CPU Cooler for Electronic Load

[Supercharged]

I'm currently working on a DC electronic load as a personal project (not intended for mass productiion), and I asked myself if it's possible to use a CPU cooler do cool the load transistors.
The rated power is high enough by far. Also the temperature a CPU can reach before it turns off is lower than the rated temperature of the MOSFETs i'm using.

[Fraser]

CPU cooling has become an art form. When I was in my teens a heatsink was normally a simple lump of black anodised aluminium with a row of fins. In this age of powerful PC processors there has been a need for efficient, yet compact heat sinks. Some are fan assisted, others passive.

If someone has gone to the trouble of designing an efficient compact heatsink for mass production (lower price) why not repurpose it for other cooling applications?

I applaud your thinking and see no issues with drilling the heatsink for attachment of the MOSFET's. Use some decent thermal transfer paste as well.

There is of course one fly in the ointment....... Some PC heatsinks are not as good as they claim. Use some common sense when selecting one. As a simplistic rule,a good heatsink needs to either be large and cooled by passive air flow, or smaller and air movement across it maintained by a fan (or a combination of size and cooling techniques) An efficient small heatsink needs a combination of large surface area and air moving over it to draw away the thermal energy. If you find a heatsink with plenty of surface area and fan forced air across such, it should shift heat well.

As I said, some PC heatsinks are a work of art these days....... Many thin fins around an aluminium core and a decent fan pushing air over the fins to shift the thermal energy. The GPU heatsinks can also be very neat and efficient.

Go for it.... Repurpose a PC heatsink

Fraser

[Vgkid]

Thanks for creating this thread. I'm building a fixed psu, and the pass transistors will need to dump quite a few watts of power, under heavy load conditions.  Since these cpu coolers are cheap enough used, so why not.  Though I do have a computer repair place a few miles away, I could always ask.

[Jeroen3]

Yes it's definitely possible. I've used old socket 775 server 3-heatpipe tower heatsinks as dummy load. You can get the pcb specs from intel. Probably from AMD too.
The only thing is that a CPU heatsink is designed to be mounted on the top of a package. And it's rather custom made for that purpose.
Most transistors you'd want to buy for this are either TO-220 or TO-247. And Infineon Optimos package is expensive and hard to do at home.

I've come up with the heatsink on the bottom. The TO-247 mosfet is sandwiched between the heatsink with aluminum oxide isolator, and the pcb. The pins are bend upward (with the metal laying down) to pcb pads. I've not made an enclosure, as this would be a challenge to keep airflow enough. Since CPU heatsinks only do a 10watts passive before becoming to hot to touch.
Obviously, the pcb part pushing on the fet will get hot. Therefore, no components are put there. Except two solder pads for a NTC glued in the TO-247 hole. Which is unused since you can't drill in heatpipes.

I've seen someone else do it before around here... Got it: https://www.eevblog.com/forum/projects/heatsink-for-dc-electronic-load/msg702710/#msg702710
For low power DC loads you can use active north/south-bridge heatsinks.

[Pjotr]

Keep in mind when using heat pipe equipped coolers, that heat pipes have a limited power transfer. For 6 mm pipes it ranges from 15 Watts at 20 degC to 30 Watts at 70 degC each pipe. The low cost Chinese coolers also use low cost grooved heatpipes that only work when the hotter part is below the cooler part.

[Vgkid]

For my useage, depending on caseI chose. I can mount the transistors on a L bracket(with insulator), and bolt the heatsink to that.

[rdl]

The heat sink/fan combos that come stock with Intel processors are not bad at all. They're also pretty cheap on ebay because of the kids that think they need to replace them with something more bling bling for their hot rod LED lit window in the side home made gaming machine.

[f5r5e5d]

"I can mount the transistors on a L bracket(with insulator), and bolt the heatsink to that."

quite a waste then, L bracket thermal path will be limit the usefulness of any attached "high performance" heat sink

try a solid rectangular bar of copper for your "bracket", then you might see some effect

[Vgkid]

^^^ Thanks, for that suggestion.

[Pjotr]

Well, do the math and calculate the total thermal resistance properly. It's no rocket science...

[Supercharged]

I've come up with the heatsink on the bottom. The TO-247 mosfet is sandwiched between the heatsink with aluminum oxide isolator, and the pcb. The pins are bend upward (with the metal laying down) to pcb pads. I've not made an enclosure, as this would be a challenge to keep airflow enough. Since CPU heatsinks only do a 10watts passive before becoming to hot to touch.
Obviously, the pcb part pushing on the fet will get hot. Therefore, no components are put there. Except two solder pads for a NTC glued in the TO-247 hole. Which is unused since you can't drill in heatpipes.
/#msg702710[/url]
For low power DC loads you can use active north/south-bridge heatsinks.

Yes i thought of something similar.

Keep in mind when using heat pipe equipped coolers, that heat pipes have a limited power transfer. For 6 mm pipes it ranges from 15 Watts at 20 degC to 30 Watts at 70 degC each pipe. The low cost Chinese coolers also use low cost grooved heatpipes that only work when the hotter part is below the cooler part.

I will look into that

thanks to everyone.

[Jeroen3]

Heatpipes do indeed seem to have a cut-off point. Say you can put in steps of 5 watts, then when you are around 50 watts while the heatsink is still touchable. You go up to 55 watts and the thermal protections kicks in. Probably because the heatpipes lose effectiveness. They are designed to work in a temperature range and require a big enough temperature difference within that range. I think that when the fin temperature goes up, the difference also shifts up, but clips at the maximum temperature of the heatpipe.
Which makes sense because they have a fluid inside. When the gas cannot condense anymore, the fluid also stops evaporating. Halting the entire process of heat transportation.

Heatpipes do not seem to care about orientation because they use capillary effect. (why water creeps upwards in toilet paper)

[Pjotr]

:LOL: Heat pipes are no toilets! There are 3 major kinds used in this area: With grooved walls, with wicked walls (with a mesh of copper wire) and with sintered walls. Only the ones with sintered walls (the more expensive) has enough capillary working to get the water back against gravity. Also bends and flattening reduces performance. Luckily the better CPU coolers use sintered heat pipes.

To get a rough indication for sintered ones without bends: http://www.1-act.com/resources/heat-pipe-calculator/

[Supercharged]

Only the ones with sintered walls (the more expensive) has enough capillary working to get the water back against gravity. Also bends and flattening reduces performance. Luckily the better CPU coolers use sintered heat pipes.
To get a rough indication: http://www.1-act.com/resources/heat-pipe-calculator/

I plan to lay the cooler flat so the heatpipes don't need to move the water against gravity (also it is more space efficient), also i'm not going to use the cheapest cooler so i would probably be fine

[Pjotr]

Usually the manufacturer states the maximum of power it can handle. But be aware that most do at 70 degC. At lower temperatures that power is less.

[Supercharged]

Usually the manufacturer states the maximum of power it can handle. But be aware that most do at 70 degC. At lower temperatures that power is less.

Correct me if i'm wrong but if it can't handle the power it will just heat up to 70°C and then it will be fine. (the transistors can handle 70°C)

[macboy]

Usually the manufacturer states the maximum of power it can handle. But be aware that most do at 70 degC. At lower temperatures that power is less.

Correct me if i'm wrong but if it can't handle the power it will just heat up to 70°C and then it will be fine. (the transistors can handle 70°C)

Exactly. But consider that if the transistor case is 70, what is the junction temperature? 70 + R_JC * Power
Also consider that unless you soldered the transistor case to the heatpipe, it won't be at 70 even if the heatpipe is. Add R_CS to R_JCabove, to account for your mounting method/insulators/etc.

[timb]

FiveFish Audio sells a DIY 240W Load kit based around two Intel CPU heatsink / fans. http://www.fivefishstudios.com/diy/eload/

They supply the board, you supply the parts. The CPU coolers can be had cheaply on eBay. Looks to be a fairly decent solution, though I haven't tried their particular kit. (I'm working on a 500W+ load based on off the shelf water cooling components.)

[sleemanj]

I like the old pentium 4 heatsinks

https://www.google.co.nz/m?q=pentium+4+heatsink&client=ms-opera-mobile&channel=new&espv=1#imgrc=o96hqePRHyr_jM%3A

Hunks of rectangular finned aluminium with an attached fan. Nothing more complicated.  Easy to drill holes where you need to screw stuff to it, or screw it to stuff.  Stick an ntc thermistor on near/on the tab(s) of the power device for fan control and over-temp shut down. Fan on at about 50 and shut down about 70 (c of course), with some hysterisis in the other direction.

[Pjotr]

Usually the manufacturer states the maximum of power it can handle. But be aware that most do at 70 degC. At lower temperatures that power is less.

Correct me if i'm wrong but if it can't handle the power it will just heat up to 70°C and then it will be fine. (the transistors can handle 70°C)

When the hot end is at 70 degC and the cool end is at 30 degC, the vapour pressure is too low to transport the max. amount of watts stated at 70 degC. The hot end will dry out and the pipe stops working (there are more mechanisms). So overrate the CPU-cooler at least 2 - 3 times to be safe. With a PC the cooler gradually heats up from the start to a more efficient temperature of 40 - 50 degC. It never starts up with all cores at full load constantly. With a dummy load that is less under control.

My own 500W dummy load is simply a common Fisher 100mm x 200 mm x 40 mm heat sink with 2 Papst AC fans screwed on. Such a thing is a testing device, so make it not too fancy: Form follows function In the past I needed arrays of dummy loads for battery testing and used low cost CPU-coolers for that with great success. I did overrate them 3 times because they had to run from 10 degC on.  Also if you use power fets as dissipating elements, overrate them at least 2 times to avoid hot spots on the die.

[SeanB]

Plain Intel non heatpipe units, or the cheaper aftermarket ones without heatpiping. Then you can drill and tap the central core for 2 TO247 packages side by side using a M3 tap. That will handle up to 65W for the high power ones with ease.

http://www.intel.com/content/www/us/en/processors/pentium/pentium-dual-core-e6000-e5000-datasheet.html?wapkw=e5300

https://www-ssl.intel.com/content/www/us/en/processors/pentium/core-2-e8000-e7000-pentium-e6000-e5000-celeron-e3000-guide.html


More than enough info for using these cheap but effective cooling solutions. Best is to ensure the fins are vertical and the hot air is directly vented out of the case, with enough slots on the opposite side for air entry so the flow is not restricted. You can also use old PC case parts like the duct to direct the air, and get good results quite cheaply.

If going for high power simply use multiple heatsink units, each one having a pair of pass transistors on it, and use either a small PCB or wire for the leads. When doing so use a gate stopper resistor on each mosfet right by the device to reduce oscillation. They make good VHF oscillators with 10cm of lead.