Multimeter calibration: what the specification actually promises
A DMM specification is a formula, not a number, and it changes with range, function and how long since the last adjustment. How to read one and what a calibration has to cover.
Ask what a multimeter's accuracy is and the answer is a formula with two terms, applying to one function, on one range, for a stated period since adjustment, within a stated temperature band. Most disappointment with multimeters comes from reading the first number and ignoring the rest of that sentence.
Reading the specification
The usual form is ± (a % of reading + b counts). The first term scales with what you are measuring; the second is fixed for the range and dominates at the bottom of it. Measure 1 V on a 10 V range with a specification of ± (0.02 % of reading + 5 counts) on a 5½-digit meter and the counts term is doing most of the work.
| Reading | % of reading term | Counts term | Total | As % of reading |
|---|---|---|---|---|
| 10.0000 V | 2.0 mV | 0.5 mV | 2.5 mV | 0.025 % |
| 5.0000 V | 1.0 mV | 0.5 mV | 1.5 mV | 0.030 % |
| 1.0000 V | 0.2 mV | 0.5 mV | 0.7 mV | 0.070 % |
| 0.1000 V | 0.02 mV | 0.5 mV | 0.52 mV | 0.520 % |
The lesson is the ordinary one — use the range that fits the measurement — but the arithmetic makes it concrete. A twentyfold degradation in relative accuracy from choosing the wrong range is not unusual.
The specification period
Manufacturers publish specifications for a period: 24 hours, 90 days, 1 year, sometimes 2 years. The shorter periods assume a recent adjustment and tight temperature control and are much tighter than the annual figure. If your calibration interval is a year, the annual specification is the one that applies to your use — quoting the 90-day figure for a meter calibrated annually is claiming a performance nobody promised.
What calibration has to cover
- Every function you use: DC volts, AC volts, DC current, AC current, resistance, and any others the work depends on.
- Several points on each range in use, not one point on one range.
- AC functions at the frequencies you use — an AC voltage specification is frequency-dependent and a meter accurate at 50 Hz may not be at 10 kHz.
- Resistance at both ends: low resistance is where lead effects dominate, high resistance is where the meter's own burden shows.
- Current ranges, which are often neglected and often the worst-performing part of a meter.
Calibrating one range of one function and treating the certificate as covering the instrument is common and indefensible. The certificate covers what was measured.
The errors that are not in the specification
The meter's specification describes the meter. It does not describe your measurement.
- Thermal EMFs: junctions of dissimilar metals at different temperatures generate microvolts. At low DC voltage this is frequently the largest error in the setup, and it is invisible to the meter.
- Lead and contact resistance: measuring a low resistance through two leads measures the leads too. A four-wire connection removes this and is the reason four-wire exists.
- Loading: the meter draws current from the circuit and changes what it is measuring, which matters when the source impedance is high.
- Burden voltage: on current ranges, the meter's shunt drops a voltage that alters the circuit it was inserted into.
- Common-mode and noise pickup, especially on long leads in an industrial environment.
None of these appear on the certificate, and all of them can exceed the certified uncertainty. The certificate tells you the instrument is sound; it does not tell you the measurement is.
Calibrator, not meter
A multimeter is calibrated against a multifunction calibrator whose own uncertainty has to be small enough relative to the meter's specification for the comparison to mean anything — the ordinary test-uncertainty-ratio question. For a high-resolution bench meter the calibrator has to be very good indeed, which is why proper DMM calibration is not something to do with a handheld source and optimism.
Frequently asked questions
- What does '± (0.02 % of reading + 5 counts)' mean?
- It is a two-term accuracy specification. The first term scales with the value you are measuring; the second is a fixed number of least-significant digits on that range. Add them to get the permitted error at a given reading. Because the counts term is fixed, its relative contribution grows as the reading falls, which is why the same specification can mean 0.025 % at full range and 0.5 % near the bottom of it — and why choosing the right range matters more than most specification sheets make obvious.
- Does the specification period matter?
- Yes, substantially. Manufacturers publish different figures for 24 hours, 90 days, 1 year and sometimes longer, and the short-period figures assume a recent adjustment and a tight temperature band. If your calibration interval is annual then the annual specification is the one that applies to your work. Quoting a 90-day figure for a meter that is adjusted once a year claims a performance the manufacturer never offered.
- What is a four-wire resistance measurement and when do I need it?
- Four-wire (Kelvin) measurement sources current through one pair of leads and senses voltage through another, so the resistance of the leads and the contacts is not included in the result. You need it whenever the resistance being measured is comparable to the lead resistance — low-value resistors, shunts, contact resistance, and resistance thermometers, where a fraction of an ohm in the leads is a real temperature error.
- Do all functions need calibrating?
- Every function and range the work depends on, yes. A certificate covers what was measured and nothing else, so a meter calibrated on DC volts alone is a meter with an uncalibrated ohms range. Current ranges are the ones most often left out and frequently the weakest part of the instrument, so they are worth asking about specifically.
- Why is my measurement worse than the meter's specification?
- Almost always because of something outside the meter. Thermal EMFs at dissimilar-metal junctions generate microvolts that swamp low DC voltage measurements; lead and contact resistance corrupt low-resistance readings; the meter loads a high-impedance source and changes it; on current ranges the meter's own burden voltage alters the circuit. None of these are in the meter's specification and none are on its certificate, because they belong to your setup rather than to the instrument.
References
- [1]EURAMET cg-15 — Guidelines on the Calibration of Digital Multimeters
- [2]ISO/IEC 17025:2017 — General requirements for the competence of testing and calibration laboratories
- [3]JCGM 100:2008 — Evaluation of measurement data: Guide to the expression of uncertainty in measurement (GUM)
- [4]JCGM 200:2012 — International vocabulary of metrology (VIM), 3rd edition
General technical guidance written against the cited sources. It is not regulatory or legal advice and does not replace the applicable standard, guideline or a qualified reviewer's judgement.
Have a question on this topic?
Ask ValiTrac AI and see the evidence and calculation behind the answer.
Ask ValiTrac AIFree calculator: Calibration interval
Paste an instrument's as-found history and see how fast it is actually drifting, how much of its tolerance that consumes, and how long the evidence supports trusting it. The strongest input to an interval is the instrument's own record — this reads it.
Open the calculatorRelated articles
Tolerance versus uncertainty
Tolerance is what you require of the instrument; uncertainty is how well the calibration could measure it. A conformity statement needs both.
What is a decision rule and why every certificate needs one
A decision rule states how measurement uncertainty is accounted for when declaring pass or fail against a tolerance. Simple acceptance, guard banding and non-binary statements are the common choices.
Building an uncertainty budget step by step
Write the measurement model, list every input, evaluate each as a standard uncertainty, apply sensitivity coefficients, combine in quadrature, expand with k. Eight steps, one table.