nRF52: "spike" after last period with inverted polarity PWM

[Franz]

Hi,

I use the a gpiote_in_event to trigger a number of PWM pulses which works fine so far. However, the exact behaviour of the generated signal is important so that the device interfaced to the signal line does also behave correctly.

I saw that if I use inverted polarity (what is what I want to do), i get a very short pulse at the end of the PWM (after the number of specified periods). It seems that after the number of specified repetitions, the HW is just about starting another PWM sequence which is then immediately aborted.

This spike is a problem in my system, how can i get rid of it?

/EDIT: see screenshot

[haakonsh]

We need more information on how exactly you've implemented "gpiote_in_event to trigger a number of PWM pulses".

Are the IN_EVENT pin or PWM pin driven high or low? 

[Franz]

Hi haakonsh,

See the code snippet below.

A falling edge on the HELPER_PIN should start the PWM

 
 
 //configure gpiote
 nrfx_gpiote_in_config_t helper_pin_cfg = NRFX_GPIOTE_CONFIG_IN_SENSE_HITOLO(true);
 err = nrfx_gpiote_in_init(HELPER_PIN, &helper_pin_cfg, NULL);
 nrfx_gpiote_in_event_enable(HELPER_PIN, false);
 
 
 //configure ppi channel
 uint32_t start_task_addr = nrfx_pwm_simple_playback(&m_pwm0, &seq, NBR_ADC_SAMPLES_AO, NRFX_PWM_FLAG_START_VIA_TASK|NRFX_PWM_FLAG_STOP);
 
 nrfx_ppi_channel_alloc(&ppi_ch_start_clk);
 nrfx_ppi_channel_assign(ppi_ch_start_clk, nrfx_gpiote_in_event_addr_get(HELPER_PIN), start_task_addr);
 nrfx_ppi_channel_fork_assign(ppi_ch_start_clk, nrfx_timer_task_address_get(&m_tmr1, NRF_TIMER_TASK_START));
 nrfx_ppi_channel_enable(ppi_ch_start_clk);

[haakonsh]

It kinda looks like the sequence is 'off by one'. Can you share your 'seq' struct? 

[Franz]

  nrf_pwm_values_individual_t seq_values[] = {0x8000|64, 0x0000|26}; //15 bit value, 16th bit = polarity
  nrf_pwm_sequence_t const seq =
  {
      .values.p_individual = seq_values,
      .length          = NRF_PWM_VALUES_LENGTH(seq_values),
      .repeats         = 0,
      .end_delay       = 0
  };

Sorry, it´s been some time - here it is!

[haakonsh]

I need the value of NBR_ADC_SAMPLES_AO as well. I want to reproduce the issue on my end.