list differing thermal conductivity and thermal resistance properties. For example,

Sil-Pad 400 TO-3 Thermal conductivity 0.9 W/m-k, thermal resistance 0.4 deg C/W
Sil-Pad 2000 TO-3 Thermal conductivity 3.5W/m-k, thermal resistance 0.2 deg C/W

I don't have any figures on Mica, but would there be any significant difference in case temperature of the device i using one over the other?

My initial though is there would only be a difference if the thermal resistance of the heatsink was less than 0.4 deg C/W. Any thoughts?

 

[djk]

Properly applied mica and grease is about 0.3°C/W

Does 0.1°C/W make a difference?

If the device is running at 50W then it will be 5°C hotter.

It starts to make a big difference between non-filled rubber at 0.8°C/W vs boron-filled at 0.2°C/W, 50W here is a 30°C difference. Some of the better insulators can cost as much as the device does (ouch!).

My last two designs had the heatsinks insulated from the chassis. That can save a lot of money in not buying fancy thermal washers, and in being able to use a smaller heatsink.

 

[smithy666]

But doesn't that still assume that your heatsink <0.3°C/W. If your heatsink was rated at 0.5°C/W, then having a 0.3°C/W or 0.5°C/W insulator shouldn't make any difference. Or am I missing something?

I assume you used different heatsinks for each output device if using a heatsink insulated from the chassis?

 

[suzyj]

I'd hope that your design isn't so marginal that the thermal resistance of a sil-pad matters. As long as the resistance of the pad is significantly lower than that of the transistor stuck to it, you should be okay.

I derived a formula for the required heatsink thermal resistance when I was designing my 100W power amp:

An externally hosted image should be here but it was not working when we last tested it.

Just plug your numbers in, and you can see the effect of different things. The greatest impact is seen by simply adding power transistors.

 

[dutch diy]

But doesn't that still assume that your heatsink <0.3°C/W. If your heatsink was rated at 0.5°C/W, then having a 0.3°C/W or 0.5°C/W insulator shouldn't make any difference. Or am I missing something?

I assume you used different heatsinks for each output device if using a heatsink insulated from the chassis?

Thermal Resistances tend to be set 'in series', where each of the components adds a temperature rise given the amount of power which needs to be dissipated.

As the ambient temp usually is fixed, adding additional thermal resistance increases the temperature of the junction which can mean you ran out of the Safe Operating Area, and thus lowers the amount of power which can be dissipated.

 

[djk]

"I assume you used different heatsinks for each output device if using a heatsink insulated from the chassis?"

I used an inexpensive stamping on each output device. The aggregate thermal performance was 0.167°C/W for a cost of about $5 worth of aluminum and no insulators.

I have some nice extruded tunnels that have about the same performance with the same fan. With the fancy insulators the tunnel cost is about $120 (for 12 TO-3 devices).

 

[Conrad Hoffman]

Attaching devices directly to the heat sink will increase the capacitance, which can matter in some circuits and with some types of devices, often MOSFETS. Back in "the day" you could get TO3 transistors in both steel and aluminum. Obviously the aluminum was preferred, as they could dissipate more power. Remember that for a proper design, ambient has to be the highest temperature the amp might ever have to perform in, say a hot day in Texas with no air conditioning, and ambient inside the chassis can easily be many tens of degrees higher than outside. I use Sil-Pads in my projects because I like the lack of mess and the convenience of changing devices when I blow it up on the test bench during development, but I also tend to use a very generous amount of heat sinking with the assumption that reliability goes up as temperature goes down.