Tesla Large Drive Unit (LDU) Motor Teardown and maintenance

[2oldteslas]

If anyone is interested here are the dimensions of my home made from scrap iron press to remove the rotor from the case. Inside dimensions: width = 12.75", length = 28", hole for rotor shaft and bearing = 3.25", height of bottle jack cradle = 2.25", length of cradle =7". This press was designed to be used horizontally on the floor or a bench. The LDU must be turned so the gearbox faces up. Shims can be added to the case end to prevent the windings from contacting the face of the press. The height of the bottle jack centerline is matched to the 5.5" centerline of the rotor splined shaft. Your jack height may vary depending on what jack you use. It takes very little force from the press compared to a hammer to remove the rotor and it is much more controlled. I have not yet tested the press to remove the inner rotor ceramic bearing but I suspect it will be just as useful once the pinion shaft is removed.

[muehlpower]

Disassembly Notes:
I cut an M10 thread into the through hole and screw in a screw. Then I press from the opposite side with an M8 screw. For blind holes I use self-made push-out screws. Also works with the case halves. The last picture shows how I push the rotor out. No hammer needed and won't destroy paint on nicely made engines.

[jrbe]

Disassembly Notes:
I cut an M10 thread into the through hole and screw in a screw. Then I press from the opposite side with an M8 screw. For blind holes I use self-made push-out screws.

Are the self-made push out screws dog pointed bolts or the threaded collar ones you showed?

Could also probably be made like this but likely stepped down more than this image shows..

That's an old Audi I5 timing belt idler "puller".

It's too bad Tesla didn't do this from the factory. Nice solution muehlpower!

[muehlpower]

I use Allen screws DIN 912. With your one-piece design, the depth of the blind hole is important, with the two-piece design, the large screw can be screwed in completely and the small screw can be adjusted or simply be very long.

[howardc64]

DELETED

[Boxster EV]

Do you have the correct coolant? Incorrect coolant or mixture will certainly corrode steel components.

I’m still on my SKF seal BTW.

[howardc64]

DELETED

[Boxster EV]

Interesting idea. I used Valvoline G48 concentrate (mixed down to 50/50 with purified water - not distilled) Probably only 2 qt is added from what was lost. G48 is the spec and what everyone seems to use.

IMG_0649.jpeg

But certainly by looking at the rust particle pattern. Downside of a multi-lipped seal is if any particles gets trapped in its interior chamber between the lips will stay there and continue to damage the shaft surface. I wonder if this is the reason Tesla went away from it? Looking at custom EV builder's effort, any LDUs that has sat idle for a long time seems to have developd a rusty coolant shaft surface. So clearly, shaft surface doesn't even like sitting in the coolant for long idle periods.

I guess another potential is the seal cage itself but I'd imagine thats stainless.

@dzolotnyuk ( post link viewtopic.php?p=46191#p46191 ) is also running the SKF FKM seal I sent him ( HMSA10 V I believe with excluder lip) with speedi-sleeve. He is probably at 10k miles now. Didn't report any problems but I'm not sure he checks speed sensor all that often. Did you use HMSA10 V or HMSAV5 (no excluder lip)?

Also what did you use to seal between speedi-sleeve and shaft?

Thanks

This is the seal I used:

https://simplybearings.co.uk/shop/produ ... CCEALw_wcB

It did leak at high RPM at circa 3000k miles however I had an issue with my cooling loop (damaged / distorted hose) so I think the pump pressure was pushing too hard on the seal. It does still weep a little but hardly anything since. I’m on about 15,000 miles now.

I used G48 coolant ready mixed (not concentrate).

I just used a thin layer of reizosil gasket maker under the speedy sleeve.

[jrbe]

At a loss on where this rust could have come from. I guess will be seeking either spray metal coolant shaft repair or SKF speedi-sleeve with with SKF FKM seal or a PTFE coated V lipped seal.

PTFE doesn't absorb much moisture, that's likely out.

The multiple lips must have trapped moisture and maybe contaminants between them.

Did you clean it with chlorinated brake clean or other spray before assembly? The cooling effect from the spray can cause condensation to form on the surface. May have been chemical residue too that got trapped between the lips.

I don't think a multiple lip seal is a good idea here. If one is used you could consider a quality oil or grease (PTFE compatible) between the lips.

[asavage]

The thing I don't like about the trapped-moisture-between-lips-caused-rust-there theory is that there is a finite amount of oxygen that can be trapped, and rusting should cease once that oxygen has been depleted by . . . oxidation.

The amount of rust I see seems inconsistent with trapped moisture . . . but the rust got there somehow (formed somehow) and I have no theory on how.

[howardc64]

DELETED

[howardc64]

DELETED

[jrbe]

The thing I don't like about the trapped-moisture-between-lips-caused-rust-there theory is that there is a finite amount of oxygen that can be trapped, and rusting should cease once that oxygen has been depleted by . . . oxidation.

The amount of rust I see seems inconsistent with trapped moisture . . . but the rust got there somehow (formed somehow) and I have no theory on how.

Agreed. But if someone hit a shaft with brake clean, didn't notice moisture build up from the cooled down surface that attracted condensation, then quickly assembled the seal they could have trapped water and chlorine between the lips (why I asked.)

There's the questions of how much was trapped at assembly, how much was trapped from break in, then how much from normal / abnormal weepage. Howard64 mentioned his prep steps,

After sanding the seal shaft surface with 600->1500 grit, I cleaned it with acetone. Probably another few weeks went by before I assembled and the surface was perfectly clean and shiny at assembly (I recall cleaning it with acetone again probably an hour before seal assembly) Didn't use any kind of lube on assembly.

so likely only humidity was trapped at assembly. That leaves normal weepage and break in leaks. Break in leaks can be a surprising amount and should not be overlooked. It's a big part of why I don't think a multi lip seal is ideal or a long term solution.

Heat signature inside coolant tube where Seal rides on the outside.

The white stuff inside may have formed or dissolved somewhat from the localized seal friction / heat.

Glysantin® G48® is an engine coolant concentrate based on ethylene glycol that needs to be diluted with water before use. Glysantin® G48® contains a corrosion inhibitor package based on salts of organic acids and silicates (Hybrid Coolant). Glysantin® G48® is free of nitrites, amines and phosphates.

White stuff is likely silicates or salts.

New seals need some time to wear in. Even when broken in they can leak slightly. Either / both would mean trapped coolant between the multiple lips. These seals are spinning very fast and are large for the speeds they run at (very high surface speeds.) Could have boiled & broken down any coolant that was trapped between the lips and left gnarly, decomposed stuff behind to attack the steel. Add to this theory weepage of coolant through the seal "refilling" the process.

There is heat signature inside of the coolant shaft where the PTFE seal rides. All of these shaft will show a whitish coating inside the hallow shaft. There is a band inside where PTFE seal rides on the outside. Don't think its so much heat the steel blued. Rather, probably just heat so the whitish coating doesn't form. Note the rust has been mostly cleaned off in the pic.

Steel will blue at temperatures that both should have melted the seal, boiled, and broken down the coolant. Might be a leftover from heat treatment. Coolant breakdown chemicals could be responsible for the bluing. I do think there's a strong possibility heat had a role in breaking down the trapped (leaked) coolant between the lips.

Did you use a vacuum to fill the system? I wonder if there's a bunch of air pockets in these areas.

IMG_2802.jpg

Leak will enter stator even with coolant manifold drain

My outer ceramic bearing was stuck to the end plate. The rusty liquid leaking from the multi-lipped seal chamber got behind the o-ring in the bearing bore (at the lowest spot) rusting the bearing in place. At this point there is no blockage to the stator.

For the the added drain hole, is there a way for air to enter above a leak to avoid vacuum holding any leaked coolant in there?
Also, the drain hole should be large enough in ID to avoid meaningful capillary action - holding the liquid in.

[howardc64]

DELETED

[jrbe]

Removed Protective Surface Finish Possible Source?


Anyway, just a possible explanation. I would guess a single lipped seal likely wouldn't see this problem even if shaft finish is gone. Wet side has ample coolant to cool. Dry side leaks migrate quickly away from the seal. Multi-lipped will trap leaks in the chamber under constant high heat.

The outer edge of the motor shaft looks ok (no rust.) I don't think it would be a problem if it's the right surface finish with good coolant and no traps.
It is possible to have the surface finish too smooth to be ideal for the seal.

I don't doubt the reman quality isn't good. Everyone is battling for low cost and usually end up cutting necessary corners to achieve their cost targets.

(vacuum coolant system before fill)
Didn't do this because Tesla's multi pump + valve design makes it a mystery on how to bleed the coolant system without their security locked diag tool.

That's what's really nice about using vacuum on a coolant system to fill it. All the air is gone, you can check for leaks before filling it. Then if it doesn't leak air in while in vacuum, it uses the vacuum in the system to suck the coolant in and leaves no trapped air. As long as there's a single path through the coolant it will get where it needs to be.
There is a danger of the vacuum destroying something though (usually about to go anyways though.) I'm not sure if it's an issue to use on Teslas or not.

Drain hole

This is on the dry side and has probably 2+mm ID with a clear drain line and see rusty residue in it. So pretty sure its draining. However, it appears the likely thicker rusty liquid doesn't drain as easily.

A few things here. A 2mm drain can cause capillary action that can hold a shallow puddle of coolant up. Also, where the drain hole is, it looks like the centering lip of the water manifold blocks a lot of the hole which could be more of a restriction, might kick off the start of the puddle.

Might be worth doing an experiment with only the motor cover and coolant manifold bolted together with the drain assembly in place and oriented as it would be in the car. Add coolant where it should drain (dry side as you call it) and see how much you need to add until it dumps. I think the results will be surprising (more liquid than you'd expect.) This also isn't including vacuum in there that could further up the level until it dumps.

On the question if any air can enter above leak. It would have to go through the seal. I suppose air is even thinner than coolant and since we tapped a drain port, ambient air pressure will enter the reluctor chamber (I didn't seal the end of drain line, maybe should have? This is probably should of O2 as @asavage was wondering where it came from). If any vacuum pressure exist on the wet side (when coolant is cooled?), seems like air can get pulled in.

The air I asked about was to allow coolant to drain out. Meaning that trigger wheel void area is able to easily suck air in from above the coolant that you want to drain. If the water manifold is siliconed in, it likely can't get air in the dry side unless it shares a breather / passageway. The vacuum of the coolant trying to drain can also hold the coolant up further, especially with a small drain ID.
The other trapped air was in the cooling system which could allow hot spots.
You want the drain line open or the leak can't escape.

The HMSA10 V (SKF) seal is a fluoro rubber and has a 3480 rpm rating (surprisingly low because of the large diameter / high surface speed.) There's likely some safety /wiggle room above that but it's well above is rated rpm here. It might work well for a while but I doubt it would do 80k+ or survive many high rpm cycles.

muehlpower's seal looks like a great option to me.

[asavage]

The air I asked about was to allow coolant to drain out. Meaning that trigger wheel void area is able to easily suck air in from above the coolant that you want to drain. If the water manifold is siliconed in, it likely can't get air in the dry side unless it shares a breather / passageway.

The assembly has a vent, though it's on the inverter side. Howard didn't seal (RTV or other compound) the inverter-to-stator passage, so the dry side has a path to the outside via the case vent, and no vacuum is formed in the area where the rotor shaft leaks.

Johan did a great job of explaining the venting:

Tesla LDU Venting

[asavage]

Deciding to leave the retaining compound out isn't a great idea. The aluminum case expands much more than the steel in the bearing as they heat up. It can get loose / spin if you skip it. In fact, heating the case is the easy way to get stuck bearings out typically.

To clarify again, Howard found what appeared to be retaining compound in the counterbore's corners, and I advised him to not intentionally place compound there, as it does/would do nothing productive there.

viewtopic.php?p=56822#p56822

[asavage]

As I have already written, I am in negotiations with a seal manufacturer in Germany. He sent me an interesting draft. I have pointed out that a Speedi Sleeve may be used. He also takes into account the slightly larger diameter and the limited length.
He also has reservations about using Loctite because of the poor heat dissipation. I did some googling and came across this glue.
https://www.sepa-europe.com/wp-content/ ... 000010.pdf

From that PDF, the HERNON 746 SET THERMAL CONDUCTICE [sic] ADHESIVE requires to not be mixed, but one component ("adhesive") placed on one workpiece, and the other component ("hardener") placed on the other workpiece. They are not allowed to be mixed, and when they are placed together, the working time is 15-30 seconds.

I cannot imagine an assembly method for installing the Speedi-Sleeve where the sliding action of installation leaves any significant amount of mixed compound where it is needed. Whatever component is placed on the rotor will be wiped off toward the rotor middle, and all of the component that is placed inside the sleeve will similarly be wiped off on the rotor end.

While a two-component thermally conductive adhesive may work well, I can't think of using one that doesn't allow pre-mixing.

Loctite/Bergquist have several competing products of this type, but I haven't waded through their datasheets adequately to comment on suitability for this purpose.

[howardc64]

DELETED

[asavage]

Fried IGBTs
Finally, I've seen this LDU->1970s American Muscle car build that fired an IGBT. Author claims can happen when power drops out while driving. I'm assuming this uses non Tesla control board that most custom EV builders are using.

Opening the HV power circuit on an induction motor whilst the unit is drawing any significant current and while it is spinning will fry the IGBTs in the inverter. It happens frequently, and we hear a story about this every month or so here. "Opening the HV power circuit" could mean hitting an E-stop switch, an unintended opening of the main battery pack contactors, or a fuse blowing. They all come to the same end: the rotating AC field and IGBTs are doing a delicate dance with lots of energy moving, and when the path to the pack is removed, the inverter's DC caps can't handle the power in play, they let go, and then the voltage spikes take out the IGBT gates and it's all over.

I don't know what the situation is with a "stock" LDU when this happens.

[howardc64]

DELETED

[howardc64]

DELETED

[asavage]

When you're installing a Speedi-Sleeve, it's to correct surface imperfections that would cause a seal to not work well. So I assume that the sleeve will not have great contact with the shaft, and will have the worst contact in exactly the area where you would want the most contact!

I'm not saying that all of any adhesive would be wiped off, only that this particular two-part-without-premixing product . . . can't work here. We need a product that has adhesive properties but also thermal transfer properties, and one that can be applied ready-to-go on the shaft, prior to pressing on the Speedi-Sleeve. The leading edge of the sleeve will push some un-mixed component out of the area it's needed, and the rotor's end will push almost all of it's component out of the sleeve. Never will the twain mix.

Put another way (beating this to death), whatever compound we apply needs to be applied to the shaft, not the sleeve, because the shaft has the grooves/divots/below surface places that will hold the compound. The sleeve doesn't, and anything you apply to the ID of the sleeve is useless, because it will all be pushed off during assembly.

The Hernon 746 can't be pre-mixed, and due to the installation method required for the Speedi-Sleeve, it won't mix during assembly; therefore, both its thermal conductive and adhesive properties would be either non-existent or sub-optimal.

[howardc64]

DELETED

[jrbe]

To clarify again, Howard found what appeared to be retaining compound in the counterbore's corners, and I advised him to not intentionally place compound there, as it does/would do nothing productive there.

viewtopic.php?p=56822#p56822

As far as I know, DIY rebuilders are just repeating Tesla's 1/2-1mm red locktite at the edge of the outer race. It would be good to have a easy foolproof way of inserting this bearing. Probably best to heat the counter bore casing which requires experience AND pulling the inverter (easy to mess up the coolant o-ring on reassembly)

There are wicking grade retaining compounds. Might be what some are seeing in odd spots on bearing bores.

Another concern for speedi-sleeve is how to avoid scoring the PTFE seal surface during seal+manifold install as the lip slides over the sharp sleeve edge.
...snip...
GFD's PTFE seal will have this same problem as any other PTFE seal. What is necessary is probably using a filler to bevel that sharp speedi sleeve edge and polish it down smooth. But will be challenging since don't want to change the surface of the speedi-sleeve right behind the edge. Probably best done on a lathe with tight control?

Could spin the rotor and use an angled dremel spinning the right way with a soft wheel to add a lead in. A lathe would be nice too.

When you're installing a Speedi-Sleeve, it's to correct surface imperfections that would cause a seal to not work well. So I assume that the sleeve will not have great contact with the shaft, and will have the worst contact in exactly the area where you would want the most contact!

I'm not saying that all of any adhesive would be wiped off, only that this particular two-part-without-premixing product . . . can't work here. We need a product that has adhesive properties but also thermal transfer properties, and one that can be applied ready-to-go on the shaft, prior to pressing on the Speedi-Sleeve. The leading edge of the sleeve will push some un-mixed component out of the area it's needed, and the rotor's end will push almost all of it's component out of the sleeve. Never will the twain mix.

Put another way (beating this to death), whatever compound we apply needs to be applied to the shaft, not the sleeve, because the shaft has the grooves/divots/below surface places that will hold the compound. The sleeve doesn't, and anything you apply to the ID of the sleeve is useless, because it will all be pushed off during assembly.

The retaining compound or thermal adhesive should be put on both pieces then cleaned up after assembly. ID & OD application will roll a bead on each leading edge as they slide together. Both gives a chance for it to fill the gaps while assembling together. Its messier but will very likely work better.