EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: cellularmitosis on March 13, 2014, 01:52:28 am
-
What rule-of-thumb / design constraint do you guys use for thermal rise above ambient when picking out a heat sink?
I've been using a 50C rise as a target to shoot for (e.g. 75C surface temp), but really, that's just a wild guess, which is informed by:
- a vague notion that Bad Things(TM) happen to silicon around 125C
- the notion that silicon lasts longer when cooler, so lower temperatures are better.
- that anything significantly lower than 75C surface temp would require a pretty large heatsink (read: expensive).
The example I'm looking at right now is sizing a heatsink for an LM317, and I'm considering the Aavid 530002B02500G ( http://www.digikey.com/product-detail/en/530002B02500G/HS380-ND/1216384 (http://www.digikey.com/product-detail/en/530002B02500G/HS380-ND/1216384) )
Conveniently, Aavid has a calculator for their 63130 extrusion profile which allows you to type in the length of the heatsink and generate performance graphs: http://www.aavid.com/products/extrusion-heatsinks/63130 (http://www.aavid.com/products/extrusion-heatsinks/63130)
e.g., for the 2.5" version ( http://www.aavid.com/products/standard/530002b02500g (http://www.aavid.com/products/standard/530002b02500g) ):
(http://i.imgur.com/zf3qblR.png)
So, using that heatsink, if I targeted a 50C rise, I'd expect to be able to dissipate about 7 watts.
But again, I don't know that 50C is a good design rule. What's your temperature rise design rule?
-
Rule of thumb is under 60deg total heatsink temp, although anything under 80deg is usually still within spec for power devices.
So, assuming it's a hot day (say 30deg) ideally you'd would want to keep the temperature rise under 30degC and under 50deg at a pinch.
-
40C rise.
Works for heatsinks, transformer, chips...
But... always check with design constraints. Obviously there ain't no 40C temp rise you can tolerate in a down-hole 200C-ambient application. Or out of room temperature, you *could* tolerate a whole hell of a lot more than 40C...... but does that leave enough room for thermal resistances? How much margin does that leave on hot days? And is it touchable? Can it be touched, does it have to be? (Hint: if you don't have finger guards -- mesh or screen or whatever -- around it, it's probably not a good idea going over 60C absolute, if that!)
Disclaimer about "length of extrusion": it's not linear with length (assuming length is in the extruded dimension): you get diminishing returns for really long heatsinks. Put differently: it's better to have a huge number of fins, than a few ridiculously long fins.
Also, fans always help. Ridiculously so. But they wear out and clog and stick and stuff. So, use with caution.
Tim
-
I work to 20C rise if I have the luxury. 40C rise max if I don't have the luxury of space etc. Above that I'd be getting the hebbie-jeebies.
-
great responses, guys, thanks all around!
-
i would go something like 45-55 degrees total heatsink temperature... Docent cause burns if you touch it accidentally but gives that snappy feeling of something is running hot...
Docent really leave much play room if i need to design something that must stay operating at hot summer heat wave (40 - 43 C)...
-
It's better to keep your electronics a good 20C+ under their rated temperature. For stuff like electrolytic capacitors, as Dave says the electrolyte often dries out (and as Dave also often points out, capacitors can degrade from simply being near hot heatsinks). For ICs - particularly very complex ones on small process nodes like modern CPUs and GPUs - naturally degrade from electromigration (http://en.wikipedia.org/wiki/Electromigration) which is worsened by heat. A voltage regulator will be far less prone to the problem than the GK110 GPU chip on the Geforce Titan, but it's still not good to run them near max temp.
Keep in mind that in natural convection, you really need to consider the overall cooling system. If you had, for example, this in a tiny plastic box with no vents that insulates the heat inside very well, your heatsink is going to become far less efficient. This is still true with forced air, but easier with forced air since you have more control over the heat radiation (until the fan dies at least, but that's what over-temperature protection is for). Not saying you necessarily need forced air for this, but just make sure to consider the entire cooling system. The manufacturer won't be testing their heatsink dipped in 1/8" of plastic, after all. That wouldn't make their numbers look very good...
Like T3 said, more fins are better. For more power dissipation you could get something like a 1u server CPU heatsink on Ebay for around $10. You shouldn't need that much here and you'd probably need to drill and tap it (and you wouldn't want to rely on that if it were production work) but I love using them for one-off higher power stuff like in class A/B audio amps or high-power linear PSUs etc.. I rarely see better deals on heatsinks for those purposes. And I hate how a lot of those extruded aluminum heatsink manufacturers usually barely put any fins on their heatsinks. One thing to keep in mind if you do use a PC/server heatsink, if the fins are closed off at the top (some server heatsinks in particular are), that means they were only designed for use with force air. If the fins are open at the top, they can be used with natural convection within their thermal capacity (problem though: good luck figuring out what that is - but that's the price you pay for a good heatsink at the prices you can buy those for).
Oh, and I agree with 30C max rise temp.
Regards,
Dan