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

[MPaulHolmes]

Oh heck ya, oscilloscope just arrived. testing to commence after 50,000,000 other family things today. haha. Probably late tonight. This sucker is going to work sooner rather than later. I'm so excited.

Yes, very minor changes to the AC control board. Should it stay completely through-hole?

[Automcdonough]

Through hole is friendlier for DIY stuff, most people won't know how to look for/avoid hairline cracks on those SMT resistors and caps. Plus a good flex in the board can cause that too.. and the larger 2225 packages get really finicky about how you solder them.

That said, with proper mechanical support and soldering it normally isn't an issue. Not only that but the reduced weight of the parts can make it more resilient against vibration. That might be TH's largest weakness unless you pot the board. Also the smaller size helps a lot, even in stuff like this where you normally wouldn't think of space as a huge constraint. With higher frequency stuff it also keeps the traces short and the drive stuff packed densely together with a solid ground plane underneath, shielded in a way that through hole makes impossible. This is all sub 100khz so not really a factor, just rambling on I suppose.

I think a mix is best.. if the board will have mechanical screw connections to the IGBT's or large caps then the board will flex, best to use TH in that area. A sub-board with the micro which is isolated from this stressing and heat would probly be better off SMT.

[dcb]

Through hole is still best if this is headed for a Kit, especially if you are close to a working thing already. I actually did my first smd on an obddiag kit the other day, pain in the butt with "regular" equipment. Any chance this will work with my mars brushless (hall effect PM jobbie)?

Is the fine resolution speed sensor there so you can detect slippage quickly and bump up the "field" power if it has dropped to much?

[MPaulHolmes]

Thanks for the suggestions Automcdonough!

dcb, I think that's it. The torque and flux components are at right angles. If the flux vector you are commanding isn't aligned with the rotor flux vector, I guess it's just a waste of power. I think in that situation, the torque component isn't happening at exactly a right angle, and it's a waste of torque. Like trying to turn a wheel by pushing straight in toward the axle rather than tangent to the wheel.

The real reason I ordered the fine resolution speed sensor was because the application note said I needed rotor velocity to compute rotor flux angle. I didn't really have a clue what any of that meant, so I just ordered it. haha. You can do sensorless, but it's harder. Heck, even AC FOC seems like it won't be easy. Everything about finding the motor characteristics to optimize the FOC equations seems proprietary. No company wants to explain how they do it.

I think it would work fine with a brushless DC. Actually, brushless DC would be easier, since the rotor angle and rotor flux angle would be the same, because of the permanent magnets. And the rotor flux angle is the horror to compute. Also, in that case I_flux is just set to 0, since you get the rotor field for free. All you need to worry about is I_torque. Then for field weakening, you actually set I_flux negative, and it fights against the magnets of the rotor, making them weaker, allowing higher rpm.

[dcb]

hmm, I was thinking for field oriented control as something similiar to field weakining, where the controller is actively looking for (i.e. with a PID) the minimum amount of current needed before slipping occurs. Maybe I'm just still confused

I get the flux angle bit, sorta, no point pulling when you should be pushing, or doing nothing. But with fine grained speed control, I guess you could detect a slowdown more quickly is all I can think of. Perhaps FOC would feel jerky if you had a chunkier speed sensor (i.e. 3 hall sensors)?

[MPaulHolmes]

I still don't really understand it. haha. It sort of makes me happy though. I think once the controller is working, and I see lots of I_torque, I_field, and who knows what all graphed out, it will make a lot of sense.

I LOVE that new oscilloscope! Holy cow! I just put that sucker on each of the 6 pwm outputs, and it made debugging a million times easier!!!!!!!!!!!!!!!

I added a pretty strong (1kOhm) pulldown resistor to the power supply of the 6 HCPL-4506 optocouplers. Now, when I create an artificial overcurrent event, all 6 pwm outputs disable! Before, the power supply was left floating, sort of disconnected from the 5v supply by turning off a p-channel mosfet. The problem was, it was floating at around 2v, which was enough to turn on some of the optocouplers, since some have a better current transfer ratio than others.

So, I think the whole stinking thing works now! I really should check to see that the 6 undervoltage circuits are working. I have every reason to believe that they are.

I could NOT have done this without that oscilloscope. The motor WILL turn by Sunday... That's the rules. I didn't write them.

This is the first time I've ever made a circuit that only had 1 thing wrong with it. hahaha. 1 resistor out of like 290 components or so.

[MPaulHolmes]

low voltage cutout works perfectly! All 6 drivers disabled when 12v supply drops to about 7.9v. The microcontroller is still functioning perfectly though at that input voltage, so it can still make decisions. Hurray! Time to put it back together now. I can't believe I made something that actually works!

Power consumption at 12v is about 620mA, so about 7watts. It's those 6 DC-DC's I think that makes it eat up more power. At 8v, it was using about 850mA.

[dcb]

Nice Job Paul! FYI, I found lots of FOC for dummies pages, and it seems a little more obvious now. current phase != voltage phase DOH! keep the field at 90 degrees. three phase to two phase transformation -> rotation -> corrections (PI) -> unrotate -> 3 phase.

[ElectricZX2]

Amazing job Paul! That's so exciting. Can't wait to see a video of the motor turning Saturday night late....

[Cancerkazoo]

A SMD controller board would be a nice option, if it was single sided.
A toaster oven works wonders if you have a solder mask layer, transparencies and an exact o knife.