Paul and Sabrina's Cheap 3 Phase Inverter (AC Controller) with Field Oriented Control

[MPaulHolmes]

It's so much more fun when there are all these trials and tribulations! You may be able to just cut the bad mosfet out of the circuit and go on like nothing happened. Once it's cut out, just check the resistance from M- bus bar (the one sticking out away from B+ and B-) to B-. HOPefully it will be a big resistance. If it is, then I say, full steam ahead!

I say just keep cutting the legs off the mosfets one at a time until you get large resistance from M- to B-. If you have to remove 5 of them, that still leaves 250amp at 125v of maximum power for a test. Just be careful with the DC throttle. If you cut out half of them, only push the DC throttle to no more than 50% of full throttle. full throttle will still try to command 500amp.

[thingstodo]

Originally Posted by MPaulHolmes

If you have to remove 5 of them, that still leaves 250amp at 125v of maximum power for a test. Just be careful with the DC throttle. If you cut out half of them, only push the DC throttle to no more than 50% of full throttle. full throttle will still try to command 500amp.

Is there no way to limit the output amps from the command line? Perhaps indirectly via battery amps?

[MPaulHolmes]

You can on the other DC contgroller I made. the identical twin to your AC controller. (3 600v 600amp igbts) I built in a programmable "max-motor-amps" variable. With the one you have, I think it could be done by changing the throttle range. Remember the 2 raw throttle values. Something like 409 and 714, where 409 was max throttle? You could change max throttle to a lower number, so that you were tricking the controller into believing it has a 10k pot instead of a 5k pot. Then, 5k (full throttle) would correspond to only 50% of the motor amps. But thats' only if you end up cutting out 5 of them. Let's wait and see how many have to be removed.

[thingstodo]

Observation:

- The main contactor has it's contacts welded together. 0.1 ohms all of the time. I removed it from the circuit. But the interesting part is that the solenoid still makes noise - click - and you can feel it when it turns on and off. But the resistance is 0.1 ohms all of the time. I will likely keep the contactor with an NFG label on it. I'll cut it open later and take pictures, or maybe a teardown video. It's useless as a contactor ...

Re-arrange the wiring.

+125v Battery Pack +--S1 switch S2---- R1   10 Ohm  R2---- AC BUS+
                                        |             |
+125v battery pack                     R1 (contactor) R2

+125v pack ground --------- AC controller ground

+ for pack, S1 and S2 are terminals on switch, R1 and R2 are terminals on pre-charge resistor

So now we are connected as Paul has suggested. I replaced a ring terminal on the precharge resistor to make sure there is good, solid contact. And I have added a meter on the DC bus to show when it is charged. This visual feedback was added since I can't trust the contactor to open!

I've add to my test procedure. I now turn on the meter at the beginning of the test set, verify that the DC bus is low, then turn on the DC breaker. I observe a precharge (it only takes a second or two) then I turn on the 12V to the controller and it does another 5 seconds of precharge, then the main contactor pulls in. More time, hopefully less equipment damage!

I went over the logs and it appears that perhaps 500 was almost too high - file 153 for those of you following along at home. The first sample is 512 at line 5. And the sample drops to 1 at line 13, just 8 samples later. It starts to oscillate a bit, from 33 - 97 counts. Since Paul said it was OK to have -60 or -50 for noise, I'll assume that 33 - 97 is OK for noise.

So I'll just use 500 * 8 = 4000 for Kp. Then I'll go through Ki at 1,2,4,8 ... until it .. umm ... not sure when I stop! I would normally stop increasing Ki when you get a nicely underdamped response (curve drops quickly, then approaches the target of 0 with no overshoot). Maybe I'll just capture and graph, then post them for Paul to review!

File 190 - Kp 4000 Ki 0. Looks good, like file 153. So file 191 will be at Kp 4000 and Ki 1. This looks weird. A bunch of negative numbers. Try Ki 2. Just as weird, but different.

Umm .. but Paul .. help?

[MPaulHolmes]

My goodness. I'm going to doublecheck the code to see what's going on. That is weird.

[MPaulHolmes]

Could you try a P like 2000, and then I of 1? I still need to check the code, and set up astro's AC controller outside to test side by side, but I would be curious to see the affect of a smaller P. It's still very agressive.

[thingstodo]

Originally Posted by MPaulHolmes

Could you try a P like 2000, and then I of 1?

Why, YES I can! It kinda freeked me out so I stopped and posted the data

I still need to check the code, and set up astro's AC controller outside to test side by side, but I would be curious to see the affect of a smaller P. It's still very agressive.

How far do you want me to back off Kp? If I still have weird stuff at 2000, do you want me to try 500 or 150 or something? Or do I back off Kp until I get to 0 closer to sample 20 or 25?

[e*clipse]

Originally Posted by MPaulHolmes

It's so much more fun when there are all these trials and tribulations! You may be able to just cut the bad mosfet out of the circuit and go on like nothing happened. Once it's cut out, just check the resistance from M- bus bar (the one sticking out away from B+ and B-) to B-. HOPefully it will be a big resistance. If it is, then I say, full steam ahead!

I say just keep cutting the legs off the mosfets one at a time until you get large resistance from M- to B-. If you have to remove 5 of them, that still leaves 250amp at 125v of maximum power for a test. Just be careful with the DC throttle. If you cut out half of them, only push the DC throttle to no more than 50% of full throttle. full throttle will still try to command 500amp.

Oh JEEze!
Perhaps I need to get mine going while someone is around with some real courage like that.!

[MPaulHolmes]

I think I found out what was wrong in the code that caused the weirdness. I'm going to go test it, in the 3,000,000degF shop. There are 2 sets of PI values. For kp-id ki-id, and kp-iq ki-iq. I had kp-iq and ki-iq set to weird values. It works to set them to the same as kp-id and kp-iq. If you don't want to wait around for me, type something like this:

kp-id 1000
ki-id 1
kp-iq 1000
ki-iq 1

run-pi-test

So, each time you change the values, change it for both of them. I'll change the code so that it happens automatically. I just wasn't sure if there was ever a situation where kp and ki for Id and Iq needed to be different. It never had to be for the 2 motors I've tested so far.

I'm going to send over another hex file. I also had set Id to 75amp, and Iq to 0amp. It would make more sense for Id to be 0 and Iq to be 75.

[thingstodo]

Originally Posted by e*clipse

Perhaps I need to get mine going while someone is around with some real courage like that.!

In my opinion, it takes no courage to turn on something and have it blow up when you are not expecting it. It takes courage when you EXPECT problems and you STILL turn it on.

We'll see how I do with that part!