Voltage calibration
DC voltage is the most accurately realised electrical quantity, thanks to the Josephson effect. A good bench multimeter can be calibrated against a reference in minutes, but thermal voltages, loading and noise can hide microvolts.
Instruments and typical accuracy
| Instrument | Typical range | Typical accuracy | Where it is used |
|---|---|---|---|
| Handheld multimeter, 3½ to 4½ digit | 100 mV to 1000 V | ±(0.05 to 0.5 % of reading + counts) | General service and field work |
| 6½ digit bench multimeter | 100 mV to 1000 V | About 30 to 50 ppm of reading on the 10 V range | Calibration labs, production test |
| 8½ digit multimeter or reference meter | 100 mV to 1000 V | A few ppm to 10 ppm of reading on 10 V | Transfer standard between labs |
| Multifunction calibrator | Microvolts to 1000 V | ±(0.002 to 0.01 % of output) | Calibrates handheld and bench instruments |
| Zener voltage reference (1.018 V, 10 V) | 1 to 10 V | Stability of a few ppm per year | Working standard, transfer standard |
| Josephson voltage standard | Up to 10 V arrays | Parts in 10⁹ | National institutes and top labs |
The more digits, the more the environment matters. The accuracy figures are typical ranges to help you compare instruments. Always use the specification of the instrument in front of you.
Reference standards and traceability
- Josephson voltage standard. A superconducting junction under microwave radiation gives voltage steps of f/KJ, with KJ = 483 597.9 GHz/V (conventional value, now exact in the SI through h and e).
- Zener references. Temperature-stabilised Zener diodes keep 10 V or 1.018 V to a few ppm per year. They are carried between labs and checked against Josephson systems.
- Calibrators and meters. A multifunction calibrator or a reference multimeter is the usual working standard, with a certificate and a stated uncertainty.
- Quality system. ISO/IEC 17025 scope listing DC voltage with the range and uncertainty.
How it is calibrated
Typical procedure
- Warm up the unit under test and the reference for the time stated in the manual, often 30 minutes to 2 hours.
- Short the inputs and read the zero offset, or null it if the instrument has an offset function.
- Connect with low-thermal-EMF leads, shielded, and keep the connections away from draughts.
- Check each range at zero, about 10 %, 50 % and 90 to 100 % of full scale, in both polarities.
- Use the integration time the data sheet assumes (for example 100 power line cycles for the best accuracy).
- Average several readings, record temperature and humidity, and compare with the specification.
- Adjust only if as-found is outside the limit, then record the as-left readings.
Where it goes wrong
- Thermal EMFs. Junctions of dissimilar metals generate roughly 0.2 to 40 µV per °C of temperature difference, a large error on a 100 mV range.
- Loading. A 10 MΩ input pulls down a source with high output resistance. Check the loading error before accepting a reading.
- Ground loops and noise. Mains pickup shows up as noise on the lowest ranges. Use guarding and star grounding.
- Short warm-up. A meter that is still drifting looks worse than it is.
- Unsafe voltages. Above 50 V, use safety leads and follow lockout procedures.
What goes in the uncertainty budget
- Reference standard uncertainty and its drift since last calibration.
- Resolution, noise and repeatability of the unit under test.
- Thermal EMFs and offsets in the connections.
- Loading error from finite input resistance.
- Temperature coefficient if the room is outside the specified band.
Combine the contributions as in the uncertainty budget calculator, then decide pass or fail with a guard band as explained on the calculators page.
Calculators
Everything runs in your browser. Nothing you type is sent anywhere.