H2O cooled infineon [was:highside FET w/o Kapton?]

O

olaf-lampe

10 kW
Joined
Mar 27, 2008
Messages
710
Location
Germany
Hi,
I'm building my own watercooled 12 FET infineon controller. The cooling bar has no connection to the housing.
Q:
Can I use the aluminium bar to feed the highside FETs with positive voltage? ( Drain is also connected to the backplate of the FET as u know )
I would only put Kapton tape between the lowside FETs and the cooling bar.

I expect lower losses and a better cooling . Also the pcb will like it when only 'half' of the current heats up the traces.

What do you think?
-Olaf
 
I'm not sure that cooling the high side fets more than the low is going to help you out a whole lot. Your "limiting reactant" so to speak is still the hottest component, which will be the low side fets.

I can't imagine it would hurt though.
 
grindz145 said:
I'm not sure that cooling the high side fets more than the low is going to help you out a whole lot. Your "limiting reactant" so to speak is still the hottest component, which will be the low side fets.
Click to expand...

You have a point here.
I now think about splitting the cooling bar in 6 parts. Three parts connect to V+, the other three connect with the phase wires, which are the drain connection of the lowside FETs. This way I could get along without Kapton tape at all and cooling would be equal for all FETs. Plus, I eliminate another load from the pcb.

Only headache is the isolation between each cooling block and still allow waterflow :(
-Olaf
 
You can isolate each of the low-side cooling bars by using short lengths of rubber or plastic hose to connect to the copper (I assume) tubing.
 
I started building the cooling bar today.
Here is a pic, where you can see the holes for the FETs have an offset of 1mm. After I saw them apart and realigned them, you see that I can now connect the 'upper' three blocks to the V+ copperstrip. The other three will connect to the phase wires.

I will use the original cooling bar to align the new blocks and then glue them together with silicone rubber. There is only a 1mm gap between the blocks, not necessary to use any fittings for the water circuit.
The copperstrip is 12mm x 1.5mm = 18mm². This should be more than enough for 240A peak?

View attachment heatsink1.jpg

-Olaf
 
your best bet would be some thermal gappad. It's higher thermal resistance but better than most of the rubbery junk they use in the chinese bikes. Capton might still be a decent option...
 
grindz145 said:
your best bet would be some thermal gappad. It's higher thermal resistance but better than most of the rubbery junk they use in the chinese bikes. Capton might still be a decent option...
Click to expand...

I don't need anything between the FETs and the cooling blocks. Maybe some silver conductive paste, but I 'll smooth the block's surface until it reflects like a mirror ;)
I only need the silikone between the blocks to glue them together and seal the water circuit
 
olaf-lampe said:
Alan B said:
Are you planning to use low conductivity cooling water?
Click to expand...

Sure!
Since my early overclocking_CPU_days, I learned my lesson.
Click to expand...

Obviously not if your systems leak and you have to use a non conductive fluid :p Having watercooled pc since
early 2000 with my own designed and built waterblocks i can tell you a properly set up
system will not leak. Using no conductive fluids such as FluidXP results in poorer cooling
than when using just water and fungicidal additive, they aren't as efficient
as conducting heat as plain ol water.

Look forward to seeing how you watercool your controller though, have you though of total
summersion using mineral oil cyled through a heat exchanger via a pump like the mp350 from Laing?
Would cool all components not just the fets ;)

Best of luck...

KiM
 
I'll use this radiator with integrated pump:
http://endless-sphere.com/forums/download/file.php?id=48446
How can I integrate a pic from another ES-thread?

Since the traces onboard are not getting hot, like when all the current runs across the pcb, the only thing I have to cool is the shunt area. I planned to use 4 shunts, therefor I have to drill two more holes in the pcb. The watercooler has a fan included, maybe I aim it towards the pcb.
But at average speeds of 30-40 km/h and topspeeds of +100km/h, I guess a fan is reduntant.

-Olaf
 
amberwolf said:
olaf-lampe said:
How can I integrate a pic from another ES-thread?
Click to expand...
Just stick in in teh image tag just like any other. ;)
Click to expand...
Code: 
[img]http://endless-sphere.com/forums/download/file.php?id=48446[/img]
See the link in my sig...
 
How many watts of heat are we thinking the controller will produce? coz that lil rad
would be useless even keeping some of the new cpus temps down under load. I was thinking at least a 240 size
heat exchanger would be needed?

KiM
 
AussieJester said:
How many watts of heat are we thinking the controller will produce? coz that lil rad
would be useless even keeping some of the new cpus temps down under load. I was thinking at least a 240 size
heat exchanger would be needed?

KiM
Click to expand...
I tested the CPU heatsink with 800W . The heatsink was 80C and the water didn't get hotter than 40C without the fan. When I switched on the fan, the temperature almost immediately dropped to 5C above ambient. A radiator works better, the higher the temp-difference is. A CPU need temps less than 50C but the FETs can handle more than 100C, before some kind of thermal runaway starts.
-Olaf
I don't know what a 240 size radiator looks like. Also I can't find the item on pollin.de anymore. It seems they are sold out.
 
Today I received the missing small parts from Lyen to complete the pcb as a 100V version.
Thanks Edward, it saved me a lot time searching for the right stuff!

I'll PM you later with some stup... errhh detailed questions, since I don't have a schematic or BOM.
Especially for the programming adapter I was searching for the right pads.
Anyone who has a picture of an eb212 with programming cable soldered in? :(

-Olaf
 
olaf-lampe said:
AussieJester said:
How many watts of heat are we thinking the controller will produce? coz that lil rad
would be useless even keeping some of the new cpus temps down under load. I was thinking at least a 240 size
heat exchanger would be needed?

KiM
Click to expand...
I tested the CPU heatsink with 800W . The heatsink was 80C and the water didn't get hotter than 40C without the fan. When I switched on the fan, the temperature almost immediately dropped to 5C above ambient. A radiator works better, the higher the temp-difference is. A CPU need temps less than 50C but the FETs can handle more than 100C, before some kind of thermal runaway starts.
-Olaf
I don't know what a 240 size radiator looks like. Also I can't find the item on pollin.de anymore. It seems they are sold out.
Click to expand...


I stand by my first comment...

best of luck...

KiM
 
I tried to sort out the placement of the small parts I got from Lyen:
View attachment eb212_capacitors.jpg

I took a picture of knuckles' 150V EB212-AS version to spy from.
I still have a 1n4007 diode left . One is on the bottom right corner beside the 5V regulator, but there are two other possibilities to put the second diode. None of them where used by knuckles build?! :?
Also a hint, where to put the programming connector would be helpful :D

-Olaf
 
Got another picture of the cooling bars.
View attachment heatsink2.jpg
I drilled 4 holes through each block for the water circuit. The end plates of block 1 and 6 will connect each hole in a way to build a maze. For the looks I'll use clear polycarbonate for it.
It will look like this CPU cooler i built years back.
View attachment PC-cooling_side2.jpg

-Olaf
 
olaf-lampe said:
I tried to sort out the placement of the small parts I got from Lyen:



Also a hint, where to put the programming connector would be helpful :D

-Olaf
Click to expand...


Below the biggest cap in this pic, there is a row of holes with white boxes printed around them. The two leftmost holes are the positive red wires. The rightmost hole won't be used. Solder in your program cable and get ready to have frustration with programming! :lol:
 
Another update:
I used a mini grinder disk on my dremel to free some copper areas on the board. Also drilled some holes to add 3 more shunts.
I decided to solder the gate pins of the FETs directly on the top layer. Therefore I drilled the through holes open. Hereby I make more room on the bottom layer to add some thick copper to the source pins of the lowside FETs.
View attachment 5 shunts.jpg
View attachment 5 shunts_rear.jpg

Next step is to make some solid copper pices that fit on the pcb traces. I hope I can heat the pices up on the stove and put them on the traces where I'll have added some SMD solderpaste before. I'll try it out on another board before I take the plunge...
-Olaf
 
Are you going to let the through holes carry the current on the source pin?

Bubba
 

Attachments

  • 5%20shunts.jpg
    50.5 KB · Views: 1,031
You must log in or register to reply here.

Similar threads

oldswamm
A Lebowski controller in a 9 FET 'Infineon' case
Replies
3
Views
2,201
oldswamm
oldswamm
Z
Add on T0247 water cooling upgrade board for a 12 FET
2 3
Replies
51
Views
7,850
crusoe
C
methods
Technical Details on Methods 100V 300A Controllers
Replies
16
Views
3,782
bionicdan
bionicdan
Reid Welch
"Extreme MY1018" archival, closed out
Replies
11
Views
15,731
GIGATT
G
Back
Top