A humidity sensor that slowly lied
In thick fog the relative humidity should read 100%, or close to it. This spring I noticed that mine topped out at 94% on mornings when I could not see the end of the street.
Finding it in the data
A single foggy morning proves little, so I took the highest reading of each month since the sensor was installed. Every month has at least a few saturated nights, which makes the monthly maximum a rough check of the top of the scale.
| Period | Highest reading |
|---|---|
| First winter | 100% |
| First summer | 99% |
| Second winter | 97% |
| This spring | 94% |
The ceiling had been sinking by two or three points a year. Nothing in the daily plots gave it away, because the daily swing looked as healthy as ever.
Why it happens
Capacitive humidity sensors use a polymer film that absorbs water. Long periods near saturation, plus dust and whatever the chimney next door emits, change how the film behaves. The maker's datasheet says as much and suggests a reconditioning bake, which I tried. It recovered about one point.
What I changed
I replaced the sensor in early September and kept the old one running beside it for a week. The new one read 5 points higher at the top of the range and about 1 point higher at 50%, so the error was not a simple offset.
I have not corrected the old readings in the data file. A correction would be a guess, and I would rather publish a known flaw than an invented fix. From now on I check the monthly maximum, which takes one line of script and would have caught this a year earlier.