When an instrument drifts out of calibration, it does not do so randomly across its range. It usually fails in one of two structured ways, and recognizing which one is happening tells you exactly which adjustment will fix it. A zero shift lifts or drops the entire reading by a constant amount. A span shift tilts the reading so the error grows toward the top of the range. The two look different in the as-found data, need different trims, and leave different fingerprints on a trend.
Zero Shift vs Span Shift in one line: A zero shift is an offset error that moves the entire calibration curve up or down by a constant amount, so the reading is off by the same value across the whole range. A span shift is a gain error that changes the slope of the curve, so the error is small near zero and grows toward full scale. Zero shift is corrected with a zero adjustment; span shift is corrected with a span adjustment.
A zero shift is a pure offset. Picture the ideal calibration line, output versus input, then slide the whole line up or down without tilting it. Every point on the range is now wrong by the same fixed amount and in the same direction. At the bottom of the range, at the top, and everywhere between, the error is constant. This is what happens when a transmitter's zero drifts: a device reading two units high at the low end reads two units high at the high end too.
A span shift is a change in slope, a gain error. Here the line is not slid but tilted, usually pivoting near the zero end, so that the error is negligible near the bottom of the range and grows steadily toward full scale. A device with a span error might read nearly correct at low input but be off by several units near the top. The error is proportional to the reading rather than constant, which is the defining signature that separates it from a zero shift.
Real instruments can suffer both at once, a curve both slid and tilted, and the as-found data then shows an error that is neither constant nor cleanly proportional but a mixture. The two can be teased apart because they have different mathematical shapes: a constant term for the offset and a proportional term for the gain. Understanding the pair as offset plus gain is what makes calibration a systematic correction rather than a game of nudging readings until they happen to line up.
The as-found data, the errors recorded at several points across the range before any adjustment, is where the diagnosis lives. Record the error at, say, the low end, the midpoint, and the high end. If the error is roughly the same value at all three points, you are looking at a zero shift, a constant offset that a single zero adjustment will remove. The tell is constancy: same size, same direction, top to bottom.
If instead the error is small at the low end and grows progressively toward the high end, roughly in proportion to the reading, you are looking at a span shift. The tell is proportionality: little error near zero, more error at mid-range, most error at full scale. When the as-found shows both a nonzero error at the low end and a growing error toward the top, the instrument has both a zero and a span problem, and both trims are needed.
The order of adjustment matters because the two interact. The standard practice is to correct the zero first, at the low end, then correct the span at the high end, then recheck the zero, because a span adjustment can slightly disturb the zero and the two are iterated until both ends fall within tolerance. Skipping the recheck, or adjusting span while the zero is still off, leaves a residual error that a proper as-left verification would catch. The as-left data, recorded after adjustment, is the proof the trim worked and belongs in the record alongside the as-found.
The two failure modes leave distinct signatures when a drifting instrument is trended against a reference in a SCADA historian. A zero shift appears as the measured line sitting parallel to a reference or check value, displaced by a constant gap that is the same whether the process is running low or high. Because the offset does not depend on the reading, the two lines stay a fixed distance apart across all operating conditions, which is the visual hallmark of a zero problem.
A span shift appears as a gap that widens with the reading. When the process sits low, the measured value tracks the reference closely; when the process climbs toward full scale, the measured value diverges, the two lines fanning apart. On a historian trend that spans a range of operating points, this fanning is unmistakable and immediately distinguishes a gain problem from an offset one, pointing the technician at a span trim rather than a zero trim before they ever reach the field.
Because Merobix historizes each point over time against whatever reference or redundant measurement is available, this diagnosis can begin before the calibration visit. Trending a suspect point across its operating range shows whether the deviation is constant or proportional, so the technician arrives already knowing which trim to expect and can capture the as-found and as-left data efficiently. Storing the calibration results back against the point closes the loop, building a history that reveals whether a given instrument tends to fail by zero drift or span drift, which in turn informs the recalibration interval and the choice of a more stable replacement.
Look at the as-found errors across the range. If the error is roughly the same size and direction at the low, mid, and high points, it is a zero shift, a constant offset. If the error is small near zero and grows toward full scale in proportion to the reading, it is a span shift, a gain error. A mixture of both shows a nonzero error at the low end that also grows toward the top.
Correct the zero first at the low end, then the span at the high end, then recheck the zero. The span adjustment can slightly disturb the zero, so the two are iterated until both ends fall within tolerance. Adjusting span while the zero is still off, or skipping the final recheck, leaves a residual error that a proper as-left verification is designed to catch.
Yes. A real instrument can drift in both offset and slope at once, so the as-found data shows an error that is neither perfectly constant nor cleanly proportional but a combination. Both a zero trim and a span trim are then required, applied in sequence and iterated, with as-left data recorded afterward to confirm the whole range is back within tolerance.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.