A packaging line stops three times a week. A textile machine faults overnight. A PLC reboots during a generator test. By the time maintenance arrives, the trends look normal, the drive shows a generic undervoltage code, and the only option seems to be replacing a drive or a control board and hoping.
In many of these cases the cause is a voltage sag (also called a voltage dip): a short reduction in RMS voltage, typically lasting from half a cycle to a few seconds. Sags are too brief for an ordinary meter to register, but more than long enough to drop out contactors, trip drives and reset controllers. This guide explains how to capture them, and — more importantly — how to work out where they came from.
The question to answer
Did the disturbance arrive from the supply, or start inside the production line? The answer decides who acts: the utility, the plant’s own electrical team, or the equipment supplier. Getting it wrong means replacing healthy equipment while the real cause carries on.
Why ordinary trends miss it
Most meters and building management systems average voltage over one, ten or fifteen minutes. A 150-millisecond dip to 60% of nominal disappears completely inside a ten-minute average — the trend stays flat while the line stops. To see a sag you need an instrument that classifies events cycle by cycle and records their depth and duration. To explain a sag you also need the waveform: Inspector keeps five cycles before and after every event, so the shape of the disturbance as it arrived is preserved.
The measurement plan
- Measure two points at once. Place one channel on the incoming supply and the other on the affected feeder, within the approved isolation and measurement ratings. Inspector V4 and the Portable V4 have two isolated channels for this.
- Document the connection — wiring arrangement, clamp direction and CT ratios. A reversed clamp turns a load step into a mystery.
- Align the clocks. Synchronise Inspector’s time with the drive, PLC and production logs so a trip can be placed on the same timeline as the electrical event.
- Monitor complete operating cycles — starts, product changes, shift changes and peak loading — not just a quiet afternoon.
Reading the evidence
Build an event timeline before naming a cause.
| What you see | What it suggests |
|---|---|
| The dip appears at both the incomer and the feeder. | A shared upstream influence — the utility supply, a large motor start elsewhere on site, or a transformer issue. |
| The dip appears only at the feeder. | A more local cause — a load step on that feeder, a loose connection, or a starting motor nearby. |
| Current rises before the voltage drops. | A local load inrush pulling the voltage down. |
| Voltage drops first, then current changes. | The disturbance arrived from upstream and the load reacted to it. |
Correlation is not responsibility. A shared timestamp establishes correlation, not blame. Wiring impedance, concurrent loads and controller timing can still change the interpretation — which is why the evidence is presented as ranked hypotheses, not a verdict.
When the stop happens during a generator test
Hospitals, data centres and factories with backup power see a specific version of this problem: a scheduled generator test “succeeds”, yet one controller reboots or a process pauses. The generator, the automatic transfer switch (ATS) and the UPS each keep a log, but their timestamps and meanings rarely line up closely enough to explain the sequence.
Plan the test with operations and reconstruct the sequence step by step:
- loss of supply or test command,
- generator start and availability,
- ATS transfer,
- UPS response,
- load recovery.
Look at voltage and frequency during the transition, not just their settled values — a generator under sudden load can sag in frequency as well as voltage. If the load drops out after a nominally successful transfer, examine downstream contactors and control supplies. And if two events appear closer together than the clock uncertainty between instruments, record the order as unresolved rather than guessing.
Never switch live just to collect data. Inspector is not the transfer controller and does not replace an approved backup-power acceptance test. Record planned tests; do not create unscheduled ones.
Acting on the evidence
Give electrical maintenance a ranked set of hypotheses — supply event, local load step, connection problem or equipment sensitivity — and investigate that branch rather than modifying every device. Depending on the finding, actions range from a utility complaint with evidence attached, to fixing a connection, staggering motor starts, adding ride-through to a control supply, or adjusting drive settings with the manufacturer.
Protective settings and ride-through changes belong to the equipment owner and manufacturer. Inspector supplies the diagnostic evidence.
Proving the fix worked
After the corrective action, repeat the same monitoring under the same conditions and keep both event records and production evidence. Track:
- unexplained stops per operating hour,
- lost production minutes,
- the share of incidents with usable evidence.
An honest way to value it
- Suppose a corrective action prevented six comparable stops during a defined observation period, each previously lasting 20 minutes.
- That is 6 × 20 ÷ 60 = 2 hours of recovered production.
- Multiply by a site-approved contribution-per-hour figure to estimate the benefit — and subtract the full cost of instruments, installation and engineering time.
Success is fewer recurring stops with a defensible explanation, not simply a quieter alarm screen.
Related reading
- What a power quality analyzer actually tells you
- VFD trips and motor stress
- What did the UPS really deliver?
Losing production to stops nobody can explain?
Tell us how often the line trips and what the drives report. In 20 minutes an engineer will tell you where we would put the two channels. Book a meeting
