Roadster PEM failure

[ion_1]

The thicker aluminum nitride or alumina, seems to be a good solution. I believe small amount of high temperature resistant thermally conductive paste has been used in some of the posts above to assist the thermal interface. This may be a good idea as none of the interfaces are perfectly flat at the microscopic level.

I'm not sure a bonded epoxy would be the most ideal either to replace the ceramic or at the interface. There is still some thermal mismatch between the aluminum sink, ceramic, and IGBT/boards. Personally, I would like the interfaces to be able to slip a bit even if a few tens of microns. If they don't something will have to break or be under much stress. That being said,a thermoset epoxy will be much more resistant to creep than the materials which was used by tesla (which seems to be a thermoplastic). Hopefully the ceramic approach by this thread's pioneers eventually demonstrates robustness and we all will have viable solution at hand.

 

[thefortunes]

Isn’t the original product some kind of paste? Or am I mistaken? I am seriously thinking of proactively performing this job and have been watching this thread in order to learn from others and just do it once.

Oem was definitely not a paste. They were some type of pliable insulating strips which had started to dry out, break down and crumble in some areas (typically at an IGBT corner near the edge of the strip). Other portions of the strips not directly under the IGBTs were lifted away from the heat sink, but remained pliable.

My observation is that the strips were actually slightly misaligned (too close to the edge of IGBTs on one side in almost all cases - you can see this in Hallvardr's pics above) and my speculation is that if the IGBTs were not perfectly flat or the boards were not progressively torqued evenly as they were installed they created a slightly hotter spot in a corner which accelerated the deterioration of the strips until a short occurred.

Having pulled the boards back apart, I am fairly comfortable recommending Gruber's solution. However, it is early in the life of my PEM after the replacement. As I mentioned when I did mine, I will be curious to see the long-term viability of both options.

Re: Xlr8's question - yes I would definitely let a prospective buyer know that I had proactively replaced a deteriorating material which could have cost them a PEM replacement.

 

[X.l.r.8]

Fair comment, but then again its a $3500-5000 retrofit, not exactly a diy repair in any shape or form.

 

[Jonathanm]

Yes I would absolutely state this to prospective buyer - in any event should it fail again, its a relatively easy fix. Worst case you may have to change out some IGBT's. (read above) There is a considerable number of people who have already done this repair...ask Gruber...
Secondly the 10K fix by Tesla is using the same flawed product that is now failing. I know what I would rather do...

If you DIY it will cost you the strips...around €200-€300 depending what option you use. If you have some toasted IGBT's around $5 for each of those...if I had to change out those I would take the boards to a specialist to re solder... again its not huge money...

 

[thefortunes]

Fair comment, but then again its a $3500-5000 retrofit, not exactly a diy repair in any shape or form.

Actually, it can be a $100 diy repair. I did it (with advice and help).

 

[X.l.r.8]

I totally agree on the process, I'm not shooting anything down, quite the opposite, the grubber solution is recommended every 7 years! that's about how long the original product has lasted so I'm not sure that's an effective solution. If the plates you spoke about a few pages back being of thicker material offering equal thermal conductivity would offer a huge step in eradicating the problem.

 

[X.l.r.8]

I could not find just the strips on Grubber Power Services website so its hard to show how that's done, again, its recommended to be changed after 7 years making it less a definitive solution potentially replacing one deteriorating material with another

 

[Jonathanm]

Flexible Thermally Conductive Adhesive Electrically Insulating Epoxy Low Thermal Expansion Cooling

Just looked at the specs on this product...the thermal conductivity seems worse than OEM...

 

[X.l.r.8]

I said a product like that, as in a compound rather than a solid strip that adheres rather than just sits there. Also I am having trouble finding the actual spec sheet on what Tesla originally used, lots of theories but no real answer.

 

[X.l.r.8]

I think a viable option is one Jonathanm already showed us in the report he already linked
https://teslamotorsclub.com/tmc/att...crew_tightening_torque-an-v1-0-en-pdf.267399/
looking through as much hard evidence as I could gather the original "must be .5mm thick at the origin"
The Gruber material is 1.0 W/mK which is equal to the model S
The backs of the TO-247 is not flat making flat plates with high Shore A hardness a problematic solution as they do not conform to the IGBT making it a poor conductor. its High W/mK is probably a good mask for that.
The TO-247 has a working temperature of 150-175 Celsius
On the tightening torque paper above, they use a product called Keratherm Red.
It due to the material conforming to the back of the IGBT it needed to be tightened to between 0.6-1.0
"The chart above showed that devices in different packages that used the Keratherm 86/82 isolation foil (60 to 70 shore A) had no significant changes in Rth(j-c) when the screw tightening torque is in the 0.6 to 1.0Nm range. Therefore a typical screw tightening torque value of 0.8Nm was recommended."

Keratherm 86/82 comes in 0.25-0.5mm thick, has an operating temperature of -40 - +200 Celcius and is rater at 6.5W/mK
It is cut by a sharp knife so its a DIY product.
The product line up looks like it has a few options
White/green/pink/red/brown thermally conductive films - Alfatec GmbH & Co. KG

Again this was the product used when addressing the tightening torque so I'm wondering it that method only applies to a material that will conform to the uneven backs of the heatsink and IGBT's
See page 12 of this document
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d46249cd10140149e0c7fe9d56c7

And yet I still worry about the life time of these products, I think as Jonathanm has detailed diagrams how to get to and replace what ever material is used it should not be a catastrophic event to repair the PEM at home.

 

[ion_1]

Keratherm 86/82 comes in 0.25-0.5mm thick, has an operating temperature of -40 - +200 Celcius and is rater at 6.5W/mK
It is cut by a sharp knife so its a DIY product.
The product line up looks like it has a few options
White/green/pink/red/brown thermally conductive films - Alfatec GmbH & Co. KG

These composites may be a great compromise. Ceramic filled with an option of fiberglass reinforcement. With a small amount of the the latter, this should vastly reduce the chance of these materials creeping and resulting in a short like the OEM materials. Should make a good interface between the IGNT and the sink.

 

[hallvardr]

why is everyone using ceramic strips? I understand the thermal conductivity is better but the failure is in the material not the properties is it not? wouldn't it be simpler to find an alternative to the original that is simply a better product. I have been looking around as to what I would use preemptively as the worry of having to spend thousands on a hundred dollar fix worries me.
Flexible Thermally Conductive Adhesive Electrically Insulating Epoxy Low Thermal Expansion Cooling
There are many solutions like this, I use this building my Lebowski type controllers, I'm sure there are other companies out there that essentially have a similar product. Has anyone had the ceramic strips stay intact without further failure?. A combination may just end up causing a heat barrier reducing the strips so its a brave person who ends up trying that. If there were more roadsters north of the boarder I could test out what works but alas I'm going to have to wait till I can move to start playing.

One of the deciding factor of me choosing ceramic is that I have heard that the European side of Gruber has used that for 20+ cars.

 

[DanielFriederich]

in my experience (which cost me some grand... :/ )
Ceramics thermal conductivity is far beyond what is needed inside the PEM and the cons, like breaking, like igbts not fully being "connected" to the ceramic stripe and therefore overheating, are a major against using Rubalit or whatever.
Using thermal paste I wouldn't suggest either since I had quite a few disruptive discharges and broken ibgts due to the thermal paste not staying at the place it should stay and causing lower insulation at the corners of the stripes and... BAM.....

in my opinion using a stripe (e.g. silicon based and fiberglass strengthened) with a slight better thermal conductivity (6W/mK) makes more sense. Heat dissipation works fine and bumpiness or not lying totally flat backplanes are leveled out by the soft material.

I have driven now 2.500 km approx. with this kind of material and the bottle neck in heat is not the PEM but the motor (no matter if digital driving 0 1 0 1 or Autobahn at 180km/h...)

 

[DanielFriederich]

Rubalit stripes with thermal paste...

 

[DanielFriederich]

One of the deciding factor of me choosing ceramic is that I have heard that the European side of Gruber has used that for 20+ cars.

True:
Hanwind is using Rubalit (Aluminium Oxide) 24W/mk
GS Technology is using fiberglass strengthened stripes 6W/mK
as far as I can judge Gruber (AZ) is using non ceramic material too

 

[DanielFriederich]

why is everyone using ceramic strips? I understand the thermal conductivity is better but the failure is in the material not the properties is it not? wouldn't it be simpler to find an alternative to the original that is simply a better product. I have been looking around as to what I would use preemptively as the worry of having to spend thousands on a hundred dollar fix worries me.
Flexible Thermally Conductive Adhesive Electrically Insulating Epoxy Low Thermal Expansion Cooling
There are many solutions like this, I use this building my Lebowski type controllers, I'm sure there are other companies out there that essentially have a similar product. Has anyone had the ceramic strips stay intact without further failure?. A combination may just end up causing a heat barrier reducing the strips so its a brave person who ends up trying that. If there were more roadsters north of the boarder I could test out what works but alas I'm going to have to wait till I can move to start playing.

This material is operating up to 125 Celsius.... operating temp of the PEM 2.X igbts is 175C max... so the insulation should be at least 200C or more...

 

[X.l.r.8]

I totally agree, the epoxy is a non starter, I separate the power phase on my sine wave controllers so I never see that kind heat. It really needs to be a product that you can torque down. The ceramic probably don’t need that kind of torque because they can never conform to the IBGT or the heat sink. The Gruber strips are right now the viable option, however I wonder how long people expect that solution to last. It may just be a fact that roadster owners keep a set of strips stored for when they need them. Just like I have to keep a spare set of headgaskets for my race car

 

[warden_clyffe]

hey folks,

after replacing all IGBT-isolators i mounted the pem again and restarted - everything works, but i do get a regular error-massege:

ID: 1145 - DMC WARNING LCMC Sensor (from Debug Text)

i can't find this error in those lists on this forum - there's only 1144 and 1146 wich both got something to do with pem-fan (i suppose). does anybody know what 1145 is ?

thanks for your support ;-)

 

[DanielFriederich]

could be like # DFC 1201

DFC: LCMC Temp Sensor Failed Yellow Fault 1 PEM - Internal Failure Replace PEM Contact Chief Technician to replace PEM due to internal PEM failure
DFC: LCMC Temp Sensor Failed Yellow Fault 2 PEM - DFC Failure Replace PEM Contact Chief Technician to replace PEM due to DFC failure

my guess is a not connected or faulty heat sink temperature sensor

what do you get as values for PEM and Motor temp?

 

[Jonathanm]

As Daniel said it could be a temp sensor...if I remember correctly it seemed possible to connect the sensor connector either way. I used the way the cable laid to determine the way it was before I got in there...this is just a guess...