TMS320F28075: FCL PI Implementation

Niloc2004

Part Number: TMS320F28075

Hi,

I am using the FCL controller and am seeing different torque outputs for the same Iq value depending on if the motor is running in quadrants 2 or 3.

This resulted in me digging into the FCL_PI implementation since there is a speed element in its implementation. However, after I teased the pieces from the CLA and ISR, it doesn't look like any PI that I have worked with before.

From my understanding, here is what the PI controller does:

At time n:

err_n = ir_ref - iq_measured

Kp = L * I * wcc / V - Assume Relatively Constant

Ki = R * I * wcc * dt / V - Assume Relatively Constant

Kerr = Ki / 2 + Kp

Kerr_old = Ki / 2 - Kp

carry_over = (err_n * Kerr_old) + Bemf *dSpeed

carry_over = err_n* Ki/2 - err_n* Kp + Bemf*dSpeed

At time n+1:

err_n_1 = ir_ref - iq_measured

Out = err_n_1* Kerr + carry_over

Out = err_n_1 * Ki/2 + err_n_1*Kp + err_n *Ki/2 - err_n*Kp + Bemf*dSpeed

Out = Ki* (err_n_1 + err_n) /2+ Kp *(err_n_1 - err_n) + Bemf*dSpeed

So sumarizing:

Out = Ki * avg_err (over 2 cycles) + Kp * (change in err) + Bemf * (change in speed)

Is there any documentation about how the FCL_PI works? Where is the integrator? Why is Kp getting applied to the change in Iq error instead of the error itself? Why is Bemf constant multiplied by the motor acceleration?

It seems like the implementation has combined a lot of logic and optimized for performance. Which is great, but makes it hard to understand. And before I am directed to Dave Wilsons blog, I've already read it and wasn't able to figure out this implementation

Thanks a bunch,

-Colin

over 2 years ago

0 Ramesh Ramamoorthy over 2 years ago

TI__Mastermind 27240 points

This implementation is basically that of an integrator. If you integrate the product of kp and difference in error, it is basically same as using the product of kp and error.

0 Niloc2004 over 2 years ago in reply to Ramesh Ramamoorthy

Intellectual 440 points

Thanks,

Where is the actual summation (integration) occurring though? It looks like Out is directly put into the Inverse Park and not getting summed with previous versions of out. Or am I missing something?

And while I'm asking questions, are there any reasons you would expect variations in torque output for the same Iq value depending on motor direction? I.e Quadrant 2 vs Quadrant 3 operation?

-Colin

0 Niloc2004 over 2 years ago in reply to Niloc2004

Intellectual 440 points

Actually, I now see where the summation occurs. I had missed the +=

#define  FCL_PI_MACRO(v)                                                       \
{                                                                              \
    v.out += (v.err * v.Kerr) + v.carryOver;                                   \
    v.out = (v.out > v.Umax) ? v.Umax : v.out;                                 \
    v.out = (v.out < v.Umin) ? v.Umin : v.out;                                 \
}

I am still wondering about quadrant 2 vs 3 torque variations.

+1 Ramesh Ramamoorthy over 2 years ago in reply to Niloc2004

TI__Mastermind 27240 points

In an ideal world, it should be same. But such variation is a possibility. It depends on the inequality of feedback sensing + direction dependent stiction / friction/ load/ ...

C2000™︎ microcontrollers

C2000 microcontrollers forum

TMS320F28075: FCL PI Implementation