Prius Gen2 Inverter DC Motor Controller

[MattsAwesomeStuff]

I can confirm that the wiki is wrong for mg1 enable as I mentioned above

K, that's on me then.

You are welcome to edit it yourself, but I would like anyone else who was working on this earlier to confirm if possible (i.e. if someone else got it to work this way, then, what we might have is 2 different wirings).

To be extra clear, the part I'll be editing is this, right?



And I am swapping the labels and descriptions for Pin #14 with Pin #20, correct? The wiring colors should stay though, right?

[Ev8]

Yes Matt that’s correct. Please do get extra confirmation on this if you want, as a relative newcomer I don’t think o should change it! I think most guys running a gen2 wouldn’t actually notice the issue as the open inverter adapter board doesn’t enable mg1 it only used mg1 for rectifying mains voltage for charging. As for this dc project we use both mg1 and 2 for drive it becomes quickly apparent.

[JackTerminus]

Ahhhhhh that makes sense, wish i would have known that when i blew MG2 and had to push my truck 5 blocks home.....

I wired up the motor to mg1 in an h bridge with my arduino board and I've got the motor spinning again.... I need to get AAA before I take it out for any testing again. Stay tuned.

but yes. can confirm. pin 20 is MG1 enable.

Its in the same wire bundle as the pwm signal wires for mg1, same way that MG2 is bundled.

[MattsAwesomeStuff]

Yes Matt that’s correct.

Thanks for confirming. Changes made:

https://openinverter.org/wiki/Toyota_Pr ... in_mapping

[RubyRR]

I'm going to be getting myself a second pair of axle stands for testing before I take it out on the road
The brake discs could do with a clean up anyway

[Ev8]

So I have killed my first gen2 today in pursuit of a working dc controller, I had it working and was testing in neutral just spinning up the motor on and off for about 20mims while I played with my code making changes to throttle pedal mapping and implementing averaging of throttle values to smooth out some jerking. Anyway gave it a Rev and just as it started to spool down bang MG2 now reporting failed, oops. And I thought I had got clever and made my setup different enough to stand a chance of working, as others have had failures of the low side igbt I was just using the high side igbts, all of them from mg1 and 2 all paralleled up to act as a single high side switch and then I installed a monster freewheeling diode where the boost inductor normally sits! Well I thought it was a good idea, I have a spare gen 2 igbt block so I will install that and maybe try dropping the frequency down from 8.8khz to 4.4 I’ve got an idea I’m to fast for the diode or something, I don’t know!

[RubyRR]

Thankyou for your sacrifice! Looks like this'll be a hard learnt lesson once it's solved.

Do you have an oscilloscope? I wonder if it could be caught in action

[Ev8]

Unfortunately I’ve not yet invested in a scope, other than a shitty ebay handheld thing! It’s on the wish list, I’m thinking as I’ve not got the low side connected the failure isn’t shoot through, and it not an over current or the Toyota desat hardware would have caught it, so maybe a voltage spike, perhaps the capacitor setup isn’t up to such a large inductive load, it’s a big motor 13” diameter, or maybe parrelling all 6 phases in the way I have is just too un balanced? I’m open to ideas to try before I sacrifice to next brick!

[Ev8]

Interesting it’s mg2 that died which is higher rated but also the shorter path for current, but in neautral I’m only pulling 100amps or so

[RubyRR]

You might know more than me, but I was thinking voltage spike also.

Just a thought: If it is that, and if the rising and falling edges of the PWM waveform are to stay the same then a lower PWM frequency might not make any difference.

What's the failure mode of the IGBTs?

[RubyRR]

I guess the MG2 path would have a higher inductance than MG1

[Ev8]

Still that’s one hell of a spike to kill 1200v igbts from a 300v pack! I’m looking at how many gen2 I’d need to sacrifice before I hit the cost of a decent scope!

[RubyRR]

Reminds me of the video Damien did on his Gen3? inverter after the capacitor exploded

[Ev8]

Looking at the diode specs the reverse recovery takes 75us during which time it’ll pull upto an extra 148amps to basically shut the diode again, at 8.8khz this is nearly half the on duration of 100% pwm, I’m starting to think that a massive freewheeling diode is actually less helpful than it might appear! It’s a good job I don’t do this stuff for work!

[Ev8]

Yeah you can definitely kill a Toyota inverter if you do stupid things to it

[RubyRR]

I think all he did was connect the HV battery at the wrong end of the bus. Correct polarity though.

[SciroccoEV]

I think all he did was connect the HV battery at the wrong end of the bus. Correct polarity though.

Damien had been jumping back and forth between Gen2 and Gen3 inverters and forgot to set a non zero deadtime for the Gen3. The resulting shoot-through currents, effectively short circuits directly across the DC bus, created high voltage spikes that exceeded the voltage rating of the DC link capacitor and caused the capacitor (actually just the small high frequency cap) to fail. The car would still drive!

The speculation about having connected the battery in the wrong physical location was a red herring. The loop area between the IGBTs and the DC link capacitor is what matters, not where the battery wires are connected.

[Ev8]

I wasn’t inferring Damien had done anything stupid with a Prius inverter, more me with stupid ideas about only using the high side switches, now I look into things more the diode is totally unsuitable just too slow reverse recovery probably to slow to the forward state also, and even if it did work it’s not as efficient as a half bridge thanks to the high reverse recovery current, I’m going to pull mine apart later and reconfigure back to normal with spares drop the frequency to 4.4khz and test from there.

[RubyRR]

The speculation about having connected the battery in the wrong physical location was a red herring. The loop area between the IGBTs and the DC link capacitor is what matters, not where the battery wires are connected.

Thanks. I vaguely recall that now you mention it

[Ev8]

Well a little reconfigured today swapped out the power stage for a good one, removed the big diode, sorry for the mess, but I’ve managed a few laps around the block and have hit 150kw according to the bms. No pops or bangs yet.

[Ev8]

Currently switching at 4.4khz, no cooling plumbed in yet just looped back into itself and filled with water but inverter temps not gone over 40c

[JackTerminus]

AHA! sounds like you might have found the right setup! How did you set it up exactly so I can do some additional testing? are you running it in a half bridge? which switches are paralleled? how many volts is your pack? and looks like no boost inductor?

Thanks, can wait to get this working!

[Ev8]

yes running as standard half bridge, all phase outputs paralleled to positive motor terminal, all phases switched together from the same pwm source,
Battery positive to positive on inverter brick , Battery negative to inverter brick negative and motor negative, capacitor modified so all positive and negatives are joined and pick up from the inverter brick terminals (as standard one part of the capacitor is linked to the boost stage only) no boost inductor or brick, I will take some pictures once ive tidyed up my bus bars and removed the legs that went to the diode i dont need.

i am Running 300v nominal 80s BMW PHEV modules (3 packs in parallel)

[Ev8]

Jack did you have your battery input at the standard connection points via the boost converter when you did yours? I am pretty sure thats not a good idea as the boost stage is only rated to 20kw

[JackTerminus]

Yes I did have it running through the boost converter as I had mg1 setup to charge at one point.... I'm running half your voltage at 40s and I've cooked 2 mg2 stages already with them wired up like yours, sounds like the key was dropping the pwm speed? let me grab my scope and see what the Damien card was outputting.