Measurement Uncertainty: A Primer
A reading means little without knowing the uncertainty involved.
With more people than ever beginning to read what’s actually printed on calibration reports these days, what was supposed to bring clarity to measurements seems to be providing more discussions and arguments. From my point of view, it seems that some of the basics of measurement uncertainty are not properly understood or are bent out of shape which leads to more problems.
In this column I assume you have not had to deal with this subject all that much and want a simple overview to retain your sanity.
What is Measurement Uncertainty?
It is a method by which one can determine how precise a measurement is to the true value of whatever is being measured. You might say “That’s why I measure the thing in the first place” but your measurement only goes so far. It is a guide as to how the measured value relates to the specified size, but how close it is to the ‘true’ value is another story since the ‘true’ value is an unknown.
It’s important to remember that a value for uncertainty does not mean the reported measurement is that far from the true value—only that it could be.
Calculating the uncertainty of a measurement takes into account all of the factors that influence the reading your measurement provides. It gives a reliable answer as to how ‘close’ your reading of size is to reality when all factors are taken into account. In short: your measuring devices give you a ‘reading’ which means little without knowing the uncertainty involved.
It matters not whether you are using a simple instrument or one that costs thousands of dollars, one could have higher uncertainty than the other irrespective of what you paid for it. And you should also keep in mind that every measurement has a level of uncertainty attached to it irrespective of who is doing the measurement.
What are those “Factors?”
The ‘factors’ that I mentioned earlier that contribute to the uncertainty are many and vary with the type of measurement involved. Some of them you are already familiar with and include variations in temperature, measuring force, the uncertainty attached to the calibrated values for the measuring instrument being used, and similar values for setting masters and related accessory items that play a key role in the process such as thread measuring wires. Some may have little cause for concern while others are of major concern. They all have to be considered.
These factors are all converted to what they mean in linear terms by using a budget format to arrive at a value for uncertainty for a specific measurement using specific devices in a specific environment.
Some folks try to cut corners by adopting budgets created by a friend or that appear in some published standards. These can give you a rough guide as to the elements or factors you have to consider for an uncertainty budget but if your lab, equipment and masters are not identical to those in the budget you are copying, you can be creating more problems than solving.
Accrediting agency assessors will want to see your worksheets and budgets before accepting any values you show for uncertainty. And they may expect you to prove some of the input values or assumptions. For example, it your budget includes a value for repeatability of an instrument, they’ll want to see the study that generated that value. This is not a situation where you can ’wing it’ with any degree of success.
Is there an “app” for all of this?
The short answer is: yes and no. There a many free apps available to deal with the number crunching but you could use a $10 calculator just as effectively. The math is not the hard part of the process. All the software in the world will not help if the person using it does not understand the metrology behind the measurement as well as the characteristics of the equipment, environment, etc. involved. Without this knowledge you won’t know what numbers to plug into the software which aligns with the old saying: ‘garbage in, garbage out.’
As I mentioned earlier, a measurement without an uncertainty statement is just a reading and thus lacks credibility or value to the person depending on it. The one thing that is certain about measurement uncertainty is that it is always present.