Power Quality Complaint Investigation

Purpose

Turn a complaint into a bounded measurement before anyone spends labor on a theory. "The lights flicker" and "my equipment keeps resetting" describe a symptom the customer noticed, not an event anybody has recorded. Without a measurement you cannot tell a utility regulation problem from a premises voltage drop from a load the customer added last spring, and all three produce identical descriptions.

The order here is not negotiable. The open or high-resistance service neutral gets ruled out first, before any logging, because it is the one condition on the list that destroys connected equipment while you are setting up and can put well over nominal voltage on a 120 V load. Everything else can wait a week for a recorder. That cannot wait an hour.

Scope

Covers a reported voltage, flicker, dimming, resetting or nuisance-tripping complaint on services at 600 V or less, through to a finding that names the problem as utility-side, service-side, or premises-side, with the measurement supporting it.

Does not cover harmonic mitigation design, filter or reactor selection, or a utility interconnection study. It does not cover corrective work either: a condemned circuit goes to the burned-receptacle SOP, a loose panel connection to the infrared survey, a service or transformer defect to the utility.

Roles and handoffs

Procedure

1. Take the complaint as data, not as a description. Get from the customer: which rooms or equipment, what time of day, how long each event lasts, what is running when it happens, and what changed in the last six months. Acceptance: at least three specific instances with a time and a coincident load, written down. Wrong looks like "flickers sometimes." Stop rule: if the customer cannot place a single event in time, install the logger first and interview afterward, because a complaint with no time anchor cannot be correlated with anything.

2. Screen for an open or high-resistance service neutral before you set up anything else. At the main panel, with the deadfront on where an accessible test point exists, measure each line to neutral and line to line under whatever load is present, then have a large 120 V load switched on and watch what happens. Acceptance: the two line-to-neutral readings stay within a few volts of one another and both stay inside the ANSI C84.1 Range A service band, which for a 120 V nominal base is 114 V to 126 V. Wrong looks like one leg climbing while the other falls, with the two summing to roughly the full line-to-line voltage. Stop rule on that pattern: open the main disconnect, tell the occupants to leave sensitive equipment unplugged, and call the utility as a service emergency; you do not repair a utility-owned service neutral, and a floating neutral has already been putting elevated voltage on the lightly loaded leg.

3. Measure at two points under a real load, not at one point at rest. Put a known load on the affected circuit, then read voltage at the service point and at the point of complaint at the same time, either with two meters or with one meter and a helper switching the load. Acceptance: both readings recorded with the load current, and the difference between them attributable to the run length and conductor size. Wrong looks like a single reading at a receptacle with nothing plugged in, which will look fine on a circuit that collapses at 15 A. Hazard clause: land test leads with the circuit isolated and the leads seated before energizing, and use a meter and leads rated for the circuit's category, because a leaded test on a live panel is where meters fail catastrophically.

4. Split the finding at the service point, and say which side it landed on. Compare the service point reading to the point of complaint reading, both under load. Acceptance: if the service point stays in ANSI C84.1 Range A while the load point sags, the problem is premises-side and stays yours; if the service point itself falls outside Range A under normal building load, it is a supply or regulation matter for the utility. Wrong looks like chasing branch circuits when the service itself is sagging. Stop rule: a utility-side finding stops premises troubleshooting and goes to the utility with the numbers, the time and the load attached, not as a phone description.

5. Quantify the premises drop against a stated basis before you go hunting. Express the drop as a percentage of the voltage measured at the source end, and name the load current it was measured at. The NEC's informational notes recommend branch-circuit conductors sized so that voltage drop does not exceed 3 percent to the farthest outlet, with the combined feeder and branch drop not exceeding 5 percent; those are recommendations in informational notes rather than enforceable requirements, so treat them as the design target that tells you whether a drop is normal or a defect. Acceptance: a drop percentage with its base voltage and its load current in the same line. Wrong looks like "it drops about ten volts" with no load stated.

6. Localize the drop by halving the run, not by replacing devices. Measure at the midpoint of the circuit under the same load, then at the quarter point on whichever half carries the loss. Acceptance: a device pair or a span identified where the drop across it is disproportionate to its length. Wrong looks like replacing the last receptacle because it is nearest the complaint. Stop rule: where a single connection carries most of the drop, stop measuring and treat it as a developing fault, because the power dissipated at a resistive joint is the voltage across it times the current through it, and that heat is inside a wall.

7. Where nothing shows at the time of the visit, log for a full cycle of the customer's operation. Install a recorder at the panel or at the affected circuit set to capture RMS voltage, current, and sag and swell events with timestamps, and leave it for at least seven days, longer if the complaint follows a weekly pattern. Acceptance: the recorder's clock is synchronized to a source the customer's own event times can be compared against, and the customer keeps a written log of events during the window. Wrong looks like three days over a holiday when the building was empty. Stop rule: do not remove the logger without at least one recorded event, or a full cycle with none, which is itself a finding.

8. Correlate, name the finding, and route it to exactly one owner. Line the recorded events up against the customer's log and the loads that were running. Acceptance: a written finding that names the location (utility, service, feeder, branch, or connected equipment), the measurement that supports it, and one owner. Wrong looks like a report listing four possibilities. Stop rule: if the log shows no event during a week in which the customer reported three, the finding is that the complaint is not a supply event, and the next step is the equipment itself, not another week of logging.

When the site does not match the assumption

The complaint tracks one appliance and nothing else. Measure the drop on that circuit while the appliance starts, then compare a second circuit at the same moment. A motor start that pulls the whole panel down briefly is a source impedance question; one that pulls only its own circuit down is a conductor or connection question.

The building has solar, a battery or a generator. Any of those can hold voltage up or push it around in ways that mask the condition. Record their state at every measurement, and take one set of readings with the inverter output opened at its own overcurrent device if the customer allows it.

The customer's equipment is the victim and they want a cause for a warranty claim. Say what your measurement establishes and what it does not. A recorded sag at a stated time is a fact you can bank; whether it caused a specific board failure is a question for the equipment manufacturer, and whether the utility is liable is one for the customer's own counsel or insurer.

The record this produces

One investigation record per complaint: the step 1 intake with its time-anchored instances; the step 2 neutral screen with all three voltage readings and the load condition; the step 3 two-point measurements with the load current; the side-of-the-service call from step 4; the drop percentage with its base and its current; the step 6 localization; the logger make, install and removal dates and the raw file; and the named finding with its single owner.

The office reads the finding to know whether it is quoting work or escalating. The customer reads the drop percentage and its basis, which is what makes a re-run make sense to someone who cannot see inside a wall. The utility reads the step 2 and step 4 numbers, and it takes a service call far more seriously arriving with two readings and a time than as a complaint.

Worked pass: 1978 two-story, complaint of dimming and a resetting router

Step 1 intake: dimming in the upstairs office, three instances, all in the evening, all while the space heater in that room was running. Nothing changed except the heater, bought in the fall.

Step 2, the screen that has to come first: L1 to neutral 121 V, L2 to neutral 119 V, L1 to L2 240 V. The two legs are within 2 V of each other, both inside the 114 V to 126 V Range A band, and 121 plus 119 equals the 240 measured line to line, which is what a healthy neutral looks like. Switching a large 120 V load on moves both by under a volt. No neutral fault, so the investigation continues rather than becoming a utility emergency.

Step 3: with the heater running at a measured 14.7 A, the service point reads 118.9 V and the office receptacle reads 108.4 V, both taken within the same minute.

Step 4: the service point at 118.9 V is inside Range A under building load, so this is premises-side and stays with the shop. No utility call.

Step 5, stated against its basis: the drop is 118.9 minus 108.4, or 10.5 V, at 14.7 A, which is 8.8 percent of the 118.9 V source-end reading. Against the 3 percent branch-circuit design target that is roughly triple, so this is a defect rather than a long run.

Step 6 finds it. The circuit has seven devices. At the fourth device the reading under the same load is 117.6 V; at the fifth it is 108.7 V. Nearly the whole loss sits between two adjacent devices about 9 ft apart, which no conductor of that size and length explains. Step 6's stop rule fires: 10.5 V across roughly 0.7 ohms of unintended resistance at 14.7 A is on the order of 150 W being dissipated at a termination inside a wall box, which is a developing fault, not a voltage-drop nuisance.

The heater is unplugged, breaker off, locked and tagged under 29 CFR 1910.333(b)(2), and the device proved dead live-dead-live per NFPA 70E-2021, 120.5, before the box is opened. The fifth device's line-side backstab is discolored and the conductor insulation is stiff for about an inch. That crosses into the burned-receptacle SOP's territory, and the release-or-condemn decision is made there rather than here.

Step 7 is not needed: an event was reproduced on demand under load, so no logger goes in. Step 8's finding reads premises-side, branch circuit 9, single high-resistance termination at the fifth device, supported by the two-point readings and the 8.8 percent drop. One owner, one work order.

References

• ANSI C84.1, in the edition your utility and specifying engineer reference, for the Range A and Range B service and utilization voltage bands used in steps 2 and 4
• NEC Article 210 and Article 215 informational notes, in the edition your authority having jurisdiction has adopted, for the 3 percent branch and 5 percent combined voltage-drop design recommendations
• IEEE Std 1159, for the definitions and duration categories of sags, swells and interruptions used when classifying recorded events
• 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5, in the edition your employer's electrical safety program has adopted, for isolation and the live-dead-live sequence
• See related: Burned Receptacle Response and Circuit Condemnation SOP, Infrared Scan of a Panel SOP