02-06-2011, 12:57 PM
|
#4321 (permalink)
|
EcoModding Apprentice
Join Date: Nov 2010
Location: Annapolis
Posts: 159
Thanks: 0
Thanked 32 Times in 27 Posts
|
Quote:
Originally Posted by flores
Even if you make the heatspeader out of copper, the base plate will still be aluminum, right? If that's true I'd say that aluminum is fine for the heatspreader. The system has to lose the heat through the base plate anyway..
Or is that an all to simple conclusion?
|
That's too simple of a conclusion.
Aluminum has about half the thermal conductivity as copper. To transfer the same heat with temperature drop we need twice the cross section area. Easy enough: just buy twice as much aluminum, which will still cost less than the copper.
But the situation is much different when we want to move the heat away from a point source rather than down a constant cross-section. Or more specifically, from a small metal patch on the back of the MOSFET to the heatsink. It becomes a 3D problem. We can't just stuff more material in the contact patch.
A solution is to use the best thermal conductor (copper) to spread the heat out from the point source, and then a lower cost bulk material (aluminum) to transfer the heat.
I haven't done an analysis on the Cougar thermal design, just a quick estimate. I believe the most thermally restrictive element is the thermal pad on the MOSFET. This paper-thin material electrically isolates the MOSFET back from the heat spreader. But even the best thermal pads are far less thermally conductive than copper or aluminum, and they are right where the heat is concentrated.
An obvious thing to do is have the MOSFETs directly contact a thin copper sheet that acts as a first-level heat spreader. But you quickly find out why Paul chose the approach he did. You have to figure out where to put the electrical isolation. The thermal pad material is expensive, and the better you do widening the heat flow path, the more you have to buy. And the mounting hardware really wants to electrically bridge your isolation.
We are designing around using "live" heat spreaders that simultaneously act as bus bars. This lets us avoid the first-level thermal pads. But we still have to get rid of the heat from the bus bar, and there isn't a cheap solution. We can't have live heat sink fins outside the box, so it's $30+ of insulating thermal pad, expensive anodizing, or oil cooling with plastic fittings.
Last edited by DJBecker; 02-06-2011 at 03:08 PM..
|
|
|
Today
|
|
|
Other popular topics in this forum...
|
|
|
02-06-2011, 01:05 PM
|
#4322 (permalink)
|
Custom EV Builder...
Join Date: Jul 2009
Location: Ohio
Posts: 57
Thanks: 8
Thanked 1 Time in 1 Post
|
Quote:
Originally Posted by DJBecker
That's too simple of a conclusion.
Aluminum has about half the thermal conductivity as copper. To transfer the same heat with temperature drop we need twice the cross section area. Easy enough: just buy twice as much aluminum, which will still cost less than the copper.
But the situation is much different when we want to move the heat away from a point source rather than down a constant cross-section. Or more specifically, from a small metal patch on the back of the MOSFET to the heatsink. It becomes a 3D problem. We can't just stuff more material in the contact patch.
A solution is to use the best thermal conductor (copper) to spread the heat out from the point source, and then a lower cost bulk material (aluminum) to transfer the heat.
I haven't done an analysis on the Cougar thermal design, just a quick estimate. I believe the most thermally restrictive element is the thermal pad on the MOSFET. This paper-thin material electrically isolates the MOSFET back from the heat spreader. But even the best thermal pads are far less thermally conductive than copper or aluminum, and they are right where the heat is concentrated.
An obvious thing to do is have the MOSFETs directly contact a thin copper sheet that acts as a first-level heat spreader. But you quickly find out why Paul chose the approach he did. You have to figure out where to put the electrical isolation. The thermal pad material is expensive, and the better you do widening the heat flow path, the more you have to buy. And the mounting hardware really wants to electrically bridge your isolation.
We are designing
|
Isn't this were the sot-227 mini bloc style mosfets come into play? I think they can be mounted directly to the surface of the heat spreader without thermal pad material and you have a larger area on the back of the mosfet itself.
Please correct me if I'm wrong as this is the approach I'm taking with my build.
|
|
|
02-06-2011, 02:26 PM
|
#4323 (permalink)
|
Joe
Join Date: Feb 2009
Location: phx
Posts: 260
Thanks: 0
Thanked 48 Times in 38 Posts
|
Quote:
Originally Posted by MPaulHolmes
I only used a copper heat spreader on Joe's I think. And maybe Adrian's. Perhaps aluminum is just fine? I haven't heard of any thermal problems from anyone. I just don't know what the difference in performance is. I'm inclined to think that aluminum is fine for most people. Bill is in Phoenix, Arizona which is another name for the fires of heck! I'm sorry for that terrible language! The only part that determines the power board is 1" or 0.75" wide.
|
i was curious about the impact of heat spreader material and width, and I happened to have my work computer at home, so I did a quick thermal model to see what happens.
Keep in mind this was done super quick with ballpark numbers and the results probably don't represent reality, but rather give insight into each configuration's performance relative to the others.
The model consisted of a 11" heat spreader on 1/4" aluminum plate. There is 250W of heat generation on the upper portion of the heat spreader on both sides to represent losses from both the mosfets and diodes. 500W total heat generation would be consistent with a current output of 300A, based on the normal calculations for mosfets and diodes. There's a tad of thermal resistance at the spreader/plate interface. Also, there's convection to 22C air applied to the bottom of the plate to represent a finned heatsink. The hole looking thing in the middle isn't a hole, but where the final temperature was read once everything reached steady state.
I ran 4 models; the two different materials and the two different widths.
Al/0.75": 86C
Cu/0.75": 78.5C
Al/1.0": 81C
Cu/ 1.0" 75C
I think the differences are significant but may not matter depending on the application. For a more demanding application that requires a lot of current output and thus, increased controller losses, going with wider copper may help the controller last longer and/or take more time to reach thermal cutback.
However, a wider heatspreader means the diodes and mosfets are farther apart, which could lead to more parasitic inductance and higher voltage spikes/noise etc.
Ah, tradeoffs...
|
|
|
The Following 3 Users Say Thank You to jyanof For This Useful Post:
|
|
02-06-2011, 03:07 PM
|
#4324 (permalink)
|
EcoModding Lurker
Join Date: Oct 2010
Location: phoenix
Posts: 42
Thanks: 0
Thanked 3 Times in 2 Posts
|
wow, thats a great model, would it make any difference in the temp, if someone placed 2 cpu fans on oppsite sides of the heat spreader long ways, one to pull air and one to suck air while the cover is on, my controller during the last part of summer here, was getting a little warm, it did not go into thermal cutback, but i thought about putting the fans in just to see if it would make any difference, and was going to do just that when I get my controller back.
|
|
|
02-06-2011, 03:32 PM
|
#4325 (permalink)
|
EcoModding Apprentice
Join Date: Nov 2010
Location: Annapolis
Posts: 159
Thanks: 0
Thanked 32 Times in 27 Posts
|
Quote:
Originally Posted by wakinyantanka
Isn't this were the sot-227 mini bloc style mosfets come into play? I think they can be mounted directly to the surface of the heat spreader without thermal pad material and you have a larger area on the back of the mosfet itself.
Please correct me if I'm wrong as this is the approach I'm taking with my build.
|
I've seen reasonable diodes in that package, but not good MOSFETs.
Avnet has a good price, $18.53, on the STPS200170TV1 diode in an ISOTOP package. That's 200A at 170V, which should be good for up to 150V traction voltage (the diode sees lower peak voltages than the MOSFET).
|
|
|
02-06-2011, 03:41 PM
|
#4326 (permalink)
|
PaulH
Join Date: Feb 2008
Location: Maricopa, AZ (sort of. Actually outside of town)
Posts: 3,832
Thanks: 1,362
Thanked 1,202 Times in 765 Posts
|
Quote:
Originally Posted by DJBecker
But you quickly find out why Paul chose the approach he did. You have to figure out where to put the electrical isolation. The thermal pad material is expensive, and the better you do widening the heat flow path, the more you have to buy. And the mounting hardware really wants to electrically bridge your isolation.
|
Actually, you quickly find out why Ian Hooper in Australia did what he did. haha. I just copied other people's work (with their permission) and bought a mill and smart people started helping me.
I like that bus bar idea!
With water cooling, is it bad if the water is in direct contact with an electrically live material? Really pure water is a terrible conductor.
|
|
|
02-06-2011, 04:01 PM
|
#4327 (permalink)
|
Joe
Join Date: Feb 2009
Location: phx
Posts: 260
Thanks: 0
Thanked 48 Times in 38 Posts
|
Quote:
Originally Posted by billhac
wow, thats a great model, would it make any difference in the temp, if someone placed 2 cpu fans on oppsite sides of the heat spreader long ways, one to pull air and one to suck air while the cover is on, my controller during the last part of summer here, was getting a little warm, it did not go into thermal cutback, but i thought about putting the fans in just to see if it would make any difference, and was going to do just that when I get my controller back.
|
thanks!
The internal fan question has been discussed before. I think it will help a small amount with the mosfets and diodes (since their surface area is small relative to a finned heatsink), but a large amount with the caps (but they don't seem to get too hot anyway). Of course you'll have a lot of dust and crap blowing around in there which may cause some trouble. I did try to model it by throwing a small bit of convection on the same surfaces as the heat generation, and the temp did drop a few degrees, but I hesitate to say it's that effective since modeling convection can be very tricky (i just guessed at what coefficients to use).
If you do give it a try, hopefully you can use the serial output and adam's RTD explorer to monitor the temperature before and after and let us know what effect internal cooling has.
|
|
|
02-06-2011, 06:41 PM
|
#4328 (permalink)
|
EcoModding Lurker
Join Date: Oct 2010
Location: phoenix
Posts: 42
Thanks: 0
Thanked 3 Times in 2 Posts
|
since I just got that working right about 2 weeks ago (operator error) i will most certinly do that, I have rtd explorer on my laptop now.
|
|
|
02-06-2011, 07:05 PM
|
#4329 (permalink)
|
EcoModding Apprentice
Join Date: Nov 2010
Location: Annapolis
Posts: 159
Thanks: 0
Thanked 32 Times in 27 Posts
|
Quote:
Originally Posted by MPaulHolmes
Actually, you quickly find out why Ian Hooper in Australia did what he did. haha. I just copied other people's work (with their permission) and bought a mill and smart people started helping me.
I like that bus bar idea!
With water cooling, is it bad if the water is in direct contact with an electrically live material? Really pure water is a terrible conductor.
|
There is no such thing as pure water. At least not for very long. And you couldn't keep a thin insulating layer in the cooling channels intact with a 150V potential between two of them.
One solution is to cool with oil. They make special oil for the purpose, but any cheap oil will do. It's non-conductive and doesn't need to be changed. The only drawback is that you have to flow about twice the volume as water for the same heat transport.
There are other insulating fluids that are better, but I don't think they are needed. Once you decide to move heat with circulating fluid instead of conduction, you can move a lot of heat. The cool plate and radiators are more important than the working fluid.
Last edited by DJBecker; 02-06-2011 at 10:08 PM..
|
|
|
02-06-2011, 11:55 PM
|
#4330 (permalink)
|
Custom EV Builder...
Join Date: Jul 2009
Location: Ohio
Posts: 57
Thanks: 8
Thanked 1 Time in 1 Post
|
Quote:
Originally Posted by DJBecker
I've seen reasonable diodes in that package, but not good MOSFETs.
Avnet has a good price, $18.53, on the STPS200170TV1 diode in an ISOTOP package. That's 200A at 170V, which should be good for up to 150V traction voltage (the diode sees lower peak voltages than the MOSFET).
|
IXYS makes these; IXFN230N20T. I think they will work nicely. At least I hope so.
|
|
|
|