Roadster PEM failure

[Jonathanm]

Part 2

Be careful with the plastic screws - mine were ok but I can imagine if they have suffered a lot of heat cycles they could become brittle....

Disconnect this small wire that disappears down into lower board. Mine had some sticky tape attaching it to 1 of the capacitors.

Lift the board free...

This is how it looks now... The left side of the black choke (I think that is what it is called on the left side of this picture) you should remove the nyloc nut holding the left side - the bolt for this nut is actually threaded into the fitting.

You should unscrew this bolt until it is flush with the threaded fitting. This will allow the board to be lifted clear.

In the photo below remove the megapole board with the removable fuse on it by removing the 2 crimp connectors screwed onto the frame and right front clamp. (2 white cables with red crimps)

Now start to undo the nuts on the clamps - I think they are 6mm. In my case all the nuts were loctited onto the studs and it was the studs underneath that started to come out. Thats ok but only do a turn or 2 on each nut before proceeding to the next 1. If you go too much on 1 nut the nut underneath will start to push up on the circuit board from underneath and may damage things...go slowly and work your way around until they are all loose. If one stud gets tight then move onto its neighbour...You can check the board to see if it is loose or not.

The board is now loose and can be lifted up. Be careful there is a cable going down to the heat sink - be careful with this. The terminal clip would not fit through the hole in the board to be able to lift it clear.

Without touching the wire take a small square sectioned file and gently file the hole slightly larger so you can get the terminal through the hole which will allow you to get the board fully clear of the PEM.

OK that's the 1st board removed!!!

Now you can look at the state of your insulation...

Now remove the smaller upper circuit board on the centre main board - same as the 1st one. Notice there are no long plastic screws on this one. Remove the small terminal connector that goes to the "choke" board. You have to remove this board as it is covering some of the IGBD's on the centre board. See the following photos...

Oops - next thread...

 

[Jonathanm]

Part 3.

Remove the small nylon screws from this small circuit board.

The other screw on this circuit board is an allen key.

This bolt needs to be removed - it is long with a plastic sleeve that fits into a metal block that supports the choke.

The board can now be lifted clear...

Remove the support cradle for the circuit board (located on the IGBT clamp)

Remove as in the previous posts all the 6mm studs on the IPGT clamps.

Same thing with the file again to remove the heat sink sensor cable.

Second board out!!!

Start of last board....

Next thread.

 

[Jonathanm]

Some of the insulation is still on the circuit board....

Voilla finished!!!!

A couple of points - I may have removed parts not necessary to do this job - as this is the first time I hope others might not be too scared off by this.
Obviously the heat sink needs to be cleaned of all the old insulation. Not sure what product I will use - maybe white spirit or something similar - acetone could damage stuff if it gets in the wrong place.

All the studs need to be removed from the boards and replaced back in the heat sinks prior to assembly. The clamp nuts should be done up with some sort of loctite - torque for these? No idea...

The new insulation are strips of CeramTec Rubalit® 708S Alumina, 96%. The lengths are 16.7cm x 3.2cm x 0.63mm. You will need 12 of these. Remember guys do it at your own risk blah blah blah...Thanks a lot to Daniel for a lot of this information. Without him pioneering this we would be still paying Tesla €10K for a repair (assuming your IPGT's are ok - I still have to test mine so I don't know how they are but there is zero evidence of any burning.) This job should only cost you a few hundred euros or dollars and some hours of your time. BTW I had this job done in 5 hours including the removal from the car. Someone who knows what they are doing would easily do it in less than 1/2 that time....

Another story is the capacitors. Mine are showing no signs of swelling or bulging. If you see any with bulged ends or distorted shape now would be the time to change them. Again this is small money to change these. If you are not handy with a soldering iron you could get any guy to do this for you. I have not tried to source the capacitors but imagine that they would be off the shelf from Digikey or other supplier...


Assembly will be the reverse of the 80 odd photos that I have taken. I really hope some of you will be able to do this job.

When I start my rebuild I will show more photos - it won't be for some months as I am away with work.

BTW don't know if the webmaster wants to combine the various posts into 1?

Jonathan

 

[augkuo]

View attachment 257169 View attachment 257170

View attachment 257171

View attachment 257172

View attachment 257174

View attachment 257175 View attachment 257177

View attachment 257178 View attachment 257179
Some of the insulation is still on the circuit board....

View attachment 257180

Voilla finished!!!!

A couple of points - I may have removed parts not necessary to do this job - as this is the first time I hope others might not be too scared off by this.
Obviously the heat sink needs to be cleaned of all the old insulation. Not sure what product I will use - maybe white spirit or something similar - acetone could damage stuff if it gets in the wrong place.

All the studs need to be removed from the boards and replaced back in the heat sinks prior to assembly. The clamp nuts should be done up with some sort of loctite - torque for these? No idea...

The new insulation are strips of CeramTec Rubalit® 708S Alumina, 96%. The lengths are 16.7cm x 3.2cm x 0.63mm. You will need 12 of these. Remember guys do it at your own risk blah blah blah...Thanks a lot to Daniel for a lot of this information. Without him pioneering this we would be still paying Tesla €10K for a repair (assuming your IPGT's are ok - I still have to test mine so I don't know how they are but there is zero evidence of any burning.) This job should only cost you a few hundred euros or dollars and some hours of your time. BTW I had this job done in 5 hours including the removal from the car. Someone who knows what they are doing would easily do it in less than 1/2 that time....

Another story is the capacitors. Mine are showing no signs of swelling or bulging. If you see any with bulged ends or distorted shape now would be the time to change them. Again this is small money to change these. If you are not handy with a soldering iron you could get any gut to do this for you. I have not tried to source the capacitors but imagine that they would be off the shelf from Digikey or other supplier...


Assembly will be the reverse of the 80 odd photos that I have taken. I really hope some of you will be able to do this job.

When I start my rebuild I will show more photos - it won't be for some months as I am away with work.

Jonathan

Great job Jonathan, I'm sure most of us will have to do this at some point. How much time do you think you had before the insulation would have been totally shot?

 

[Jonathanm]

Well its totally shot now - the car won't work...I understand its the IGBT's shorting out on the heatsink. The insulation is no longer working...AFAIK it only needs 1 IGBT to short out and the car is for a flat bed...
People smarter than me please correct....Daniel who has done this job has found nice improvements with the cooling of the PEM...

 

[DanielFriederich]

ok... about the cooling
with those alloy oxide stripes as insulation and a resulting better heat transfer the PEM is no longer the bottle neck.
Today I went on the Autobahn running 15km with almost top speed (13 rpm, approx. 100 kw, outside temp 12 C)
and guess what:
PEM not above 37 and when leaving the Autobahn temp of the PEM rapidly falls to below 30!
Battery went up from 28 to 38 and engine now is becoming the bottle neck going up to 120 (still in blue range)

 

[DanielFriederich]

and btw:
Jonathan, you are my new nr. 1 hero! fantastic work!

 

[Jonathanm]

Nice work, Jonathan. I think others would benefit from you posting your photos and describing your procedure so please do. Any indication where the noise was coming from? Seeing the condition of your insulation motivates me to replace mine sometime this winter.

My guess is it is the relays in the PEM...right side...I can even see them now! ....

 

[augkuo]

ok... about the cooling
with those alloy oxide stripes as insulation and a resulting better heat transfer the PEM is no longer the bottle neck.
Today I went on the Autobahn running 15km with almost top speed (13 rpm, approx. 100 kw, outside temp 12 C)
and guess what:
PEM not above 37 and when leaving the Autobahn temp of the PEM rapidly falls to below 30!
Battery went up from 28 to 38 and engine now is becoming the bottle neck going up to 120 (still in blue range)

View attachment 257205

The alloy oxide stripes are the ones Jonathan mentioned correct? Thanks!


The new insulation are strips of CeramTec Rubalit® 708S Alumina, 96%. The lengths are 16.7cm x 3.2cm x 0.63mm. You will need 12 of these.

 

[DanielFriederich]

The alloy oxide stripes are the ones Jonathan mentioned correct? Thanks!

YES

 

[gregd]

ok... about the cooling
with those alloy oxide stripes as insulation and a resulting better heat transfer the PEM is no longer the bottle neck.
Today I went on the Autobahn running 15km with almost top speed (13 rpm, approx. 100 kw, outside temp 12 C)
and guess what:
PEM not above 37 and when leaving the Autobahn temp of the PEM rapidly falls to below 30!

So, this makes me curious. Why would transferring heat from the IGBTs to the heat sink more efficiently make the heat sink better able to cool? The temperature is being read on the sink side of the insulators, not the IGBT side, right? It would seem to me that the IGBTs themselves are probably running hotter, not cooler, which would be bad. Think of the limiting case, where the IGBTs aren't even mounted on the heat sink. The sink would stay nice and cool, and would be indicated as such on the VDS, but the IGBTs would probably fry.

What am I missing?

 

[ML Auto]

So, this makes me curious. Why would transferring heat from the IGBTs to the heat sink more efficiently make the heat sink better able to cool? The temperature is being read on the sink side of the insulators, not the IGBT side, right? It would seem to me that the IGBTs themselves are probably running hotter, not cooler, which would be bad. Think of the limiting case, where the IGBTs aren't even mounted on the heat sink. The sink would stay nice and cool, and would be indicated as such on the VDS, but the IGBTs would probably fry.

What am I missing?

You are not missing anything. That is exactly what I was stating in my previous post #119.

 

[gregd]

You are not missing anything. That is exactly what I was stating in my previous post #119.

Oh, right. But in post #120 he was happy because the base temp went up 10C. Now not so much?

DanielFriederich, I'm confused...

 

[hcsharp]

So, this makes me curious. Why would transferring heat from the IGBTs to the heat sink more efficiently make the heat sink better able to cool? The temperature is being read on the sink side of the insulators, not the IGBT side, right? It would seem to me that the IGBTs themselves are probably running hotter, not cooler, which would be bad. Think of the limiting case, where the IGBTs aren't even mounted on the heat sink. The sink would stay nice and cool, and would be indicated as such on the VDS, but the IGBTs would probably fry.

What am I missing?

I have a different take on this than MLAUTO. Transferring heat more efficiently to the heat sinks means there is less thermal insulation between the IGBTs and the heat sink. In other words heat is more easily transferred from the IGBTs to the sink. So the heat sink gets warmer as it draws more heat from the IGBTs which in turn get cooler. In Daniel's case the sink is almost 10C warmer and the IGBTs are likely to be that much cooler than the OEM setup.

As for your question, "Why would transferring heat from the IGBTs to the heat sink more efficiently make the heat sink better able to cool?" I've already answered some of that but it's also true that a bigger temp differential between the cooling air and the heat sink will allow more heat to be removed from the heat sink, and therefore from the IGBTs.

 

[gregd]

I have a different take on this than MLAUTO. Transferring heat more efficiently to the heat sinks means there is less thermal insulation between the IGBTs and the heat sink. In other words heat is more easily transferred from the IGBTs to the sink. So the heat sink gets warmer as it draws more heat from the IGBTs which in turn get cooler. In Daniel's case the sink is almost 10C warmer and the IGBTs are likely to be that much cooler than the OEM setup.

As for your question, "Why would transferring heat from the IGBTs to the heat sink more efficiently make the heat sink better able to cool?" I've already answered some of that but it's also true that a bigger temp differential between the cooling air and the heat sink will allow more heat to be removed from the heat sink, and therefore from the IGBTs.

Ok, all that makes sense, except that the pictures show the measurement of the temperature is not on the IGBTs, but on the heatsink. And things do cool better (as in, more heat is transferred) when there is a higher differential, but Daniel's latest report (post #166) implies that the measured temps are lower.

To me that says that he's only succeeded in fooling the car into thinking the PEM was running cooler when in fact the IGBTs are probably running hotter. The only way everything would run cooler is if he's swapped out the IGBTs for a different part that has a lower "on" resistance, resulting in less heat being generated in the first place.

 

[DanielFriederich]

ok... that thought about the sensors I can understand. But indeed temperature rises faster but also declines faster now with a better heat transfer from the IGBTs to the cooling plates.

My 2 cents on that:
- IGBTs stay (overall) cooler due to a faster heat transfer to the plates (thats a guess but most likely true)
- due to cooler IGBTs avg. temperature we see less thermal loss due to lower resistance in operation of the IGBTs (thats a guess!)
- the insulation used by Tesla has about only one third of thermal conductivity compared to the alloy oxide stripes (thats a fact)
- max. operation temp of the Infineons is 175C... so if the car is "fooled" that PEM plates are cooler than before and IGBTs are hotter the performance of the PEM should decline, esp. after hard 0 1 driving due to the thermal operation curve of the IGBTs (in fact its not)
- there are 4 sensors for temp in the PEM (one directly on each Megapole board and one inside the PEM), so higher avg. temps could also be a result of more heat transfered from e.g. the clamps mounted to the circuit boards to the inner environment of the PEM (thats a real guess!)

I might make a video of 0 1 or high speed driving with the VDS screen as "main actor" and dash as "special guest appearance"

Appendum: it never gets boring with that car somehow love it!

 

[Scene]

My PEM has recently failed too (UK Roadster 2.5 with Desat errors).

I’m seriously considering repairing the PEM myself because Tesla (UK) can’t tell me if their expensive PEM refurbs have the insulation issues addressed - otherwise it will fail again at some point. The other alternative is to ship the PEM to a 3rd party repairer in Europe (if I can find one).

I have some questions:

1. Thanks for the info regarding the CeramTec strips. Probably a stupid question, but is additional heat sink compound/paste also required between the strips and the IGBTs? If so, what is recommended?

2. I am considering changing all the IBGTs as well for alternative devices with a lower saturation voltage. What are your recommendations? If there are no better alternatives, would it be a good idea to replace all the IGBTs in any case, even though it’s a lot more work than just testing and replacing failed devices?

Thanks.

 

[DanielFriederich]

I’m seriously considering repairing the PEM myself because Tesla (UK) can’t tell me if their expensive PEM refurbs have the insulation issues addressed - otherwise it will fail again at some point. The other alternative is to ship the PEM to a 3rd party repairer in Europe (if I can find one).

It seems like the original insulation type is used in refurbished PEM. GS Technology Switzerland mentioned that they had a few refurbished ones with disruptive discharges!

1. Thanks for the info regarding the CeramTec strips. Probably a stupid question, but is additional heat sink compound/paste also required between the strips and the IGBTs? If so, what is recommended?

not between the stripes and the plates, maybe only a very thin film between the IGBTs and the alloy stripes to avoid "hot spots" on the IGBT backplane. the higher the thermal conductivity the better

2. I am considering changing all the IBGTs as well for alternative devices with a lower saturation voltage. What are your recommendations? If there are no better alternatives, would it be a good idea to replace all the IGBTs in any case, even though it’s a lot more work than just testing and replacing failed devices?

nor Hanwind in Germany or GS Technology are changing them as long as they are good.... this also applies to Gruber in US as far as I know.
They seem not even to be matched like in hifi amplifiers....

Would be very interesting how the driver boards correspond to an IGBT with lower saturation voltage. I considered that in the beginning but discarded it due to the unknown results and behavior of the PEM...

 

[hcsharp]

Would be very interesting how the driver boards correspond to an IGBT with lower saturation voltage. I considered that in the beginning but discarded it due to the unknown results and behavior of the PEM...

I thought about doing this pro-actively to get less heat overall and more efficient operation - including more range. But when I ran the numbers it didn't seem like the benefit was worth all the work taking my PEM apart and soldering. That is especially true given that my PEM is working fine.

But if I had my PEM apart anyway to replace a few IGBTs, I would replace all of them with the newer low saturation voltage versions. I confess I didn't think there would be any issues with the driver.

 

[Jonathanm]

I am buying my new insulation strips (CeramTec Rubalit® 708S Alumina, 96%) at the moment - if anyone want some let me know. The cost is not the material but the time in setting up the laser cutting machine. I have found a supplier in Paris. It will probably cost €100 or so more for another 12. If interested PM me.