Sensor Drift: When Every Control Decision Downstream Is Wrong

Sensors don't fail cleanly. A failed sensor is easy — it reads open, or shorted, or pinned at a rail, and something alarms. A drifted sensor is far worse, because it keeps producing plausible numbers that are simply wrong, and the entire control system believes them.

Temperature sensors commonly drift on the order of a degree or two per year. That sounds trivial. It isn't, because a sensor doesn't just report — it drives. Every loop, reset schedule, lockout and interlock downstream of that sensor inherits its error.

The multiplier. A supply air temperature sensor reading three degrees low means the air handler produces air three degrees colder than the sequence intended — continuously, every hour, in every operating mode. Nothing alarms. The BAS reports normal. The building simply costs more to run than it should, forever.

Check sensors in order of blast radius

You will never calibrate every sensor in a building. So don't try. Check them in the order of how much damage a wrong reading does.

Outdoor air temperature

The highest-leverage sensor in most buildings and often the worst-mounted. It gates economizer changeover, drives reset schedules, and can lock out free cooling entirely. A drifted OAT sensor reading warm will hold the economizer closed through hundreds of hours of perfectly good free-cooling weather, and the only symptom is a slightly higher chiller bill.

Look for: mounting location first — a sensor in direct sun, above a roof membrane, or near an exhaust discharge is measuring something other than outdoor air, and no calibration fixes that

Discharge / supply air temperature

Drives the primary loop on the air handler. Error here propagates to every zone the unit serves. This is the sensor most likely to cause the continuous overcooling described above.

Mixed air temperature

Used for economizer control and freeze protection. Also the easiest sensor to sanity-check without any instrument at all, because you can calculate what it should read (see below).

Zone temperature

Lowest individual impact but highest count, and one drifted zone sensor can hold an entire supply-air-temperature reset hostage. If the reset logic serves the worst-case zone, a single sensor reading warm makes the whole building chase a zone that isn't actually hot.

Chilled and hot water supply

Drives plant-side decisions. A chilled water supply sensor reading high makes the chiller work harder than it needs to, every hour it runs.

Validating without a reference instrument

A calibrated reference thermometer is the right tool. When you don't have one in your hand, you are not out of options — you have cross-validation.

Compare sensors that should agree

Multiple outdoor air sensors across a campus should read within a degree or two of each other. When one reads five degrees off its neighbours in the same weather, you have found your suspect without measuring anything.

Calculate what mixed air should be

Mixed air temperature is a weighted blend of return air and outdoor air based on damper position. At 100% outdoor air, mixed air should approach outdoor air. At minimum position, it should sit close to return air. If your mixed air reading doesn't move sensibly between those two endpoints as the damper strokes, either the sensor is wrong or the damper isn't where it says it is.

Use the building as its own reference

On an unoccupied weekend with equipment off, zone temperatures across a floor should converge as the space equalises. Sensors that remain stubbornly offset from their neighbours after hours of no conditioning are worth a closer look.

One caution on cross-validation. This finds sensors that disagree. It does not find sensors that agree with each other and are all wrong together — which happens when a whole batch was installed from the same lot, or when a template applied the same wrong scaling everywhere. Cross-validation narrows the field; it doesn't replace an actual reference measurement.

Scaling errors look like drift and aren't

Before assuming a sensor has drifted, rule out the cheaper explanation: the sensor is fine and the software is misreading it.

Symptom	Drift	Scaling / type error
Error size	Small, a few degrees	Often large, or wrong across the whole range
Error behaviour	Roughly constant offset	Grows or shrinks across the range
Onset	Gradual, over months or years	Immediate, from the day it was configured
Other sensors on the same controller	Usually unaffected	Frequently affected the same way

A sensor configured as the wrong thermistor curve, or with a 3-wire input wired as 2-wire, produces a reading that is wrong in a way that scales with temperature. Genuine drift usually looks like a stable offset. If a sensor has been wrong since the day it was commissioned, it never drifted — it was never right.

What to actually do about it

Record the offset before you correct it. If a sensor is four degrees off, that is data about how long it has been wrong and what it has been costing.
Check the mounting before the calibration. A correctly calibrated sensor in the wrong location is still giving you a wrong number.
Fix the highest-leverage sensor first, even if it isn't the one generating complaints.
Write down what you changed. A sensor offset applied in software and never documented becomes a mystery for the next technician.

Drift is the clearest example of a point that reads plausibly and lies completely. The system has no way to know. Only you do, and only if you go and check.

Go deeper

The method behind this, chapter by chapter

Every sensor, drive and control loop in the building — how it works, how to prove it with a meter, and how it fails. BAS Field Fundamentals is the method this article uses, with 20 field cases and a browser diagnostic simulator to practice on.

Get BAS Field Fundamentals

Found a wrong number or a broken link? Tell me and I’ll fix it.