BLE Data Throughput regression on NUCLEO-WBA65RI (Server) with STM32CubeWBA v1.8.0 vs v1.7.0

[SH288972]

Hello,

I'm running the BLE_DataThroughput_Server Application on a NUCLEO-WBA65RI (server, notifying) connected to a NUCLEO-WBA55CG (client, receiving notifications), and I'm seeing a significant throughput regression between STM32CubeWBA v1.7.0 and upper versions, with strictly identical configuration.

 

Setup (identical in both cases):

• Server: NUCLEO-WBA65RI, BLE_DataThroughput_Server Application, from STM32CubeWBA v1.7.0, v1.8.0, and v1.10.0

• Client: NUCLEO-WBA55CG,  BLE_DataThroughput_Client Application, from STM32CubeWBA v1.10.0

• Connection interval: 11.25 ms
• PHY: 2M
• MTU size : 244
• Build configuration: Release
• I added :

const uint8_t a_BLE_rx_pckt_count_and_rx_pckt_len[2] = { 0x07, 0x00 }; /* rx_pckt_count , rx_pckt_len */ aci_hal_write_config_data(CONFIG_DATA_LL_RX_ACL_CTRL_OFFSET, 2, a_BLE_rx_pckt_count_and_rx_pckt_len);

as described in the "How to get the best throughput" section of the wiki (rx_pckt_count=7, rx_pckt_len=0)

 

Results:

• BLE_DataThroughput_Server Application from STM32CubeWBA v1.7.0 → ~130 KB/s (consistent with the ~120 KB/s figure in the throughput measurement wiki page)
• BLE_DataThroughput_Server Application from STM32CubeWBA v1.8.0 and v1.10.0 → ~58 KB/s, same hardware, same config, no other code change

 

Questions:

I see in the v1.8.0 release note :

BLE:

• Support of BLE 6.0
• Man-In-The-Middle (MITM) management is enhanced in the BLE stack to provide more flexibility and better interoperability.
• Add BLE Host Stack events FIFO buffer mechanism

1. Is this a known regression from v1.8.0 ?
2. Is there a fix planned, or a recommended workaround to recover v1.7.0 throughput level, while staying on v1.8.0 or upper versions ?

 

Happy to share full traces / project files if useful.

Thanks in advance.