Correct procedure for uninitializing/reinitializing app_pwm?

[Chris]

Here is my power consumption (3.0V):
no app_pwm: 0.010mA
after app_pwm_init(): 0.295mA
after app_pwm_enable(): 0.891mA

If I do app_pwm_uninit(), power consumption goes back down to 0.010mA as expected.

The problem is that pwm will not work correctly after I run app_pwm_init()/app_pwm_enable() again.

APP_PWM_INSTANCE(PWM1,1);
static volatile bool ready_flag;

void pwm_ready_callback(uint32_t pwm_id) {
    ready_flag = true;
}

void pwm_init() {
    app_pwm_config_t pwm1_cfg = APP_PWM_DEFAULT_CONFIG_1CH(1000L, 28);
    pwm1_cfg.pin_polarity[0] = APP_PWM_POLARITY_ACTIVE_HIGH;
    APP_ERROR_CHECK(app_pwm_init(&PWM1,&pwm1_cfg,pwm_ready_callback));
    app_pwm_enable(&PWM1);

    //Adding the following four lines don't give the same behavior as not having them at all.
    APP_ERROR_CHECK(app_pwm_uninit(&PWM1)); //does app_pwm_disable
    pwm1_cfg.pin_polarity[0] = APP_PWM_POLARITY_ACTIVE_HIGH; //this should not be needed again and does not do anything the second time around
    APP_ERROR_CHECK(app_pwm_init(&PWM1,&pwm1_cfg,pwm_ready_callback));
    app_pwm_enable(&PWM1);
}

void test_pwm() {
    while(app_pwm_channel_duty_set(&PWM1, 0, 5) == NRF_ERROR_BUSY);
}

This is not my actual code, but it demonstrates my problem. The code without the extra four lines in pwm_init() works as I expect. Why would adding those four lines not give the exact same behavior? For some reason, the pwm then acts as if its polarity is active_low.

Thanks.

[Hung Bui]

Hi Chris,

We found a bug with the app_pwm.c that when you uninit and then init it again the polarity may changed. I attached here a patched file for SDK v9.0. It's a temporary workaround, we will need to do further test before it comes to the next version of the SDK. app_pwm_SDK9_0_patched.c

[Chris]

I forgot to mention that I am using SDK8.1.0. I ran a quick test with the above patched version and it seems to pass that simple uninit/reinit scenario, but wanted to make sure that it is ok to run it in SDK8.1.0?

Also, I was using the patched pwm_20150806 version from here: devzone.nordicsemi.com/.../

Have those 'fixes' been incorporated into your patched version above?

[Hung Bui]

Hi Chris,

No, this patch only taking care of the inverting issue when disable&enable and when uninit&init. What I would suggest is to do a diff between the stock app_pwm.c in the SDK with this patch, and then add the changes into the patch by Koppel in the other thread.

Basically it's the code to disable ppi channels in app_pwm_disable() and app_pwm_enable() function. And the call for pan73_workaround() moved from app_pwm_init() to app_pwm_enable and app_pwm_disable.

The full patch for both problem will be available in the next version of the SDK.

[Chris]

Also, are two APP_PWM_INSTANCEs completely independent from each other?

APP_PWM_INSTANCE(PWM1,1);
APP_PWM_INSTANCE(PWM2,2);

I should be able to app_pwm_uninit()/app_pwm_init()/app_pwm_enable() one without affecting the other, right?

[Hung Bui]

Hi Chris,

No they should not affect each other since they use different timers for controlling.