Inzichten · 6 min lezen
IoT and telemetry sample CSVs: device IDs, timestamps, and gaps
Engineers often snapshot device telemetry as CSV before Parquet or time-series stores. Here is how to read bursts, missing readings, and clock skew in flat files.
Gepubliceerd op 19 maart 2025 · Table
Firmware and gateway teams dump sample windows to CSV for support tickets and algorithm tuning. Rows usually include device ID, sensor channel, timestamp, and reading. Real-world files mix UTC and local time, duplicate samples, and null placeholders that are not true numeric gaps.
What to inspect
- Timestamp format and timezone: ISO-8601 with offset is safest.
- Device ID cardinality versus row count to spot accidental cartesian explosions from joins.
- Sentinel values like
-999that mean offline instead of measurement. - Gap detection by sorting on time per device and eyeballing jumps larger than the expected sampling interval.
Table for ad hoc slices
Paste or load a CSV slice into Table, sort by timestamp, filter to one device ID, and scroll without standing up a notebook. Export a cleaned subset once outliers are removed.