LAUNCHXL-F28069M: Building a Three Phase VSI - Generating the Reference Waveform

Hello

You're right, the correct way to do it is to set up an interrupt. There is an example in control suite (C:\ti\controlSUITE\device_support\f2806x\v151\F2806x_examples_ccsv5\cpu_timer) how to set up an interrupt based on a CPU timer.

In that interrupt you can do smth like:

// Speed ramp with limitation
speed = speed + 1 * INTERRUPT_PERIOD;	// INTERRUPT_PERIOD= 0.001 for 1 ms interrupt
if (speed > maxSpeed)
	speed = maxSpeed;

// Calculate angle for this speed
referenceAngle = referenceAngle + speed * INTERRUPT_PERIOD;

// Wrap angle in [0; 2pi]
if (referenceAngle > _2pi) referenceAngle -= _2pi;
else if (referenceAngle < 0) referenceAngle += _2pi;

// Calc duty cycles
dutyCycle1 = sin(referenceAngle);
dutyCycle2 = sin(referenceAngle + _2pi_div_3);
dutyCycle3 = sin(referenceAngle - _2pi_div_3);

// Load CMPAs
EPwm1Regs.CMPA.half.CMPA = (Uint16)(dutyCycle1 * (float)TBPRD);

You would like to create a macro "INTERRUPT_PERIOD" to convert time-related physical values like seconds, rpm's to discrete domain. I hope this helps.

Thanks for the speedy reply! I'll check out that example and see how I get on. I had been thinking about how I could implement a slow startup but your speed ramp would seem to be an elegant solution.

I'll report back if I get stuck ;)

Update....

I have set up a CPU timer to interrupt at 20kHz and trigger the following ISR:

interrupt void TINT0_ISR(void)                  // PIE1.7 @ 0x000D4C  TINT0 (CPU TIMER 0)
{
    PieCtrlRegs.PIEACK.all = PIEACK_GROUP1;     // Must acknowledge the PIE group
    GpioDataRegs.GPATOGGLE.bit.GPIO0 = 1;     //toggle

    theta = theta + (2 * Pi * outputFrequency * 0.00005); //where duty cycle is proportional to sin(theta)
    // Wrap angle in [0; 2pi]
    if (theta > (2 * Pi)) theta -= (2 * Pi);
    else if (theta < 0) theta += (2 * Pi);

    // Calc duty cycles
    float A = ((modulationIndex * sin(theta)) + 1) / 2;                 //Reference waveform magnitude for Phase A
    float B = (modulationIndex * sin(theta + ((2 * Pi) / 3)) + 1) / 2;  //Reference waveform magnitude for Phase B
    float C = (modulationIndex * sin(theta - ((2 * Pi) / 3)) + 1) / 2;  //Reference waveform magnitude for Phase C

    Uint16 dutyA = (Uint16)(A * (float)PWM_HALF_PERIOD);
    Uint16 dutyB = (Uint16)(B * (float)PWM_HALF_PERIOD);
    Uint16 dutyC = (Uint16)(C * (float)PWM_HALF_PERIOD);

    if (dutyA > PWM_MIN_DUTY) dutyA = PWM_MIN_DUTY;
    else if (dutyA < PWM_MAX_DUTY) dutyA = PWM_MAX_DUTY;
    if (dutyB > PWM_MIN_DUTY) dutyA = PWM_MIN_DUTY;
    else if (dutyB < PWM_MAX_DUTY) dutyA = PWM_MAX_DUTY;
    if (dutyC > PWM_MIN_DUTY) dutyA = PWM_MIN_DUTY;
    else if (dutyC < PWM_MAX_DUTY) dutyA = PWM_MAX_DUTY;

    // Load CMPAs
    EPwm4Regs.CMPA.half.CMPA = dutyA;
    EPwm5Regs.CMPA.half.CMPA = dutyB;
    EPwm6Regs.CMPA.half.CMPA = dutyC;
}

My theory here is this:

1) Toggle a GPIO pin every time the ISR is called so that I can see it on the scope and reassure myself that the ISR is being called!

2) The variable theta is initialised outside the ISR to zero. I then create a new value for theta based on the elapsed time.

3) As suggested in a previous post, I then constrain theta to between 0 and 2pi

4) I multiply by a modulation index to control the maginitude of my output waveform and I take sin of theta. I then add one and divide by two such that the result varies between 0 and 1.

5) I constrain the duty cycle between 10% and 90%

6) I update the ePWM compare registers

This code would appear to me to make sense however the output from my inverter is far from sinusoidal. I end up with a pretty nasty sawtooth waveform with a large DC offset. Are there any obvious flaws in my logic above? I am struggling to understand what is going on and I don't see any way to verify that my PWM signals are correct without applying them to an inverter leg and seeing what happens!

I think we need a "bang head off wall emoji". That's what happens when you spend too long staring at a problem.

I changed that piece of code and passed the three top gate PWM signals through a simple RC low pass filter. This was the result:

4152.TEK00001.TIF

Nice waveforms =) Glad to hear you got it working =)