Extended sleep mode disable/re-enable

I had set the default_operation_adv = NULL, rather than user_app_adv_start as all the other examples did. Once its put back to user_app_adv_start then it works fine in the flash while the advertising is running. If I wait until the advertising is complete to press the button then it again does the freezing. Based on the current consumption when it freezes, it looks like its not actually in extended_sleep mode, and actually still in active mode, just stuck somewhere.

Do you have any ideas on why this is? From my perspective all this changed is it advertising at the beginning versus not, but it seems to affect more than just that.

The module is configured in ext_sleep_modewithoutOTP copy, user_app_init is used as the .app_on_init, and the wakeup controller is configured to use a GPIO as the wakeup from the button. When the chip wakes up it does not start advertising, because I have purposedly removed that aspect. The user_app_adv_start() callback is only called when someone holds the button for 10 seconds.

I can wake it up the first time from the sleep with the GPIO trigger. Once it wakes up it will freeze after the 3-4 seconds, unless advertising is enabled, in which case everything will work fine.

The problem with running in debug mode is that everything works fine in debug mode, but not when running from the flash.

I understand the aspect of default_operation_adv = user_app_adv_start automatically starting the advertising at the beginning. Im just not sure why that seems to change the behavior of the code in terms of getting frozen after the defined period, versus working as expected.

In terms of a base project that it is similar to, it would be a combination of the user_peripheral one and the proxy one for the sleep modes.

The code also works fine if I use the app_easy_timer() instead of setting up the timer0 as the counter. I know the timer0 is disabled in extended sleep mode but im using arch_disable_sleep() upon wakeup from the button so I dont see why that would be a problem.

Upon wakeup from the button the timer is fully reinitialized/configured so it seems to be starting properly, but after some time the code just stops.

Edit: This time from button wakeup until freeze is 2.7 seconds, very consistently. And its definately not sleeping, the run current goes up from 200uA to 300uA when the freeze occurs.

When using the flash to boot, the device definately goes to sleep when I'm not pressing the button or inbetween advertising, sleep current drops to ~10uA.

I am using printf in order to determine that it is frozen. When the button is pressed, sleep is disabled, and the timer0 is started and generates an interrupt every 100ms. All the interrupt handler does is print an incrementing index. The incrementing index works for 2.7 seconds, until it just stops, that's how I know it freezes consistently after 2.7 seconds.

这是或多或少地得到解决。I don't exactly know why it was freezing after the 2.7 seconds, or why it was working when advertising is enabled, but it works properly now when I've removed the

if (arch_ble_ext_wakeup_get()) in the button press callback. I think it's in the example projects because the button starts the advertising, and it only evaluates true when the advertising is complete. So for my project it makes sense for it to not be there.

The wdog is not defined in the config file, but those times seem too close to be a coincidence. The CFG_DEVELOPMENT_DEBUG is defined. The problem occurs when running from flash, and occurs whether or not the debugger is attached.

So I tried enabling the watchdog and that seems to solve it also, rather than removing the check on the arch_ble_ext_wakeup_get in the wakeup callback. I'll summarize the combinations here for people:

If you have

if (arch_ble_ext_wakeup_get())
{
arch_ble_force_wakeup();
arch_ble_ext_wakeup_off();
app_easy_wakeup();
}

And you want to disable sleep for some period, then re-enable sleep at some later point, then the watchdog must be enabled. Otherwise after waking the code will freeze after the watchdog timeout, still not sure why exactly. There's also the case when advertising is enabled then it will work while the device is advertising, then will freeze again when finished, therefore with infinite advertising you dont see this problem.