← Smart Buildings Studio
Concept Explainer · Smart Buildings

Drift, Offset & Span

The three ways a BAS sensor silently goes wrong — and why commissioning-day accuracy doesn't last forever.

A sensor that was verified accurate at commissioning does not stay that way indefinitely, even if nothing ever visibly fails. No fault alarm fires. No wire comes loose. The sensor just gradually starts telling the control loop something slightly different from the truth. Understanding why requires separating three distinct ideas that get lumped together under "the sensor is off": offset error, span error, and drift — which is really just offset and/or span error changing slowly over time.

The Setup

Two different ways a measurement curve can be wrong

Plot a sensor's measured reading against the actual value it's sensing and, for a perfect sensor, you get a straight 45° line: measured always equals actual. Real sensors deviate from that line in two structurally different ways. An offset (zero) error shifts every reading by the same constant amount — the line stays parallel to the ideal, just moved up or down. A span (gain) error is different in kind: the sensor may be correct at one calibration point, but its slope is slightly wrong, so the gap between measured and actual grows as the actual value moves away from that point. Most real sensors have some combination of both.

Actual vs. measured — three error types

Calibration Errors
Offset onlyactual value →measured →const. gapSpan onlyactual value →measured →growing gapOffset + spanactual value →measured →shifted & growing- - - ideal (measured = actual)— actual sensor curve
Offset (zero) error
Additive, constant
e.g. a temperature sensor reading a steady 2°F high at every actual temperature across its range.
Span (gain) error
Multiplicative, growing
e.g. correct at 32°F, but reading 5°F too high by 200°F — the slope, not the zero point, is wrong.

The same sensor, one year apart

Drift
actual value →measured value →ideal (measured = actual)at commissioning — close to idealone year later — offset + span driftgrowing gapno fault, no alarm, no physical damage — just aging, thermal cycling, and contamination, slowly
What causes drift
Aging · thermal cycling · contamination
Slow component-level changes present even in an undamaged, correctly-installed sensor.
What fixes it
Periodic recalibration
Checked against a traceable reference standard on a defined interval — not a one-time commissioning step.
Why this works

Drift isn't a fourth error type — it's offset and/or span error, moving.

A sensor calibrated correctly a year ago can develop offset drift, span drift, or both, purely from ordinary aging: thermal expansion and contraction cycling a strain gauge or thermistor thousands of times, slow chemical or electrical changes in a sensing element, contamination building up on an optical or capacitive sensing surface, or connector and solder-joint resistance creeping upward. None of that requires physical damage or a wiring fault — it's the normal, expected behavior of a real physical measurement device over time. That's precisely why periodic recalibration against a traceable reference standard (not a one-time commissioning check) is a real, recurring maintenance requirement in any BAS program, not a box to check once and forget.

Common misconception
"It was calibrated correctly at commissioning, so it'll stay accurate unless it physically breaks."

False, and it's a quietly expensive assumption. Drift is a normal, expected aging and environmental effect that shows up in undamaged, correctly-installed sensors as a matter of course — it isn't evidence of a fault, and it usually produces no alarm at all, because nothing is actually broken. This is exactly why calibration schedules— recurring checks against a traceable reference, not a single commissioning-day verification — exist in serious BAS maintenance programs. A control loop that silently trusts an un-recalibrated sensor for years doesn't fail loudly; it develops increasingly wrong setpoint tracking, comfort complaints, or energy waste that slowly gets worse with no obvious root cause, precisely because the sensor feeding the loop never threw a fault — it just quietly stopped telling the truth.

Related Concept Explainers
Common-Mode Noise & Differential Measurement
Read it →
Why FDD Rules Flag Sensor Bias Before Anyone Notices
Coming soon

Drift, Offset & Span — Concept Explainer

Explains the three distinct ways a BAS sensor's accuracy can degrade — a constant offset (zero) error, a growing span (gain) error, and drift, which is either of those changing slowly over time — and why periodic recalibration against a traceable reference is a recurring maintenance requirement rather than a one-time commissioning step.

Offset vs. Span — Two Structurally Different Errors

An offset (zero) error shifts every reading by the same constant amount regardless of the actual value — plotted on an actual-vs-measured graph, the sensor's curve stays parallel to the ideal 45 degree reference line, just shifted. A span (gain) error is multiplicative: the sensor may read correctly at one calibration point (often the low end of its range) but its slope is slightly wrong, so the gap between measured and actual grows as the actual value moves further from that point. Most real-world sensor error is a combination of both, and distinguishing them matters because they typically require different correction: offset error is corrected with a simple additive adjustment (a zero trim), while span error requires adjusting the sensor's gain across its range.

Drift Is Offset and/or Span Error, Changing Over Time

Drift is not a separate error mechanism — it is the gradual change of offset error, span error, or both, over weeks, months, or years, driven by component aging, repeated thermal cycling, and contamination or degradation of the sensing element. A sensor with excellent as-installed accuracy can develop meaningful offset or span drift within a year or two of continuous service, with no visible damage and no fault condition to alarm on, because from the BAS controller's perspective the sensor is simply reporting a number — it has no independent way to know that number has quietly become wrong.

Why Calibration Schedules Exist

Because drift is a normal, expected property of real sensors rather than evidence of a defect, one-time commissioning calibration is not sufficient on its own. BAS maintenance programs schedule periodic recalibration against a traceable reference standard (a calibrated reference instrument with documented accuracy) at intervals appropriate to the sensor type and criticality, precisely to catch and correct drift before it meaningfully degrades control loop performance, comfort, or energy efficiency. A control loop that trusts an un-recalibrated sensor indefinitely will not necessarily throw any fault — it will simply track an increasingly wrong setpoint, quietly, for as long as no one checks.

Frequently asked questions

Can offset error and span error be corrected with the same adjustment?

No. Offset error is a constant additive shift, corrected with a single zero-point (additive) trim. Span error is a change in slope/gain, corrected by adjusting the sensor's gain so its response matches the reference across the full range — typically requiring at least a two-point calibration (a low reference point and a high reference point) to characterize and correct both errors independently.

How often should BAS sensors be recalibrated?

It depends on sensor type, criticality, and manufacturer guidance — commonly ranging from annually for critical control sensors to every few years for less critical monitoring points. The key principle is that a documented, recurring interval exists at all, checked against a traceable reference standard, rather than relying solely on the accuracy verified at initial commissioning.

Would a sensor with drift trigger a BAS fault alarm?

Usually not, at least not directly. Drift produces a gradually wrong reading, not a hard failure (like an open circuit or out-of-range signal) that a typical fault detection rule would catch. Drift is more often surfaced indirectly, if at all, through fault detection and diagnostics (FDD) rules that compare a sensor's readings to a redundant sensor, a model-predicted value, or long-term trend behavior — which is why scheduled recalibration matters even in buildings with FDD in place.

What's a traceable reference standard, in practice?

A traceable reference standard is a calibration reference instrument whose own accuracy has been verified through an unbroken chain of comparisons back to a recognized national or international measurement standard, with documentation at each step. Using one to recalibrate a field sensor is what gives the recalibration itself a defensible, quantifiable accuracy — as opposed to comparing against another uncalibrated instrument of unknown accuracy.

🎓

Try our Smart Buildings Studio

More calculators, simulators, and guides for this discipline.

Related tools & guides

Common-Mode Noise & Differential Measurement — Concept ExplainerBAS Design Reference — Interactive ReaderHVAC Fault Injection SimulatorFault Detection and Diagnostics (FDD) for Building Systems