Toyota inverter IPM(IGBT module) thermal grease replacement

[Lwerewolf]

Hello,

Hopefully this is the right place - I'm not building an EV (well, yet... you know how this goes ), but I have a gs450h daily with a throttling inverter, and I suspect that it's a somewhat common condition for old, used inverters (spoiler: degraded thermal compound, for the ones that use it). Didn't manage to find a discussion of this here, so here we go:

Background (how I got to the grease in the first place):

In my case, when MG2 is under specific conditions (~2500-4500rpm, tall gear - ~50rpm per kph... so say from 50 to 90kph) and ambient temperatures are "high" (say, >10c - basically not chilly), the car throttles when given full power - mg2 IGBTs hit >120 and throttle, no DTCs or "bad things happened" logs in the HV ECU, no throttling at any other speeds (temps stay around 60 even in 30deg. ambient). Quick checks - nothing else appears to be overheating, radiators are clean, fresh coolant, no way the cooling fins are blocked (I have a "blown" spare - I've torn it down - the passages are not exactly tiny), water pump works. I then started worrying about either Toyota messing up the shift logic (EU-spec 450h's have a taller final drive) or the solder connecting the IGBTs to the heatsink deteriorating (which doesn't exist - the ls600h apparently uses TIM too, but... you know what happens when you fix your mind on something, I guess).
(VERY LATE EDIT: there are solder layers, read everything that I've written with a grain of salt ^_^)

At any rate, while reading the oak ridge articles again (I was focused on the combined motor-inverter efficiency graphs (l110 transaxle and camry inverter) and wondering whether I can override the shift-down to low gear earlier to move away from that "inefficient" area, or supercool the system during the summer using a water->AC heat exchanger or peltier elements or whatever - condensation worries), I finally noticed that the IGBT baseplates interface with the cold plate via thermal grease.

Quick observations:
-ls600h is the first inverter (to my knowledge) using double-sided cooling and SiC IGBTs (EDIT: nope, not SiC, again - something got in my head... still - double sided) - higher thermal limits, bigger cooling capacity. No reported failures, to my knowledge. No reported failures (again, to my knowledge) on any of the newer inverters using this design, either.
-Prius gen2, rx400h/HiHy, 1st gen gs450h, Prius gen 3 - single-sided cooling. Plenty of reported failures on all models, TSBs for rx400h/HiHy and rius 3.
-Anecdotal observations - inverter failures tend to happen with old and/or high mileage cars, other than the prius 3.
-It's thermal grease/paste... it is susceptible to pump-out/dry-out/ and so on over time.
-Prius 3's IGBTs don't have base plates, or a thermal interface material between them and the coolant loop - the IPM is integrated with the cold plate.
-EDIT: rx450h's inverter appears to be using double-side cooled IGBTs too:
https://www.denso.com/-/media/jp/ja/inn ... v14/08.pdf

７. 結論 両面放熱パワー素子と積層冷却を用いて高密度実装を実現し, 従来モデルより高い出力密度を持つ LS600h用 PCU を開発した．熱抵抗は従来設計の半分以下となり, シリコンチップの使用数を大幅に削減できた．本冷却方式を用いた PCU は, 09 年に RX450h にも搭載された．

Conclusion so far - good idea to attempt replacing the IGBTs' thermal grease on a potentially >10 year old inverter, given that it's possible.

Main questions - what thermal compound to use, how to apply it (pattern, thickness).

Sources so far:

Here's the rx400h TSB:
https://static.nhtsa.gov/odi/rcl/2013/R ... 6-9944.pdf - page 24 (pdf page 35)
White grease, shin-etsu g747, silicon based, low volatility, low viscosity, no solvents.
Application - essentially manual screen printing. Thickness - unknown, can extrapolate if I manage to get the masking plate (have queried, awaiting an answer), or by guesstimating based on the base plate sizes, the illustration of the plate in the PDF and the information that 5 tubes (the toyota part number is for 20g tubes, msrp $45 >_<) precisely fill both IGBT "sheets" per base plate.

Here's the 2010-2014 (gen3) Prius TSB:
https://static.nhtsa.gov/odi/tsbs/2019/ ... 6-9999.pdf
The TSB references the workshop repair manual for instructions in IPM (IGBT module) removal. This PDF is useful:
https://f01.justanswer.com/xsvectOR/7fe ... _prius.pdf
Key points:
1) IPM includes the heat sink, you can't "repaste" it here. I wonder why they fail so often as to require a TSB... Maybe the lack of thermal mass (no baseplates & huge heat sink)?
2) Masking seal has a pattern on it - it's "stencil printing".
3) Grease is shin-etsu X23-7884-4 - seems toyota-specific, couldn't find anything on it other than its color (the oak ridge national lab article) and a safety data sheet in Japanese. Doesn't look too interesting.

Further readings so far:

https://www.fujielectric.com/company/te ... 9-2010.pdf
Mostly discusses thermal grease application methods.
Does mention some names:

The following types of thermal grease were used : Electrolube’s HTC, Dow Corning’s SH 340, American Oil & Supply’s AOS 340 and Shin-Etsu Silicone’s G747.

https://www.semikron.com/service-suppor ... ev-00.html
Discusses thermal grease application in more detail - selection criteria, application methods (& tools & potentially how to make/get them), consequences, etc.

https://www.mitsubishielectric.com/semi ... note_e.pdf
-General application notes on certain Mitsubishi IGBTs. This caught my attention (page 15):

Typical value given in datasheet is measured by using thermal grease producted by Shin-Etsu Chemical Co.,Ltd. (G-746, which has not issued in Shin-Etsu's publications, is almost the same as G-747.)

https://americas.fujielectric.com/produ ... ctor-faqs/
Question 39 lists some "recommended thermal greases" - potential G-746 (supposedly resp. G-747) replacements.
The mentioned chapter: https://www.fujielectric.com/products/s ... 4b_06a.pdf

https://www.infineon.com/dgdl/Infineon- ... 00d9fd2ea2
Nothing in particular, just the first thing that I came across. The "temperature development" graph is a good lesson for starters

https://www.hitachi-power-semiconductor ... method.pdf
Comparison of different shin-etsu greases.

http://pdd.hitachi.eu/documents/product ... ote_r2.pdf
page 26, table 4.2 - another name, also G747 is mentioned yet again.

http://www.files.e-shops.co.il/multisit ... reases.pdf
More names, equivalents.

http://www.ohm.com.tr/doc/Mitsubishi-El ... HVIGBT.pdf
G746 mentioned again, along with Dowsil 340 & G747.

I guess that's it for today, starting to lose my mind somewhat. Plan is to round-up thermal compound options and figure out how to apply- roller, diy laser cut 0.1mm steel masking plate with some pattern emulating the newest developments, etc.

Hope this is of some use to somebody. Would love to hear your thoughts

[Kevin Sharpe]

Would love to hear your thoughts

I have transistor based power electronics that's been working everyday for over 50 years. I really don't think thermal grease will be an issue with something as 'new' as a 10 year old Toyota inverter that's sealed from the elements and operated within a very 'safe' power band (remember that Toyota will automatically protect the inverter).

If you're convinced the issue is the inverter then just replace it with another. They're cheap as chips

[Lwerewolf]

LHD gs450h inverters are in the ~1k eur range in Europe, if you can find them at all

I'm not convinced in anything other than this being worth a try.

[PacEmaker]

Interesting that you seem to be experiencing degraded cooling of your GS450h IPM, so I can see why you are looking at replacing the thermal grease. I see degraded, dried out thermal grease in my audio repair business all the time - I also see plenty of older gear that is still going strong. Initial application, conditions of use (or abuse) etc. all factor in to how long the things keep their smoke in.

I for one would be interested to see whether you are able to 'remedy' your situation by replacing the thermal grease using the best practice described in the articles.

[Kevin Sharpe]

LHD gs450h inverters are in the ~1k eur range in Europe, if you can find them at all

Gerardo in Spain typically sell LHD for 400 euro 'retail' on eBay. If you look around you can pick them up for half that price at local scrap yards. Buy one for 200 euro, test it in the car and sell the one you don't want for 200 euro. Will cost you a couple of hours effort which is a lot less than stripping down the inverter power stage.

[Isaac96]

Okay here's my 2 cents:
I'm sorta a DIY computer enthusiast so I've messed with a few thermal pastes (never anything industrial though). It seems like in general, you can't apply too much -- as long as there is enough mounting pressure to squeeze it out.
This is reflected by this Semikron application note I found - https://www.semikron.com/dl/service-sup ... 19-rev-00/
Thermal resistance increases very quickly if there is not enough grease, but only slowly if there's too much.

Application techniques - it sounds like you're going for industrial style, but a simple X shape will probably work fine.
https://www.pugetsystems.com/labs/artic ... iques-170/
(again, from a DIY computer point of view).
When I tore down my DMOC inverter and replaced the heatsink I used Noctua thermal paste, which is the standard for the DIY computer people. (These are CM600DY IGBTs, 600A 600V, half-bridge modules). By just applying an X and then bolting the modules down tightly, the paste squeezed out and covered the entire module.

For specific pastes - no idea. Of course you want longevity and reliability.

Best of luck with your repairs!

-Isaac

[Lwerewolf]

I do have a few tubes of kryonaut laying around...

The mounting pressure is unlikely to be high - the rx400h igbt assembly is bolted down to 6nm. I'll try with a low viscosity paste (sub-100 Pa.s - stock shin-etsu is ~50 at... 25deg?) and a roller - see how the pattern goes. Can try heating up the cold plate & paste beforehand too - should help with flow, even if applied to the IGBT baseplates first. I have my old blown inverter to practice/experiment on - IGBT assembly base plates & the heatsink itself are intact

Things yet to choose/acquire - wet film gauge, suitable roller, good torque wrench or bit driver, thermal paste.

[Martin1775]

Hello, I want to share my experience with IGBT modules.
I professionally repair rectifier devices with 9kw to 18kw in which IGBT modules 7mbr120sb50 or 7mbr120sb75 are installed. After about 6 to 10 years, the internal temperatures of some modules become higher because the module appears to be detached from the base plate. Renewing the thermal paste does not bring much improvement. New thermal paste or film makes a maximum of 5 to 10°C difference on the internal temperature sensor under load. Old modules go up over 120 °C and fail, new ones are below 80 °C under the same conditions.

[Lwerewolf]

Renewing the thermal paste does not bring much improvement. New thermal paste or film makes a maximum of 5 to 10°C difference on the internal temperature sensor under load. Old modules go up over 120 °C and fail, new ones are below 80 °C under the same conditions.

Ah well, I guess it won't hurt to try either way. Can you share what thermal greases are used, and what the considerations are? Looking at Wacker P-12, Dow 340, Electrolube HTSP, Kerafol KP-97 and KP-12 (conductive but doesn't seem to be an issue...?) - basically anything with low-ish viscosity that is "resistant to dry-out/separation" (several studies indicate that this is probably the wrong criteria, given G747's apparent degradation over time, but alas). Haven't looked into any films yet.

I see degraded, dried out thermal grease in my audio repair business all the time - I also see plenty of older gear that is still going strong. Initial application, conditions of use (or abuse) etc. all factor in to how long the things keep their smoke in.

On that note, while looking at this article again:
https://techno-fandom.org/~hobbit/cars/ginv/i1mech.html

Specifically, this picture:
https://techno-fandom.org/~hobbit/cars/ ... 3rmipm.jpg

When I pulled apart the inverter that I blew up in 2018, the "grease" literally looked like transparent silicone lube... Who knows. The ORNL teardown of the camry inverter also had grease in, well... at least to me - good looking condition. I guess, on mine, the filler may have completely separated from the base fluid/matrix/polymers/so on.

Now, about the LHD vs RHD inverters - upon further investigation, it looks like the primary differences on the board are the "bus bars" running through the IPM - the ones above the IGBTs themselves. If you look at teardowns, the MG1 IGBTs are connected on one side, and MG2 on the other - page 14 (actual pdf page 24) of the oak ridge camry teardown, located here:
https://www.osti.gov/biblio/928684-eval ... ive-system
...you can see the base layout. It appears that LHD vs RHD cars have the bus bars rotated so that, respectively, MG2 (motor) connections are duplicated on the right, and MG1 (generator) connections are duplicated on the left - you can see the effect in the second picture of this post:
viewtopic.php?f=14&t=205&start=110#p4501
Should be adaptable by bending the bus bar and attaching custom adapter brackets. These cross-connecting bars don't appear to be shown anywhere, unfortunately... maybe I'll tear down my blown IGBT board for educational pics... and, well, to show you the IGBT carnage, of course

Time to sleep.

[Martin1775]

For the heat transfer we use a special metal foil which is covered with a kind of black wax. We get it from the manufacturer of the rectifiers, there is no name on it, but I think it is the fischer WFQ 25. From the factory when new, the igbt modules were installed with a melting thermal paste that then solidified. For installation of it the module have to run at maximum power for at least half an hour, but for repair we use the foil, it is better to handle.
We use the for general applications the fischer WLP and if we had no foil it works as well.

[Lwerewolf]

Pump up the jam

Yeah, the grease totally hasn't become translucent and hasn't been pumped out of the hot regions...



More incoming, time to regrease & assemble.

Using electrolube htsp, some thin rubber roller, 2.5-25nm torque wrench (clicktorque a5), tqc sp4000. Torque values from the rx400h TSB. Ambient temperature is 25 degrees celsius. Target TIM thickness - 125um.

[Lwerewolf]

Initial results - no change, or at least - still throttling under the mentioned conditions, no improvements after multiple runs. I'm waiting for some tools to ease the inverter removal/installation procedure, then I'll check if I've messed something up during paste application - visibly too much/too little/etc. Have some other thoughts in the meantime.

[xp677]

I've owned GS's with over 200,000 miles, as old as 13 years. Never had an issue with inverter temperatures in any of them.

The grease is just there to fill voids in the mating surfaces, it doesn't just disappear, it will be fine with the original stuff.

[Kevin Sharpe]

The grease is just there to fill voids in the mating surfaces, it doesn't just disappear, it will be fine with the original stuff.

Agreed.

I'm waiting for some tools to ease the inverter removal/installation procedure, then I'll check if I've messed something up during paste application - visibly too much/too little/etc.

I don't think the thermal paste is the issue.

Can I suggest you fit thermocouples on both the IGBT heatsink and the chill plate to measure the temperature differential? I suspect this will show the problem is not related to the interface between the two.

One thought, have you verified that the chill plate cooling channels are clear of deposits and don't have any air locks?

[Lwerewolf]

I agree - the thermal interface should be fine, and definitely shouldn't be causing such a spike in temperatures, even if it weren't. Still, I saw thermal grease, and immediately decided that it was worth a shot

I can disassemble the chill plate, but I don't have the correct RTV (three bond's "long life coolant" resistant one - toyota exclusive, >$100 per tube, and it does seem to matter - if you look at the clublexus thread on the 2UR-GSE engine's valley plate leaks, a lot developed leaks after a service campaign related to a... heat exchanger(or was it ?) due to dealerships using the regular RTV). At any rate - I've torn down the one that I originally blew up, and... there's no way in hell that these passages are getting clogged. I'll upload some pictures later, but they are anything but tiny.

As for air locks - might be a possibility, I haven't used a vacuum fill device.

At any rate, next idea - smoothing capacitor degradation. The inverter essentially operates as an indirect AC/AC converter, and it just so happens to throttle at full power when one of the motors (mg1, acting as a generator in this case) is locked at 12400rpm (until the ICE can reach 6400RPM - then it slows down, around 140kph or so), and the other motor is at ~2500-5000RPM.

I have a lot more to read (and basic concepts to grasp - enthusiastic newbie here ), but I think checking the smoothing capacitor bank next might be a good idea. I also have a feeling that since the inverter won't really be used as an AC-AC converter in a full-on EV project (unless someone decides to utilize it as a CVT for another electric motor, for some reason), this won't be that much of an issue for that use-case

Ordering a DER DE-5000 tonight.

[Isaac96]

At any rate, next idea - smoothing capacitor degradation. The inverter essentially operates as an indirect AC/AC converter, and it just so happens to throttle at full power when one of the motors (mg1, acting as a generator in this case) is locked at 12400rpm (until the ICE can reach 6400RPM - then it slows down, around 140kph or so), and the other motor is at ~2500-5000RPM.

That's a super interesting idea!
How would the capacitor affect power output? Possibly voltage spikes are causing overvoltage events?

-Isaac

[xp677]

There shouldn't be a need to disassemble the unit to clean inside the cold plate, just mix up some hot water with washing soda (sodium carbonate), pour it in, swish it round, pump it through, forwards, backwards, however you like, let it sit for a bit (but not too long), and flush it out a few times with distilled water.

[Lwerewolf]

Still haven't attempted to flush the cold plate with soda, wanted to do all that with the inverter out of the car for measuring the caps, but alas - the DE-5000 hasn't arrived yet - managed to cancel the order as it was stuck on "payment pending" (i.e. "we've taken your money but it's still pending for whatever reason"). Reordered from a different seller 2 days ago - we'll see. I still have doubts that this is a potential issue, though - the one that I originally blew had literally been sitting outside for two years, and the water passages were free from any corrosion, blockage or contamination when I opened them up - they weren't plugged.

I've been eyeing ls600h inverters for awhile. The gs450h and ls600h appear to be almost identical - wouldn't be surprised if the command protocols are compatible as well, I've read the relevant threads here and on the diyelectriccar forum. The wiring diagram is basically the same, the connectors are the same (unobtainium ) as well. Internals differ, attachment points differ (easily solvable), the 3-phase connectors differ (most annoying bit - have to take out the engine and transmission for a clean install). Most important bit - no recorded instance (that I've read about - I've been looking) of a toyota inverter with double-side cooled IGBTs ever failing. This includes the ls600h, rx450h, is/gs/rc300h, 4th gen gs450h, 4th gen prius. Plenty of recorded cases for the rx400h, third gen gs450h, prius 2, especially prius 3 & the derivatives. From what I've read on double-side cooled IGBT assemblies, they shouldn't really suffer from solder thermal fatigue (since there is none), and there are no bond wires to worry about either. At any rate, this is just me pondering, the "lack of failed units" bit seems quite solid.

[Lwerewolf]

Well, it's not the caps, as expected - at least judging by what I've been able to measure with the DE5000 at 120hz. Capacitance and dissipation factor seem to be practically what ORNL posted in their teardown articles - that's for all three HV capacitor banks. At this point it's either an attempt at fitting an ls600h inverter, or getting the freshest gs450h/crown inverter that I can find - probably a crown gws204 variant from Japan - and then converting the IPM to LHD. I'm definitely changing the final drive (well, rear diff) to the US/Japan-spec 3.769 as well, or perhaps to something even higher - more power at usable speeds and... other reasons. This car supposedly does 275kph on the autobahn (limited at 250.. yeah, right), a reduction to 238kph still puts it in the beyond silly category - can't imagine it hitting these speeds at any track day event (other than the nurburgring straight or something similar), let alone on public roads.

[trunks9us]

Have apart my inverter atm to do wiring upgrades. Does anyone happen to know what metal material the heat sink is made out of on a Prius 3rd Like is it aluminum or nickel etc. i am considering using liquid metal since the heat sink isn’t touch any thing and it’s all metal but I can’t use it if it’s aluminum from the gallium tearing it down. Liquid metal usually decreases temps by about 10 degrees. I was also gonna mod some notcua fans inside pcb board areas to. Being that there is 12v connections it wbt be difficult to make happen as long as I can find points to make it fit.

[AdvancedEnergySys]

Lapping the contact surface of the IGBT copper base plate and mounting surface of inverter housing would give you the best results. I believe warping of the inverter housing over time would cause the igbts to separate and any uneven heat would more than likely be the result of that and not bad thermal paste.