bsncubed/aes67-ESP32-P4
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Step 6.3: PTP hybrid mode (unicast Delay_Req/Resp), step 6 done

Browse Source 
- As TimeReceiver in ptp.mode hybrid: Delay_Req unicast to the GM (IP
  from its Announce, MAC recorded by the RX hook from its Sync frames),
  unicastFlag set.
- As TimeTransmitter: a Delay_Req with the unicastFlag gets a unicast
  Delay_Resp; multicast requests get multicast replies.
- EMAC timestamps every received frame (en_ts4all): its PTP filter left
  unicast Delay_Req unstamped ("no HW RX timestamp").
- A unicast Delay_Resp's logMessageInterval 0x7F is ignored (use
  ptp.log_delay_req); log_us() clamps to -7..6 (the shift was undefined).
- Verified vs ptp4l --hybrid_e2e 1: board hybrid TimeReceiver 18
  Delay_Req in 20 s, all answered, locked; board TimeTransmitter
  (p1 100): tcpdump shows Sync/Follow_Up/Announce to 224.0.1.129,
  Delay_Req 192.168.192.233 -> .244 [unicast] and Delay_Resp .244 ->
  .233 [unicast] within 0.4 ms; ptp4l s2.
- Docs: TimeTransmitter/hybrid notes; step 6 ticked.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Ben Nicholson
2026-09-25 08:21:09 +10:00
parent dcea39b216
commit 5bd1a20e43
5 changed files with 67 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/aes67-core-base.md
+5
View File
@@ -60,6 +60,11 @@ Reference devices for UI and defaults: Riedel Bolero (PTP status), Riedel Direct
- RX timestamps: a hook on the driver's info input path (`esp_eth_update_input_path_info`) records port-319 event timestamps and forwards every frame to lwIP. TX: Delay_Req is a raw Eth/IPv4/UDP frame sent with `esp_eth_transmit_ctrl_vargs` for its HW timestamp. Needs `CONFIG_ETH_TRANSMIT_MUTEX`.
- Servo: step on the first Sync from a new GM (frequency seeded from the measured rate), then linuxptp-style PI (kp 0.7 / ki 0.3 at 1 Sync/s, scaled by the interval); re-step above 1 ms. Locked: 8 Syncs below 1 µs; unlocked after 3 above. GM loss keeps the frequency (holdover).
- Path delay is corrected for offset drift between t2 and t3 until the clock is syntonised; delay statistics start at lock.
- TimeTransmitter/BMCA (step 6): own dataset from the role; a foreign GM is followed only if better; LISTENING -> MASTER after the announce receipt timeout. Two-step Sync (raw frame, HW TX time in Follow_Up), Announce with the PTP timescale flag, Delay_Resp with HW RX time. Hybrid: as TimeReceiver, Delay_Req unicast to the GM (IP from Announce, MAC from its Sync) with the unicastFlag; as TimeTransmitter, a unicast Delay_Req gets a unicast Delay_Resp.
- The EMAC's PTP filter only timestamps multicast PTP; unicast Delay_Req got none. The EMAC now timestamps every received frame (`emac_ll_ts_all_enable`); the RX hook picks the port-319 PTP event messages.
- A unicast Delay_Resp carries logMessageInterval 0x7F; then ptp.log_delay_req is used.
- As GM without an earlier lock the clock starts at 0 (1970): no RTC/NTP. Media timing is unaffected; NTP seeding could follow with the internet interface (phase 2).
- Sync send times jitter by ~10 ms (FreeRTOS 100 Hz tick); accuracy is unaffected (two-step).
- Measured vs ptp4l (Intel i210 GM, non-PTP switch): lock in ~35 s cold, ~22 s after GM loss; offset within a few hundred ns; board crystal -39.8 ppm. Link asymmetry (1G GM / 100M board through a store-and-forward switch) adds a constant offset error of a few µs that no receiver can see.
### PTP status (`status.ptp`) — main panel modelled on Riedel Bolero "PTP Status"