LDU Overcurrent issues [SOLVED]

[Boxster EV]

I'm having a few challenges and have spent the past few evenings fault finding to no avail. To help diagnose, here's what changed in the time leading up the current issue:

1) I updated the esp8266 to show the inverter parameter index (inverter.js, index.js, index.html).
2) I was working with Jon Volk adding some idlemode CAN commands via a Teensy 3.2
3) I had a mechanical overcurrent whereby the wheels shuddered and hit OC (this was expected as the Teensy idlemode code wasnt quite correct).
4) Returned tot he car a little while later and it's now refusing to run.


The issue is that when I pull a direction high, the inverter instantly hits overcurrent in software (doesnt even attempt at spinning the wheels). I can just hear a click from the inverter at exactly the time the OC fault is displayed.

See here:



I've since gone back to firmware v4.90R, but the fault is the same as it was with more recent release

DIN spot values at OC.:

Param file:

params (9).json

(1.38 KiB) Downloaded 93 times

I have removed the CAN set-up to more easily fault find. No change.

I'm trying to do every diagnosis I can before considering pulling the LDU from the car. Would appreciate some suggestions at what else I could test.

[johu]

As soon as you switch to forward the PWM is actually started and will reveal blown IGBTs. Unfortunately on Tesla boards both desat and over current are reported as the same fault.
Looks like the idle experiment did some damage. I suggest next time you experiment lower ocurlim to something like 300A. Thats how I worked on the Leaf inverter.

[Boxster EV]

I was hoping you wouldn’t say that.

Pull the DU for further inspection?

[johu]

Yeah sorry
with that track record it's hard to conclude something else. Pull it.

[Boxster EV]

That’s cool. All part of the fun.

[Boxster EV]

Chatting to Jon about this, it’ll be interesting if the IGBTs are damaged as there was very little load on the DU when this happened (wheels were in the air).

[johu]

It's the transients.
E.g. the Leaf inverter happily pumps out 650A RMS under acceleration. But if I accidentally spin it up in idle, it pulls desat. Luckily it is more robust than the Tesla one.
That said I had produced lots of transients in my Polo development which didn't even have desat detection on the gate drivers. But the conservative current limit of 250A for 400A IGBTs made it virtually indestructible.

[jon volk]

So, if Im following, you're saying that the rate of change in current is the issue rather than the actual load?

[johu]

Not quite, it is still over current that we are seeing. And that is produced by a high change rate of voltage, in either direction. If you modulate frequency or amplitude very fast, so output voltage is no longer sinusoidal but rather jumps between high and low. An oscillating speed controller can cause this. That is why we have an IIR filter and a throttle ramp, but apparently that still isn't always enough.

[jon volk]

Thanks for the explanation. When I went searching for transient currents it was mostly limited to articles about power regulator ICs. There also seems to be some variability between LDUs. On the one that replaced my drive with the blown up differential, I had to back off throttle ramp values to prevent an OC if I whacked the throttle real fast from low speed. Interestingly from a dead stop or higher rpm I could still hit it with a fast ramp with no ill effects apart from mechanical driveline loading.

Paul and I were going back and forth trying to pinpoint what was causing odd behavior prior to the latest fault. The idle changes in question lower the idle throttle value to 0 in relationship to pot 2. In practice this ramps idle throttle value in as you take you foot off the brake for a smooth start/hill holding/parking lot maneuvering without the throttle pedal. If it works incorrectly, the resulting behavior I've observed is the same as normal brake-off idle. Using the same json settings and CAN code I run, Pauls car would exhibit OC faults on gear change, but starting the inverter in F or R would work as intended I believe.

Paul can jump in and correct anything I may have missed.

Also, I thought that a desat fault from the gate driver would show up as a DIN OC input and a software OC would not. Is that incorrect?

[Boxster EV]

Here's a video of my car from last Wednesday evening before the failure.

Note the first part of the video, the CAN mapping is working correctly. POT2 is controlling idlemode and RPM ramping as expected.

In the second part of the video, upon power cycling the board, the input from POT2 is ignored which then dumped a load of current into the DU upon selecting a direction and caused an overcurrent.

[Boxster EV]

In other news, I pulled the DU yesterday afternoon and ran some initial tests. I believe I should be seeing a diode drop on each phase. However on one of them there seems to be a short circuit.

Can someone please confirm that I've tested this correctly?

[tom91]

Yea thats a dead igbt section.

Seen plenty of those on various drive units/inverters

[johu]

What is a sane current limit for the LDU?
I realized I program them with the aggressive "Jon Volk" parameters which have the current limit disabled. Thats ok for Toyota inverters but obviously not for Tesla ones.

[jon volk]

I think that’ll be an interesting conversation. Once upon a time I had played with lowering the ac current limits by a fair amount and it seemingly caused desat faults whereas the -2500 or so would not.

[Boxster EV]

What is a sane current limit for the LDU?
I realized I program them with the aggressive "Jon Volk" parameters which have the current limit disabled. Thats ok for Toyota inverters but obviously not for Tesla ones.

Jon's parameters are also based on CAN mapping and wouldn't be suitable for someone not running that set up.

[Jack Bauer]

Did the contactors drop out at the time of the overcurrent?

[Boxster EV]

Did the contactors drop out at the time of the overcurrent?

No they didn’t.

[jon volk]

[/quote]

Jon's parameters are also based on CAN mapping and wouldn't be suitable for someone not running that set up.
[/quote]

The parameters hes referring to are before I started the CAN mapping. It was more suited to max power and still exhibits some low throttle quirks. The parameters were reasonably safe on my unit, but seeing the variation between drives, perhaps it might be reasonable to back off the slip values 10% and/or raise fweak by the same.

[Jack Bauer]

If its any help I think I have one of Jon's inverter sections left.

[Boxster EV]

Thanks Damien. That's really good of you. I'll send a PM.

[Jack Bauer]

Good news/bad news. Good news is I do indeed have one section left. The other I gave to someone who never used it of course! Anyway bad news is the 8 way jst connector got a knock at some point. I was able to remove it just now with no pad damage and can include a new one. Let me know if that works and I can post tomorrow.

[Boxster EV]

Good news/bad news. Good news is I do indeed have one section left. The other I gave to someone who never used it of course! Anyway bad news is the 8 way jst connector got a knock at some point. I was able to remove it just now with no pad damage and can include a new one. Let me know if that works and I can post tomorrow.

That definitely works for me - thanks a million! Perhaps I can do a little write-up during the changeover that others can follow in future.

[muehlpower]

Thanks Damien. That's really good of you. I'll send a PM.

it looks like the whole gate driver can be exchanged after loosening 3 screws and re-soldering 10 wires!

Sooner or later someone has to deal with the renewal of the IGBTs, I would like to have the destroyed one to experiment with. I pay the postage, of course.

[Boxster EV]

Thanks Damien. That's really good of you. I'll send a PM.

it looks like the whole gate driver can be exchanged after loosening 3 screws and re-soldering 10 wires!

Sooner or later someone has to deal with the renewal of the IGBTs, I would like to have the destroyed one to experiment with. I pay the postage, of course.

I can send you my old bank of IGBTs to experiment with.