Code analysis problem: the PWM time diagram

[GeorgeBayern]

today I carefully studied the PWM time sequence and digged little deep about the the function
“ControlLoop_IRQHandler”. I mad a diagram about the interrupt sequence for better understanding and communication. A few information maybe wrong in this diagram, I would improve it after disscusion.

I just put motor M1 here for better presentation, and here is my understanding. point it if I am wrong :

1. tim 1_8 frequency is 168Mhz, according the timer set up： Prescale=0; Periode =TIM_1_8_PERIOD ( 3500), we can get the frequecy count up/down to 3500 = 168Mhz/3500 = 48Khz. That also how the black triangle wave been drawn in the diagram.

2. The " RepitionCounter = TIM_1_8_RCR (2) ", which means the interrupts happens 16Khz (the black square in the diagram) .

3. the callback functions ( current_meas_cb(); dc_calib_cb(); PWM_update_cb(); ） all updated in every timer interrupt (after timestamp3) ?

4. For update 0, if "current_meas_cb() " happened in timestamp 0, then the corresponding "dc_calib_cb() " happened 1 timestamp later, then “PWM_update_cb()” 3 timestamp later?

in ControlLoop_IRQHandler():

motors[1].current_meas_cb(timestamp, current1);
motors[1].dc_calib_cb(timestamp + TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1), current1);
motors[1].pwm_update_cb(timestamp + 3 * TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1));

and I red the post Oska posted many years ago, feels some data don’t fit. @madcowswe

The motor timing diagram

So a few people have asked me about the motor timing diagram, and what exactly is going on. I kind of just stuck it there for making sure it doesn’t get lost, and I apologise that there is no legend or any other explanation. So here goes. First: the diagram reproduced here for easy viewing: [timing diagram] So some background: we are using Center Aligned PWM, which looks something like this: [CPWM] And that triangle wave is the counter value in the PWM module, and is the same triangle wave…

[madcowswe]

Diagram looks good, except we only measure current on the “bottom corner” of the triangle wave, and only measure calibration on the “top corner”.

[GeorgeBayern]

The mistake I made is: I didn’t notice the following code is only called when counting down:

if (!counting_down) {

    TaskTimer::enabled = odrv.task_timers_armed_;

    // Run sampling handlers and kick off control tasks when TIM8 is

    // counting up.

    odrv.sampling_cb();

    NVIC->STIR = ControlLoop_IRQn;

}

so I improved the diagram like following:

But one thing is not totally straight:

motors[1].pwm_update_cb(timestamp + 3 * TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1));

PWM update is “3 timstrape” latency (one and half current-measurement period) than " sampling()", it does not apply with the saying you mentioned:

So the control latency (current measurement to PWM goes live) is exactly one current-measurement period.

Did I understand something wrong this part?@madcowswe

thank you.

[madcowswe]

The current is sampled, then the CPU uses the data to compute the control loops etc, then the resulting PWM timings are loaded into the shadow registers and go live on the “current sampling” corner shown.