cooling heat sinks / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

cooling heat sinks

1 2 next

Move Thread

LAN_403

IamSmooth

Sun May 02 2010, 10:47PM

Registered Member #190 Joined: Fri Feb 17 2006, 12:00AM
Location:
Posts: 1567

What methods do all of you use to cool down your heat sinks for high-current igbts or mosfets? Forced convection? Water-coolant jacket attached to the heat sink? Submerge it in a bath of liquid nitrogen?

Now that it is summer I am finding it harder to keep my heat sinks cool for my induction heater. Blowing 95F air does not buy me much time.

EDIT:

As an additional point, it seems the 1200v rated IGBT heats up much more than the 600v IGBT at the same power settings. When i swapped out the 1200v chip for a 600v one (with similar current ratings), the 600v chip was cool while the other chip was hot.

KLH

Mon May 03 2010, 12:43PM

Registered Member #1819 Joined: Thu Nov 20 2008, 04:05PM
Location:
Posts: 137

I do have quite a bit of information on device cooling from part of my recent research for computer parts, including processor heatsink-fan coolers.

After looking at many different designs, I noted that many of these are worthwhile not just for cooling of high-speed and high-power integrated circuits, but also for discrete device cooling. With some slight modding (drilling mounting holes, obtaining heatsink clips, using thermal transfer paste), many of these powerful and cost-effective coolers could be used to cool semiconductor devices so powerful to the point that the average little aluminum heatsink would be completely inadequate. (I personally could imagine attaching a device in a package such as the ISOTOP or SOT-227 to a cooler like the Hyper 212+ from Cooler Master.)

Because most CPU dies are actually much smaller than the actual carrier substrate that they reside on, you could make a heat spreader (basically a piece of sheet metal) much in the same manner as a CPU heat spreader to disperse the heat from a small semiconductor device to the heatsink base plate.

The possible results that can be obtained from using modified CPU coolers appear to be much more cost-effective than some actual solutions designed to cool discrete semiconductors. Mouser sells a "high-performance" dedicated IGBT heatsink for over $100; this heatsink is made of aluminum (as far as I remember) and contains no special innovations such as heat pipes or even a fan. Compare this heatsink and its price to that of the ~$35 Hyper 212+ cooler.

While I could not find any direct thermal dissipation ratings for this cooler, I am certain that the TDP (thermal design power) is far above 130W because this is the TDP rating for the next smaller model, the Hyper TX3. However, another thing to note is that the Hyper 212+ cooler is almost 160 mm tall and uses a 120 mm cooling fan; this means that the heatsink is FAR larger than any standard discrete device heatsink.

This is just my minor input on possible cooling solutions. I hope it is helpful.

Also, to answer your other questions:

-I prefer forced air cooling, as it provides high-power cooling (far more effective than passive convection) with high reliability (greater than that of water cooling). If you don't mind noise, a very powerful cooling fan such as the 120 mm 250 CFM (!!!) PFB1212UHE by Delta Electronics in conjunction with the Hyper 212+ will provide a very effective (and likely overkill) cooling solution for almost any thermal management problem.

-If my memory is vaguely correct, higher voltage devices generally have a higher voltage drop, resulting in higher losses depending on the usage conditions.

hak8or

Mon May 03 2010, 12:53PM

Registered Member #2025 Joined: Fri Mar 13 2009, 05:39PM
Location:
Posts: 16

Personally, I allways stick to CPU HSF's for cooling high power packages. They are designed to be as efficient as they can, due to space constraints in the case of a computer.

You should check PC review sites for information on HSF(Heat sink fan) and how well they do under what load. Toms hardware, Overclock.net, and xtreemsystems.com are the best from what I see.

,2370-5.html <-- an example
There are other sources that say how much watts of heat a cpu gives off based on the voltage, speed, and load.

Don't go for any liquid cooling kits, they are god awful and leak often.

Sulaiman

Mon May 03 2010, 02:25PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3141

I too use CPU heatsink/fan for cooling discrete components,
I do it because a fan/heatsink for an obsolete processor is VERY cheap.
The (watts per degree)/(cost) is much better than a simple convection heatsink.
(If there's a 12 Vdc supply conveniently available - which is usual).
For some things (e.g. an amateur radio transceiver) I use convection cooling as I don't want the noise of the fan.

I am slowly 'migrating' to a "standard" 12 Vdc supply for all my designs due to the cheap availability of 12 V VRLA batteries, 12V 'nominal' Solar chargers,12Vdc fans on heatsinks, and use in automobiles.

KLH

Wed May 05 2010, 01:44PM

Registered Member #1819 Joined: Thu Nov 20 2008, 04:05PM
Location:
Posts: 137

This thread seems to have stagnated somewhat, but I have just found something recently that might be of great use fo cooling individual discrete devices. The Xigmatek Porter-CN881 (Product Number CAM-S8HH1-U01) is a heatsink designed to cool the North Bridge chip on computer system mainboards; the design of this cooler, with a single direct-contact heatpipe, along with the mounting holes, appears to be ideal for cooling SOT-247 (ISOTOP) packages as well as leaded device packages such as the TO-220 and and TO-247 (when using a pressure clip attached to the cooler's mounting holes). I have not yet searched where this heatsink may be purchased, but since it's from a company that also sells retail products, I'm sure there's a distributor readily available.

Mounting options may be strange; this is a tower cooler, so using through-hole leaded devices in a vertical configuration will result in unusual heatsink placement. The devices may be bent downwards (with the heat spreader plate facing up) so the heatsink may be placed vertically as normal. Device packages such as the SOT-247 must be mounted upside down relative to a circuit board, but the bolt-on terminals of this type of device package may provide other options.

In addition, I would recommend mounting an 80 mm cooling fan unless the cooler is mounted horizontally (although this is only an option, as stated in the product specifcations). If you don't mind noise, a Vantec Tornado Series fan or a Delta PFB or FFB Series fan would provide more cooling than you probably would ever need.

I hope this suggestion is useful.

condor

Thu May 06 2010, 12:20AM

Registered Member #2780 Joined: Mon Apr 05 2010, 12:39AM
Location:
Posts: 7

Look at the data sheet on that celem you were asking about on the other thread. It is designed for a sub base on the back.

Make a base with a hollowed out space between your two outlet pipes in the back. use one front nipple for supply the other for return.

You have the water there why not use it? Those celems are made to be stacked.

Mount your igbt or fet right on that sub base. ?thats what Im planning.

Tiberius

Thu May 06 2010, 04:52AM

Registered Member #1484 Joined: Wed May 14 2008, 03:24PM
Location: Cary, NC, USA
Posts: 27

Although it is usually overkill, I like to use water cooling whenever feasible. It gives you some headroom, in case things go terribly wrong, to scope the circuit or remove power before your components release the magic smoke.

You can make practically commercial quality cold plates for components by soldering some copper tubing to a machined copper base, and you can acquire a suitable base by cutting the fins off of a copper heatsink, or removing the copper slug from a copper/aluminum heatsink. You may want to crush the tubing a bit in a vice to give it more contact area with the base. Several "passes" with the tubing are good for best performance, and you can use off the shelf U-joints for this if you are unable to bend your tubing at a sharp enough angle, though your work with copper tubing looks pretty good from what I have seen

For TO-247/TO-220 devices I have soldered some copper tabs to a larger, flattened (1/2") copper pipe, and even this relatively inefficient design extracts much more heat than any air-cooled sink I have been able to fit on or near the same boards.

The down side is that you will almost certainly need a separate cooling loop, unless you are using fresh water for cooling. If using fresh water, you could probably hit the switches with it first, and then your work coil. The switches will not produce all that much heat and will benefit more from lower temperatures.

If you are using the transistor case as a conductor, you may want to use mercury or another liquid metal as your thermal interface material. Liquid metals much more effective thermal interfaces than any thermal grease, and wick into the interface after the surfaces have been wetted filling all voids. There are some concerns with mercury vapor, but as long as it remains relatively cool (and it should on your cold plate!), then it does not vaporize quickly enough to be a significant concern. If you have a small and/or poorly ventilated work area there are liquid metal alloys that are safer than mercury, such as Galinstan.

GhostNull

Fri May 07 2010, 10:10AM

Registered Member #2648 Joined: Sun Jan 24 2010, 12:45PM
Location: Australia
Posts: 291

IamSmooth wrote ...

Submerge it in a bath of liquid nitrogen?

Maybe not, electronics also have a minimum work temp. too remember

Steve Conner

Fri May 07 2010, 10:39AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

There's a research paper out there about how power semiconductors perform at cryogenic temperatures.

I seem to remember they found that MOSFETs work better under liquid nitrogen, but IGBTs don't like it at all.

Coherent

Fri May 07 2010, 06:44PM

Registered Member #1886 Joined: Sun Dec 28 2008, 02:55AM
Location:
Posts: 73

You can always try submerging them in mineral oil like people do with transformers.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob