Highly scientific FOC Simulation

[Candas1]

Here is the spammer again.
Wanted to share something I used last few days for learning purpose.
It’s clearly not a scientific tool but gives a good intuition about how FOC works.
It’s a very simplified FOC simulation in a gsheet

In the Phase current to DG tab:

• If you enter a different gain for phase A (e.g. 1,1), this makes D/Q currents wobbly, the PIs will have to compensate this. With good hardware I guess it shouldn’t happen.
  
  
• If you enter a delay (e.g. 0,1), it introduces a lag between the rotor angle and the sensor angle, this shows not measuring the current and the position at the same time can introduce an error in D and Q (a little) current measurement, reducing the efficiency.
  
  

in the DQ current to PWM tab:

• increasing D current (e.g. 0,5) not only introduces a shift in phases, but also increases the amplitude
  
  

[Candas1]

And another more advanced motor simulator

[Candas1]

You must be thinking, my phase currents don’t look like that. Fair point, let’s add some noise:

Someone proposed to filter the phase current, great idea !!! let’s do it:

It looks much better now. But why did the D current increase ? Oh no we introduced lag !!!

The same would happen if filtering Alpha and Beta… What about filtered the D and Q currrents ?

Ah, much better !!!

[Candas1]

But my sensor is slow, it’s not updating often!!!

Oh no !!! D axis current is off and wobbly! More work for the PI controller…

Lukily we can use sensor smoothing !!!

OK I stop bothering you

[o_lampe]

Very informative and entertaining.
But we need filtering,unless we have stable signals without glitches.
IMHO these glitches are the bigger problem.

[Candas1]

Yes I am showing filtering D Q currents works better, that’s what SimpleFOC is doing. But it’s not perfect either…

[Candas1]

Changing phase C current doesn’t impact Alpha/Beta and D/Q currents, how come ?

Because Ia+Ib+Ic= 0 (Kirchhoff’s law), only 2 phase currents are enough for FOC.
But why do we even measure 3 currents then ?

In reality, the total of phase currents don’t add up to 0 because of inaccuracies in hardware, and noise during sampling. One of the reasons why it’s useful to have 3 phase currents is that this can be used to slightly reduce the noise by sharing this error between the 3 phase currents:

[Candas1]

Added different modulation types and the effect of deatime on line to line voltage in the DQ to PWM tab if you are interested: