A 5-point calibration check is the standard way to see how a transmitter behaves across its whole range rather than at a single convenient point. You test it at five points, going up and then coming back down, because a transmitter can be accurate at zero and span yet bow in the middle, and it can read one way climbing and another way falling. Running both directions at five points reveals the two most useful error characteristics, linearity and hysteresis, and comparing the worst error to the calibration tolerance tells you whether the transmitter is still fit to leave in service or needs adjustment. It is a check first; it becomes an adjustment only when the numbers say so.
Do a 5-Point Calibration Check in one line: To perform a 5-point calibration check, apply inputs at 0, 25, 50, 75, and 100 percent of range ascending, then step back down through 75, 50, 25, and 0 percent, recording the transmitter output error at every point without adjusting anything. From that data you read linearity as how far the output bows from a straight line, hysteresis as the difference between the up and down readings at the same point, and total error as the worst deviation. Compare the worst error to the calibration tolerance: if it is inside, record it as an as-found pass, and if it exceeds tolerance, adjust and then repeat to capture as-left values.
The procedure is a disciplined sequence, and the order is the point. You apply a precise input at 0 percent of the calibrated range and record the transmitter's output, then step up to 25, 50, 75, and 100 percent, recording the output at each. Then, without overshooting, you step back down through 75, 50, 25, and 0 percent, recording again at each. The reason you approach each point from the correct direction, always increasing on the way up and always decreasing on the way down, is that overshooting and backing off would mask hysteresis, which is exactly the error the down pass exists to expose. A clean up-and-down run gives you two readings at each intermediate point, one ascending and one descending.
At every point you record the error, meaning the difference between what the transmitter output should be for that input and what it actually is. It helps to record the error in the same terms your tolerance is stated in, whether that is percent of span, engineering units, or milliamps, so the comparison at the end is direct. You do this as-found, before touching any adjustment, because the as-found record is the evidence that answers the real question: was this transmitter within tolerance during the time it was in service. Adjusting first and recording second destroys that evidence and leaves you unable to say whether the measurements it produced last month were trustworthy.
The reference matters as much as the procedure. Your calibration source and your readout must be enough more accurate than the transmitter's tolerance that the check actually proves something, a relationship expressed as the test uncertainty ratio, the ratio of the transmitter's tolerance to the accuracy of your standards. A comfortable ratio means the standard is accurate enough that its own error does not muddy the verdict; a poor ratio means a borderline pass or fail could be your equipment rather than the transmitter. Confirming the reference is adequate before you start is what keeps a 5-point check honest.
Linearity is how much the transmitter's output deviates from a straight line across the range, and the five points are chosen to reveal it because a two-point check at only zero and span cannot see a mid-range bow. If a transmitter reads correctly at 0 and 100 percent but sits high at 50 percent, that mid-range error is a linearity error, and it is invisible unless you test the middle. Plotting or tabulating the error at all five points shows the shape of the deviation, and the largest departure from the ideal straight line is the linearity error you report.
Hysteresis is read directly from the spread between the up pass and the down pass at the same input. If the transmitter reads a certain output at 50 percent on the way up but a different output at 50 percent on the way down, that difference is hysteresis, caused by mechanical or sensing lag inside the instrument. The up-and-down structure of the test is what makes hysteresis measurable; a single-direction sweep would report the transmitter as more accurate than it really is because it never sees the return path. The largest up-versus-down difference across the five points is the hysteresis you record.
Total error is the bottom line, the single worst deviation of the transmitter output from ideal across the entire up-and-down run, combining linearity, hysteresis, and any zero or span offset into one number. That worst-case error is what you compare against the calibration tolerance, the error band the transmitter is allowed to occupy and still be considered good. If the total error is inside tolerance at every point in both directions, the check passes as-found and no adjustment is warranted. If any point exceeds tolerance, the check has become a call to adjust, and you then trim zero and span, address the offending points, and rerun the 5-point test to capture the as-left values that prove the adjustment brought it back inside the band.
The as-found and as-left pair is the backbone of a calibration record because together they tell the whole story: as-found is the transmitter's condition when you arrived, as-left is its condition when you finished. A worksheet for a 5-point check lays out the five input points down the side, columns for the ideal output, the as-found up reading, the as-found down reading, and their errors, then a second block of the same for as-left if an adjustment was made. A transmitter that passed as-found needs no as-left because nothing changed, and recording that it was found in tolerance is itself valuable evidence. This layout makes the linearity, hysteresis, and total error fall out of the table rather than requiring a separate calculation.
The check confirms accuracy at one moment, but a monitoring system watches the transmitter between checks, which is what makes the calibration interval defensible. When a cloud SCADA platform such as Merobix trends the transmitter's output continuously, a slow drift that would push the next as-found reading out of tolerance is visible as a gradual bias long before the scheduled check. That lets you shorten or lengthen the interval based on how the transmitter actually behaves, and it flags the transmitters worth checking first because their trends have started to move.
The trend also gives the as-found and as-left values context they lack on paper. An as-found reading that was found out of tolerance is more meaningful when you can see in the history exactly when the drift began and whether it was gradual or a step change, which points to different causes. And after the adjustment, the recorded trend should show the transmitter's output settle back onto the reference, an independent confirmation that the as-left values on the worksheet translated into a real improvement in the signal the process and its operators rely on.
Two points at zero and span cannot see two important errors. A transmitter can read correctly at both ends yet bow in the middle, which is a linearity error only the intermediate points reveal, and it can read differently climbing than falling, which is hysteresis only the down pass exposes. Running five points in both directions captures linearity, hysteresis, and total error, giving a true picture of how the transmitter behaves across its whole range.
Hysteresis is the difference between the ascending and descending readings at the same input point. At 50 percent, for example, note the output on the way up and the output on the way down, and their difference is the hysteresis at that point. The largest up-versus-down difference across the five points is the hysteresis you report. This is why each point must be approached from the correct direction without overshooting, since backing off would hide the very lag the test measures.
You keep it a check, recording as-found values without touching anything, as long as the worst total error stays inside the calibration tolerance. It becomes an adjustment only when a point exceeds tolerance, at which stage you trim zero and span, correct the offending points, and then rerun the 5-point test to record as-left values proving it is back inside the band. Adjusting a transmitter that was already within tolerance adds no accuracy and erases the as-found history that showed it was trustworthy.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.