nRF51822 BLE_ERROR_NO_TX_BUFFERS issue

[Tosa]

I'm using the nRF51822 DK + SDK10.0.0 + s110. I modified the app_ble_hrs example to send 2x 20B packets of data (via a modified heart_rate_meas_send() function) upon a m_battery_timer_id timeout (the other timers are disabled). My goal is to reach 40B (2x 20B packets) every 6ms. The client is a 2013 Asus Nexus 7 which allows 3-4 packets per connection event @ 7.5ms connection intervals. I'm using sd_ble_tx_buffer_count_get() and BLE_EVT_TX_COMPLETE evt.count to keep track of the availble TX buffers per the doc.

I'm seeing really poor performance in that at data intervals of <15ms, I quickly reach 0 available buffers). 15ms for 40B is only 21kbps. Even at 15ms data intervals or higher I see the number of TX buffers available slowly drop over time (starting off at 7 buffers). I'm running it now at 20ms data intervals and the number of TX buffers was 4 after 20 minutes and has now actually started missing data after 30 minutes and has only 2 TX buffers available :( . I also noticed that this calculation of available buffers does not necessarily match the number of buffers really available, since the counter that I decrement in heart_rate_meas_send() may decrement more than the total number of TX buffers without BLE_ERROR_NO_TX_BUFFERS being triggered.

Any help is appreciated! Would an nRF52 help with this issue?

[Tosa]

I verified with the 7.5ms connection events with a sniffer. As expected, the sniffer snows mostly 2x packets every connection event and sometimes 3x. I'm using SDK10.0.0 which has S110 8.0.0. I'm not sure about the chip rev; how do I tell? I'm using a DK marked v1.1.0.

The sniffer logs looks good for the most part (I'm using a TI sniffer which is not a versatile as a Frontline BPA LE so I haven't been able to run it long enough to see what happens when a BLE_ERROR_NO_TX_BUFFERS error occurs).