This corroborates some of take-aways from the "sensor party" readings above:
"Desk" seems to be a few points high
"BA" and "Desk 3" are a few points low
"Desk 4" is broken / 30-50% too low
the gap between lowest and highest sensors (excluding "Desk 4") seems to be 7-8%
I don't know if these numbers seem reasonable or normal for these sensors; they are advertised as having Β±2% accuracy in the [5%,95%] range, so it doesn't seem too far off-base.
The trouble with "Desk 4"
It seems clear that the humidity (but not temperature) reading from desk4 are wrong (30-50% too low).
I first used this sensor for ~5d at my in-laws' house over Thanksgiving:
I thought it was weird that it was in the 40s at first, and then spent most of the next few days in the 10-20% range. The house's temp+humidity system thought it was relatively stable around 40%, which makes me think that the multi-day decreasing arc above depicts desk4 becoming uncalibrated (for some unknown reason).
Another possibility is that desk4 was correct above (and my in-laws' house did dry out over that period) but then broke on my journey home: it was in my backpack and I got rained on a bit, so that the outside of the backpack was damp, and the inside was likely very humid.
In any case, it's been obviously wrong since I got home.
Report findings on adafruit forums
TODO: post on the adafruit forums and see if I should e.g. expect a refund on one or more of these, or what tips they might have.
I wanted to test the concordance of my 7 HTU21D sensors, so I put them close together in a relatively-stable-climate area of my apt:
Sensor Party!
There is a pretty consistent bias in humidity readings for each sensor:
The most skewed pair of sensors is about 7% apart.
Additionally, one sensor not pictured above,
desk4
, hovers about 40% below the others (including dipping below 0):Temperature readings show similar persistent per-sensor biases, but the largest magnitude is only about 1ΒΊF (and
desk4
is middle-of-the-pack):Boveda calibration packs
I also rotated the sensors through Boveda calibration packs (a 75% pack I'd had for a year, as well as a new one, and a new 32% pack):
This corroborates some of take-aways from the "sensor party" readings above:
I don't know if these numbers seem reasonable or normal for these sensors; they are advertised as having Β±2% accuracy in the [5%,95%] range, so it doesn't seem too far off-base.
The trouble with "Desk 4"
It seems clear that the humidity (but not temperature) reading from
desk4
are wrong (30-50% too low).I first used this sensor for ~5d at my in-laws' house over Thanksgiving:
I thought it was weird that it was in the 40s at first, and then spent most of the next few days in the 10-20% range. The house's temp+humidity system thought it was relatively stable around 40%, which makes me think that the multi-day decreasing arc above depicts
desk4
becoming uncalibrated (for some unknown reason).Another possibility is that
desk4
was correct above (and my in-laws' house did dry out over that period) but then broke on my journey home: it was in my backpack and I got rained on a bit, so that the outside of the backpack was damp, and the inside was likely very humid.In any case, it's been obviously wrong since I got home.
Report findings on adafruit forums
TODO: post on the adafruit forums and see if I should e.g. expect a refund on one or more of these, or what tips they might have.