Calibration

A bakery sells products by weight: if its scale reads high, customers may receive less than the stated amount, and if it reads low, the bakery may give away product. Calibration provides evidence about its measurement performance.

Choose the measurement that matters, because a temperature sensor for food safety, a pressure gauge for a process and a retail scale serve different purposes and their acceptable errors cannot be chosen from one generic list. Identify the range of use, since a thermometer that performs well near room temperature may not be reliable at freezing or cooking temperatures and test points should cover the work it actually does.

Use a suitable reference whose performance and uncertainty are good enough for the decision, because a reference with unknown history does not establish a trustworthy comparison. NIST describes metrological traceability as an unbroken documented chain of calibrations relating a result to a reference, each contributing uncertainty, so the phrase traceable to NIST by itself does not prove a result is fit for use.

Define conditions, since temperature, humidity, setup and operator method can affect readings, and understand uncertainty, because a calibration result is not perfectly exact. Uncertainty describes a range supported by the method and must be considered alongside tolerance when deciding if equipment is acceptable.

Set acceptance limits before testing, drawn from product requirements, regulation, risk or an agreed specification, and do not change limits after seeing an inconvenient result. Keep calibration and adjustment distinct: a lab can measure an instrument as found, adjust it, then measure it as left, and only the before reading reveals whether earlier work may have been affected.

An illustrative reading error is device reading minus reference value, so if a scale reads 10.08 kg against a 10.00 kg reference, the indicated error is +0.08 kg at that point. This alone does not establish a pass or fail without tolerance and uncertainty.

Check the certificate for measured values, uncertainty, reference information, conditions and any statement of conformity with its decision rule, and make sure it carries the right serial number, location and date, since an attractive stamp alone is not enough. A quick check against a known point between formal calibrations can detect drift, but it does not necessarily replace a full calibration across the required range.

Choose an interval based on evidence such as manufacturer recommendations, stability, frequency of use, handling and consequences of failure, because there is no universal annual requirement for every instrument. ILAC's guidance on calibration intervals addresses how to review and adjust intervals using risk and performance information, and it is guidance for a method of thinking, not an automatic legal timetable for all businesses.

Protect equipment between checks, because damage, contamination or a knocked setting can invalidate a result before the next scheduled date, and if a device fails, remove it from service, label its status and assess work completed since the last reliable check. For owners, calibration is evidence that measurements can support a decision, and repeated drift in the same direction may call for a shorter interval or a better device, while stable results may support a longer interval if the applicable system permits.