I had an interesting conversation today about calibration of stuff.

So, first I have to take a step back. If you have something that you can reliably measure and that if repeatable you can calibrate and control for it. If you can accurately measure something, there is a correct measured value of whatever it is you’re trying to calibrate.

The corollary is that there is a single correct answer.

For instance, if you wanted to calibrate a clock you can compare the time to a known good source such as NIST (National Institute of Standards and Technology). Now we enter into the other interesting topic of the right answer. There’s no concept of averaging a bunch of different times — like the old saying goes “a man with one watch knows what time it is, a man with two watches is never quite sure.” You simply have to know what the right value is. After that, you can compare the drift characteristics of your clock to the known good source. Once you have that, you can bring your clock’s rate to match the standard.

In this example you can measure the time — it’s given by your clock. You can compare it to other sources. You can reliably reproduce the measurement — a second measured by your clock may not be the same as a second from NIST, but it should be consistently different. (if it’s not consistent you can determine that by comparing to your standard and then try to make that better)

The key takeaway is that with things you can measure there should be no question about what is right. Once you determine “right” you know that everything that isn’t right, is, in fact, wrong. You can’t fudge right when you’re measuring it.