[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Cooling MOFSETS and IGBT's.



Original poster: "sundog by way of Terry Fritz <twftesla-at-uswest-dot-net>" <sundog-at-timeship-dot-net>



 Hi All!

 After owning it only a month, the cpu fan on my new machine comes to a
rattling, buzzing halt.  The "high quality" aluminum heatsink stood no
chance on it's own against a P3 1ghz crunching numbers for an additional 70
hours, so I shut it all down.  After a bit of fiddling, I decided to
implement a water-cooler for it, and it's been at a high of only 100 deg F
all weekend.  Normally it would hit 100 degrees just sitting idle (where it
now sits at about 3 deg above room temp), and peak out hot enough to fry
bacon on.  No more.  In addition to that, it's *quiet*  No more noisy fan
whirring 24x7.

  Marc Metlicka brought it to my attention that IGBT users can also benefit
from a water cooler.  The cost was cheap for me, only $20 for a small
submersible pump, $5 for tygon tubing.  The 1x1x9" copper stock I had lying
in the garage, and the 3lb coffee can was one of the emptys I keep hanging
around. Oh yeah, and 1/4" copper tubing.  Any self-respecting coiler will
have some of this stuff ;)

 I cut off a 1/2" thick slice, and drilled 4 holes in it with the
drillpress, soldered some 1/4" copper tubing to it, and plugged the
remaining holes in it.  I then lapped the surface of it smooth so it'd make
good contact with the surface of the chip.

  I had serious doubts about the ability of this home-made waterblock's
ability to remove heat.  So I hooked it all up, grabbed my trusty
thermometer probe for the DMM.  I kicked the pump on, let all the air get
forced out, and applied a lighter flame to the bottom of the block.  Then a
pencil torch.  Then a pencil torch for a solid minute.  Even at the end of a
1 minute run, I could put my thumb immediately on the block where the flame
had been, and it was barely warm, with condensation on it.  The water temp
in the can went from 77 to 82 degrees with all that testing.  No fancy
radiator, just about half a gallon of water.

 The waterblock only took me an hour to produce.  For multiple blocks, I'd
recommend a high-volume inline pump and barb fittings (or hoseclamps), as
each block drops a fair amount of pressure across it.  I don't know how much
of a problem heat is with IGBT's, but I'm 99.99% sure a waterblock on each
one would put and end to any heat problems you have with them.  And it'd get
rid of those bulky heatsinks (I replaced a heatsink half the size of a coke
can with a 1x1x1/2" block, and it's far more efficient).  For the low useage
the IGBT sees, you could use a 5 gal bucket for a resevoir, and simply dump
in a bag of ice to really chill them.  For that matter, any solid-state
device that runs hot will benefit from water-cooling.  Keep that in mind for
those MOFSET driven SS coils.  From what I recall they run hot.

  I'll try to post pics of the water cooler on my page, but time has been
very much so against me lately, and I'm now behind in everything.  Again :(

 On a thought, I do *NOT* recommend trying to watercool electrodes.  It'd
take a pump insulated to 30kv or so.  The $20 home-depot-special will most
assuredly arc over and fry (and put HV on the mains).  So don't try it.

Thanks Marc for thinkin' of using this on IGBS's!

   Comments, questions and snide remarks welcome!
											Shad