Which clock is used for timing after BLE establishes a connection?

[ethan]

Hi,

I am analyzing the issue that the BLE module can maintain a relatively short connection distance. Besides the transmission power and frequency offset, I also take into account the factor of the clock, because the connection is a synchronous event and must have a certain relationship with the clock. 

I would like to ask, in the connection event, what type of clock is used to determine the next connection time? Is it a hardware TIMER or a software RTC? In other words, is it related to a high-frequency clock or a low-frequency clock?

[Susheel Nuguru]

Hi Ethan,

I would like to ask, in the connection event, what type of clock is used to determine the next connection time? Is it a hardware TIMER or a software RTC? In other words, is it related to a high-frequency clock or a low-frequency clock?

Since you have tagged nrf5 sdk and nrf52, I am assuming you are using softdevice and also assuming that you are asking about the timing for the next "connection event" and not the next connection. It is not possible to anticipate the next connection. But the "connection event" For tracking next connection event softdevice have REM (Radio Event Manager) using LFCLK based hardware RTC (RTC0) to do timekeeping for the softdevice for all roles (including connections)

[ethan]

Hi,

 Your understanding is correct. I used softdevice, and the time I expressed refers to the time of the connection event, that is, the connection interval.

According to your statement, the connection interval was calculated using RTC(LFCLK) instead of TIMER(HFCLK). Then I would like to ask further, LFCLK has a 32k external crystal oscillator and an internal RC circuit as the clock source. Under normal circumstances, a 32k crystal oscillator should be more accurate. Then, if RC is used as the clock source and the error is relatively large, will it affect the connection distance?

[Susheel Nuguru]

ethan said:

Then, if RC is used as the clock source and the error is relatively large, will it affect the connection distance?

RC is normally not used for LFCLK while using BLE, better is to use LFSYNTH deduced from HFXO for better accuracy. None of our examples for BLE uses RC INT for LFCLK.

[Emil Lenngren]

It will not affect the connection distance, but the energy consumption (if the device is a peripheral). The reason is that with a more inaccurate clock, the software must turn on the radio a bit earlier, to handle the case where the clock is slower than it should, to be sure to catch the connection event from the central.

[ethan]

Hi,

Do you mean that the software will automatically detect the accuracy of the LFCLK and then start the radio in advance or delay based on this error to ensure reliability?

[ethan]

Hi Susheel Nuguru,

Do you agree with Emil Lenngren's statement? That is, the LFCLK accuracy will not affect the connection distance but only increase power consumption.

[Susheel Nuguru]

No, the details of accuracy of the clock is given to the softdevice at the time of softdevice init using nrf_clock_lf_cfg_t . Please check nrf_sdh_enable_request and all the defines below that are defined in sdk_config.h file

    nrf_clock_lf_cfg_t const clock_lf_cfg =
    {
        .source        = NRF_SDH_CLOCK_LF_SRC,
        .rc_ctiv       = NRF_SDH_CLOCK_LF_RC_CTIV,
        .rc_temp_ctiv  = NRF_SDH_CLOCK_LF_RC_TEMP_CTIV,
    #if defined(S140) && (NRF_SD_BLE_API_VERSION < 6)
        .xtal_accuracy = NRF_SDH_CLOCK_LF_XTAL_ACCURACY
    #else
        .accuracy      = NRF_SDH_CLOCK_LF_XTAL_ACCURACY
    #endif
    };