RE: PPI related problems

leo.lau — Wed, 10 Feb 2021 02:52:43 GMT

Thank you for the quick reply.

I am planning to increase sensors' sampling rate from 100Hz to 800Hz which combine 8 sets of data into one sd_ble_gatts_hvx() packet in order to utilize ATT_MTU throughput. It will increases the power consumption if awaking CPU every timer interrupt to update pointer address. As your suggestion may be it is better to initiate TWI transfer using CPU to avoid the mentioned errata (which is my original approach).

Regards

RE: PPI related problems

Kenneth — Mon, 08 Feb 2021 15:35:23 GMT

Hi,

I am not sure how autonomous you want this, but you should be aware of this errata:
https://infocenter.nordicsemi.com/topic/errata_nRF52832_Rev2/ERR/nRF52832/Rev2/latest/anomaly_832_109.html

To avoid the above somewhat tricky or power increase workaround, it may be easier to just always initiate the transfers from the CPU, this will ensure the above errata does not occur, and you also will have full control of which sensor you read and write to. The delay due to CPU wakeup will be very (few us) short compared to the TWI transfer time.

Best regards,
Kenneth