MSPM0G3507: Simple PI compensation code using MSPM0G3507

Hi Yogesh!

Are you referring to PI as in Proportional-Integral?

-Matthew 

Yes Matthew.

Hi Matthew,

I have tried using "mathacl_mpy_div_op_LP_MSPM0G3507_nortos_ticlang". Need your help to understand better. 

I am planning some kind of PI simulation to check whether the calculation  is correct or not.

Set point is 3500 and ADC measured values are 3400 ,3600,  3450  and 3550. First iteration results as below:
error = 3500-3400=100
error31 = (100 << (31-12))-1 = 52428799 (0x031FFFFF)
integral = 0+(0x0CCCCCCC*0x031FFFFF >>31) = 5242879 (0x004FFFFF)
Proportional = 0x26666666*0x031FFFFF >>31 = 15728639(0x00EFFFFF)
output = integral + Proportional = 20971518 (0x013FFFFE)
outputPI = 0x013FFFFE*300>>31 = 2

First iteration results are as expected. Second iteration results as below: 
error = 3500-3600=-100
error31 = (-100 << (31-12))-1 = -52428801(0xFCDFFFFF)
integral = 5242879 +(0x0CCCCCCC*0xFCDFFFFF>>31) = 429496726(0x19999996)  -- Need explanation (Ideally zero)
Proportional = 0x26666666*0xFCDFFFFF>>31 = 1272761547(0x4BDCCCCB) -- Need explanation
output = integral + Proportional = 1503238553(0x59999999)  -- clamped to higher value
outputPI =  0x59999999*300>>31 = 209  -- which is wrong

#include "ti_msp_dl_config.h"

/* Ki = 0.1, 0.1 x 2^31 = 2,147,483,648 = 0x0CCC CCCC */
#define Ki (0x0CCCCCCC)
/* Kp = 0.3, 0.3 x 2^31 = 644,245,094 = 0x2666 6666 */
#define Kp (0x26666666)
#define setpoint  3500
const DL_MathACL_operationConfig gMpyConfig = {
    .opType      = DL_MATHACL_OP_TYPE_MPY_32,
    .opSign      = DL_MATHACL_OPSIGN_UNSIGNED,
    .iterations  = 1,
    .scaleFactor = 0,
    .qType       = DL_MATHACL_Q_TYPE_Q31};

float res=0.0;
volatile int32_t outputPI=0,errorQ31=0,integral = 0,Proportional=0,output=0;
int array[4]={3400 ,3600,  3450 , 3550}, measured_value=0,error=0;

int main(void)
{
    SYSCFG_DL_init();
    DL_TimerA_startCounter(PWM_0_INST);
    int i = 0;

    while (1) {
        measured_value=array[i];
        error = setpoint-measured_value;
        errorQ31 = (error << (31-12))-1;
        DL_MathACL_startMpyOperation(MATHACL, &gMpyConfig, Ki, errorQ31);
        DL_MathACL_waitForOperation(MATHACL);
        integral = integral + DL_MathACL_getResultOne(MATHACL);
        if(integral>=0x59999999){
            integral = 0x59999999;
        }
        else if(integral <0){
            integral = 0;
        }
        DL_MathACL_startMpyOperation(MATHACL, &gMpyConfig, Kp, errorQ31);
        DL_MathACL_waitForOperation(MATHACL);
        Proportional = DL_MathACL_getResultOne(MATHACL);
        output = integral + Proportional ;
        if(output>=0x59999999){
            output = 0x59999999;
        }
        else if(output < 0){
            output = 0;
        }
        DL_MathACL_startMpyOperation(MATHACL, &gMpyConfig, 300, output);
        DL_MathACL_waitForOperation(MATHACL);
        outputPI = DL_MathACL_getResultOne(MATHACL);
        DL_TimerA_setCaptureCompareValue(PWM_0_INST, outputPI, DL_TIMER_CC_0_INDEX);
        i++;
        if(i>3){
            i=0;
        }
        //res = _IQ30toF(output>>1);
    }
}

Please go through and let me know if any corrections. 

Regards,

Yogesh

Hi Yogesh!

How are you taking time into account when calculating your integral?

-Matthew

Hi Matthew,

Do you have any sample code. 

Regards,

Yogesh

Hi Yogesh!

We don't have an example of a PI compensator; however my suggestion here is to remove the MATHACL from your application, and test your algorithm in its simplest form. The MATHACL can be implemented after the algorithm is working correctly. 

-Matthew 

Hi Matthew,

Let me know the modifications required in my code.

I am not converting continuous time domain to discrete. Normal PI calculation only i am targeting.

Regards,

Yogesh

Hi Sal Ye,

Sorry for the delayed response.Will check and update you on IQ math

Regards,

Yogesh