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

[thingstodo]

Originally Posted by MPaulHolmes

So with a little data you can get a ball park figure!

The motors that I have data for are closer to the MGR voltage than anything I'll be putting in a car. I wonder if any of this information is useful?

If you find a test motor, let me know the rated current and I will see if I have a motor in the same range. It could give some starting numbers?

When I get my controller, I will match up one of the 75 HP motors from the day job (30 HP @ 208, should be about 75 HP @ 575). The oldest and least efficient one that I can find - my 30 HP surplus test motor is about 20 years old.

[MPaulHolmes]

So, following Edward Cheeseman's lead, a guess for the rotor time constant of my motor is as follows:

Zmotor = Rmotor + i*(2*pi*freq*Lmotor).
Now, ||Zmotor|| = 230Vrms / 20.5Amp = 11.2195 Ohm. (230Vac and 20.5Amp is from the nameplate.)

The power factor is 0.78, which is also from the nameplate. So, Rmotor = ||Zmotor|| * 0.78 = 11.2195 Ohm *0.78 = 8.7512 Ohm. The frequency on the nameplate is 50Hz.

If anybody remembers imaginary numbers in math class,
Z = a + b*i means
||Z|| = sqrt(a^2 + b^2)

So,
||Zmotor|| = sqrt(Rmotor^2 + (2*pi*freq*Lmotor)^2)
11.2195Ohm = sqrt((8.7512Ohm)^2 + (2*pi*50Hz*Lmotor)^2)

**mathy mathy math math**

Lmotor = 0.02431Henry

**It was assumed that Lrotor = 1/2 * Lmotor **

Lrotor =1/2 * Lmotor = 1/2 * 0.02431 Henry = 0.01216 Henry

************************************************** ****
To find Rrotor, Edward did the following:
Assume the motor runs at 92.5% efficiency at full load. The motor is rated at 5.5kW at 230V. The full load current is 20.5Amp. So,

PowerLossTotal = (1-0.925)*5500Watt = 412.5Watt
PowerLossPerPhase = PowerLossTotal / 3 = 137.5Watt
PowerLossRotor = PowerLossPerPhase / 3 = 45.83Watt **It was assumed that the rotor loss was 1/3 of total losses.**

PowerLossRotor = currentRotor*currentRotor*Rrotor, So
45.83W = 20.5Amp * 20.5Amp * Rrotor
So, Rrotor = 0.10906Ohm
************************************************** *****
So, the almighty rotor time constant is Lrotor / Rrotor = 0.111 seconds in my case!

[thingstodo]

Originally Posted by MPaulHolmes

230Vac and 20.5Amp is from the nameplate

The power factor is 0.78, which is also from the nameplate

That's a bit lower power factor than my old beast. I'll see if I can find a motor that's around 20.5 amps and 0.78 power factor. Just to see what the VFD reports for the R1, R2, l1, L2 and X

[e*clipse]

Paul, just wondering - I can be no help here with the ACIM magic.

Did my explanation make any sense regarding the resolvers? Would you have any interest in using the MGR as a test subject?

If so, I could help with resolver code. I could also lend you one. (either a resolver or the whole MGR) I mean, the only difference between a system that uses an encoder and one that uses a resolver may be a bit of code and a small daughter board.

- Thoughts?

- E*clipse

[MPaulHolmes]

Yes, it sure did. I read through the PDF you posted, and see what you mean about the sin and cos envelope. I'm not sure of how to extract the peaks. I was reading up on that, and it seems non-trivial. I think that resolver to encoder chip sounds like a good idea. The board could be really little and cheap. I would very much appreciate borrowing one of those MGRs. It wouldn't need to be this week or next week, as I would have to make a board to deal with the resolver issue. Can you calculate what shipping would be?

OK! I searched for the resolver to encoder chip, and I can't find your link or p/n anywhere. Could you repost it?

edit: I found an envelope detector which seems simple. It involves a capacitor though. Care would have to be taken so that it doesn't cause too much lag in the detection.

edit again: It could be made to work, but the chip would be better.

[e*clipse]

So far I've found two chips that could do the job. Neither of them are cheap.

1) Tamagawa AU6802N1 "Smartcoder" resolver to digital
2) Analog Devices AD2S1210 10-bit R/D converter with reference oscillator

Both are made for vehicle reliability and temperature ranges

I guess either method will require a small daughter board. Both IC's are high pin count (48) surface mount chips that will require a bit of support circuitry. Making a sine wave generator and envelope filter will require a bunch of analog stuff.

I've gotten most of the way through a envelope filter and sine wave generator, if you'd like.

Other than UPS, can you think of a cheap method of shipping a 100lb thingeee?

E*clipse

Originally Posted by MPaulHolmes

Yes, it sure did. I read through the PDF you posted, and see what you mean about the sin and cos envelope. I'm not sure of how to extract the peaks. I was reading up on that, and it seems non-trivial. I think that resolver to encoder chip sounds like a good idea. The board could be really little and cheap. I would very much appreciate borrowing one of those MGRs. It wouldn't need to be this week or next week, as I would have to make a board to deal with the resolver issue. Can you calculate what shipping would be?

OK! I searched for the resolver to encoder chip, and I can't find your link or p/n anywhere. Could you repost it?

edit: I found an envelope detector which seems simple. It involves a capacitor though. Care would have to be taken so that it doesn't cause too much lag in the detection.

edit again: It could be made to work, but the chip would be better.

[z_power]

jddcircuit drom diyelectriccar forums is a guru on resolver reading; quite a lot of info in his thread: Toyota IPM Motor Controller Design Details - DIY Electric Car Forums

[MPaulHolmes]

Wow e*clipse, I didn't know you had been so busy solving that problem. Whatever you think is best sounds good to me. I'm going to check out jddcircuit too. Thanks, z_power!

[e*clipse]

Well, um... I don't know if it's that good. I've put it on a back burner; I'll need to revive things and remind myself what I was doing.

One thing you may notice from the DIYEV thread is how just about every 'Yota hybrid uses these things. They're really simple and robust - to me that's a nice reason to use the technology. Oh, it's also well proven on CNC equipment.

One thing that may be helpful is to get you something that plugs into the encoder ports. This thing would put out a sine wave and (with a pair of envelope filters, return a sine and cosine wave that have a center of 2.5V and minimum of 0 (or 0.5V) and a maximum of 5V (or 4.5V) - thus the signal would be very similar to those Honeywell rotary position sensors. I could prolly make a quick n' dirty proto board thing that can be cleaned up later into the daughter board we're talking about.

I can focus on that without getting too distracted with the code necessary to determine position from the AtoD conversion numbers. If for some reason I get ahead of schedule, I can work on some PIC code that may be useful.

Um, I did check shipping costs on a motor, and UPS is one of the best options at about $100. I'm wondering if there is a Pick n' Pull in the Phoenix area that has a Highlander Hybrid or Lexus RX400h??

- E*clipse

Originally Posted by MPaulHolmes

Wow e*clipse, I didn't know you had been so busy solving that problem. Whatever you think is best sounds good to me. I'm going to check out jddcircuit too. Thanks, z_power!

[MPaulHolmes]

Having 2 waves, centered about 2.5v, 90 degrees out of phase would solve the problem completely for me. If you want to focus on that, it would be all I need. They don't even need to be perfect sine and cos waves. Just garbage such that they are the same waves every time, and (sin(theta1),cos(theta1)) == (sin(theta2),cos(theta2)) if and only if theta1 == theta2. Then we could do up a "asin" lookup table which would reduce the angle to 2 possibilities, and then look at cos(), and go with the angle that's closest to the cos() value. badabing badaboom. There's plenty of memory on the dspic, so a big lookup table for lots of resolution wouldn't be a problem. It would be one array access, and then an if..then, and we'd know the angle. That would be fast at 30MIPS.