TWI causing nRF52 not to sleep

RE: TWI causing nRF52 not to sleep

Susheel Nuguru — Wed, 31 Aug 2016 11:27:23 GMT

RK, you are talking about this case?. That issue was in 1.0.0.3_alpha

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 10:27:31 GMT

So after much investigation it was a float calculation that was being performed immediately after the TWI reading that was causing the nRF52 to stay awake. Adding the following power manage function solved it

static void power_manage(void)
{

    uint32_t err_code;
    __set_FPSCR(__get_FPSCR()  & ~(FPU_EXCEPTION_MASK));      
    (void) __get_FPSCR();
    NVIC_ClearPendingIRQ(FPU_IRQn);

    err_code = sd_app_evt_wait();
    APP_ERROR_CHECK(err_code);
}

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 10:24:15 GMT

Yes indeed. nRF52 header wasn't included. Now included, and power_manage() now working. Thanks RK.

RE: TWI causing nRF52 not to sleep

RK — Wed, 31 Aug 2016 10:10:47 GMT

well that's rather unfortunate then isn't it as we might have solved this yesterday. FPU_IRQn is defined by the nrf52 header file so of course you have it.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 10:02:21 GMT

Oops I think I know why the log was showing no interrupt. I am running an LED timer initially and that may have been the trigger at the time, not this issue.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 09:59:58 GMT

Yes, see above my original question about the flag clearing. The issue I have is:

FPU_IRQn

does not appear defined from SES's perspective. So I was wondering how I could define it.

Also I don't know about the logs, I'm almost certain the second one was at a break point on a printf immediately below sd_app_evt_wait() (i.e., immediately having returned from whatever took it out of the sleep function). I can take another snapshot shortly.

RE: TWI causing nRF52 not to sleep

RK — Wed, 31 Aug 2016 09:53:35 GMT

There's a suggested flag clearing to do before you call sd_app_evt_wait(). If that does turn out to be the case however the register dumps you posted earlier were very misleading, in fact wrong, as they show there are no exceptions pending.

RE: TWI causing nRF52 not to sleep

RK — Wed, 31 Aug 2016 09:50:14 GMT

put FPU in the search box and see if anything we've talked about before gives a hint. There's soemthing about FTU in the release notes or the documentation somewhere too.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 09:49:01 GMT

If I simply do this in main above the loop:

float divider = 6.66;
float test = 10 / divider;

It's enough to keep the nRF52 awake. Very strange.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 09:41:31 GMT

132 v 2 for SD

RE: TWI causing nRF52 not to sleep

RK — Wed, 31 Aug 2016 09:37:41 GMT

The constant waking is usually when there's an FP error, as far as I remember from the release notes and the other threads about this. The FP interrupt is raised and never cleared. What SD version?

@aryan am I remembering this correctly, or was there another FPU issue.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 09:35:18 GMT

I don't know either. It's just with that calculation commented out, sleep is maintained. With that calculation in, continual waking is observed. I also did a little tester float divider calculation and the same thing happened.

RE: TWI causing nRF52 not to sleep

RK — Wed, 31 Aug 2016 09:29:31 GMT

well that makes no sense, I know I suggested FPU might be an issue however there's nothing wrong with that calculation and the flags in the text you posted earlier said there were no FP exceptions.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 09:27:18 GMT

Ok, so it does indeed look like a floating point calculation I was doing immediately having read from the sensor. So the continual wake has noting to do with the TWI itself. I have tried to play around with casting to get the calculation with the resolution I need without luck. Essentially I need to achieve this with the nRF52 being able to go to sleep afterwards:

float tempRH = ((125 * float_humidity) / 65336) - 6;

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 08:28:12 GMT

Just in case it is FPU I thought I would clear any interrupts, however FPU_IRQn is showing as not defined in SES. Amended question above.

RE: TWI causing nRF52 not to sleep

RichieJH — Wed, 31 Aug 2016 07:39:02 GMT

Thanks. How to I set-up breakpoints in that file using Segger Embedded Studio? It seems to be set up as a package file for SES and when I remove it from the project macros to add the file separately to enable me to add the breakpoints SES throws up an error. I have a delay between writing and reading from the sensor and its a shared bus with another sensor which I am not using in this application. Just for further information.

RE: TWI causing nRF52 not to sleep

Susheel Nuguru — Tue, 30 Aug 2016 11:31:07 GMT

breakpoint in interrupt_handlers should clarify that. Also you can try to see if any fault exception happened by putting breakpoints in the default handlers defined in arm_startup_nrf52.s file

RE: TWI causing nRF52 not to sleep

RK — Tue, 30 Aug 2016 11:13:31 GMT

I assumed that register state was after the sd_app_evt_wait() has returned so that would be after the interrupts (if there were any) had been serviced. If one wasn't cleared then it wouldn't ever get out of the interrupt handlers and return, so it's not that. It would have to be a pretty fast interrupt to keep the thing awake, does the twi driver use a timer for anything, wondered if it was a fast timeout timer which wasn't getting suppressed and was constantly waking it up

RE: TWI causing nRF52 not to sleep

Susheel Nuguru — Tue, 30 Aug 2016 11:05:51 GMT

I cannot see any pending interrupts either. Are you sure that you are sniffing the registers before the interrupt is serviced?

RE: TWI causing nRF52 not to sleep

RK — Tue, 30 Aug 2016 08:39:03 GMT

I don't see anything there either. I thought it might be the FPU exception but it's not, there's none shown. So I don't know why it keeps bumping out of sleep

RE: TWI causing nRF52 not to sleep

RichieJH — Tue, 30 Aug 2016 07:24:38 GMT

Thanks RK. OK checking the Registers in Segger Embedded Studio shows that there are no obvious interrupts causing the nRF52 not to hold in sd_app_evt_wait(). I attach the txt outputs in my original question. The first one is when no sensor reading is begun. The second is after a sensor reading at the point immediately after sd_app_evt_wait()

RE: TWI causing nRF52 not to sleep

RK — Tue, 30 Aug 2016 00:49:42 GMT

What interrupts are pending when it comes out of sleep mode? Look at the interrupt registers, all of them, including the standard cortex ones (like FPU).