Help with heatsink calculations

[Cicero]

So I'm looking into heatsinks, and I think I'm a little rusty.

I'm thinking of using a 2N3055 (TO-3 package) to sink some decent current (0-1A) over a decent voltage range (0-60V). Or thats the idea anyway.
https://www.onsemi.com/pub/Collateral/2N3055-D.PDF

From the datasheet:

• Total max power dissipation is 115W with a derating of 0.657W/degC from 25deg onwards
• Max junction operating temp is 200degC
• Thermal resistance (junc to case) = 1.52degC/W

I'd just like for you guys to double check if I'm on the right track.

I dont want the junction temp to reach 80% of the max, so I set Tj = 160degC
Ambient will be around 40deg (and I plan to use a small fan to try keep it there or there abouts), Ta = 40degC

If I want to run it at 60W, which is 60V @ 1A, then I worked out I'd need a sink with a thermal resistance of 0.48degC/W!!!!! Which is ridiculous, or is it understandable? (this is with negligible case to sink resistance, which we know isn't negligible).

I can find a pretty small but decent looking sink with a 6degC/W rating, but that allows the transistor to only run at around 15-16W if I'm not mistaken.

Now am I doing this stuff correctly? To come up with their 115W they state, what do they do...cool the thing in liquid nitrogen or something?

 

[ronsimpson]

Now am I doing this stuff correctly? To come up with their 115W they state, what do they do...cool the thing in liquid nitrogen or something?

YES.

There is a theatrical infinite heat sink. Some of the numbers are derived from this infinite heat sink.

 

[Diver300]

So I'm looking into heatsinks, and I think I'm a little rusty.

• Thermal resistance (junc to case) = 1.52degC/W

I dont want the junction temp to reach 80% of the max, so I set Tj = 160degC
If I want to run it at 60W

60 W * 1.52 °C/W = 91.2 °C rise, so you heat sink has to be kept below 68.8 °C

You might be much better off to use more semiconductors. If you used say 3 off 2N3055 transistors, with suitable current balancing, you would then be looking at 20 W * 1.52 °C/W = 30.4 °C rise, so your heat sink has to be kept below 129.6 °C. The heat sink has to dissipate the 60 W, but it can then be a much more reasonable (129.6 - 40) / 60 = 1.5 °C/W

It will be too hot to touch at 60 W

Something like:- https://uk.farnell.com/abl-heatsinks/520ab1000mb-t03/heat-sink-to-3-1-4-c-w/dp/253728

 

[Cicero]

60 W * 1.52 °C/W = 91.2 °C rise, so you heat sink has to be kept below 68.8 °C

You might be much better off to use more semiconductors. If you used say 3 off 2N3055 transistors, with suitable current balancing, you would then be looking at 20 W * 1.52 °C/W = 30.4 °C rise, so your heat sink has to be kept below 129.6 °C. The heat sink has to dissipate the 60 W, but it can then be a much more reasonable (129.6 - 40) / 60 = 1.5 °C/W

It will be too hot to touch at 60 W

Something like:- https://uk.farnell.com/abl-heatsinks/520ab1000mb-t03/heat-sink-to-3-1-4-c-w/dp/253728

Haha, yeah, I was gonna do that kind of thing with 3 transistors to get 3A actually! Murphy's law. But if it seems like I'd need 12 or more for 3A, so I think I'll rather drop some good power over a few power resistors or something and work some compromise out.

Pity though....but at least my calcs are right it seems, so I've got that going for me. I guess power is power, its gotta go somewhere.

Another option I'm looking into is pulling a big heatsink off an older piece of equipment. My mate has a whole bunch apparently that he's hoarded, so I'll probably look into that route. That sink you pointed out is decent for sure, but its too pricey for me, especially if I need to buy multiple units.

Thanks for the help guys.

 

[crutschow]

Using a fan cooled heat-sink can significantly reduce the needed size of the sink. Look for those used to cool PC CPUs.

 

[Cicero]

Using a fan cooled heat-sink can significantly reduce the needed size of the sink. Look for those used to cool PC CPUs.

I thought about this a bit, but maybe I'm missing something. How much help do they actually give, is there a way to quantify it? Or some theoretical way to factor them into the maths?

I was planning on using a small one to keep the ambient temperature on the heatsink stable, but I thought it wouldn't reduce the theoretical heatsink needed, only allow the calculations to work out because you're keeping the ambient temperature (Ta) stable(ish) and not allowing heated air to build up? But does it actually reduce the thermal resistance because its forcing ambient further into the sink?

 

[audioguru]

A heatsink heats up. A fan literally blows away some of the heat.

 

[tvtech]

I love you AG

You need to know that.

tvtech

 

[crutschow]

Moving air greatly increase the heat dissipation of a heat sink as compared to normal convection. If you look at the thermal resistance for those heat sinks with fans you will see how much lower they are.
The thermal resistance depends upon the velocity of the moving air so it's somewhat complex to quantify.

Here's a paper that shows the reduction in thermal impedance due to various air flow rates.

 

[Cicero]

Moving air greatly increase the heat dissipation of a heat sink as compared to normal convection. If you look at the thermal resistance for those heat sinks with fans you will see how much lower they are.
The thermal resistance depends upon the velocity of the moving air so it's somewhat complex to quantify.

Here's a paper that shows the reduction in thermal impedance due to various air flow rates.

Champion, cheers mate.

For some reason I had the idea that a heatsinks rating was spec'd at a value when the direct ambient was kept absolutely constant by some means, be it 'artificial' even, and not at normal convection. Stupid when I think about it now.

Thanks

 

[ronsimpson]

Here is a heat sink. At 1000 feet/minute of air flow has 0.4C/watt resistance.
At 100 feet/minute the resistance is 1.3C/watt.
I wanted to find a graph the goes to zero air flow but I don't have time. I remember the curve heads up steep as you approach zero air flow.

 

[gary350]

The hardest thing to know is BTU rate the transistors heats up and how much cooling you get from a fan.

Here is what you do. Weigh the transistor on a gram scale. Get a piece of metal like aluminum 20 times heaver attach it to the transistor and turn it on. Time the rate at which the metal heats up. Keep notes and plot a graph for every 30 second up to about 150 degrees. Now you have data to work with you can calculate how much larger or smaller the aluminum needs to be to keep the temperature from going over 150 degrees.

It is very interesting aluminum and copper both have a high heat transfer rate but very low heat radiation rate. Cast iron has the highest heat radiation rate but very poor heat transfer rate.

A small fan will increase efficiency several times. Do an experiment with a fan to find out how fast a fan will cool a certain mass size piece of metal then do the math to find what size metal or fan you really need.

 

[Flyback]

you are right in your top post with 0.48degc/w
You need junction-to-ambient thermal resistance of 2 degC/w.
But really 160degc is too much for a junction of transistor.
You've obviously heard about good thermal coupling using eg heatsink paste etc.
Stand the heatsink up so you get better convection.
By a heatsink with a hose connection on it for cooling liquid.

 

[carbonzit]

It is very interesting aluminum and copper both have a high heat transfer rate but very low heat radiation rate. Cast iron has the highest heat radiation rate but very poor heat transfer rate.

Hmm, interesting, didn't know that. But doesn't radiation rate depend on the exterior color of the radiator? Seems to me I've read about "black body" radiation which is the (ideal) highest possible, so what if one were to black-anodize an AL heatsink?

 

[Flyback]

that's interesting about iron because usually the heavier the metal, the easier it can conduct heat?

 

[Cicero]

you are right in your top post with 0.48degc/w
You need junction-to-ambient thermal resistance of 2 degC/w.
But really 160degc is too much for a junction of transistor.
You've obviously heard about good thermal coupling using eg heatsink paste etc.
Stand the heatsink up so you get better convection.
By a heatsink with a hose connection on it for cooling liquid.

Yeah thanks, was just amazed actually at the low number actually. Most of the stuff I do is very low power (max 2W devices), so I dont do much heatsink design. But when I do, its normally super small and just to check it'll operate at the temperature limits. Never once used a fan as well, no moving parts allowed.

But this is just for a bit of bench testing equipment, sinking some good current, it doesn't justify any sort of serious engineering like liquid cooling etc.

 

[crutschow]

that's interesting about iron because usually the heavier the metal, the easier it can conduct heat?

No. Heat conductivity is generally (but no always) related to electrical conductivity in metals so iron is much worse than aluminum or copper for conducting heat as well as electricity. Look at this table, for example.
One notable exception to this I noticed is tungsten.