Calculate how much heat an IC produces

I'm building a device containing a large number of ICs. Now, I'm 99% certain heat won't be an issue. (Considering the volts and amps involved, I doubt you could accurately detect the amount of heat involved.) But, just for argument's sake, how would you go about calculating this stuff?

My first thought was to look at the datasheet. But it doesn't seem to say anywhere "this chip will produce X units of heat in normal operation". (Presumably because there's too many different variables that affect it, so they can't easily come up with a definitive number.)

The only relevant thing I can see is a section on "thermal resistence". If I'm understanding this currently, this is a measure of how quickly any heat generated would be able to escape the casing. (Presumably depending on how hot inside vs how hot outside; it seems to be expressed in units of °C/W.)

Clearly thermal resistence is part of the equation. But without knowing how much heat per second the IC produces in the first place, I'm not sure where to start with this.

Here is rather simplified answer: with few exceptions (like light emitting or RF radiating) the electronic circuits convert all incoming power into heat. So, if you measure the power consumed by your assembled device you can get pretty good approximation of the heat to be dissipated.

Of course, if you want to calculate it in advance or predict temperature in various conditions you need all those things described in @SpehroPefhany's answer.

• According to electronics.stackexchange.com/questions/65200/… it seems you can just multiply the number of Watts a device uses by the °C/W rating on the datasheet to discover the temperature rise. (Presumably assuming everything is steady-state.) I was expecting the formulas to be far more complicated than that... – MathematicalOrchid Sep 10 '18 at 20:44
• As @SpehroPefhany mentioned in his answer, many devices use different power under different conditions. Changing copper plating from recommended in datasheet will change C/W rating and so on. In short, prediction of the temperature of various components can be quite more complicated that that. However the basic premise of this answer stays correct - if you measure the power consumed by your device and device does not radiate it in any other forms, then you can safely assume that all consumed power will be eventually converted to heat. – Maple Sep 10 '18 at 21:34
• Using the biggest numbers I can find on the datasheet, I get +2.6 °C. Seems like this can only be a problem if I stack so many ICs together that the air can't circulate properly. – MathematicalOrchid Sep 11 '18 at 8:17

Power is measured in watts (joules per second). You calculate the dissipation by understanding the IC and how much current it draws from what voltage(s) under the conditions you will be using it. Some parts draw current just sitting there that is significant, others use almost no power until you clock the part, then the power may be almost proportional to the clock frequency.

Once you know the power, you have to understand how it is transferred out to the environment. One component of that is how it gets from the die to the outside of the IC package. Eventually it has to find its way to the air, water, earth, be radiated into space, etc, or the temperature will continue to rise until the part fails.

Whole careers are built on thermal management, and of course books are written. Intuitively you can get an idea by looking at devices around you. Something the size of your fist that draws a few watts may be okay if the power is not too concentrated inside, for example. It also may fail in the blink of an eye if the power is concentrated in a small chip that is poorly coupled to the outside (too much thermal resistance). Thermal resistance implies a linear relationship between temperature rise and power, which may or may not be true. Radiation and convention are far from linear, but conduction is and mass transport (eg. water cooling) may be.

My first thought was to look at the datasheet. But it doesn't seem to say anywhere "this chip will produce X units of heat in normal operation".

Remember that

$$P = I \times V$$

That is, power (and, thus, thermal dissipation) is equal to current multiplied by voltage. The datasheet may not explicitly list thermal dissipation, but it should list a maximum current draw; multiply that by your working voltage and you have your number.

• Wait, are you saying heat produced exactly equals Watts of electricity consumed?? – MathematicalOrchid Sep 9 '18 at 8:26
• @MathematicalOrchid that's exactly what two other answers told you already. duskwuff must have done something right to make it sink in :) – Maple Sep 9 '18 at 21:23
• @MathematicalOrchid Yes, so long as the chip isn't delivering power to another part. (And if that were the case, the datasheet would probably mention thermal dissipation explicitly.) The power has to go somewhere, after all. – duskwuff -inactive- Sep 9 '18 at 21:30
• Of course, this only applies to circuits that do not expel power in some other forms, like electromagnetic radiation (light, RF etc). But it covers digital circuits nicely – Maple Sep 9 '18 at 21:31