fix(repeater/ble): widen conn interval 15-30 -> 60-80 ms; pktfeed ring 32 -> 64

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
tthandClaude Fable 5 committed 2026-07-21 13:43:34 +02:00
1 parent 75b4434d61
commit 5afeca1b76
2 files changed
+15 -6

No files matched your search

+9 -2
View File
@@ -50,8 +50,15 @@
// freeze is still caught by the 20 s hardware watchdog (16 s < 20 s). Like the
// companion, request the params explicitly on 'secured' (some centrals ignore
// the advertised PPCP).
#define RPT_BLE_MIN_CONN_INTERVAL 12 // 15 ms (1.25 ms units)
#define RPT_BLE_MAX_CONN_INTERVAL 24 // 30 ms
// Interval 60-80 ms (was 15-30 ms): a controlled A/B test (21 Jul 2026) showed
// ~40 % of DIRECT-route LoRa forwards silently lost with an IDLE BLE
// connection at 15-30 ms — the SoftDevice's per-event preemption disturbs the
// SX126x TX path (startTransmit fail / send-complete timeout). Fewer
// connection events = proportionally fewer collisions; the Dispatcher TX
// retry (MESH_TX_RETRIES) covers the rest. Console throughput drops ~3x,
// which the duty-cycled bridge tolerates fine.
#define RPT_BLE_MIN_CONN_INTERVAL 48 // 60 ms (1.25 ms units)
#define RPT_BLE_MAX_CONN_INTERVAL 64 // 80 ms
// latency 0 (not the companion's 4): the mast link is marginal (RSSI -75..-90)
// and the disconnects are HCI reason 0x3E (lost sync), not supervision timeout.
// With latency>0 the node may skip connection events -> the central loses sync
+6 -4
View File
@@ -94,11 +94,13 @@ build_flags =
; neighbour advert period even on inversion evenings.
-D MAX_HEARD_NODES=128
-D RXLOG_SIZE=512
; PKTFEED_SIZE 32 (~268 B/rec, 8.6 KB): observer feed ring -- like rxlog but
; PKTFEED_SIZE 64 (~268 B/rec, 17 KB): observer feed ring -- like rxlog but
; with full raw frame bytes (`pktlog` cmd) for the meshcoretomqtt-compatible
; community analyzer feed. 32 records ≈ 4+ min of typical traffic between
; 5 s bridge polls; drop detection via the E-trailer catches any overflow.
-D PKTFEED_SIZE=32
; community analyzer feed. The bridge now polls in a connect->drain->disconnect
; duty cycle (BLE connections disturb LoRa TX, see MESH_TX_RETRIES), so drains
; are ~3-4 min apart: 64 records ≈ 10+ min of typical traffic buys margin for
; busy evenings; drop detection via the E-trailer catches any overflow.
-D PKTFEED_SIZE=64
; -D MESH_PACKET_LOGGING=1
build_src_filter = ${SenseCap_Solar.build_src_filter}
+<../examples/simple_repeater/*.cpp>