Power-quality datalogger for long-term logging of intermittent faults

The guide on the oscilloscope for fault analysis covers capturing a brief transient during an attended, targeted measurement session. The [guide on measuring harmonics with a power analyzer](/guides/practical/harmonischen-meten-netanalyzer-k-factor) covers a snapshot of the harmonic load at one point in time. This article covers a third deployment scenario: the power-quality datalogger that, unattended, keeps logging for days to weeks per IEC 61000-4-30, specifically to catch an intermittent or unpredictable fault pattern that would likely be missed during a brief, attended measurement.

Why a different instrument for a different problem

An oscilloscope captures a single event at the moment a technician is present and has set the trigger — suitable for a reproducible phenomenon or one that can be induced on demand. A power analyzer gives a representative snapshot of the harmonic and power-quality situation at the moment of measurement — suitable for characterising a structural load pattern. Neither is suited to a complaint such as "the drive trips randomly, a few times a week, at unpredictable times" — this requires unattended, long-duration logging that captures an event just as readily at three in the morning as at three in the afternoon.

IEC 61000-4-30: measurement classes A and S

IEC 61000-4-30 defines the measurement methods for power-quality parameters (voltage dips, swells, interruptions, flicker, harmonics, unbalance, frequency) and assigns two relevant measurement classes:

Class A: the highest accuracy class, with strictly defined measurement algorithms and uncertainty limits. Measurements from two Class A instruments from different manufacturers on the same signal must agree within the specified uncertainty. Class A is required where the measurement is contractually binding or may serve as evidence in a dispute — for example, a dispute with the grid operator over the cause of repeated dips.
Class S: a survey class with looser tolerances, intended for statistical exploration and trend analysis, not for dispute resolution.

For long-term diagnostic logging to substantiate a complaint to the grid operator, Class A is the appropriate choice; for an exploratory, internal investigation of a suspected problem, Class S may suffice.

Practical deployment

A power-quality datalogger is typically connected temporarily at the point where the fault is suspected (main distribution board, a circuit feeding a specific machine, or the grid operator's connection point) and left logging continuously for days to weeks. Key points for a meaningful deployment:

Time synchronisation: the datalogger's clock must be accurately synchronised, so a logged dip or interruption can be correlated exactly with the fault log of the affected equipment (PLC alarm, drive trip, protection relay operation) or, in a complaint to the grid operator, with its own records.
Threshold settings: the trigger thresholds for dip, swell, and interruption should match what the affected equipment actually perceives as disruptive (see also the guide on voltage dips and interruptions per EN 50160), not merely a generic factory default.
Sufficient logging duration: for a fault occurring a few times a week, a logging period of only one or two days is often too short to capture the pattern with statistical reliability; a period of at least one to two weeks is more typical.
Correlation afterwards: the actual diagnostic value lies not in the logging itself, but in lining up the timestamps of logged network events with the timestamps of the equipment's fault messages.

Difference from the power-analyzer snapshot

A power analyzer deployed for a harmonics measurement typically measures during a limited, attended session and produces an average or representative picture of the load at that time. A power-quality datalogger is primarily set up for event-driven capture of abnormal events (dip, swell, interruption, transient) over a long period, including the exact time each event occurred — the time component here is just as important as the measured value itself.

Common mistakes

Deploying Class S equipment for a measurement intended as evidence in a dispute with the grid operator, when Class A is required for that purpose.
Choosing too short a logging period for a fault that occurs only sporadically, so the event is simply missed.
No time synchronisation between the datalogger and the fault log of the affected equipment, so no reliable correlation can be made afterwards.
Using generic trigger thresholds instead of thresholds matched to the actual sensitivity of the affected equipment, so relevant events go unrecorded or the logger is flooded with irrelevant entries.