How to Decide Whether an Instrument Is Still Fit for the Decision

Why this matters

An instrument almost never fails by going dark. It fails by continuing to produce numbers that look exactly like the numbers it produced when it was right. So the question worth asking is not "is this meter in calibration." It is "can this meter settle the specific question I am standing here asking." Those are different questions with different answers on the same morning, with the same instrument, on the same machine. A meter that cannot tell you whether a motor is running two percent over its nameplate can tell you with total confidence that it is nowhere near locked-rotor current. Techs who never ask the second question condemn parts on a margin their instrument cannot resolve, and then argue with the supply house about a returned motor that tested fine on the bench.

The gate

Compare the instrument's total uncertainty for that one reading against the margin between the reading and the accept/reject line. The unit of analysis is one decision, not one instrument and not one visit.

The convention most measurement work uses is 4 to 1: the instrument's uncertainty should be no more than a quarter of the tolerance band you are judging against, so the instrument is not a meaningful share of the answer. That ratio reaches you through your own shop procedure, a customer's specification or a manufacturer's service instruction. It is not law and nothing obliges a field-service shop to adopt it. Adopt it anyway, because it converts a feeling into a number you can compute in ten seconds standing at the panel.

Note what the gate does not contain: a date. A calibration sticker is one input to the uncertainty figure, not a substitute for it.

Before the reading this rests on

The worked case below is a clamp-on current reading taken on a running motor with the disconnect enclosure open, so the whole gate sits behind the energized-work gate at 29 CFR 1910.333(a)(1), which requires live parts to be de-energized before work on or near them unless de-energizing is infeasible, and whose note treats a test that can only be performed energized as one of those cases. Instrument, leads and probe tips must all carry an IEC 61010-1 measurement category and voltage rating at or above the point in the system you are clamping, because an under-rated instrument across a line-voltage transient vents at your hands instead of failing quietly. Wear the electrical protective equipment 29 CFR 1910.335(a) requires for that exposure, keep the free hand outside the enclosure, and stand out of the line of the panel opening.

A clamp jaw needs one conductor by itself. If the conductors are bundled and you would have to separate them to get a clean clamp, that is no longer a test that can only be done energized: de-energize, apply electrical lockout under 29 CFR 1910.333(b)(2) (or 29 CFR 1926.417 if the work is construction), prove dead using the live-dead-live sequence in NFPA 70E-2021, 120.5, in the edition your employer's electrical safety program has adopted, and separate the conductors with the circuit dead. 29 CFR 1910.147 is the wrong standard for this act; it excludes electrical hazards on utilization equipment at 1910.147(a)(1)(ii)(C).

Step 1: Write the accept/reject line and the margin before you look at the display

Name the number that decides, and the number you are comparing it to, in that order. "Nameplate full-load amps is 24.0" is the line. The margin is your reading minus that line. Writing the line first stops the display from setting your expectation, which is the mechanism behind almost every reading a tech remembers as "about right."

Skip this and you have no denominator. The gate is a ratio, and without the margin you fall back on whether the number felt high, which is exactly the judgment the instrument was supposed to replace.

Step 2: Build the uncertainty for that reading, not the instrument in general

An accuracy spec is not a single number. It is normally written as a percentage of reading plus a fixed count term, and both halves matter. Say the specification for the clamp function reads plus or minus 2 percent of reading plus 5 counts. A count is one step of the least significant digit on the range in use, so on a range that displays tenths of an amp, 5 counts is 0.5 A. Add the two terms arithmetically, the way the manufacturer intends them to be read.

Three conditions the specification assumes, all of which route to the instrument's own manual rather than to a number anyone can quote generically: the instrument is inside the reference temperature range the spec was written for (outside it, a temperature coefficient per degree gets added), it has been powered long enough to warm up, and it is inside its stated calibration interval. Miss any of the three and the printed spec is not the spec that applies.

Step 3: Add the uncertainties that are not on the spec sheet

The instrument's own uncertainty is usually the smallest term in a field reading. The larger ones come from what you did with it: where the jaw sat relative to the conductor and to its neighbours, whether the reading had settled, whether the load was in the state you assumed. Those are separate articles and this one does not re-derive them (see References), but they belong in the same budget. When you are being conservative, add them; combining independent contributions in quadrature is defensible and produces a smaller total, which is exactly why the additive version is the safer default in the field.

Step 4: Field-verify if the ratio lands anywhere near the line

A verification is not a calibration. It tells you the instrument agrees with a known value at one point, which is enough to catch the failures that actually happen: a damaged jaw, a bad range, a dying battery, a meter dropped last week.

For a clamp, the honest field method is a turns multiplier. Pass a conductor carrying a known steady current through the jaw N times and the display should read N times that current. Build that loop on a bench with a dedicated test lead and a bench source, energize it only after the loop is closed and clear of your hands, and never make it by wrapping a live branch conductor around the jaw inside an enclosure, which turns a measurement into an unplanned reconfiguration of an energized circuit. Ten turns of a conductor carrying 1.00 A should read 10.0 A; a display that reads 9.2 A is telling you something a sticker cannot.

Step 5: Decide, and write the ratio next to the number

Record the reading, the line, the margin, and the ratio. That is four short fields, and it is the difference between a ticket that supports a recommendation and a ticket that contains a number.

What the calibration sticker actually settles

A calibration establishes agreement with a reference at the points tested, on the day tested, under laboratory conditions. It carries three pieces of information a field tech should know how to use.

• As-found data. If the certificate reports what the instrument read before adjustment, you can tell whether it drifted a little or a lot. An instrument returned as-found within spec every cycle is a candidate for a longer interval; one returned out of tolerance is a reason to go back and look at decisions you made on close margins since the last certificate.
• The interval. Intervals are set by the manufacturer's recommendation, by your own procedure, or by the contract you work under, and there is no universal number. What is universal: the interval is a schedule, not a guarantee. An instrument can go out of tolerance the week after it comes back.
• The points tested. A certificate covering the voltage function says nothing about the clamp function, and a certificate at one range says less than you would like about a range four decades away.

One instrument, two decisions on the same motor

The clamp reads 24.6 A on a running motor. Nameplate full-load amps is 24.0. Nameplate locked-rotor amps is 150. Both figures are illustrative, but the arithmetic is the point.

Uncertainty first, since it is the same for both decisions: 2 percent of 24.6 A is 0.49 A, and 5 counts on the tenths range is 0.5 A. Total is 0.99 A, call it 1.0 A. That figure does not change when the question changes, because neither the reading nor the instrument changed.

Decision A: is this motor running over its nameplate full-load current? The margin is 24.6 minus 24.0, which is 0.6 A. The ratio is 0.6 to 1.0, or 0.6 to 1. The gate wants 4 to 1, which at this reading would require the total uncertainty to be at or under 0.15 A. The instrument is not fit for this decision and no amount of re-reading fixes it, because taking the same reading again with the same instrument does not shrink the uncertainty. Whether a motor above nameplate current is actually in trouble also depends on its service factor and the supply voltage, which a sibling article owns; the point here is that this instrument cannot even establish that it is above nameplate.

Decision B: is this motor anywhere near locked-rotor? Same reading, same 1.0 A of uncertainty. The margin is 150 minus 24.6, which is 125.4 A. The ratio is roughly 125 to 1. Fit, with about thirty times the headroom the gate asks for. You can say the motor is not stalled and defend it.

What flips Decision A. Nothing about the instrument. The margin. At a reading of 29.0 A against the same 24.0 nameplate, the uncertainty is 2 percent of 29.0, which is 0.58 A, plus the same 0.5 A, so 1.08 A, and the margin is 5.0 A. That ratio is 4.6 to 1 and the instrument is fit. In plain terms: this clamp can prove a 20.8 percent overload and cannot prove a 2.5 percent one. That is a real and useful capability, stated honestly.

The failure mode. A tech reads 24.6 against 24.0, writes "drawing above nameplate," and a motor gets quoted. The number was not wrong. The claim built on it was, because the entire 0.6 A of evidence fits inside the instrument's own uncertainty band, and a second instrument with a different offset would have read 24.1 and produced the opposite recommendation on the same machine on the same day.

How to verify you got this right

Re-run your own ratio before you write the recommendation. Three specific checks, because these are the three that fail:

• The margin and the uncertainty are in the same unit and from the same reading. Percent against amps is the most common way this goes wrong.
• The counts term used the range actually displayed, not the range you assumed. Moving up one range multiplies the counts term.
• The ratio was computed against the tolerance band, not the reading. A 2 percent instrument on a 24 A reading looks precise until you notice the question only left you 2.5 percent of room.

If the ratio comes out under 4 to 1, you have three honest moves and none of them is rounding: get an instrument with a tighter spec for that function, change the question to one with a wider band, or say plainly on the ticket that the reading does not settle it and name what would.

References

• 29 CFR 1910.333(a)(1) and 1910.335(a) - the energized-work gate and the electrical protective equipment required for the exposure
• 29 CFR 1910.333(b)(2) and 29 CFR 1926.417 - electrical lockout in general industry and in construction
• NFPA 70E-2021, 120.5 - the live-dead-live verification sequence, binding through your employer's electrical safety program or an adopting jurisdiction
• IEC 61010-1 - measurement category and voltage ratings for instruments, leads and probe assemblies, binding through the product's listing rather than on its own
• See related: What a Meter Category Rating Is Protecting You From; Why Lead and Probe Condition Is Part of the Measurement; How to Choose the Right Instrument for the Question