A sensor went quiet, and the card already knows why

The wet bay sensor has not reported since Tuesday. It is Friday, there is a hard frost forecast, and the freeze watch you thought you had has not existed for three days. Nothing told you, because from the hub’s point of view a sensor that has stopped talking and a sensor with nothing to report look exactly alike: silence.

Silence has three causes and they want three different Saturdays. The sensor is out of range and the hub needs moving. The cell is dead and needs changing. Or somebody picked the thing up, and you want to know who.

The good news is that you almost never have to guess. Every clue that separates the three is already sitting on the sensor’s own card, recorded in the last frames before it went quiet — and this article is mostly about reading them in the right order.

Stale and offline are two different failures

A broadcaster that stops broadcasting shows as stale. A device the hub has to connect to and cannot reach shows as offline. Both land in the same place on the dashboard and they do not mean the same thing at all.

That matters most for the devices where offline is normal. A Micro-Air EasyStart only answers while the air conditioner has power, so it is unreachable between compressor cycles as a matter of routine. A Power Watchdog will read as unreachable for as long as somebody has the Hughes app open holding the connection. Neither is a fault, and treating either as one is how people end up ignoring the state that does matter.

For a genuine broadcaster, the working rule is that minutes are radio noise and an hour is a sensor that has gone. A sensor at the edge of range does not fail cleanly, it frays — it reports, misses a handful, reports again — so anything that pages you on a single missed reading is training you to mute it, and a muted system is worse than no system, because you have stopped checking.

Reading the card, in order

a. The last battery reading

First, because it is the most common answer and it is right there. A last reading under 20 per cent, or a steady slide over the preceding weeks, is a dead cell. So is a reading of 90 per cent taken on a warm afternoon by a sensor that went quiet on the first frost — a coin cell reads high when it is warm and still cannot deliver the current for a radio burst when it is cold.

Treat the number as a hint rather than a measurement. A percentage is a manufacturer’s curve laid over a voltage, flat for months and then a short cliff, so a comfortable figure proves very little. A RuuviTag reports volts instead of a percentage, which is more honest and takes more knowing. The Jinou‘s battery comes from a stable byte position rather than something we checked against a reference, so read it as indicative. And a ThermoPro TP357 broadcasts only a low-battery flag rather than a level, so no battery figure appears for it at all and you go straight to the next clue.

b. How loudly the hub was hearing it

Every device card carries a signal strength in dBm — the hub’s own measurement of how loudly it hears that sensor, not something the sensor sends. Roughly: −50 to −70 is comfortable, −85 and worse is the edge. A sensor that sat at −90 for a week before it went quiet was a range problem waiting to happen, and no cell you fit will change that.

Once a sensor goes stale the figure freezes at the last frame received, which is exactly the reading you want. Two things move it in ways people find unintuitive: aluminum siding bounces the signal, so nudging a sensor six inches sometimes fixes everything and sometimes changes nothing at all, and the refrigerator is a large insulated metal box sitting in the middle of most floorplans, directly between a hub at the front and a sensor at the back. Why sensors drop out at the back of the rig is the longer version.

The one worth knowing about because it is seasonal rather than structural: water absorbs 2.4 GHz. A full fresh tank between the hub and a bay sensor is a real obstacle, which is why a sensor can report perfectly on the way to the dump station and drop out on the way home.

c. What it was reading when it stopped

The last temperature the sensor reported is the tie-breaker between the first two. Below freezing, gone overnight, back by the middle of the morning — that is a cold cell, not a range problem, and the daytime percentage is lying to you. It will do the same thing every cold night until the cell is replaced, and lithium rather than alkaline is the fix in anything that lives outside the heated space.

d. Whether anybody touched it

A RuuviTag counts its own movements, so a tag that has sat at the same count for a month and then ticks up ten times in a minute before going silent was picked up. That is the one clue on this page that distinguishes moved from died, and it is a good reason to put a RuuviTag on the sensor you would least like to lose. Driving moves the whole rig and slamming a bay door moves what is stuck to it, so read the count with the clock.

Failing that, a Shelly BLU door or motion sensor near the same place answers a narrower question: a bay door opening a minute before the sensor in that bay went quiet says somebody was in there. It does not say what they did, and it will not tell a thief from a helpful neighbor.

The combinations

• Stale, battery healthy, last signal −85 or worse — range. Move the hub to a central outlet at waist height, out of any cabinet. Neither the sensor nor the cell is at fault.
• Stale, battery under 20 per cent or sliding since spring — change the cell, and write the date on the tape while you are there.
• Stale overnight, back by noon, last temperature below freezing — a cold cell. It has enough to run warm and not enough to run cold, and it will repeat every frost until you replace it.
• Stale, and the movement count jumped or a bay door opened in the minute before — somebody handled it. Go and look, or ask.
• Every sensor on one hub goes stale in the same minute, and the hub is still online — not the sensors. The hub’s radio or its position changed: shut in a cabinet, wrapped in something metal, or the Wi-Fi router ended up sitting on top of it.
• The slow sensor goes stale several times a day and the fast one beside it never does — your stale window is shorter than the slow sensor’s broadcast interval. That is a setting, not a fault.

What none of this tells you

The hub knows when a sensor went quiet and never why. It cannot tell a dead cell from a stolen sensor from one that fell down the back of the sofa, and no threshold you set will make a sensor at the edge of range reliable — only moving the hub, or adding one, does that.

Which is the honest reason this whole family matters more than it looks. Every other article in the cookbook assumes the readings are arriving. This is the one that checks.

What to do about it

Let stale be the trigger and the card be the diagnosis, in the order above: battery, then signal, then the temperature it was reading when it stopped. Three glances, and most of the time you will know before you have put your boots on.

If it is range, try the free fix first — a central outlet at waist height, out of the cabinet, away from the fridge — because it works more often than people expect. If that fails, a second hub at the far end of a long rig settles it in one step, and hub count is not something we meter: every plan stores data for two hubs, and a third is added on request at no charge.

And do the thing the wet bay sensor at the top of this page needed: before the season you are relying on it for, look at the timestamps rather than the values. A number that has been reassuringly steady for a very long time is the one to be suspicious of.

Where to get one

Only one purchase in this article is really about this problem: a RuuviTag, on the sensor whose disappearance you would most want explained. The compatibility database shows which of the rest report a battery level at all, which is worth a look before you buy the cheap one for somewhere cold.