Electricly insulating heat transfer material

[Mathurin]

Ok, my controller is so designed that the back of the fets must be electricly insulated from the case in order to work. The problem is that the case is where the heat goes, and adding a strip of material between the fets and case obviously hinders heat transfer.

I'm going to have to open up my controller yet again to beef up the motor phase wires and put heat shrink tube instead of the electric tape I put on two wires, while I'm at it I'd like to optimise the heat transfer of the Fets.

So I'm wondering what people think is the best material for this application.

 

[Reid Welch]

I think you want heat sink compound.

Zinc oxide in grease--it's not electrically conductive.

Heat sink compound is typically a white paste made from zinc oxide in a silicone base.
http://www.pcguide.com/ref/cpu/char/coolCompound-c.html

Curiously, most particulates in grease are not conductive: they can't make contiguous paths for current, owing to the dielectric nature of oil or grease,
at least not at relatively low voltages such as we see.

I've got an 8oz bottle of "Dow Corning 11 Compound", of which I've used about half its contents in a period of over thirty years (!). We used this stuff in the USN for plug valve lubrication. I mention it because it was an old-stock bottle even then (when I,,,liberated it from active duty, ahem).

It's a stiff-bodied clear silicone, and absolutely time-stable and heat resistant to high degree. If, for instance, I were to mix some zinc oxide into this grease: first rate heat sink compound.

But even better (i'd bet), would be if I made it a paste with lead carbonate or lead monoxide (litharge) which I have leftover from my paint-making days (antique paints technology).

But for general public use, zinc is non toxic and lead is enviro-nasty.

So I won't do that.

But you can make your own heat sink grease if you like.
It's low tech indeed.

_____________just looked up in Wikipedia, I didn't know any of this:

History

In 1942 moisture in aircraft engines and corona formation made high-altitude flight all but impossible. Dr. Shailer Bass developed Dow Corning's first product, a simple silicone grease (Dow Corning #4 Compound) that solved the problem. Dow Corning was formally established in 1943 specifically to explore the potential of silicones. Dr. E.C. Sulllivan was named president and Dr. William R. Collings was named general manager in 1943. Dr. Collings latter became president from 1954 until 1962.

A large, majority-owned subsidiary of Dow Corning Corporation is the Hemlock Semiconductor Corporation. Founded in the 1960's before the computer revolution, it is still one of the world's leading manufaturers of high-purity polycrystalline silicon which is sold in varying purity grades for use in both semiconductor silicon wafer manufacture and photovoltaics applications as solar cells.

 

[Mathurin]

Grease won't cut it, has to be something more substantial.

 

[Lessss]

http://www.bergquistcompany.com/ts_spec_apps_isoedge.cfm

http://www.ferrotec.com/technology/thermoelectric/thermalRefAC.php

 

[Mathurin]

Well, you can see there's a sheet of material between the fets & case:

I did try grease on the fets in one of the iterations, and it blew a capacitor, you can see one of them is still "wet" with it in this pic:

So, last time I put it back together I put a strip of electric tape figuring it was thinner and would therefore transfer heat better. Well, going up extremes of hills yesterday it would appear the yellow & green phases have become connected together. I'm ready to bet their fet's drains got connected through the case maybe after melting or wearing the electric tape...

So the only solution I see is to get a thermal pad as in picture 1, the kind they stopped using in computers 10 years ago. The modern type sorta melts onto the processor untill the heatsink and processor touch so there's a minimum thin film left. The old kind is like a rubbery pad that gets clamped between the processor and heat sink, quite less then optimal. But that's the way the thing was packaged originally, hey? Too bad I no longer have that strip of material.

And Lesss, well that thing is pretty interesting but it would appear to be unavailable to mere mortals...

 

[Lessss]

Sorry was a 30 second google search before heading out to work.

 

[Reid Welch]

Pardon me--I did not realize that the sinks need to be electrically isolated from the case.

Do they need to be electrically isolated from each other?

If not, how about bolting a strip of copper or al. to them, and blacken it,
to both increase the sink area, and increase its emissivity?

 

[Mathurin]

Well, I appreciate the effort.


Copper blackening seems like it could be a winner. I guess the idea at this point would be to get a thin strip of copper from the hobby shop, cut it to size, blacken it, install with heat sink paste & something cushy on the metal bar that clamps the fets so they all get good pressure... Sounds almost too good to be true, if this works it ought to suck the heat out better then a thermal pad could. Gonna have to read up more about it.

 

[Lessss]

You sure you don't have any solder on the other side of teh board from teh fets touching the case. I still find it odd that the fets have to be insulated like that.

 

[Mathurin]

Well, I'm not 100% shure there isn't any solder touching anywhere, though if there is it's much more likely to be on the positive side. I doubt it, because when the first phase shorted to case, I think I would have noticed it. The middle legs are connected to the phase wires, and also to their metal plate thing in the back too. If you don't belive me, feel free to look'em up, hey.

The electric tape I used was rated to 80celcius, in hindsight rather a poor choice, but then hindsight is 20/20. Twice I got the controller hot enoufgh so it smelt like burnt plastic. Both times after straining for a while, and both times the phase wires were kinda warm, but the case stayed freezing. The second time it got to smell like burnt plastic it went dead. I thought the bike was amazingly hard to get up the rest of that hill, when it got flat I realised something was wrong, unplugged the phase wires and man did that thing ever glide! Two of the phases were electricly connected together.

It's been now two days and I haven't been arsed to re-open it yet, it's not super critical because I got a backup vehicle that's my miyata touring bike. (it's got generator lights, to think some people cheat using battery powered lights... Feh!) But obviously, untill I open it I won't know fer shure what happenned.

 

[Mathurin]

Well, an actual engeneer told me to use mica and heat sink paste. So that's what I got, and I plan to use it.

Theese are very thin plastic-ish things shaped like the case of the fet (220?), they're transparent and they sound like they're made out of glass if you let 'em drop on a desk, except they're flexible. If you flex them too much they turn white. 5cents per, and you need the heat sink paste to hold them on the fets while you slide the controller into the case.

Sounds good enoufgh.

 

[fechter]

The original rubber stuff is silicone. It actually does a really good job of tranferring the heat and not melting despite it's thickness. I usually put some heasink compound (grease) on them as well. Mica is also good, but is brittle and must be handled carefully. Grease that stuff too.

Even better is to replace the FETs altogether with ones that have a lower "on" resistance, so they don't need to dissipate as much heat.

I saw in the other thread where you beefed up the trace on the bottom of the board. One of these was bridging the current measurement shunt, which is what gave you the dramatic increase in current limit. It's also what melted.

That capacitor was way too small for the increased current, which is why it failed also. I would suggest multiple capacitors and get ones that have a low "ESR" or equivalent series resistance.

If you replace enough parts, you can make it pretty bulletproof.

 

[fechter]

Another approach is to bolt the FETs directly to a hefty bar of aluminum, then use the rubber stuff to insulate the bar from the case. This allows heat transfer over a much larger surface area and doesn't allow the FETs to heat too rapidly. In your controller, this looks like it would be hard to implement because there's not much room for a bar. Just an idea.

 

[Mathurin]

Thanks for the insight.


I've read in some newsgroup that a properly anodised aluminium is sufficiant.
The guy could have been full of it, though.

If I can get this controller working again then I'm probably going to use the mica + grease since that's what I have on hand, but I see that thick mica is sometimes used as an insulator, so that can't be too good.



While looking up blackened copper I came across foxhole radios.
It's not really related, but it's fascinating stuff:
http://bizarrelabs.com/foxhole.htm