Principles of long-term logging
A long recording almost always fails for logistical reasons, not because of the measurement itself. If power is lost or the memory fills up halfway through, the test is wasted. That is why preparation counts for more than fine configuration.
How to choose the sampling interval
It must be fast enough to capture the changes of interest, with margin based on the frequency content of the variable, and not so fast that it generates unnecessary data. A common approach is to measure faster and store averages or maximum and minimum values for each interval, so that brief peaks are not lost. Averaging also reduces mains noise.
Reliable setup
- Power: backup for the data acquisition system and the computer, or a standalone logger with its own memory.
- Connections: firm terminals and secured wiring; intermittent readings during the night are hard to explain.
- Sensors: suitable for staying on site for a long time, stable and free of self-heating from excessive current.
- Clock: synchronized with a reference, because the drift of a unit’s clock accumulates error over time and makes it harder to correlate with other events.
Data
- Calculation of the space required: channels, size of each reading and number of samples.
- Periodic writing to disk and intermediate backups, so as not to lose everything in a failure.
- Logging of events and conditions, such as alarms, range changes and ambient temperature.
An example of pre-run verification
Before leaving the system alone, do a short test run, review the saved data, simulate a power cut to see whether it recovers, and confirm that the channel names and units are correct. It is best to review part of the log the next day, not at the end.
If the sensors are on power or high-voltage circuits, the installation must be done by qualified personnel following the applicable regulations.