Unable to set different PWM rates on adjacent pins

[Topo-png]

When using analogWrite or ap3_pwm_output on pins 12 & 13 I'm unable to set different PWM rates without board freezing.

Using Artemis Redboard ATP and Arduino IDE for packages

This code freezes

void setup()
{

}

void loop()
{

  ap3_pwm_output(12, 0, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  ap3_pwm_output(13, 0, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  delay(1000);
  ap3_pwm_output(12, 50, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  ap3_pwm_output(13, 100, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  delay(500);
}

This code works

void setup()
{

}

void loop()
{

  ap3_pwm_output(12, 0, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  ap3_pwm_output(13, 0, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  delay(1000);
  ap3_pwm_output(12, 100, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  ap3_pwm_output(13, 100, 100, AM_HAL_CTIMER_HFRC_12MHZ);
  delay(500);
}

By changing pins 12 to 32, both pieces of code function as expected.

edit: fixed typo and markdown

[nseidle]

I hit this too. I'll have a look.

[nseidle]

We're hitting and infinite loop here.

[nseidle]

Happens on pins 12/13, 5/6, but not 4/5, 6/7, 11/12

Something to do with segments switching...?

[oclyke]

Thanks for investigating @nseidle - I will keep my eye on this and try to roll in any findings to the v2.0.0 effort

[nseidle]

Whew, I've got a gnarly fix:

// if timer is running wait for timer value to roll over (will indicate that at least one pulse has been emitted)
AM_CRITICAL_BEGIN // critical section when reading / writing config registers
    if ((segment == AM_HAL_CTIMER_TIMERA && * ((uint32_t *)CTIMERADDRn(CTIMER, timer, CTRL0)) &(CTIMER_CTRL0_TMRA0EN_Msk)) ||
    (segment == AM_HAL_CTIMER_TIMERB && * ((uint32_t *)CTIMERADDRn(CTIMER, timer, CTRL0)) &(CTIMER_CTRL0_TMRB0EN_Msk))
    )
{
    uint32_t current = 0;
    uint32_t last = 0;
    do
    {
        last = current;
        current = am_hal_ctimer_read(timer, segment);
    } while (current >= last);
}
AM_CRITICAL_END // end critical section

Basically, only enter the while loop if there is a segment that is already enabled that is also the segment that the user is asking us to enable.

This needs lots of testing but it fixes my code for now.

[nseidle]

@oclyke - Thanks! I'd like your feedback on the viability of this fix. I'm not entirely confident in the PWM arena but I'm learning as I go.

[nseidle]

PR #229 should fix this issue. We'll leave open until merged.

[oclyke]

Closing since #229 was merged