Datasheet Jargon

Datasheets are a nightmare. The terms used are confusing, and often not clearly defined which makes it impossible to compare devices from different manufacturers. In this guide I’m going to use EvF as an example, with occasional reference to filaments. The terms are equally applicable to the measurement of anything else (temperature for instance).

The most commonly used (and mis-used) terms are:

  • Accuracy
  • Repeatability
  • Reproducibility
  • Linearity
  • Resolution
  • Precision

“How accurate is it?”

Probably the most frequent question, and the most difficult to answer.

Definition of accuracy: “The degree to which the result of a measurement conforms to the correct value”.

Here’s a hypothetical graph from a “perfectly accurate” force measurement system with the true force on the X axis and the indicated value on the Y. The values are deliberately arbitary at this stage (hence the axis title) and the point is that there is no difference between the true reading and the indicated reading at any of the seven calibration points (including zero).

The manufacturer could therefore quite reasonably claim that his device shows no errors between 1 and 6 arbitrons. Note first that we don’t know what happens below 1 arbitron and I deliberately didn’t put a dot at zero-zero to make this point; the device might have an offset or non-linear characteristic (more later) when it measures near zero so the graph, if extrapolated, would not go through the zero-zero point. This has implications if you are only intending to use the device near the bottom of its range; the quoted “accuracy” may not apply if you are below the bottom calibration point. It is not uncommon to quote an accuracy figure and to then note, in the small print, that the figure applies “down to 10% (or 5%) of the measuring range”. 10% of the measuring range in the case above is 0.5 arbitrons, or half way between the last calibration point and zero and as we will see, there are a number of effects which can degrade the first few percent of the range.

Real Numbers: Using force as an example, an electronic von Frey system is a force transducer. If, for instance, the calibrated force range of the device is up to 100 grams force (often abbreviated to gf and also, incorrectly, to just g), then we should certainly check where the bottom calibration point is are before using it to measure forces lower than 10gf or 5gf. If the range of the device is greater (500gf or even 1000gf), we should be even more careful. We should also not be unduly placated by other quoted parameters such as repeatability and particularly, resolution. These numbers are relevant, but only part of the story.

Repeatability: To me, the most important number is one that is not generally quoted; it is the repeatability under real measurement conditions (it’s sometimes wrapped up with reproducibility, and there’s more about this further down). So what is repeatability?

Repeatability is the ability of a measurement system to indicate the same measured value each time it is given the same quantity to measure.
(my own definition)  

The point is that many measurement systems have very good repeatability (or, to put it the other way round, very low scatter or spread) when they are calibrated under favourable conditions. Staying with my example of the EvF force transducer, these are most easily calibrated against another, higher grade, force transducer or against a set of known masses. You can do this with your kitchen scales and a bag of sugar; put the mass on a number of times and look at the variation in the number of grams indicated. Your scales will probably read exactly the same value each time, and so you would be entitled to quote a repeatability (at this force level) equal to the smallest change that your scales can indicate (this is the “Resolution” of the system).

And that’s fine for kitchen scales; they’re intended to be used on a bench. But now try holding the scales in your hand and putting the bag of sugar on…You’ll probably get a different answer each time now, and you’ll also be able to make the reading change by moving the scales up and down. The repeatability has degraded because of the additional forces from the motion (more accurately from the accelerations produced).

This, while an unfair test for the kitchen scales, is completely relevant to a force transducer that is intended to be handheld and used to apply a force to a rodent’s foot. The force transducer may show a very high repeatability while mounted on the bench, but that’s not how it will be used.

So how do we specify MouseMet? We do it two ways. First of all we calibrate with masses, on the bench, (because if you can’t get that right then there’s no point in doing anything else) but we then also use the instrument to measure forces generated by Mousecal. This is real life because the transducer is hand-held during the test. Finally, we know (and will share with you) the spread or scatter that the system should give, in the lab, with mice (or rats). This value has been derived from experimental data from several MouseMet systems in the field.

So what causes lack of repeatability? Waving the instrument around, as in the example above) is a bit of an extreme case. I used it because that is really what happens with some of the EvF probes available right now. More commonly, it’s caused by:

  • electronic noise
  • off-axis loading
  • temperature sensitivity

Some of these contributors will appear as a percentage of the reading. Importantly, others, like electronic noise, tend to be a constant amount so, as the reading gets smaller at the bottom of the range, their significance becomes greater.

Reproducibility is the next term to consider. I define repeatability as the variation you get in one calibration session. If you do the same thing another day, or if someone else runs the calibration, then the difference between their results and yours is reproducibility. It can’t be better than the repeatability…and its often worse.


—More to come, if you’ve read this far and need a quick answer to a measurement question, then please send me an email.