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

[MPaulHolmes]

Holy crap, I think after you type (lowercase p)

p 3844
i 62

all our wildest dreams are going to come true. hahaha

By the way, there is a significant dead zone for the 2.5v being "zero throttle", so it doesn't have to be too exact.

I should have just disabled the throttle. Sorry about that!!! I can send a new hex file with the throttle disconnected, so it doesn't get in the way of the rotor and pi test.

[MPaulHolmes]

Definitely make sure the contactor stays closed. If it opens, and you use some power, it could run the capacitor all the way to 0v, and then when the contactor closes again, the capacitor will get hammered with like 10,000amp for an instant.

[MPaulHolmes]

I think we may have another issue to work through. If the motor was spinning, but the list of RPMs after run-rotor-test was all zero, then I think the motor was running sensorlessly. It will still turn without an encoder, but not very fast. It turns slowly, like you describe. So.... I'm starting to think that the controller isn't receiving the encoder ticks. The values that were output from run-rotor-test was a bunch of RPMs, for different rotor time constant values. Then, the idea is to take the rotor time constant that corresponds to the fastest rpm, since it works the best. But all your RPMs were zero, which means there was no ticks coming from the encoder.

[thingstodo]

Originally Posted by MPaulHolmes

Holy crap, I think after you type (lowercase p)

p 3844
i 62

all our wildest dreams are going to come true. hahaha

By the way, there is a significant dead zone for the 2.5v being "zero throttle", so it doesn't have to be too exact.

I should have just disabled the throttle. Sorry about that!!! I can send a new hex file with the throttle disconnected, so it doesn't get in the way of the rotor and pi test.

I have the throttle in the dead zone - it just surprised me a bit.

No need for a new file.

[thingstodo]

Originally Posted by MPaulHolmes

Definitely make sure the contactor stays closed. If it opens, and you use some power, it could run the capacitor all the way to 0v, and then when the contactor closes again, the capacitor will get hammered with like 10,000amp for an instant.

I think that turned out to be low voltage on the 12V battery.

And I agree that it would be .. *BAD* .. if the controller continued on the test. I don't think that happened (or I would have another welded contactor)

I think what happened with the first one is that I left the high voltage on during a hex file load. That cycles power to the outputs a few times ... and if the bus was not high enough, it would have welded the contacts.

A low voltage cut-out may be something to consider in the future. If the controller does not lose power but the contactors drop out ...

[thingstodo]

Originally Posted by MPaulHolmes

I think we may have another issue to work through. If the motor was spinning, but the list of RPMs after run-rotor-test was all zero, then I think the motor was running sensorlessly. It will still turn without an encoder, but not very fast. It turns slowly, like you describe. So.... I'm starting to think that the controller isn't receiving the encoder ticks. The values that were output from run-rotor-test was a bunch of RPMs, for different rotor time constant values. Then, the idea is to take the rotor time constant that corresponds to the fastest rpm, since it works the best. But all your RPMs were zero, which means there was no ticks coming from the encoder.

I don't have the encoder mounted on this setup, just yet.

I guess that's my next task. The plastic mount I had made for the motor shaft to encoder will work ... but I need to fab up a bit of a mount to take the weight of the encoder and some way of shimming things so that they are basically level/aligned.

After the run-rotor-test ... what do I do with the results? I'm sure that there a command to set the rotor constant, but I don't think I have seen it as yet.

BTW - is there a way to list the present settings? If I had done that, I would have caught the typo on k 3884 instead of p 3884.

Sorry I did not respond last night. I was trying to edit video .. load drivers for obsolete phones .. trying to zoom a window of gopro video ... not a successful evening!

[MPaulHolmes]

I can just have the software scan through the rotor test results, and pick the best one, and then save it to EEProm. Also, I can have it automatically save the best PI values too. If we want to get really fancy, we could have the controller turn on, and if there's no valid EEProm data, it could automatically do the PI test and then rotor test, and then save it. That way you wouldn't need to communicate through the serial port. All a person would have to do is, connect the controller to the motor, go have a bowl of cereal, come back, and it's ready to go!

There is a way to list preset settings. I think I had it commented out or something. I'll include that too. What word should it be? Fran had used "config" for the 144v 500amp DC controller. How about "settings"? Or we could allow both to do the same thing.

Also, I'm going to add a new "bad command or file name" (yay dos!) error message when a command doesn't exist.

Well, I think a serial connection is necessary.

I"ve been reading up on sensorless FOC, and it doesn't look that bad. It requires knowledge of the line to line resistance and line to line inductance. Here's a way to compute that:
If you put 100% duty across phase 1 to phase 2 (phase 3 off), then, you can measure

I1 = Vb/R * (1 - e^(-t1*R/L))
I2 = Vb/R * (1 - e^(-t2*R/L))

Where
Vb = battery pack voltage
R = resistance of phase 1 (or 2 or 3)
e = 2.7182818
1 = 1
- = -
^ = exponent
L = inductance of phase1 (or 2 or 3) stator coil.
t1 = time #1 (first measurement time)
t2 = time #2 (2nd measurement time)

then you have 2 equations with 2 unknowns, and you can solve for R and L. But the stator coils aren't really isolated inductors, but primaries of transformers, with the rotor being the secondary. I bet that changes how that works. dang it.

[MPaulHolmes]

Oh heck, I'll add "config" and "settings" and we can go from there. I'll automatically save the best rotor time constant once it's found. The command is
rotor-time-constant-index 13
(or whatever number) For example, if you got these results from the test:

+00000
+00345
+00400
+00402
+00404
+00402
+00400
+00120
+00000

You would type
rotor-time-constant-index 4

Since the 4th one was the best (starting at 0).


Whenever I try to do too many weird things nothing gets done. You are keeping me honest! haha. We'll go for full automation later.

[thingstodo]

Originally Posted by MPaulHolmes

I can just have the software scan through the rotor test results, and pick the best one, and then save it to EEProm. Also, I can have it automatically save the best PI values too. If we want to get really fancy, we could have the controller turn on, and if there's no valid EEProm data, it could automatically do the PI test and then rotor test, and then save it. That way you wouldn't need to communicate through the serial port. All a person would have to do is, connect the controller to the motor, go have a bowl of cereal, come back, and it's ready to go!

Plug it in and go is good for 'eventually' but I'd like to make it work with the command line first

There is a way to list preset settings. I think I had it commented out or something. I'll include that too. What word should it be? Fran had used "config" for the 144v 500amp DC controller. How about "settings"? Or we could allow both to do the same thing.

Also, I'm going to add a new "bad command or file name" (yay dos!) error message when a command doesn't exist.

Good idea - bad command would have shown me that I made a typo!

Well, I think a serial connection is necessary.

Me too

I"ve been reading up on sensorless FOC, and it doesn't look that bad. It requires knowledge of the line to line resistance and line to line inductance. Here's a way to compute that:
If you put 100% duty across phase 1 to phase 2 (phase 3 off), then, you can measure

I1 = Vb/R * (1 - e^(-t1*R/L))
I2 = Vb/R * (1 - e^(-t2*R/L))

Where
Vb = battery pack voltage
R = resistance of phase 1 (or 2 or 3)
e = 2.7182818
1 = 1
- = -
^ = exponent
L = inductance of phase1 (or 2 or 3) stator coil.
t1 = time #1 (first measurement time)
t2 = time #2 (2nd measurement time)

then you have 2 equations with 2 unknowns, and you can solve for R and L. But the stator coils aren't really isolated inductors, but primaries of transformers, with the rotor being the secondary. I bet that changes how that works. dang it.

The primary versus secondary of the transformer problem seems like a pretty big deal. Industrial drives (all of the ones we deal with) do a small amount of rotation during the 'rotating tune'. Perhaps this helps with the coupling between primary and secondary? Take readings L1 to L2, L1 to L3, L2 to L3 ... rotate the rotor a little bit .. take the readings again .. rinse and repeat until you the rotor is back where it started (1/2 turn for 2 pole, 1/4 turn for 4 pole, 1/6 turn for 6 pole, whatever).

Maybe it would make sense to do the whole thing twice? How much room do you have for data in the controller?

The industrial controllers can also do a 'static tune', where they do not rotate the shaft. It is not as accurate, and your starting torque suffers (125% to 150% instead of 200%) but if you can't de-couple your motor from your load (like a compressor, for example) it gets you decent values for control.

BTW - none of these numbers is displayed anywhere.. for any of the vendors. If the tune works, the values are stored. If not, there is a 'fail' message and you try again or use Volts per Hz open loop control.

[thingstodo]

Originally Posted by MPaulHolmes

Oh heck, I'll add "config" and "settings" and we can go from there. I'll automatically save the best rotor time constant once it's found. The command is
rotor-time-constant-index 13
(or whatever number) For example, if you got these results from the test:

+00000
+00345
+00400
+00402
+00404
+00402
+00400
+00120
+00000

You would type
rotor-time-constant-index 4

Since the 4th one was the best (starting at 0).


Whenever I try to do too many weird things nothing gets done. You are keeping me honest! haha. We'll go for full automation later.

I guess I should have read all of your posts before replying

rotor-time-constant-index 4 sounds good .. but what would it show when you type in "config"?