Diagnostics

Everything a host (or a person with a bus adapter) can read to answer "is this module healthy, and if not — why".

STATUS (0x0006)

Bit	Name	Meaning
0	CAL_OK	every channel calibrated. set from a marker stored with the calibration when it is saved; see also `CAL_SRC`
1	SENSOR_OK	measurement path alive (measurement frames arriving)
2	MAINS_OK	mains present: channel-0 voltage above ~50 V (not a frequency check; FREQ reads 0 out of band)
3	LOCKED	reserved (production lock)
4	TIME_VALID	clock has been set since boot
5	NVM_FAULT	energy store FAULT/CORRUPT — energy not persisting
6	—	service mode indicator

RESET_CAUSE (0x00CC) — why this boot happened, raw reset flags: bit0 PIN · bit1 POWER · bit2 SOFTWARE (reset command, update activation, address change) · bit3/4 watchdog · bit5 LOW_POWER · bit6 OPTION. PIN accompanies most internal resets — judge by the other bits.
BOOT_COUNT (0x00CE) — persisted boot counter (low 16 bits). A master that missed a restart sees it step; pair with EVENT_FLAGS bit 5 (RESET_OCCURRED).
UPTIME (0x08A0) — seconds since boot; decreasing between polls = restart.

Reg	Healthy	Meaning when not
FRAME_CNT (0x0020)	strictly increasing	frozen = acquisition stopped (SENSOR_OK will drop)
LOST_FRAMES (0x012E)	flat (small constant steps during firmware updates are normal and energy-compensated)	growing in normal operation = acquisition problem
RING_DROP (0x0106)	0	sample loss under load — report it
WIN_NSAMP / WIN_PERIOD (0x0110/0x0112)	stable; WIN_NSAMP ≈ 5 mains periods per window at 50 Hz / 6 at 60 Hz	unstable = mains lock problems
CRC_ERR (0x0024) / FRAME_OK (0x0104)	0 / 1	measurement-link errors

Reg	Meaning
LTA_STATUS (0x0870)	energy store: 0 OK · 1 new (first life) · 2 FAULT · 3 CORRUPT. 2/3 → STATUS bit 5; metering continues, persistence resumes when the store recovers
LTA_SRC (0x0871)	where the counter came from at boot: 0 none · 1 NVM · 2 retention · 3 legacy (4 reserved)
LTG_STATUS (0x0873)	per-channel counters: 3 = restarted at 0 (new/foreign storage or layout change); transient — clears after the first write (~1 s)
CAL_SRC (0x0872)	3 = stored calibration not for this firmware → defaults in use, needs service
EE_RD_ERR / EE_WR_ERR / EE_RECOVER (0x0874–0x0879)	store I/O error/recovery counters
LTA_AGE_S (0x087A)	seconds since the last successful energy write. A write happens only when the counters changed, so large values at idle / very low load are normal (saturates at 65535) — judge store health by LTA_STATUS / LTA_FAIL_CNT, not by age
NVM_REJ_CNT / NVM_REJ_ADDR (0x0898/0x089A)	records rejected as not written by this device
SEC_ROLLBACK (0x0BED)	reserved (reads 0)

LAST_EXCEPTION (0x0158 / holding 0x830A) — the last Modbus exception with function code and address: the first thing to read when a SCADA integration "sometimes gets errors".
I2C_DIAG (0x013A) — bus health counters (address matches / bus errors / overruns).
CMD_SEQ / CMD_STATE — command flow sanity (Commands).

LED	Pattern	Meaning
green	short flash every ~2 s	metering runs
green	double flash	first ~8 s of every boot (until the image confirms) or an update session is open
green	off	module dead · LEDs disabled · no measurement frames for ~0.5 s
red	short blink	a request addressed to this module
red	steady	fault: no measurement frames, or energy store FAULT/CORRUPT
both	alternating ~5 Hz for 10 s	IDENTIFY command (0x0E) — locating the module

CFG_LED = 0 turns the LEDs off (IDENTIFY still blinks).

Poll cheap, alert on:

STATUS: cal_ok, sensor_ok, mains_ok set; nvm_fault clear; time_valid set (else re-feed time).
FRAME_CNT advancing; RING_DROP = 0; LOST_FRAMES flat.
BOOT_COUNT / UPTIME — unexpected restarts; after one: re-arm EVENT_MASK/thresholds, re-set time, re-anchor energy deltas (max(0, new − last)).
LT_GEN unchanged (else re-anchor energy, check the event log for who reset it).
LTA_STATUS ≤ 1, LTA_FAIL_CNT flat, CAL_SRC ∈ {1, 2}.
Event log (HIST_PERIOD = 2) drained periodically — resets, adopts, calibration saves, firmware commits, rejected-record events, with before/after values.