Wrong readings

"Wrong readings" almost never means the sensor is lying. Modern SHT3x and BME280 chips are accurate to ± 0.4 °C and ± 2 % RH straight off the reel. What is wrong is usually where the probe is.

1 · Is it the probe placement?

The single most common report on the helpdesk: "the fridge is set to +4 °C but OpenSense says +1 °C".

Cold air sinks. Hot air rises. If your probe is at the bottom of a walk-in fridge, it reads colder than the food on the top shelf. If your probe is near the evaporator, it reads −2 °C while the door is open and food on shelf 3 is at +9 °C. Move the probe to the warmest stored item and trust the reading.

For a display chiller with two shelves, the upper shelf is warmer than the lower one by 2–4 °C. For a chest freezer, the top is warmer than the bottom by 3–6 °C.

2 · Did someone tape the sensor to a cold surface?

The Shelly H&T's plastic shell conducts heat slowly but its mounting magnet conducts well. If it is mag-mounted to the steel rail of the evaporator coil, the reading drifts toward the rail temperature, not the air. Re-mount with Velcro on a non-conducting surface (plastic shelf, the door's inside skin, the underside of a wire shelf).

3 · Is the cadence the problem?

If readings appear to lag the real temperature — door opened, you expect a spike, you do not see one for two minutes — the cadence is too slow. Default is 60 s on Shelly, configurable up to 30 s. For HACCP set 60 s. For monitoring door-open events, set 30 s.

LoRaWAN devices have a minimum cadence of 5 minutes in EU868 due to duty-cycle limits; do not buy LoRaWAN for door-event monitoring.

4 · Is the value drifting?

If readings track a real-world reference at first and then slowly drift (e.g. one degree per month), the chip is contaminated. Common cause: volatile solvents in the room (cleaning sprays, perfume) hit the polymer membrane on the humidity sensor and shift its zero point. Sensirion SHT3x can be rebaked: leave the sensor at 60 °C in dry air for 24 h; recovery is usually 80 %.

If you cannot bake or do not have time, calibrate-out the drift in the channel's offset field. Audit-trail-safe: the offset is recorded as a configuration change with a timestamp.

5 · Is it the wrong sensor entirely?

A surprisingly common scenario: someone bought a "Shelly H&T" that turned out to be the older Gen1 unit with a different chip (HDC1080 or DHT22 in some counterfeits). Both are factor-of-2 worse on humidity. Symptom: RH reading is high, the room actually feels dry.

Check the device's Device info page: it lists the chip model. If the chip is not an SHT3x, replace the unit.

6 · Is it the rule, not the reading?

A rare but worth-it check: the channel's operating range may have been edited to a narrower band, so the same in-bounds reading is now considered "out". Look at the channel history; the right panel shows the last 10 changes to thresholds.

When the reading really is wrong

In the rare case that the sensor itself is faulty:

• Shelly: shrug, replace. They are €30 retail and the warranty path is Amazon's, not ours.
• Aqara: refresh battery, repair to gateway. Has been observed to recover after a deep-discharge and one pairing cycle.
• Efento: contact Efento's support — these are €100 sensors and the vendor is responsive.
• DIY ESP32: re-solder the BME280 connections, check the I²C pull-ups (the breakout boards have them but the bare chip does not).

For HACCP compliance: if the reading was wrong and an alarm was issued, leave the alarm in the audit trail. Add a free-text note explaining the diagnosis and the corrective action. Auditors prefer this to a clean log that you "edited".