Current consumption difference between GNU and IAR builds

RE: Current consumption difference between GNU and IAR builds

ovrebekk — Mon, 19 Feb 2018 11:45:54 GMT

No problem Robert. If you need more help with this just let me know.

RE: Current consumption difference between GNU and IAR builds

Robert — Tue, 13 Feb 2018 13:25:53 GMT

Sorry at this time, my focus has been changed to other features. I will have to look into this again soon though.

RE: Current consumption difference between GNU and IAR builds

ovrebekk — Tue, 13 Feb 2018 07:36:41 GMT

Hi Robert

Did you ever get around to running it again, or did you figure out the problem some other way?

Best regards
Torbjørn

RE: Current consumption difference between GNU and IAR builds

Robert — Fri, 26 Jan 2018 16:05:55 GMT

BLE event type AdvertisingPeripheral connection
Voltage V

BLE interval ms

TX payload bytes

TX power 40-4-8-12-16-20-40 dBm

Test setup
•Chip: nRF52832QFAAB0 
•Softdevice: s132 2.0.0 
•32MHz crystal (HFXO) 
•32kHz crystal (LFXO) 
•Sleep clock accuracy master and slave: 20 ppm 
•DCDC enabled 

Estimated values

BLE event total charge: 12.3 uC 
BLE event total length: 4.6 ms 
Average BLE event interval: 2005.0 ms 
Total average current:  8.0 uA

These setting were run on the pc10056 dev kit for both comipilations. I am going to run this again, to make sure that it wasn't something in my setup. Before measuring current, all the cables were disconnected and the board was reset.

RE: Current consumption difference between GNU and IAR builds

ovrebekk — Fri, 26 Jan 2018 15:24:46 GMT

Are you sure that the IAR test was done on the same hardware and firmware?

Can you plug your advertising settings into the online power profiler, and see how the numbers compare? 200uA could be normal for advertising, if the advertising interval is short.

Best regards

RE: Current consumption difference between GNU and IAR builds

Robert — Fri, 26 Jan 2018 13:54:43 GMT

The full code is ble_app_pwr_profiling/main.c. I didn't want to have to push any buttons so I made a minor change.

int main(void)
{
    //    ret_code_t err_code;
    //    bool is_notification_mode    = false;
    //    bool is_non_connectable_mode = false;

    timers_init();
    buttons_init();
    
#if 0
#if BUTTONS_NUMBER > 2
    // Check button states.
    // Notification Start button.
    is_notification_mode = bsp_board_button_state_get(NOTIF_BUTTON_ID);

    // Non-connectable advertisement start button.
    if (!is_notification_mode)
    {
        is_non_connectable_mode = bsp_board_button_state_get(NON_CONN_ADV_BUTTON_ID);
    }
    else
    {
        // Un-configured button.
    }
#else
    is_notification_mode = true;
#endif
#endif
    
    // Initialize SoftDevice.
    ble_stack_init();

#if 0
    if (!is_notification_mode && !is_non_connectable_mode)
    {
        // The startup was not because of button presses. This is the first start.
        // Go into System-Off mode. Button presses will wake the chip up.
        err_code = sd_power_system_off();
        APP_ERROR_CHECK(err_code);
    }
#endif
    
    // If we reach this point, the application was woken up
    // by pressing one of the two configured buttons.
    gap_params_init();
    gatt_init();

#if 1
    // Notification button is pressed. Start connectable advertisement.
    connectable_adv_init();
    service_add();
#else
    if (is_notification_mode)
    {
        // Notification button is pressed. Start connectable advertisement.
        connectable_adv_init();
        service_add();
    }
    else
    {
        non_connectable_adv_init();
    }
#endif
    
    advertising_data_init();
    advertising_start();

    // Enter main loop.
    for (;;)
    {
        power_manage();
    }
}

The numbers taken for the IAR were done by someone else, and I don't recall the exact amount of current. The current for the version compiled with gcc was around 200uA.

RE: Current consumption difference between GNU and IAR builds

ovrebekk — Fri, 26 Jan 2018 08:12:49 GMT

Seeing the numbers would also be helpful, they can some times indicate what is running or not.
As Turbo is saying, the choice of compiler shouldn't have such a big impact on current consumption. If you see a huge difference I am inclined to think that the power management is not working properly with one of the compilers.

RE: Current consumption difference between GNU and IAR builds

Turbo J — Thu, 25 Jan 2018 15:27:13 GMT

In other words: Show us code.

RE: Current consumption difference between GNU and IAR builds

Turbo J — Thu, 25 Jan 2018 15:26:57 GMT

That seems unlikely. With correct power management advertising current is dominated by the radio and not the CPU core.