Timestamps read dd/mm/yyyy hh:mm on this clock; the column
header names it.
The clock the export is ordered by, and the one averaging buckets are cut on. A raw export still carries both timestamp columns.
Averaged per device, on the wall clock of the timezone chosen above.
N/A readings (-1) are left out of every average.
Blank keeps every row. Filters run before averaging, so a dropped reading enters no bucket and no MEAN. A row missing the value a filter tests is dropped.
Scans one component's readings, per device, for values outside the physical range, runs of 3 or more consecutive zeros, and runs of 5 or more identical (stagnant) readings — plus any of the row and continuity checks below that you fill in. A device passes when at most 1% of its readings are affected and nothing is missing. Click a result row for the readings behind its numbers.
Prefilled per component; leave a side blank for unbounded.
Blank leaves a check out. These look at the row itself, so they apply whatever component is selected above.
Reports every break longer than interval + tolerance between one measurement and the next. One interval covers the whole run, so select devices that share a cadence; the fleet's is 5 minutes.
| Name | MAC | Latitude | Longitude | Upload Token |
|---|
Per-device, versioned correction coefficients. The active set corrects new readings as they arrive; any set can recompute a past window via Backfill. Raw measurements are never changed.
| Ver | Label | Active | Created |
|---|
Older loggers that have no GYA board: they write into a Google Sheet, and the AirThess pipeline reads those sheets and uploads the readings here under a synthetic MAC. From then on they behave like any other device. One API key authorises the whole set — the uploader holds it, and it also guards this page (the sheet URLs are effectively capability links to the data).
| # | Device MAC | Name | Lat | Lon | Sheet URL | Polled |
|---|