Bug in the 16-bit PWM implementation in dimming examples

RE: Bug in the 16-bit PWM implementation in dimming examples

DeveloperZ — Wed, 13 Feb 2019 14:43:12 GMT

Hi,

Any update on the issue?

Thanks.

RE: Bug in the 16-bit PWM implementation in dimming examples

DeveloperZ — Tue, 22 Jan 2019 16:23:43 GMT

Hi Joakim,

Were you able to reproduce this issue and check what the exact reason might be?

Thank you.

RE: Bug in the 16-bit PWM implementation in dimming examples

DeveloperZ — Thu, 17 Jan 2019 21:20:23 GMT

Hey! Thanks And no, its not a trace issue, the CURRENT_LIGHT_INTENSITY variable gets affected and it doesn't reach same max values every time... and the "Jumps" are quite random too, so are we leaning towards some BLE transmission, interruptions or something related...?

RE: Bug in the 16-bit PWM implementation in dimming examples

awneil — Thu, 17 Jan 2019 18:33:27 GMT

Or it could just be a limitation of the Trace.

The Trace is not good for spitting out very large volumes of stuff; it is quite easily overloaded - which could explain the "gaps" that you see...

RE: Bug in the 16-bit PWM implementation in dimming examples

hmolesworth — Thu, 17 Jan 2019 17:44:51 GMT

Looks like you have another process interrupting the trace; instead of 5 or 6 increment in the trace timestamp it jumps from (first example) 1052181 to 1052360 = 179 which is about 32 light levels or suspiciously close to the actual light levels "jumped" (19403-19371)=32. If a delay of 1mSec fixes the "jump" then you can assume that the interrupting process takes less than 1mSec to complete; perhaps it's a BLE transmission, or a serial port interrupt for a trace debug message .. ie it's not a light level jump but a jump in trace messages

RE: Bug in the 16-bit PWM implementation in dimming examples

DeveloperZ — Thu, 17 Jan 2019 16:59:33 GMT

Hi,

Below is the code where I am recording the "Current Dimming" values:

while (m_dim && On_OFF)
{
    if(m_dim_state == 0)
    {
		CURRENT_LIGHT_INTENSITY++;
		if(CURRENT_LIGHT_INTENSITY > (UINT16_MAX/2))
        {
			m_dim_state = 1;
            m_dim = false;
            break;
		}
	}
    else if(m_dim_state == 1)
    {
		CURRENT_LIGHT_INTENSITY--;
        if(CURRENT_LIGHT_INTENSITY < 0)
        {
			nrf_delay_ms(10);
            while (app_pwm_channel_duty_set2(&PWM0, 0, 0) == NRF_ERROR_BUSY);
            CURRENT_LIGHT_INTENSITY = 0;
            m_dim_state = 0;
            m_dim = false;
            break;
	    }
    }
    while ( app_pwm_channel_duty_set2(&PWM0, 0, CURRENT_LIGHT_INTENSITY) == NRF_ERROR_BUSY);
    
    __LOG(LOG_SRC_APP, LOG_LEVEL_ERROR, "Current dimming: %d\n", CURRENT_LIGHT_INTENSITY);
}

The code for the function that sets the duty cycle for PWM is:

ret_code_t app_pwm_channel_duty_set2(app_pwm_t const * const p_instance,
                                  uint8_t channel, app_pwm_duty_t duty)
{
    uint16_t ticks = ((uint32_t)(duty) * m_pwm0_max) / UINT16_MAX;
    return app_pwm_channel_duty_ticks_set(&PWM0, 0, ticks);
}

As I said, I added EnOcean functionality, the "m_dim" indicates switch press; 'On_OFF' indicates LED On/Off; "m_dim_state" indicates increase/ decrease of intensity and CURRENT_LIGHT_INTENSITY indicates the level.

The image that was posted in the question was taken when I was trying to increase the intensity of the LED. The jumps are also vague as the jump will not be at the same value next time and also the amount of jump is always different.

More or less, it can be said that, when I loop through the values from 0 to 32767, I observe numerous jumps and rarely catch the exact limits(0,32767)

Thank you.

RE: Bug in the 16-bit PWM implementation in dimming examples

Joakim Jakobsen — Thu, 17 Jan 2019 14:47:13 GMT

Hi.

Where and how did you add these logs to output the "Current Dimming"?

The three highlighted outputs (Current dimming: 19402/19410/19491) that stands out from the rest is always where the jump in value happens.
Are these values collected the same way as the rest?
Why does the output look different?

Maybe you could upload the code where this is implemented?

Best regards,
Joakim