Heat dissipation - back to basics

[arhi]

hm .. when the sw engineer goes up too deep into hw some basics constantly pop up.. good thing is there's bunch of you I can bother every time I notice that...

Heat dissipation ... I have no idea what's this all about ... (ok, I know "why" the heat is dissipating) ... in the sense that, if it is getting hot - I put one with higher rating - if it is warm I add big heat sink + fan if needed... being surrounded with computer equipment most part of my life, having different CPU heat sinks w/o fan's a fact.

Now, to ask the darn question ....

P = V * I

so simple formula .. (I saw it in few threads before) ... or:

Ppsu = Pout*(1-EFF)/(EFF) //power supply heat, EFF == efficiency

Pq = Logic Supply Voltage x Logic Supply Current + Load supply Voltage x Load Supply Current // Device power consumption under quiescent, no load, conditions

Pcond = 2 x Irms^2 x Rds(on) // Conductive Power Dissipation

Psw = (Eon + Eoff) x F // Switching Power Dissipation

( **broken link removed** )

....

so, bunch of formula's ... all of them (most at least) get the result in W .. interesting ..

so "element" (ic for example) X is producing 6.5W of heat in realtime working conditions, that's important information. In data sheet I can see that "element" operating temperature is -5C - 80C for e.g.

WHAT NOW?

Ther's Heat Sink Temperature Calculator ... not sure I get it completely but seams to me that Thermal Resistance is "how good the element or better say package will tranfer heat to "case" or to "air" ... the Junction is not clear term for me here (yes, english is kinda 5th language for me, sorry) ... sounds like "point where package and heatsink/air meet" but I'm not sure I'm right. Lower the termal resistance, faster the heat will transfer (in order to reach max entropy / make both objects (in this case package and heatsink/air) have same temperature - cooling the hotter one, heating the colder one) ... the amount of newly generated heat is that 6.5W we calculated but how hot is that, with thermal resistance of 3°C/Watt (TO220) how big the heatsink need to be ?! If I do not put heatsink how hot will the TO220 get how fast?

 

[stevez]

I recall finding some excellent information on a heat sink manufacturer's website. I'd go to Mouser, Newark or similar electronics distributors catalogs to get the names of heat sink mfrs' then look on their websites for applications information. Sorry, I cannot recall the name of the mfr who produced the information but I am sure you'll find it.

 

[duffy]

This can only mean you never heard the one about the mathematician, the physicist and the engineer.

They were all asked to calculate the volume of a red rubber ball.

The mathematician measured the ball and calculated the volume. Because the ball was not a perfect sphere and the calipers tended to press into the ball, the mathematician never got the same answer twice.

The physicist submerged the ball in a graduated beaker full of water to measure the displacement but because the ball wanted to float, he had to keep pushing the ball down with a stick and would sometimes be measuring the volume of that as well. He had to average his results with a plus or minus tolerance.

The engineer looked up the model and serial number in the vendor's Red Rubber Ball Catalog and referenced the volume stated by the manufacturer.



So go look at heat sink specs from the manufacturer. You get thermal resistance numbers (both still air and moving) or power dissipation for a given temperature rise.

 

[arhi]

duffy

 

[JimB]

not sure I get it completely but seams to me that Thermal Resistance is "how good the element or better say package will tranfer heat to "case" or to "air"

Yes, heat transfer is a flow of energy, and the medium through which the energy flows offers resistance to that flow.

... the Junction is not clear term for me here (yes, english is kinda 5th language for me, sorry) ... sounds like "point where package and heatsink/air meet" but I'm not sure I'm right.

No, you are wrong.
The junction is the connection within the lump of silicon between the emitter and base of the transistor.

Lower the thermal resistance, faster the heat will transfer (in order to reach max entropy / make both objects (in this case package and heatsink/air) have same temperature - cooling the hotter one, heating the colder one) ...

Yes.

the amount of newly generated heat is that 6.5W we calculated but how hot is that, with thermal resistance of 3°C/Watt (TO220) how big the heatsink need to be ?! If I do not put heatsink how hot will the TO220 get how fast?

The figure of 3°C/Watt is probably from junction to case (transistor housing), to calculate how hot the transistor will get we need to know the thermal resistance from case to ambient. At the moment we dont know that.

Assume that we mount the transistor on a heatsink with a thermal resistance of 2°C/Watt, the thermal resistance from junction to ambient will be 3 + 2 = 5°C/Watt. (I have ignored the thermal resistance beween transistor and heatsink).
So, if our transistor is trying to dissipate 6.5watts, the temperature rise above ambient will be 6.5 x 5 = 32.5°C.
If the ambient temperature is 25°C, the junction temperature will be 25 + 32.5 = 57.5°C.
If the maximum junction temperature is 150°C, the transistor will be OK.

However, if we remove the heatsink, the case to ambient thermal resistance could be of the order of 60°C/Watt. Now the thermal resistance from junction to ambient will be 3 + 60 = 63°C/Watt.
With 6.5 watts dissipation the temperature rise above ambient will be 6.5 x 63 = 409.5°C.
If the ambient temperature is 25°C, the junction temperature will be 25 + 409.5 = 434.5°C.
At which point we let the magic smoke out of the transistor and it will not work any more.

I hope this helps.

JimB

 

[arhi]

oh yes, 10000% clear ... I got it now .. it is actually so simple and obvious (when someone else point that out to you, just like 5*1.2=6)... thanks