LHT65N Outage Recovery: Keep Historical Readings Out of Live Alarm Rules
After a LoRaWAN outage, a burst of arriving LHT65N messages can contain old measurements. If an application timestamps every arrival as “now”, yesterday’s temperature can appear to be the current condition. The practical answer is to preserve measurement time separately from receipt time and give recovered records an explicit historical-data path. Restoring communications and recovering evidence are different acceptance checks.
Recognise the recovered message before using its value
Dragino’s LHT65N/S documentation describes automatic recovery of previously unacknowledged readings when the relevant datalog settings are enabled. It also supports retrieving stored values for a requested timestamp range. Its datalog record contains a Unix timestamp and message flags; the documented record length is 11 bytes. Firmware, external-sensor mode and message type still need checking against the exact device.
Two flags answer different questions. The documented No ACK Message flag identifies a reading previously sent without an acknowledgement; the Poll Message Flag identifies a polling reply. Do not collapse them into one “new measurement” event. A recovery transmission is new radio traffic carrying evidence that may belong to an earlier observation time.
The pictured LHT65N is listed by ProSense Instruments with an external-probe option. The housing photograph identifies the product, not its selected radio plan, installed firmware, connected probe or datalog configuration. Confirm those separately before applying a decoder or recovery procedure.

Keep two clocks and a record classification
For each accepted record, retain the sensor identity, channel, decoded value and unit, device observation timestamp where available, server receipt timestamp and the recovery or polling flags. Preserve the sanitised raw payload alongside the decoder revision in a controlled diagnostic record. Security keys do not belong in ordinary support exports.
Use observation time for the historical trend when it has been validated. Keep receipt time for investigating delivery delays. Convert timestamps to Australia/Adelaide for the display using timezone-aware software, while retaining the underlying time value. Do not apply a fixed local offset throughout the year or silently replace an unverified device timestamp with receipt time.
Define what happens when the device clock is uncertain. Mark the time quality and retain the record for investigation instead of presenting it as a trustworthy current sample. Dragino documents time-synchronisation status separately from the temperature value; a plausible temperature does not establish a correct clock.
Prevent history from becoming a false live alarm
This is an application-design requirement, not a claim that every dashboard implements it automatically. Agree which messages may update the live-value tile, which enter historical storage and which can activate a current-condition alarm. A recovered high temperature may justify a retrospective excursion review without proving that the temperature is high at the moment of receipt.
Also protect the opposite case: an old in-range record must not clear a current alarm or make an offline device appear healthy merely because the record arrived recently. Communications recovery, measurement freshness and process condition need separate indicators. Set an application-appropriate freshness rule with the responsible team; there is no universal safe age for every installation.
Work through an isolated acceptance test
Hypothetical example: a device records 12 °C at 10:00, communications fail, and that record reaches the server at 10:40 after recovery. A fresh measurement at 10:40 is 6 °C. These are illustrative values, not recommended storage limits or a field test result.
The historical chart should place the validated 12 °C observation at 10:00 and retain its 10:40 receipt time for diagnostics. The live tile should follow the agreed freshness and classification rules. The retrospective review should remain visible; a later in-range sample should not erase the earlier event from the record.
Use a spare device or replay sanitised packets through an isolated application test facility. Include a normal current reading, a polling reply, a recovered unacknowledged reading and an uncertain-time record. Deliver the samples out of observation order, then repeat a historical sample. Confirm that reprocessing does not create duplicate trend points or duplicate retrospective actions under the agreed deduplication policy.
Define a deduplication key from the verified data model rather than assuming a network frame counter is the original measurement identity. Include the device and relevant channel or record type when assessing a possible duplicate. Retain conflicting records for investigation instead of dropping them solely because two temperatures happen to match.
Close the recovery with evidence
Common mistakes are plotting everything at receipt time, allowing the last packet received to win the live tile, interpreting a poll reply as a fresh alarm, and claiming a complete record because the graph has resumed. Reconcile the expected observation period with recovered records and document any remaining gaps. Do not invent readings to close them.
Before deployment, ask ProSense to confirm the exact LHT65N variant, firmware, probe, datalog settings, decoder and receiving-platform requirements. Supply a sanitised example with both timestamps and the expected live-versus-historical behaviour. Integration with ProSight or another platform needs separate verification; recovered data does not replace a monitoring response that was unavailable during the outage.
Need help with your application?
ProSense Instruments - explore instruments, sensors, data loggers and test equipment.
ProSense Solutions - discuss complete monitoring and automation systems.
ProSight Software - explore cloud dashboards, alarms and reporting. Device integration must be confirmed for your application.
Instrument enquiries: sales@prosenseinstruments.com.au | 1800 560 854
Project enquiries: sales@prosense.com.au. Software enquiries: info@prosightsoftware.com.




Comments