docs: update README to current lora-refactor state

README.md

@@ -10,6 +10,8 @@ Firmware for LilyGO T3 v1.6.1 (`ESP32 + SX1276 + SSD1306`) that runs as either:
 - LoRa link uses explicit CRC16 frame protection in firmware (`src/lora_transport.cpp`).
 - Sender batches up to `30` samples and retries on missing ACK (`BATCH_MAX_RETRIES=2`, retry policy `Keep`).
 - Sender meter parsing is decoupled from LoRa ACK waits using a dedicated FreeRTOS reader task + queue (`src/main.cpp`).
+- Batch payload codec is schema v3 and uses a sparse `present_mask` over a 30-second window (no schema v2 compatibility).
+- Sender derives epoch timestamps from meter Sekundenindex (`0-0:96.8.0*255`) using an epoch anchor when time is synced.
 - Receiver uses STA mode when config is valid, otherwise AP fallback with web config.
 - No debug auto-reboot timer is active in normal firmware loops.
 
@@ -29,11 +31,14 @@ Transport is chunked (`batch_id`, `chunk_index`, `chunk_count`, `total_len`) and
 
 Payload codec (`src/payload_codec.cpp`) currently uses:
 - `kMagic=0xDDB3`
-- `kSchema=2`
-- metadata: sender, batch, timestamp, interval, battery, fault counters
+- `kSchema=3`
+- metadata: sender, batch, `t_last`, `present_mask`, battery, fault counters
 - data arrays: `energy_wh[]`, `p1_w[]`, `p2_w[]`, `p3_w[]`
 
-`n == 0` is valid and used for sync request packets.
+`present_mask` is a 30-bit second map in the `[t_last-29, t_last]` window.
+Only set bits carry samples, so missing seconds are explicitly represented.
+
+`n == 0` is valid for sync request packets (`present_mask == 0`).
 
 ### `AckDown` (7 bytes)
 
@@ -66,14 +71,21 @@ Implemented in `src/meter_driver.cpp` + sender loop in `src/main.cpp`:
 
 - UART: `Serial2`, RX pin `GPIO34` (`PIN_METER_RX`), `9600 7E1`
 - ESP32 RX buffer is enlarged to `8192` bytes to survive long LoRa blocking sections.
-- Frame detection: starts at `'/'`, ends at `'!'`, timeout protection included (`METER_FRAME_TIMEOUT_MS=20000`).
+- Frame detection: starts at `'/'`, ends at `'!'`, timeout protection included (`METER_FRAME_TIMEOUT_MS=3000`).
 - Parsing runs in a dedicated sender task and is handed to the main sender loop via queue.
 - Parsed OBIS values:
+  - `0-0:96.8.0*255` (meter Sekundenindex, 4-byte hex)
   - `1-0:1.8.0` (total energy)
   - `1-0:16.7.0` (total power)
   - `1-0:36.7.0`, `56.7.0`, `76.7.0` (phase powers)
 - `1-0:1.8.0*Wh` is automatically scaled to kWh
 
+Timestamping/validation on sender:
+- Anchor when time is valid: `epoch_offset = epoch_now - meter_seconds`.
+- Derived sample time: `ts_utc = meter_seconds + epoch_offset`.
+- Meter-time checks: monotonicity and `delta_meter_seconds` vs elapsed wall time (with tolerance) plus anchor drift checks.
+- On detected jump/drift, sender records meter fault and resulting timestamp discontinuities propagate through `present_mask` to receiver.
+
 Sender samples every second and transmits batches every 30 seconds.
 
 ## Receiver Behavior
@@ -82,8 +94,9 @@ For valid `BatchUp` decode:
 1. Reassemble chunks and decode payload.
 2. Send `AckDown` immediately.
 3. Drop duplicate batches per sender (`batch_id` tracking).
-4. If `n==0`: treat as sync request only.
-5. Else convert samples, log to SD, update web UI, publish MQTT.
+4. Track duplicate stats per sender: absolute duplicates, total received, duplicate percentage, last duplicate timestamp.
+5. If `n==0`: treat as sync request only.
+6. Else reconstruct timestamps from `t_last` + `present_mask`, preserving skipped seconds, then log to SD, update web UI, publish MQTT.
 
 ## MQTT Topics and Payloads
 
@@ -109,6 +122,8 @@ Home Assistant discovery is enabled (`ENABLE_HA_DISCOVERY=true`) and publishes c
 - AP fallback SSID prefix: `DD3-Bridge-`.
 - Default web credentials: `admin/admin`.
 - SD logging enabled (`ENABLE_SD_LOGGING=true`).
+- Sender-specific web status includes duplicate-batch counters and the last duplicate time.
+- Sender-specific OLED page shows duplicate rate as `pct (absolute)` and last duplicate as `HH:MM`.
 
 ## Build Environments