Hello, ive been building an IGBT bridge, as shown in the picture attached, one thing which came to concern me, these devices switch fairly high current, and the 8" by 8" heatsink will heat up fairly fast, so to aviod this, Ill install a few fans around the sink.. one thing which I wondered though,
when I place a fan on the bottom where the fins are, will the heatsink be cooled better if the fan was "sucking" the air away from the sink? or if it was blowing air towards the heatsink?

thank you.

 

Hello,

Blowing to the heatsink will be more effective.
Take a look at processor coolers, they all blow toward the heatsink.

Bertus

 

strange, my CPU fan blows outwards, I can feel the air comming out of the fan guide away from the CPU heatsink

 

The velocity of the air stream directed at the fins breaks up the surface layer of air. Sucking is less effective, regardless of how somebody else did it.

 

Have you done a power dissipation assessment for your application? Looks like you should be able to get somewhat below 1C/W for that sink when fan forced, but if you need something lower then you may need to do a thermal test before you get your hopes up too high.

Ciao, Tim

 

Hello,

I think this page will give you some more idea's for the heatsink and cooling:
http://homepages.which.net/~paul.hills/Heatsinks/Heatsinks.html

This is a link from this page of the EDUCYPEDIA on Heatsinks, fans and cooling:
http://www.educypedia.be/electronics/heating.htm

Bertus

 

Hi
The heatsink will be best cooled by fans blowing on it , but bear in mind if you fit this assembly in an enclosure the enclosure fan sucks air out.

 

these devices switch fairly high current, and the 8" by 8" heatsink will heat up fairly fast,

Those look like 50 Amps or so per device? Even with a fraction of an Ohm on resistance they will get toasty.

If you have access to something like shop compressed air, build an emergency overload cooling valve that blows compressed air on the heatsink.

Because of the adiabatic heat loss during expansion the compressed air will be very cool and it will have a high flow rate. It shouldn't be used to waste compressed air, just to keep those big IGBT getting smoked by a momentary overload.

 

Are you using thermal transfer compound between the IGBTs and the heatsink?

If not, you should.

Also, If you are going to be pushing those IBGTs, You may want to consider more area between each, or separate sinks for each.
Remember copper is also available. A forced air, copper fin device works quite well. Blow air INTO the sink. You do not want to be pulling very hot air across the fan motor.

 

If the heatsink heats really fast, it should have good contact with the transistors, and i think i can see thermal grease in the picture. I think it's safe to assume that the heatsink is too small for this.
What's the power dissipation of all the transistors and the thermal resistance of the heatsink?

 

Keep in mind that aluminum is roughly 58% as effective at heat transfer as copper is.

Water is 1.4 times as effective at heat transfer; however using water cooling gets rather complex. If your enclosure is limited in size, you might not have other options.

Heat pipes are quite interesting. A basic heat pipe can be constructed from copper tubing using a wicking element like Litz wire or Solder-Wick running the length of the pipe. The pipe is sealed and evacuated (purged of air) and then a small amount of isopropyl alcohol is introduced. Heat applied on one end of the pipe causes the alcohol to boil; the vapor travels to the cool end of the pipe where it reverts to liquid form. The capillary action of the wicking element transfers the liquid back to the hot end of the pipe, where the cycle repeats. The speed of transfer is quite high; even supersonic.

Of course, the heat still needs to be dissipated on the cooling end by a suitably large radiator. Heat pipes are used in aviation and spacecraft. They are simple, reliable, and have no moving parts. The toughest part is achieving a hermetic seal on the closures; as if there is even a very small leak, the device will fail.

 

sorry, I was away for a bit,

so Ill try to answer as best as I can to the comments,

yes, there is thermal grease for the heatsink, lots of it, I just had a bit of a hard time whiping off the exess

also, I should mention the specs of this,
what you see are 4 IGBTs, on the top and bottom, and in the middle are a pair of dual diode modules (with 2 diodes inside 1 package)

the IGBTs are rated 600V 200A, the diods 1200V 60A (per diode)
both devices rated for 600A pulsed (1mS)

I keep these IGBTs in saturation, so switching losses are not as high, and switch them within thier frequency range (if not below)

this is not a continous system, it is pulsed,
the whole bridge sees roughly 500A pulsed for 50uS, so fairly low duty for the IGBTs. (low on time, high off time)

as for liquid cooling, I have considered this, but do not have any heatsinks on hand that can properly cool these devices (I do have copper heatsinks, but are smaller, and a few watercooled ones, but yet again are smaller.

another option I lapsed the thought of, was peltier cells, but I belive that is too inneffecient and overkill for this, I dont want frost on my IGBTs

another thing about this design, it is open frames, bieng the enclosure has quite a bit of holes to allow lots of air circulation, yet not enough for some idiot to stick thier hand in it.

hope this helps you understand better, and if you have further questions, please dont hesitate to ask.

 

Can you measure the Vce when the IGBTs are in conduction? How long is the off-time?
Is the diode current also 500A? If that's the case, the diode drop can reach 1v easily, meaning 500 watts per diode, and maybe reverse recovery time has some extra influence. What's the part number of the IGBTs and the diodes?

Thermal grease should be applied in a thin layer. It's better than the discontinuous surface of a heatsink filled with air, but isn't a great thermal conductor, either.

How hot is the heatsink, anyways, and what is it's thermal resistance?

 

Ive found out it is not too hot, about 50C on average, and usually reaches 70-80C after a minute of run time.
as for its thermal resistance, I dont know, I do not have the datasheet for this heatsinks, and got it from someone else.

the offtime is 250uS

also, here are the part numbers:

IGBTs: APT200GT60JR by Microsemi
Diodes: STTH12012TV1 by ST

these IGBTs seem fine with me overexeeding thier current ratings, but are prone to overvoltage, which is also why the bridge has transient protection for its C-E, and zeners for its G-E.

 

Well, it's a lot of power there, it would be wise to ask someone that is familiar with such a kind of power management, to make sure if the circuit is safe. If something fails, it's going to be quite a fireworks show.

Not really sure from the datasheet what is the max pulsed current it can handle at 80°C, but if something does fail, it will be very dangerous. Better to seek expert help, and keep everything the safest way possible.

 

I figured I might as well show you guys a little bit more of the total system,

now that grade 10 is over, and summer brake has starter, I can get more progress done, but in the picture shown, you see the controller modules, that provides 5V+ 12V+ and 24V+ to the logic circuitry (not shown, still working on it) and that is for the gate drivers themselves (im using a pair of UCCs, they drive a mosfet driver for the gates, running at 24V, (ill go higher if I want to keep the IGBTs in saturation) the controller is still far from complete (note that the top is missing, I need to make an interface for it, and finish the frame itself in the front where fans are.)

Thanks for your help everyone, Ill consider some method of shielding for the bridge itself.. most likely some type of swiss cheese like cage to still allow plenty of cooling and some protection.

 

Given the cooling topic - it would be good to see a thermal assessment based on your V-I profile during conduction, and the datasheet transient thermal Rjc and Rc-hs. That would then give you a design Tjmax to compare with datasheet Tjmax and whatever margin you have determined appropriate.

Ciao, Tim