Journey to the center of the Roadster UMC

[Panicopticon]

I looked around the forum and didn't find anyone else who had undertaken this teardown, so I decided to give it a go. I'm curious about the header visible here, what the failure mode is, and the general design and operation of the device.

A fellow Roadster owner (I'm unsure if they want to be identified publicly ATM..) shipped me their dead UMC controller, found the box on the porch this morning and immediately set to work...

Given other Tesla Roadster stuff I've taken apart and analyzed I suspect the header is a Pickit ISP, but wont know for sure until later...

Without further ado, the photos.

The case is glued shut, without any screws. The internals are potted in a vulcanizing foam rubber that I've seen in other automotive applications. The short version is that its a royal PITA to take apart. The foam is chemically resistant to everything I have on hand (various alcohols, MEK, acetone, paint thinner, etc) and anything stronger typically damages the boards. If folks have suggestions, I'm certainly interested.

My early guess is that the failure mode in this unit may be the result of poor manufacturing process. Specifically an interaction between the foaming rubber and the components or the solder and flux used in assembly resulting in corrosion, and eventual failure. Of course, this is only an early guess. As this is the only device I've looked at I won't speculate if I think this is the cause of the failure in other units.

Normally I'd do extremely high quality imaging the boards with my scanner, but given the potting this may not be possible...

If folks have specific questions I'll do my best to answer them.

1310 EST: I'm about 2 hours in and I've stopped my disassembly for the moment has my hands are tried -- you have to slowly and carefully cut away the foamed rubber. I'll resume sometime later on and continue to add photos to the link above.

 

[wiztecy]

Here's a sliced & diced one:
Roadster UMC Alternative - Page 3

I hate that UMC, its total junk. I was using mine as a mobile charging solution, kept in the trunk, and had my MC240 as my dedicated home charger. It never failed me. But the cable/connector end, well the clamp failed, and the rubber protector & insulation was pulling away from the connector. I tried to clamp it back down using a hose clamp, total fail. It eventually started cutting into the wire but I removed it before it cut totally through. So with that side I ended up cutting the connector off but its unsalvagable and good for parts only. Sad since that charger was super great. So I resorted to newer UMC for my dedicated charger and bought a Tesla S charger. Very light, very compact and nice for a mobile solution. But it again is junk in quality.

So I lucked out and got a super deal on a 70amp Roadster Clipper Creek connector with the long ass heavy duty cable to the box. CC/Tesla didn't skimp there that's for sure. I'm either going to go and put that into the MC240 to get that up and going again, or pick up a JuiceBox 50+amp charger and run the Roadster Clipper Creek cable so I can utilize its thickness when needed for a higher amp charge and equip it with the wireless connection.

I also have the 70amp HPC but never installed it, I need to upgrade the power feed to the house before that can go in.

 

[Panicopticon]

Interesting, I hadn't seen that thread. Thanks!

I still plan to continue my current approach...

 

[hcsharp]

I looked around the forum and didn't find anyone else who had undertaken this teardown, so I decided to give it a go. I'm curious about the header visible here, what the failure mode is, and the general design and operation of the device.

A fellow Roadster owner (I'm unsure if they want to be identified publicly ATM..) shipped me their dead UMC controller, found the box on the porch this morning and immediately set to work...

Nice work! That potting foam is hard to remove and you've done a better job than I've ever seen anyone do. I'm quite looking forward to following your progress! I have my own theories as to why they fail so often and I will enjoy hearing your analysis.

 

[Panicopticon]

I've already seen some corrosion on the top board (controller?) near what appears to be a electrolytic cap (I don't have the halves separated yet, so this is still a guess). The leads on it are completely gone.

There is also some poor post board assembly cleaning practices on the bottom (relay side) of the board. Tons of left over flux, much of which shows signs of corrosion and contamination. Typically I'd expect that to be cleaned off up before potting.

Still have a ton of work to do. Overall the construction is quite robust, the potting helps significantly...

 

[hcsharp]

I've already seen some corrosion on the top board (controller?) near what appears to be a electrolytic cap (I don't have the halves separated yet, so this is still a guess). The leads on it are completely gone.

There is also some poor post board assembly cleaning practices on the bottom (relay side) of the board. Tons of left over flux, much of which shows signs of corrosion and contamination. Typically I'd expect that to be cleaned off up before potting.

Still have a ton of work to do. Overall the construction is quite robust, the potting helps significantly...

That cap might be part of the high-voltage side of the power supply (for +-12v, 5v). Amazing the leads have corroded completely.

Note to self: clean your boards after assembly.

 

[thefortunes]

It's my old UMC (if you search my posts you'll see I bought an OpenEVSE to replace the controller, but salvaged both ends).

Very curious what the end diagnosis turns out to be.

 

[Panicopticon]

Well. That's some progress. I've located and main relay and have partial separation by following along the top of it. Some part numbers visible, I suspect with a tiny bit of work someone could track down the relay, I suspect its a Potter & Brumfield, looks like they're owned by AMP/TE now.

Board to board interlinks using thin cable, and some evidence of blue wire repairs (possibly from one board to the other? kynar wire that seemed to go from one board to the other...)

Added some more photos.

I'll keep at it.

Boards are separated. Must admit I'm curious about what the coil is up to around the output wiring. Its clearly some sort of sensor, presumably inductive, but its around both cables. I'm used to seeing current clamps just around the hot line. I'm assuming the current sensing is under the can -- sadly the interior of that is also foam potted... sigh.

Yet more photos.

The failed cap is on what I think is the low voltage (control) stage.

 

[Panicopticon]

I have the control board largely cleaned off, haven't taken any photos yet. I believe the failed component is a solid state (tantalum? cap), its very clearly smoked. The corroded part from earlier is actually an inductor, not a cap. As I'd guessed the CPU is a PIC, specifically a 16F676. Some quick testing makes me extremely confident that my earlier guess of the header being the ISP for the CPU is correct. I'll post a proper pinout tomorrow.

 

[hcsharp]

The coil on the output wiring is a current sense for detecting ground fault leakage. That's why the two power wires go through it but not ground or pilot wire. It doesn't measure current going to the car. Normally the current going through both wires will cancel each other out, but not if there is ground leakage. Nice work! Thx for the update.

 

[Panicopticon]

Added some more photos.

Believe the programming header pins out as follows:

(J1) is silkscreened on the board.

| 1 | 2 | 3 | 4 | 5 | (J1)

1 = CPU Pin 12, ICSPCLK
2 = CPU Pin 13, ICSPDAT
3 = CPU Pin 14, VSS
4 = CPU Pin 1, VDD
5 = CPU Pin 4, MCLR

Confident this is correct, as it matches up to the PICIT3 pinout. Just read out the contents with my Pickit3. The lock bits aren't set, readout works just fine. IDA shows a clean disassembly, doesn't look overly interesting.

For completeness, I tore down the Marinco connectors. On the 14-50 side, no ID resistor, just a jumper between purple wire (sense line?) and ground. One surprise, on the Marinco socket side, there is a fuse inline with the purple wire. The fuse is marked:

LF (LittleFuse Logo) F125mAH 250VP

Its a minifuse, nothing too unusual.

Added some photos of the connector internals.