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

[thingstodo]

BTW - I have not powered up the new AC boards yet. It looks like it will not happen until the new year.

I've chosen an Allen-Bradley 1333 AC Controller, 5 HP, 230V, that has a bad current sensor for my first AC Controller brain transplant. It looks like I can run the old controller (it runs for about 10 seconds before it faults out on the current sensor) to verify the caps and such are working OK. I do have a couple of the same model that have working current sensors if it makes more sense to try one of them first.

Disconnecting the old controller looks like removing ribbon connectors for the IGBTs and the current sensors

Connecting the new control board looks like plugging in ribbon connectors and one new current sensor. The voltages appear to match. I guess we'll see.

The controller is set up to run from single-phase 240V or 3 phase 208V input. The control board doesn't really care what voltage the DC bus is, right? It assumes a battery bank but it doesn't really CARE. As long as the controller has the isolated 24V input, the output drives IGBTs, and the hall effect current sensors accept 5V, Gnd and the signal comes back proportional to the current measured .... we're good?

[MPaulHolmes]

Originally Posted by thingstodo

The controller is set up to run from single-phase 240V or 3 phase 208V input. The control board doesn't really care what voltage the DC bus is, right? It assumes a battery bank but it doesn't really CARE. As long as the controller has the isolated 24V input, the output drives IGBTs, and the hall effect current sensors accept 5V, Gnd and the signal comes back proportional to the current measured .... we're good?

Yes that's right. There's DC voltage monitoring now, but the only use of it at the moment would be to verify precharge. But I'm not even doing that right this second.

I have NOT added the V/Hz to this code yet. It's a good idea though. It's pretty easy to get working. IT was version 0.001 a few years ago. I just need a lookup table, and to race through the table at a given frequency, with all 3 phases 120deg apart from each other.

[thingstodo]

Originally Posted by MPaulHolmes

Yes that's right. There's DC voltage monitoring now, but the only use of it at the moment would be to verify precharge. But I'm not even doing that right this second.

Hmmm ... the existing pre-charge (when powered with AC) uses the brute-force method. The full wave bridge has a resistor in series, connected to the positive bus. When the DC bus voltage charges through that resistor and gets high enough for the DC/DC converter circuit to work properly, the micro powers up, and turns on the main contactor. That contactor bypasses the pre-charge resistor.

Since I'm not using that (old and inefficient) DC/DC converter, and since I will be using mostly whatever voltage pack I have available for testing ... maybe I can implement the DC voltage monitoring for pre-charge?

I have NOT added the V/Hz to this code yet. It's a good idea though. It's pretty easy to get working. IT was version 0.001 a few years ago. I just need a lookup table, and to race through the table at a given frequency, with all 3 phases 120deg apart from each other.

Would it be a different call to a different library to put the PWM out to the IGBTs?

The method to switch to V/Hz from Field Oriented Control would take a bit of fool-proofing (we fools are very creative!) since I think your throttle is a torque command for FOC and it would be a frequency output in V/Hz?

I haven't seen code for the new board as yet. Is it time for a backup email or are you doing version control through git?

[thingstodo]

Originally Posted by arber333

Finally someone figured cruise control for this inverter! Which is what V/Hz is yes?

I guess it would work. That's not really why I was wanting it though. The throttle is a torque signal to the controller. So fixing the throttle signal does not *REALLY* work for cruise control. You need a PID around the present speed. The throttle on a gas engine is actually a torque signal as well ... so cruise control should work similar for gas or electric.

Well sort off... I suggest the primary use of V/Hz would be to implement cruise control. You have PID regulation working. You frame it for V/Hz to move found one chosen frequency, fixed by a button switch (on dash) and PID would try to follow this frequency - speed control! Throttle would still be effective upwards from chosen speed, so you could tap the button with throttle and set a higher speed or just tap button with throttle loose and car would deccelerate some. To break the link would require pulse on brake switch line

I think all of this works with the throttle/torque signal as well. No need to switch to the less efficient V/Hz output.

, you have this wired yes?

I have not seen a Brake switch into the controller as yet. It's the lower part of the wig-wag throttle .. so far .. in my testing at least and what I have seen of Paul's Leaf motor cart.

Limp home mode could effect itself automaticaly if encoder signal would be lost. That way you could have some control if sensor was out. It happened to me with my controller. I use optical RPM sensor for arduino (a glorified mouse sensor) and it caught moisture and corrosion. I replaced it with better protected one but it stays a weak link im my car. My 5c...

The switch between a Torque input and a speed input is going to give you a sudden change, I think. If you are cruising at highway speed, going down a small incline for example, you could have your throttle close to 1/3. If the encoder quit, your speed would drop to 1/3 of your maximum speed, which could be a rapid deceleration.

Or you could be accelerating gently from a traffic light, lose the encoder, and now your car wants to keep accelerating to 3/4 speed instead.

I'm kinda worried about the transition. Right now, if you lose the encoder, the motor decelerates rapidly to maybe 1/10 of the speed you were going ... not a perfect solution if you are on a freeway!

If we can come up with a way to do something reasonable and the speed changes gradually then you have time to get to the side of the road, or use your brakes, or let off the throttle. That's sort of what I'm looking for. The car would definitely handle differently in V/Hz than it does in Field Oriented Control. Acceleration and deceleration would be less aggressive and less efficient. But the car should still be drive-able.

I wonder if the throttle input for V/Hz could be interpreted as amps? That is close to a torque setpoint, and there should be no big change when switching from Field Oriented control to V/Hz and back again ....

I need to do some reading ... Google to the rescue!

[thingstodo]

Originally Posted by thingstodo

If we can come up with a way to do something reasonable and the speed changes gradually then you have time to get to the side of the road, or use your brakes, or let off the throttle. That's sort of what I'm looking for. The car would definitely handle differently in V/Hz than it does in Field Oriented Control. Acceleration and deceleration would be less aggressive and less efficient. But the car should still be drive-able.

I wonder if the throttle input for V/Hz could be interpreted as amps? That is close to a torque setpoint, and there should be no big change when switching from Field Oriented control to V/Hz and back again ....

I need to do some reading ... Google to the rescue!

A summary of what I've read, mixed with a bit of my own experience:
- The amps for an AC motor don't really go much below 25%, even if it is spinning slowly and unloaded, unless you are moving to regenerative braking from motoring.
- So the amps go from 25% to about 115% for the throttle signal, but it depends almost totally on the acceleration/deceleration
- Keeping the amps at 75% the car speeds up slowly (relatively) compared to keeping the amps at 100%. The controller needs to increase the output voltage as the frequency rises to keep the amps at 100%. At some point you run out of voltage and it can no longer maintain 100% current.
- Using Paul's wig-wag style throttle arrangement, and the amps as a setpoint from the throttle, there is a deadband from 25% down to -25% where the motor is not exactly coasting but is not really motoring or regen braking. Perhaps that works OK as a control philosophy?

Switching to using the throttle as a current command when the encoder signal goes away sounds interesting to me. I think Klaus (is that his name?) does a similar thing with the controller that Damien Maguire is using. I think Klaus refers to it as slip control instead of field oriented control. And it uses encoder feeedback .. so maybe I'm off base there?

Does this re-interpreting -the-throttle-signal-and-switching-to-V/Hz-control -with-amps sound like something that we should pursue? Comments?

Besides the math and tracking the rotor position/angle ... Field Oriented Control appears to allow for the voltage to adjust auto-magically as required to get the commanded torque. Where the V/Hz has a specific voltage programmed as an equation for each output frequency. If you want the current to rise, you increase the output frequency with your PID loop. Depending on the motor load, you could be applying 'TOO MUCH' voltage and wasting some power as motor heat .. or you could be applying 'TOO LITTLE' voltage and not getting the current (and the torque) that you want.

I'm getting excited about seeing the difference in the control schemes, and trying to determine how much farther you can go, or how much faster you can accelerate, with Field Oriented Control ... or specifically the encoder signal .. compared to essentially open loop control at V/Hz.

My my ... I do get long winded!

[MPaulHolmes]

The encoder signal breaking is pretty catastrophic. So maybe if that happens, the controller could go into a coast mode where all IGBTs go off. Then whether it's permanent magnet or induction, it would just be like you lost power in a normal car, and are coasting. Maybe then you could assume the motor is about the same RPM as it was before "broken encoder" happened. Then you could set the frequency/voltage to match what the frequency was before, and officially be in V/Hz mode for the purpose of limping home. Or if it's a Leaf sort of motor, just coast to a halt with all igbts off. One reason I like the Leaf resolver is it's very reliable and is largely immune to noise/dirt. It would have to be a cut wire to cause a problem.

The newest board has a variable brake input. Just like a throttle. I'm not sure how useful it is. Tesla, for example, uses normal brakes, and has the throttle in a wig wag, with programmable max regen, but just stops the regen once the car has stopped. I don't know if everybody does that or what. Arber, do you have reverse from the transmission? I wonder how standard it would be for people driving a car with no reverse except for a button on their dashboard.

[MPaulHolmes]

That's good to know! I'll make sure I include a reverse button then.

[thingstodo]

Originally Posted by arber333

Well in short, yes i use Johannes inverter in my car, and it is excellent.

My apologies to Johannes - I'm not sure why I thought his name was Klaus?

[thingstodo]

Originally Posted by MPaulHolmes

That's good to know! I'll make sure I include a reverse button then.

If you are using a manual gearbox, it is still nice to reduce available torque and max speed when you are in reverse.

Can you make the reverse switch configurable, so that the motor direction can be switched (for a single speed gearbox) or not (if you want to switch to a reverse gear) but the max speed and motor torque are reduced?

[thingstodo]

Originally Posted by MPaulHolmes

The encoder signal breaking is pretty catastrophic.

Sounds like the Leaf has the encoder well protected. The Siemens motor I tested with has the encoder internal to the motor casing so I expect it is a bit more difficult to damage.

My external encoders are really easy to damage. Alignment, vibration, coupling ... my DIY stuff is much less robust!

So maybe if that happens, the controller could go into a coast mode where all IGBTs go off. Then whether it's permanent magnet or induction, it would just be like you lost power in a normal car, and are coasting.

That sounds pretty good. You may need some extra code to figure out whether the motor has no encoder when it begins to move or perhaps you are on a hill and need a bit more throttle to get the motor to rotate?

Maybe then you could assume the motor is about the same RPM as it was before "broken encoder" happened. Then you could set the frequency/voltage to match what the frequency was before, and officially be in V/Hz mode for the purpose of limping home.

The switch from Field Oriented Control to V/Hz is sounding like a risk. Coasting to a stop, then starting out in limp mode on V/Hz should be an acceptable solution for as often as this should happen.

I'm not sure what I was thinking .. switching control methods while the motor is still rotating ... !!! .. my normal paranoia seems to have taken a short vacation.

The newest board has a variable brake input. Just like a throttle. I'm not sure how useful it is. Tesla, for example, uses normal brakes, and has the throttle in a wig wag, with programmable max regen, but just stops the regen once the car has stopped. I don't know if everybody does that or what. Arber, do you have reverse from the transmission? I wonder how standard it would be for people driving a car with no reverse except for a button on their dashboard.

There are a couple of things to watch for with regen. When you are doing regen from the throttle, and then press on the brake, the brake needs to ramp up the regen from that regen point you are already at, not release the regen to 0 and start ramping again. That feels like pressing the brake gives you acceleration ... very disturbing!

And the regen only pumps power back into the batteries to a certain speed. After that you are using battery power to slow down your car and would be better off letting the friction brakes heat up a bit instead. That may be a bit hard to determine by calculations ... perhaps experimentally would be best and just enter the low speed cutoff into the configuration?

That could be tricky, as you have no control on when the friction brakes engage ... how much HIP you need to put into the brake pedal before the friction brakes start helping to slow the car.

Maybe I'm getting carried away again? Heck, we KNOW I'm getting carried away! That's what I do.