The calibration process begins with the design of the measuring instrument that needs to be calibrated. The design has to be able to “hold a calibration” through its calibration interval. In other words, the design has to be capable of measurements that are “within engineering tolerance” when used within the stated environmental conditions over some reasonable period of time. Having a design with these characteristics increases the likelihood of the actual measuring instruments performing as expected.
The exact mechanism for assigning tolerance values varies by country and industry type. The measuring equipment manufacturer generally assigns the measurement tolerance, suggests a calibration interval and specifies the environmental range of use and storage. The using organization generally assigns the actual calibration interval, which is dependent on this specific measuring equipment’s likely usage level. A very common interval in the United States for 8–12 hours of use 5 days per week is six months. That same instrument in 24/7 usage would generally get a shorter interval. The assignment of calibration intervals can be a formal process based on the results of previous calibrations.
The next step is defining the calibration process. The selection of a standard or standards is the most visible part of the calibration process. Ideally, the standard has less than 1/4 of the measurement uncertainty of the device being calibrated. When this goal is met, the accumulated measurement uncertainty of all of the standards involved is considered to be insignificant when the final measurement is also made with the 4:1 ratio. This ratio was probably first formalized in Handbook 52 that accompanied MIL-STD-45662A, an early US Department of Defense metrology program specification. It was 10:1 from its inception in the 1950s until the 1970s, when advancing technology made 10:1 impossible for most electronic measurements.
Maintaining a 4:1 accuracy ratio with modern equipment is difficult. The test equipment being calibrated can be just as accurate as the working standard. If the accuracy ratio is less than 4:1, then the calibration tolerance can be reduced to compensate. When 1:1 is reached, only an exact match between the standard and the device being calibrated is a completely correct calibration. Another common method for dealing with this capability mismatch is to reduce the accuracy of the device being calibrated.
For example, a gage with 3% manufacturer-stated accuracy can be changed to 4% so that a 1% accuracy standard can be used at 4:1. If the gage is used in an application requiring 16% accuracy, having the gage accuracy reduced to 4% will not affect the accuracy of the final measurements. This is called a limited calibration. But if the final measurement requires 10% accuracy, then the 3% gage never can be better than 3.3:1. Then perhaps adjusting the calibration tolerance for the gage would be a better solution. If the calibration is performed at 100 units, the 1% standard would actually be anywhere between 99 and 101 units. The acceptable values of calibrations where the test equipment is at the 4:1 ratio would be 96 to 104 units, inclusive. Changing the acceptable range to 97 to 103 units would remove the potential contribution of all of the standards and preserve a 3.3:1 ratio. Continuing, a further change to the acceptable range to 98 to 102 restores more than a 4:1 final ratio.
This is a simplified example. The mathematics of the example can be challenged. It is important that whatever thinking guided this process in an actual calibration be recorded and accessible. Informality contributes to tolerance stacks and other difficult to diagnose post calibration problems.
Also in the example above, ideally the calibration value of 100 units would be the best point in the gage’s range to perform a single-point calibration. It may be the manufacturer’s recommendation or it may be the way similar devices are already being calibrated. Multiple point calibrations are also used. Depending on the device, a zero unit state, the absence of the phenomenon being measured, may also be a calibration point. Or zero may be resettable by the user-there are several variations possible. Again, the points to use during calibration should be recorded.
There may be specific connection techniques between the standard and the device being calibrated that may influence the calibration. For example, in electronic calibrations involving analog phenomena, the impedance of the cable connections can directly influence the result.
All of the information above is collected in a calibration procedure, which is a specific test method. These procedures capture all of the steps needed to perform a successful calibration. The manufacturer may provide one or the organization may prepare one that also captures all of the organization’s other requirements. There are clearinghouses for calibration procedures such as the Government-Industry Data Exchange Program (GIDEP) in the United States.
This exact process is repeated for each of the standards used until transfer standards, certified reference materials and/or natural physical constants, the measurement standards with the least uncertainty in the laboratory, are reached. This establishes the traceability of the calibration.
See metrology for other factors that are considered during calibration process development.
After all of this, individual instruments of the specific type discussed above can finally be calibrated. The process generally begins with a basic damage check. Some organizations such as nuclear power plants collect “as-found” calibration data before any routine maintenance is performed. After routine maintenance and deficiencies detected during calibration are addressed, an “as-left” calibration is performed.
More commonly, a calibration technician is entrusted with the entire process and signs the calibration certificate, which documents the completion of a successful calibration.