multiple packets per connection interval between nrf52 central and peripheral

[blasagna]

I am evaluating the throughput of BLE communication between 2 NRF52832's, 1 acting as a central and 1 as a peripheral. I've modified the ATT_MTU Throughput Example, and I'm attempting to configure the peripheral to send more than 1 notification per connection interval. I believe I've successfully increased the `hvn_tx_queue_size` GATTS configuration following the instructions here and the I'm avoiding `NRF_ERROR_RESOUCES` errors from `sd_ble_gatts_hvx` by tracking the available space in the queue as described in variant #2 of the GATT Server handle notification message sequence chart below.  Despite all of this, the count reported on each BLE_GATTS_EVT_HVN_TX_COMPLETE is always equal to 1.

I'm using SDK v15.0.0 and softdevice S132 v6.0.0, and queueing notifications of 223 Bytes every 4 milliseconds using `sd_ble_gatts_hvx`. The connection interval is 20 milliseconds, MTU is  247 Bytes, PHY is 2 Mbps, GAP event length 320, and 

Is there any additional configuration needed on either the central or peripheral to see more than 1 notification per connection interval?

[blasagna]

Adding some updated information. After reading the s132 migration notes for v4.0.3 section BLE bandwidth configuration, I think I have a better idea of how bandwidth configuration works. I've updated my parameters to the following:

connection inteval: 50 milliseconds

MTU: 247 bytes

PHY: 2 Mbps

GAP event length: 50 milliseconds

notification length: 223 bytes, each containing an incrementing counter

sd_ble_gatts_hvx called every 4 milliseconds

I am still always seeing the count from each BLE_GATTS_HVN_TX_COMPLETE equal to 1, but the throughput I'm observing suggests that more than 1 notification is being sent in each connection interval. By that I mean that if only 1 223 byte packet was being sent every 50 milliseconds, I'd expect to see only about 20 packets per second, but on the receiving side I am seeing the full 250 packets per second, arriving in order without losing any based on the incrementing counter in each packet.

Am I incorrect in assuming that there should be a BLE_GATTS_HVN_TX_COMPLETE once for each connection interval?

[bjorn-spockeli]

HI Bob, 

I apologize for the late reply. 

I sniffed the traffic between two nRF52 DKs running the ble_app_att_mtu_throughput example modified to print the count variable passed in the hvn_tx_complete struct. I observed multiple packets per connection interval and the  hvn_tx_complete  count is always 1, i.e. you get  the BLE_GATTS_EVT_HVN_TX_COMPLETE event per notification and not per connection event. 

Best regards

Bjørn