Calibration 101: Two Unavoidably Abstract Ideas

I have extensive experience with calibration practice as well as the theories behind them. At the same time, I feel comfortable in labelling myself as a “newcomer” and addressing other newcomers to this field. I identify myself as a newcomer because it is so common for me to see an old concept in a new light or to gain a new understanding of an idea that I first encountered decades ago. I may even have at times convinced myself in the past that I understood it completely. 

By starting with a title that includes “Calibration 101”, I want to take us to a fundamental discussion of how calibration actually works.

Calibration includes many physical objects and results. In addition, there are two abstract concepts that any calibration technician will encounter in this work. The technician will encounter these two abstractions in a very definite order. The first one is a precondition and the other follows in its wake but is just as universal, and just as important. We can use a variety of tools to describe and discuss these two concepts, but none of them will allow us to avoid the fact that they will remain abstract forever, no matter how much we as humans may shrink from abstraction.

Traceability is a Precondition

The concept of traceability comes first in any discussion of what a technician actually does on any day at work, whether that is your very first day on the job or on the day that you retire. I say that traceability comes first because traceability is “an attribute of data” [i]. We must grasp this idea from the very beginning. Companies pay calibration technicians to produce data, no matter what their other duties include. Technician cannot produce something called “data” without having their measuring tools previously linked to a structure of traceability. Without traceability, all that any technician can produce instead of data is “fiction”. Traceability contains an abstract structure that captures the performance of the data from our current measuring device by linking it to the “unbroken chain” of standards that lead to the data from the best standards we possess anywhere on the globe. Without the presence of this chain, we have no way to navigate and no way to make safe comparisons. Also, it is this chain that allows us to monitor our measuring devices to see if they are still reliable.

There are concrete elements sprinkled throughout this picture: Typically, the technician may rely on a “calibration sticker” or label of some kind that assures him that the device with which he is about to make some measurements is reliably “traceable” at that moment. When used, the sticker should identify this measuring tool in a unique way so that it cannot be confused with similar nearby tools. The sticker should also portray a definite period within which the technician and his organization may safely assume that the tool is still usable. At the end of that period, the tool is no longer technically capable of producing “traceable data” until someone recalibrates it.

This scenario contains abstract and concrete elements, both of which are essential to the calibration technician’s success. If that sticker is supposed to be there, then it had better be right there, where we can see it and touch it, or nothing further has any practical meaning. But at the same time, the sticker is just a small piece of paper with permanent writing on one side, and glue on the other side if the abstract element is missing. Has someone in our organization already performed the actions necessary to link this measurement tool to the chain of traceability? If the answer is “yes”, we can proceed to produce and record some data. Otherwise, we need to stop right there and sort things out before we can proceed.

Measurement Uncertainty comes next

Every chance that we have to measure anything presupposes the concept of traceability. That is why I have labelled it a precondition. Once we have agreement among ourselves, our practice may pick up some convenient concrete elements such as a sticker that allows us to easily verify the traceability of a measurement device. The process may also include other visible results, such as paper records that we can scan or bury in a file.

Next, a new calibration technician confronts another fundamental calibration concept which may, in many ways, seem even more abstract.

Any measuring device that a technician may choose is either traceable or not. One or the other, there is no grey middle area. A technician may choose to verify their instruments traceability before proceeding, but that choice has no effect on whether or not their measuring activities afterwards are technically traceable.

With the performance of the very first measurement, the technician enters a framework in which the concept of “measurement uncertainty”[ii], immediately applies. No sticker or any other physical object will ever notify us technician that we have entered this state.

Unlike traceability, which is “either / or”, measurement uncertainty is always present and always applies once we make even a single measurement. Unlike a calibration sticker or paper record, uncertainty may be more difficult to see and really grasp, unless we know what to look for.

New technicians listen up, please! The moment that you make a traceable measurement, the resulting measurement value contains an uncertainty element that has an “unknown magnitude and an unknown sign”. This means that there is no possible mathematical process that can reliably remove this uncertainty element to leave a “more accurate”, “more correct” result. So far, we need only discuss a single measurement value. The “dispersion” of results occurs when we continue to keep measuring the same thing (assuming that we have adequate resolution) and accumulating more “data”. We will see our measurements group themselves in some way into a dispersion. As this measurement result group grows, it gradually becomes less risky for us to judge the most likely value.

Let’s peek back at our unbroken chain now that we have these two concepts to work with. The most popular portrayal of this chain is that it consists of ever more accurate artifacts as we ascend the chain, all level producing traceable data. I have slowly changed my understanding of this chain. I find it more powerful to imagine that each element in the chain consists of an element that has a decreasing amount of measurement uncertainty, not an increasing amount of accuracy. 

I have come to prefer this preferential view of uncertainty behavior over accuracy behavior because it allows me take one further step toward recognizing the central position that uncertainty takes in the measurement process. The fact that I have worked for 30 years in the Pharma business sector has had a big influence on my perspective. From the outside, one might well assume that Pharma is critically dependent on measurement. From my viewpoint, Pharma measurement practices ignore the realities of measurement uncertainty at nearly every opportunity. If this weren’t true across the entire business sector, this behavior would have been severely punished. Instead with the cooperation of various regulators, it is universal.

In conclusion

I have not tried to fully explain these two abstract concepts. What I have tried to do instead is to prepare new technicians for the ambiguities surrounding them as they enter the calibration workplace. 

You may very well hear or read impossible statements like “traceable to NIST”. You may encounter entire organizations whose unspoken policy is to ignore measurement uncertainty, under unspoken conditions that no one else can audit. It isn’t the presence of abstract concepts in this area that allows these things to survive. Math and probability can be quite abstract and yet we continue to understand that “2 plus 2 equals 4 and not 5”. The mixture of concrete and abstract in calibration practice is what masks and allows these mistakes.

Nothing can ever be “traceable to NIST”. Measurement uncertainty always attaches to measurement. Keep these rules in mind especially when you choose to say nothing in the presence of someone trying their hardest to violate them. There may well be an opportunity later to speak up, provide some leadership and be a Hero! Meanwhile you are still getting paid.

I will leave you with this proof of two things: 1) not all abstractions are bad…2) it is possible to write without using the passive voice[iii].

[i] The “VIM (The International Vocabulary of Metrology) say that right away. Please check other sources for other definitions of traceability.

[ii] VIM says that “Measurement uncertainty is a parameter characterizing the dispersion of the quantity values being attributed to a measurand”.

[iii] Set Word Spell Check to trap “passive voice” and then practice makes perfect (0%).