Paul & Sabrina's cheap DIY 144v motor controller

[jyanof]

i've missed a bunch of stuff! that's what i get for not checking the board in a while.

Paul - i've been helping 64jeep a bit with his controller and we did a similar setup for the 3 drivers.

each IGBT has its own VLA-503 driver and DCDC. We swapped out the NAND gate on the control board for a higher current version and use its output to drive the 3 optocoupler inputs to the VLA-503's in parallel (each has its own resistor in series built into the driver). i ended up disabling the desaturation detection on each driver instead of figuring out how to incorporate it into the control logic.

We were also worried about how the timing worked out when driving each IGBT like this. we ended up scoping the outputs of the drivers on a dual input scope and they were all really close, within a few ns or so of each other. a sample size of 1, but we're only building 1 =P

We also didn't optimize the gate resistors, so the switching times might be a little slower than ideal - I thought that might be a good way to accommodate deviations in switching. like, what's a few ns delay out of a few hundred ns total switching time? (or whatever the real times were). so, I forget what gate resistor we have in there now, it might be 5 ohm while the datasheet suggests going down to 3.3 ohm or something. Whatever we started out with, it seemed to work ok so we didn't mess with it.

so, it seems totally doable to have separate drivers. we haven't maxed out the current yet, but it's been good so far to 75% current. if you have have control of each PWM signal down to the counter, that's pretty slick.

[jyanof]

Originally Posted by MPaulHolmes

Nah, 8kHz just seems to be the favorite for high performance controllers.

Just my opinion, but when i tried the 8khz software on my controller, i found it very annoying, moreso than the 1.5khz noise from the Curtis 1231 i used to have. 8khz was quieter, but more piercing (to my ears).

The curtis would drop from 15khz to 1.5khz if the temperature went above about 50C (also if the PWM duty cycle was really low, but that was for a different reason). You might consider dropping the frequency only under certain conditions, like high temperature or current...

[MPaulHolmes]

Are you guys using 16kHz on 64jeeps IGBTs?

[jyanof]

Originally Posted by MPaulHolmes

Are you guys using 16kHz on 64jeeps IGBTs?

yep, it's just the latest version of software for the open revolt. (forgot to mention, we swapped out the current sensor for the HASS 900, so commanding 500A would actually be 833A).

He's done a couple of brisk accelerations while I was watching the RTD data and the coldplate would only rise about 10-15F. Again, we're only at 600A max output for now, and we're water cooling, but so far so good.

[jyanof]

Originally Posted by MPaulHolmes

Are you guys using 16kHz on 64jeeps IGBTs?

oh, and he's at 230V (70 cells), and waste heat would obviously be worse at higher voltages.

[64jeep]

For anyone who’s interested in some of the details of my controller, here’s the build photos:

Porsche 924 Electric Conversion: Motor Controller Part 3: Building and Testing
Porsche 924 Electric Conversion: Motor Controller Part 4: More Testing

The IGBTs were turning on and off within 20 ns of each other. I thought I took pictures of the scope, but they’re not on the blog.

jyanof, it's actually the HASS500 that was put in place of the HASS300. Max current is 500/300 x 500A = 833A without changing the code. I currently have it set at 625 amps.

[MPaulHolmes]

Holy mackerel that was awesome!!! My brother from another mother! Great job!!!!! Very tight packaging. I could mill channels like that. Maybe I should try water cooling at some point. Where does the heat go when the water "takes it away" (sorry thermodynamics teacher!)? Do you use a radiator of some sort?

[64jeep]

I took the old heater core out and run the water through there. I rigged up one of those big computer fans to blow air through it. I've got a picture here at the bottom.

Porsche 924 Electric Conversion: Batteries!

Being in AZ I was concerned about a fish tank pump not working at high temperatures, so I found a pump designed for solar panel cooling that's good up to 80 C. I got a friend with a CNC router to mill out all those channels. If you want to make one just like it, I have the CAD model that you could use to generate the tool path.

[jyanof]

Originally Posted by 64jeep

The IGBTs were turning on and off within 20 ns of each other. I thought I took pictures of the scope, but they’re not on the blog.

jyanof, it's actually the HASS500 that was put in place of the HASS300. Max current is 500/300 x 500A = 833A without changing the code. I currently have it set at 625 amps.

ah ha! thanks for the actual numbers!

[MPaulHolmes]

I used a random sequence generator based on atmospheric noise (so about as random as you can get) to spread the frequency from 6kHz to 10kHz. I found out that the dsPIC can change pwm frequency glitchlessly on the fly no problem, so I'm just going to do the spread spectrum switching. This will be the best of all worlds, except for when you want to put an oscilloscope on it. haha.
These are the maximum PWM duties:
RANDOM.ORG - Sequence Generator

EDIT: Oh crap, the averge period is 1/7.5kHz, since more time is spent on the low end. Now I've got to figure what some bounds should be such that 1/8KHz is the average period. Since I'm fixated on 8KHz. haha

EDIT AGAIN: OK, so max duties from 1600 up to 2400 give an exact average period of 1/8KHz. In terms of frequency, it's 10KHz down to 6666.666KHz.