Help selecting best IGBT insulating material

[Perrin21]

I am trying to locate the best thermal insulating material for my Tesla inverter megapole boards.  The inverter is very expensive to replace (£10k) and has an inherent problem that the thermal insulation breaks down over time and shorts out the IGBTs.  I would like to rectify this issue with a new and improved insulation.  Is there anything available that has improved thermal properties over the stock compound and that's automotive grade?  Ideally an improvement in thermal properties would be great as the power electronics do get quite hot and at that point can start to thermal throttle power output.  Any suggestions?  Have there been any advancements in this area over the last decade?  water cooling maybe etc?

Here is what happens to the old compound

[T3sl4co1l]

Use weaker clamps.  Holy shit whoever designed that thing, wasn't thinking...

Tim

[Perrin21]

Use weaker clamps.  Holy shit whoever designed that thing, wasn't thinking...

Tim

It was built by Sheng Chien.  Heres a Teardown

[David Hess]

I am sure the stock solution was selected for ease of assembly and not thermal performance.  I would just use mica insulators and zinc oxide thermal grease.

[Perrin21]

ive heard good things about alloy oxide plates have you heard of these?  apparently they offer a significant upgrade to the heat transfer.  https://www.voelkner.de/products/59255/QuickCool-5061-00581C-Isolierscheibe-L-x-B-20.5mm-x-17.5mm-Passend-fuer-TO-247-1St..html

[T3sl4co1l]

They're fine, but rather thick so the advantage isn't huge.  Al2O3 is an okay thermal conductor.

AlN however is very good, comparable to (I forget if slightly better or worse) BeO but without any of the toxicity.

They're very stiff, so you need very flat surfaces as well.

Hm, I also wonder how flat they are to begin with.  They don't look ground/lapped.  Ideally all four surfaces would be ground flat, but yeesh...

There's also hard anodized aluminum tabs, which can additionally serve as heat spreaders, and can be grounded to shunt EMI.  Anodize isn't as hard as solid ceramic, so it can be damaged by scratching or bending, and doesn't have as high a voltage standoff rating.

Tim

[WattsThat]

Sort of OT but if anyone wants an easy lesson on why you should buy the correct tools for the job at hand, have a look at about 13:00 in the video in reply #2. Crappy tools make the job so much harder to do right but so easy to inflict direct and indirect damage.

I also concur with Tim’s assessment about the lack of heat sink flatness. In fact, that heatsink is part of the problem. It appears to have had a standard aluminum brush finish applied, usually done with about a 100 grit belt sander. Nothing flat or smooth about it. That surface irregularly adds to the lack of contact area/heat transfer and just accelerates the failures.

Kinda scary thinking that module is in a car on the open road, I certainly wouldn’t want to be following behind it at highway speed.

[David Hess]

They're fine, but rather thick so the advantage isn't huge.  Al2O3 is an okay thermal conductor.

AlN however is very good, comparable to (I forget if slightly better or worse) BeO but without any of the toxicity.

They're very stiff, so you need very flat surfaces as well.

Alumina is fragile and subject to cracking.

There's also hard anodized aluminum tabs, which can additionally serve as heat spreaders, and can be grounded to shunt EMI.  Anodize isn't as hard as solid ceramic, so it can be damaged by scratching or bending, and doesn't have as high a voltage standoff rating.

Anodized aluminum is fragile because the insulated coating is so easily scratched.

Hm, I also wonder how flat they are to begin with.  They don't look ground/lapped.  Ideally all four surfaces would be ground flat, but yeesh...

I also concur with Tim’s assessment about the lack of heat sink flatness. In fact, that heatsink is part of the problem. It appears to have had a standard aluminum brush finish applied, usually done with about a 100 grit belt sander. Nothing flat or smooth about it. That surface irregularly adds to the lack of contact area/heat transfer and just accelerates the failures.

I think the lack of surface flatness explains their insulator selection.  They saved money but not machining a flat surface which required using the elastic pad which also makes assembly cheaper.

[wraper]

It appears to have had a standard aluminum brush finish applied, usually done with about a 100 grit belt sander. Nothing flat or smooth about it. That surface irregularly adds to the lack of contact area/heat transfer and just accelerates the failures.

Surface smoothness makes zero difference to heat transfer. If you use hard insulating material like mica or ceramic pad, you need thermal paste anyway. If you use elastic material like silpad, it fills any gaps under pressure by itself. The only thing you care is surface flatness. If it is significantly convex or concave, then you have issues because even thick layer of thermal paste cannot cope with that decently.

[T3sl4co1l]

Also, I wonder if what they used is a phase change compound instead, and the heatsink is subject to temperatures where the stuff stays molten.  Or there's something about heat cycling the stuff that causes it to expand and ooze out cyclically.

They could just as well opt for a thermal epoxy, bonding it once and forever without any worry of shifting.  An electrically insulating compound can be selected, and glass beads can be added to ensure consistent application thickness -- and thus breakdown voltage.

Service might be more annoying of course, but by gluing to a heat spreader, replacement is just as easy, or even easier perhaps.

Heh, for that matter, the kind of component-level repair possible on an assembly like this, need not be all that much.  If a couple transistors die with arc flash, exposed connections on the board will be eroded.  You can clean up the soot and the burn marks, but it's not going to be a brand new board.

Tim