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

[MPaulHolmes]

Oh crap!! Not good. I had used the clare ixdi604pi last time, and it could go up to 35v, so I assumed it was the same for the mic4451. Thank you for helping me to avert a disaster! Fortunately, it's a standard pinout, so I just need to find a clare 35v driver. I just found one, but it was out of stock at digikey. I'll go back to searching. YOu saved my bacon!!!

OK, the IXDI614PI will do very nicely. It can go up to 40v. It's out of stock, but I have written to Clare, and a couple other places, and should hear back soon. You saved me!!! Thank you for preventing disaster!

[roguethunder]

Any luck tracking down the part ya need? ^.^

[MPaulHolmes]

Yes!

[drbobwoolery]

Originally Posted by Alcibiades Ramos

I was looking for something that would have a little more amps and I found this but they are $53.90

dev.ixys.com/DataSheet/98606.pdf


I also found this one for $149.00 1200v 650 amps and it's a module:

FZ400R12KE3B1

FZ400R12KE3B1 Infineon Technologies IGBT Modules

If the current sensors are not too high, actually measuring all three, and seeing that they add to zero may be the easiest way to detect a number of parts failure faults at once. Whenever the three currents do not sum to zero, a general fault condition would be generated.

[thingstodo]

Originally Posted by drbobwoolery

If the current sensors are not too high, actually measuring all three, and seeing that they add to zero may be the easiest way to detect a number of parts failure faults at once. Whenever the three currents do not sum to zero, a general fault condition would be generated.

Interesting. I agree that the 'vector sum' of the currents should be 0. And I agree that if the currents do not add up, there are a number of things that could be wrong.

How fast do you need to read the 3 phase currents so that they'll add to 0? Or do you need to average them, since you are sending a bunch of pulses?

I know that you can get a current signal (magnitude) from a hall effect sensor, or three of them I guess - but how do you calculate .. or measure, I guess, the phase angles?

I've seen a lower tech method used - put all three 'output' conductors (or bus bars, I guess) through a current transformer and see if anything is measured. If so, bad things are happening and current is 'escaping' somewhere. I think the alarm reads as 'ground fault'.

I have taken apart some older industrial VFDs that use this lower tech method. None of my employer's more modern VFDs have blown up yet, so I haven't had the opportunity to take them apart and see how it's done now.

[MPaulHolmes]

I'm using the dsPIC, which can read 4 channels simultaneously, so you don't have to worry about the lag time between different phase currents. I'm just reading 2 of the currents on this board. Someone recommended to me that at relatively low powers (60-70kW haha) 2 sensors is OK. He did another one that was closer to 150kW and used 3 sensors.

[nlc]

Originally Posted by thingstodo

How fast do you need to read the 3 phase currents so that they'll add to 0? Or do you need to average them, since you are sending a bunch of pulses?

In FOC, you always work with instantaneous phase currents, there is no averaging. It's why it's mandatory to have a microcontroller which can do simultaneous sampling on at least two channel.

It's true the controller sends voltage pulses on the motor phases, but the current is smooth, with just a small ondulation, it's why the better is to sample the current in the middle of the voltage pulse, to be in the middle of the ondulation and have the exact current value.

[nlc]

Originally Posted by MPaulHolmes

Someone recommended to me that at relatively low powers (60-70kW haha) 2 sensors is OK. He did another one that was closer to 150kW and used 3 sensors.

In my opinion, the real reason to use 3 sensors on big controller is because the current sensor price is really low compared to the global controller cost, thus it become really interesting to completely protect the third phase semiconductor too.

Because with 2 sensors, the 3 phase outputs can be short circuit protected between each other, but only two of them are short circuit protected between phase and +vbus ou gnd. With the third phase current, all phase can be completely protected (short circuit with an other phase, vbus ou gnd)

[thingstodo]

Originally Posted by MPaulHolmes

I'm just reading 2 of the currents on this board.

So three phases of current add to 0. If you measure 2 you are calculating the third.

Is there a library call for FOC that compares the magnitudes and gives you a 'ground fault' result?

[thingstodo]

Originally Posted by nlc

In my opinion, the real reason to use 3 sensors on big controller is because the current sensor price is really low compared to the global controller cost, thus it become really interesting to completely protect the third phase semiconductor too.

Because with 2 sensors, the 3 phase outputs can be short circuit protected between each other, but only two of them are short circuit protected between phase and +vbus ou gnd. With the third phase current, all phase can be completely protected (short circuit with an other phase, vbus ou gnd)

I think that what Paul said is that he's measuring 2 currents at the same time and that this is enough information.

Perhaps I should describe what I think the measurements can do and then you can correct that. I must have something messed up here.

- the hall effect current sensors are on the three output phases, after the IGBTs
- phase to phase issues can be detected in the motor by comparing phase currents
- phase to ground issues can be detected in the motor by having the currents not add to 0
- V+ to V- cannot be detected, since the sensors are isolated from the bus by the IGBTs?
- V+ or V- to ground cannot be detected, since it is not a complete circuit?
- measuring 2 currents gives most of that protection but you have to do some math

Where did I go wrong?