Figuring power dissipation increase w/ heatsink

[applefat]

Hello!

I'm trying to make sense of what is probably simple. Imagine I have an IC package that is rated to dissipate at most 1 Watt. The thermal resistance from case to ambient is not known (could maybe be generalized).

I have a heatsink that has a thermal resistance of 20°C/W. Can I estimate what the new power dissipation of this package is?

I've looked at the wiki page on thermal resistance and it compares a source of power to a current source, a temperature differential to a voltage differential, and thermal resistance to a resistor. That analogy confuses me in this scenario however, because wouldn't adding a heat sink put another thermal resistance in series with the junction-case resistance, thus increasing the overall thermal resistance of the system?!?

 

[berkeman]

You are correct. Thermal resistances add in series. Wikipedia.org is helpful for some things, but can be confusing (and wrong) sometimes.

You would add Theta(junction-to-case) plus Theta(case-to-ambient via heatsink) to get your total Theta(junction-to-ambient), in general.

 

[russ_watters]

Note that the power dissipated doesn't change, just the temperature gradient.

 

[berkeman]

Note that the power dissipated doesn't change, just the temperature gradient.

Interesting point. It got me thinking, though... It's not exactly true for heatsinks of semiconductor devices.

Quiz Question for the Original Poster (OP) -- Why not? Does it matter if the semiconductor device is a bipolar transistor versus a MOSFET? Does the circuit topology matter?

 

[russ_watters]

Interesting point. It got me thinking, though... It's not exactly true for heatsinks of semiconductor devices.

Not sure what you mean by that. Is there some variation in how they work if there are large temperature variations? Or is it different for different kinds of semiconductors? Typically, if you take a PC processor (for example) and put a bigger heat sink on it, you get a lower operating temp, but its heat dissipation remains the same.

 

[berkeman]

If the sinked transistor is supplying the current for the load, and the load stays constant while the efficiency of the supplying transistor varies with temperature, then the power dissipated by the supply plus load vaies with the sinking thermal resistance. That efficiency has opposite temperature coefficients for bipolar versus MOSFET transistors.

If the "load" is everything, then your comment is still correct, I think. But if the load is being supplied by a transistor on a heat sink, then the total power consumed depends on the heat sink and type of transistor (slightly).

Different Quiz Question for anybody -- how can you tell qualitatively when a heat sink is sized correctly?

 

[Topher925]

If the sinked transistor is supplying the current for the load, and the load stays constant while the efficiency of the supplying transistor varies with temperature, then the power dissipated by the supply plus load vaies with the sinking thermal resistance. That efficiency has opposite temperature coefficients for bipolar versus MOSFET transistors.

So a cooler BJT will output less heat at cooler temperatures due to an increase of efficiency while the opposite is true for MOSFETs?

Different Quiz Question for anybody -- how can you tell qualitatively when a heat sink is sized correctly?

There are only small to no variations in temperature with increased current loading?

 

[russ_watters]

I did not know that their efficiency varied with temperature - how much/over what temparature range?

 

[berkeman]

I did not know that their efficiency varied with temperature - how much/over what temparature range?

Vbe has the same temperature coefficient as a diode, so about -2mV/degreeC. As the junction temperature increases, the bipolar transistor becomes more efficient, so it dissipates less heat for the same power transferred. MOSFETs look mostly resistive, so their Vds voltage drop increases with temperature, and their efficiency drops at high temperatures.

Pretty small effects, but they do enter into sizing and margin calculations for circuit power calculations.