BLE: intermittent connect, repeated disconnect reason 0x3E during connection attempts

[BadreddineNebhen]

Hi ST Community,

I’m working with STM32WBA (STM32WPAN / BlueNRG-style stack APIs). I’m using Extended Advertising on the peripheral (connectable + scannable, secondary PHY = Coded) and a STM32WBA central that scans and then calls aci_gap_create_connection().

Problem
Connection is intermittent: sometimes it connects, but often the peripheral prints multiple times:

HCI_DISCONNECTION_COMPLETE_EVT_CODE
- Connection Handle: 0x01
- Reason: 0x3E
(repeats 4–5 times)

The central does retries very often, so the peripheral sees repeated disconnect events before it finally connects.

Setup details

Peripheral (Extended Advertising)

• Advertising set configured as connectable + scannable:

  • u16Property = HCI_ADV_EVENT_PROP_CONNECTABLE | HCI_ADV_EVENT_PROP_SCANNABLE;

• Primary PHY = 1M, Secondary PHY = Coded (0x03):

  • aci_gap_adv_set_configuration(..., primaryPHY=0x00, secondaryPHY=0x03, ...)

• Address type used in advertising set:

  • u8AddressType = GAP_STATIC_RANDOM_ADDR;

• Advertising channels 37/38/39 enabled

• TX power set via aci_hal_set_tx_power_level(1, CFG_TX_POWER)

Central (Extended Advertising enabled)

• Scan finds target via HCI_LE_EXTENDED_ADVERTISING_REPORT_SUBEVT_CODE

• Central then calls aci_gap_create_connection()

Here is the current connect code (simplified):

result = aci_gap_create_connection(
    SCAN_INT_MS(500), SCAN_WIN_MS(500),
    GAP_STATIC_RANDOM_ADDR,       // Peer_Address_Type  <-- (my current code)
    g_nSelection.pu8TargetAdr,
    CFG_BD_ADDRESS_TYPE,          // Own_Address_Type
    CONN_INT_MS(7.5), CONN_INT_MS(7.5),
    0,
    u16ConnSupTimeout,
    CONN_CE_LENGTH_MS(10), CONN_CE_LENGTH_MS(10)
);

I also tried less aggressive intervals (e.g. Min 30 ms, Max 250 ms) and it got “better”, but the repeated 0x3E disconnect still happens sometimes.

Questions

1. For Extended Advertising connections, should the central use the peer address type from the advertising report (public/random) instead of constants like GAP_STATIC_RANDOM_ADDR?

   • My suspicion: I’m misusing Peer_Address_Type (should be 0=public / 1=random), and sometimes the stack interprets the peer address type incorrectly, leading to fast failure/disconnect reason 0x3E.

2. Is there a recommended stable set of parameters for:

   • LE_Scan_Interval / LE_Scan_Window during connection establishment

   • Conn_Interval_Min / Max for LE Coded / extended advertising

   • Minimum_CE_Length / Maximum_CE_Length (should these be 0?)

3. Any known issue/errata with STM32WBA + Extended Advertising + Coded PHY that can cause repeated disconnect reason 0x3E during initial connection?

Additional info

• Peripheral prints repeated disconnect reason 0x3E before a successful connection happens.

• Central is retrying connection frequently.

• Addresses themselves look correct (I see the target address from scan and use it for connect).

Any guidance on correct use of aci_gap_create_connection() with Extended Advertising (especially peer address type handling) and recommended parameters would be greatly appreciated.

Thanks!