A metrologist wants to know that any variations in measurements are the result of variations of the parts being measured, not variations in the measuring devices or their users. Subjective interpretation of inspection or measurement devices is a complex variable that can influence the quality of the results drawn from that interpretation. Therefore, automatic inspection-and-measurement devices that take users out of the process must be more accurate. Well, maybe.

Interpretation of the device or the data
In this issue of Measurement Matters I am talking about variability in the interpretation of what the measuring device is communicating to the user. This is at the data collection part of the process. Another obvious source of error is from misinterpretation during the analysis of the collected measurement data, but that’s further into the process. Data recorded for future analysis are subject to sources of error due to user misinterpretation of what the measurement device is indicating. In other words, the measuring device may be working with great precision, but the measurement itself may be misinterpreted and lead to errors later.

Sources of variability
Any measurement is actually a snapshot in time of any number of things that can be varying. One thing that can vary and affect measurements is temperature. Even a temperature-controlled environment is subject to some level of temperature variability. A thermostat allows the temperature to vary within a range that can be a few degrees, or even fractions of a degree. In any case, the temperature can’t be held perfectly constant (the inspector’s body heat is a source of temperature change, for instance). I alluded to operator alertness as a variable last month. Time of day can influence decisions operators make based on subjective analyses such as visual inspections. Studies have shown that inspectors pass more bad parts on Mondays and Fridays than on other days of the week, for example.

Then there’s the measuring device itself. Take a gauge with an analog scale where a needle points to values on a dial attached to the face of the device. The needle is separated from the surface of the dial by a small distance. This design characteristic is a source of parallax error. In simple terms, parallax error occurs whenever the dial reading is not made by looking at the needle from directly above. You can simulate the effect by holding a finger out at arm’s length and aligning it with the vertical edge of a doorjamb or a window on the wall in front of you (this is easier to do with one eye closed). Now move your head to the left and right without moving your finger. You’ll see that your finger appears to be in front of different areas of the wall than when looking straight ahead. This exaggerated effect is what can happen with that small space between the needle and the dial of a gage when it’s not viewed from directly above. This is an example of what I meant by the device working with total precision, but the interpretation being in error. For a critical reading, parallax error might mean the difference between pass and fail, or be the source of erroneous data.

Mis-use vs. mis-reading
Let’s say your measuring device has a digital readout. No parallax error there. Unless the digits are changing, there is little chance of a reading error (although there is the potential for transcription error when writing the displayed values). Regardless of the readout, it’s possible to use a measuring device improperly, yet read it accurately. Examples of possible improper use of gauges and instruments depend on the devices themselves.

Consider a dial caliper. If a piece of cylindrical metal stock being measured isn’t perpendicular to the jaws of the caliper the reading could be too large (from measuring an ellipse instead of a circle). Or what about a stylus profiler? If the tracking force is too great or too small the reading may not be correct. Or a measuring microscope. If a non-telecentric lens is improperly focused, a measured reading can be too large or small. As these few examples show, training in the proper use of such devices can be critical to the integrity of the measurement data.

Automatic vs. manual
Let’s get back to the subject of measurement variability. Assume the operators are properly trained so they all follow the same procedures in exactly the same ways. As pointed out earlier, no two operators are the same. Even mundane differences such as how well each slept can affect the quality of their measurements. If you envision a perfect world where ten operators were equally trained to operate ten identical measuring devices used in an unvarying environment with equal lighting measuring identical parts, their results would vary because the operators are human.

Now let’s automate those ten measuring devices in that perfect world. In other words, remove the operators, or replace them with identical robots. All variability would be gone. Or would it?

Eliminating one source of variation in a measurement allows for analysis of other sources. What you’ll usually find is that an automatic measuring device will have some amount of measurement variation, but it will be of much lower magnitude than the variation from operator subjectivity of a manually operated device.

Of course this is simplistic. It ignores the fact that the magnitude of any variation in measurements due to differences in the operators can be less than the acceptable tolerance specifications for the parts being measured. In other words, if the measured tolerance is coarse enough, the operator-to-operator variance might be low enough to be statistically insignificant. For now, assume the operator-to-operator variation is large enough to matter.

It comes down to the process.
The good news is that well-designed measuring equipment takes into account all sources that might lead to measurement variation. Line filters and conditioners address variations in the power line entering the device. Granite and steel structures react slowly to minor temperature variations, ensuring long-term stability. Various forms of vibration isolation minimize the influences of activity near the measuring device. Even software algorithms used in the measuring machine might use smoothing or averaging to minimize measurement variability.

What’s important is understanding that proper use of any measuring device is critical. No matter how accurate, how expensive or how reliable the device might be, a poor environment, unskilled users, a lack of maintenance and a lack of consistent processes can result in measurement data that is not as reliable as it might be. Think of measurement and inspection as processes and consider the effects of all the steps in those processes on the outcome. Understanding the major sources of error and doing something to minimize them means you can have total confidence in your measurements for making important product and process decisions, no matter what measuring device you’re using or who’s using it.