nrf52833 BLE Packet Loss Rate High

[soloman]

Hi,

    I use nrf52833 Development board PCA10100 to do loss test as peripheral and speed is 1Mbps, tx power is MAX. It's loss rate is very high, when nrf52833 sends a packet every millisecond to Upper computer. What are the possible reasons for this problem？And the upper computer proved to be fine.

Sincerely!

[Vidar Berg]

Hello,

Data transmission in BLE is reliable. That is, if the packets fails to be sent after x retransmits (number of retransmit attempts are determined by the connection parameters), the connection will be terminated by the BT stack. So, I'm wondering where you are seeing the packet loss?  Do you check the return code from the notification send function?  Also, which SDK are you working with?

Thanks,

Vidar

[soloman]

Hello,

Yes, the return code of sd_ble_gatts_hvx is still SUCCESS when slave nrf52833 transmits packet per millisecond, while the master is a bluetooth module. I find master not receive packet enough, and there is no proper tool to capture the packet to determine whether it was sent from slave. So, you said nrf52833's BLE is reliable, that means slave nrf52833 transmits packet per millisecond is reliable, right? In addition, what configuration may occur the slave machine fails to send even if the packet sending function returns success, when the SDK is nRF5_SDK_17.1.0_ddde560.

Thanks,

Soloman

[soloman]

Hi Vidar,

There didn't trigger  the BLE_GATTC_EVT_TIMEOUT or BLE_GATTS_EVT_TIMEOUT event in ble_evt_handler, when setting breakpoints to this two position.  And this problem was seen with different central.

Thanks,

Soloman

[Vidar Berg]

Hi Soloman,

In that case, please double check that you are reading the return code correctly. The log message will print the hexadecimal value while the error codes are listed with their decimal values. Error 0x13 is the same as NRF_ERROR_RESOURCES ('19').

[soloman]

Hi Vidar,

Yes!Yes! The return code is 0x13, that means NRF_ERROR_RESOURCES.

Thanks,

Soloman

[Vidar Berg]

Hi Soloman,

Thanks for confirming. So, this problem can be mitigated by increasing the notification queue size as discussed in this thread:  Increase hvn_tx_queue_size to maximize throughput 

Best regards,

Vidar 

[soloman]

Hi Vidar,

I try to set que_size，but the board still reset when the value is not 1. It will  run for a while and then reset, when the value is 1.

Thanks,

Soloman

[Vidar Berg]

Hi Soloman,

Please check the debug log to see what error led to the reset.

Best regards,

Vidar