Calibration is the process of comparing a measuring instrument with a measurement standard to establish the relationship between the values indicated by the instrument and those of the standard. Certificates showing the results from such a calibration commonly list pressure values determined by a standard, the corresponding pressure values indicated by the instrument being tested and the differences between these values at a number of pressures. In some cases the reported results are not in terms of pressure values but another parameter - for example certificates of calibration for pressure balances often report the effective area of an instrument's piston cylinder and also a pressure dependent term that shows how the effective area changes with pressure.

To provide confidence in the accuracy of calibration results the measurements must have demonstrable traceability. This means that all results associated with a calibration - including those relating to the calibration of the measurement standard used - must be traceable back to standards held at a national measurement institute, such as NIST, through an unbroken chain of comparisons and where each link has stated measurement uncertainties. In addition, it is important that appropriate equipment and procedures are used in the calibration process, and that they are used by trained and authorized personnel operating in an adequate experimental environment. Essentially, to be able to demonstrate formal traceability of measurements, the calibrations should either be undertaken by a national metrology institute such as NIST and/or a laboratory that has been independently third-party accredited.

Instruments usually need to be calibrated, whether they are simple devices with modest performance or state-of-the-art systems, as it is only by this process that their pressure measuring properties can be determined. This often means that the whole system needs to be calibrated and not just the sensor itself, as any associated electronics are just as likely to change characteristics as the sensor.

The frequency with which calibrations should be carried out is an important, if sometimes difficult, question; there are two main considerations. Firstly, all measuring devices - whether they are simple, 'fundamental', or sophisticated - change characteristics with time; the issue is how much do they change? New devices should be calibrated relatively frequently in order to establish their reproducibility - essentially their metrological stability or the change in their measuring ability between calibrations. Initial estimates of reproducibility are sometimes made using type-test data from earlier calibration results of similar instruments but the resultant uncertainty of measurement has to be cautiously higher, until real data is available.

Secondly, the required uncertainty of measurement should be assessed. If the instrument's reproducibility is shown, by successive calibrations, to be substantially better than the uncertainty required then the interval between calibrations can be extended - perhaps even up to 5 years or so, but at the other extreme - where the instrument's reproducibility approaches the uncertainty needed - the calibration intervals should be much shorter, perhaps even daily in some cases.

There are situations where an instrument need not be calibrated, for example where its readings are 'for information only' and their accuracy has little or no impact on the process or service being provided. But in these circumstances it is important to be careful that false assumptions are not accidentally built into the hand-waving generalizations that sometimes accompany arguments for not calibrating an instrument. Non-calibrated instruments can appear to be working properly whilst being in error by large margins and manufacturers' specification sheets should certainly not be taken as a reliable guide.

The cost of a calibration is sometimes the main factor in deciding not to have a device calibrated and clearly it is important to take economic issues into account. But there can be hidden costs and significant risks taken through not calibrating an instrument and hence not controlling or understanding a process adequately - that ought to tip the balance. For example using a calibrated instrument may reduce the number of end-products rejected because they are outside acceptable tolerances. It may also be that more products can be sold through having reduced and more competitive tolerances, better reliability, or that a wider customer base, including quality-controlled markets, can be better accessed. An assessment of risks can help the decision; for example it might be appropriate to calibrate even the most stable pressure balance more than once every three years if by not doing so you are potentially making yourself liable for large sums of money. There are, of course, many health and safety, legal, and regulatory issues that that should be considered too.

Instrument calibrations should be made at an appropriate level. It is not always necessary or desirable to have an instrument calibrated against a national measurement standard but the importance of being able to demonstrate traceability and understanding the degree of measurement uncertainty needed in a particular application should always be taken into account. There are several routes to obtaining a calibration but only two are recommended: through a NIST-accredited laboratory or directly from a national measurement institute such as NIST. There are non-accredited calibration services available but these are not recommended because they cannot give the degree of confidence provided by a third-party accredited laboratory. Of course you might decide to undertake the calibrations in-house yourself but again the confidence that can be placed on the results will be much greater if your system is formally third-party accredited. Whichever route you take it is well worth reviewing your expectation of an instrument before requesting a calibration - just to ensure that the calibration is likely to meet your needs.