Zone Temperature Sensors: Why They Lie and How to Catch Them in the Act
- Alex Khachaturian
- Sep 29
- 6 min read
Promise: If a room “feels wrong” but your graphic says it’s fine, this playbook shows you exactly how to prove the truth, fix it fast, and stop callbacks.
TL;DR
Zone temperature sensors lie for repeatable reasons: bad placement, wrong sensor type/curve, scaling/offsets, stale BACnet values, drafts/solar gain, and smoothing gone wild.
Win fast: bring a trustworthy reference probe, trend the right points, and run a 15-minute “truth test” (baseline → nudge → recovery).
Fix in order: physics → wiring/curve → object flags → polling/trends → logic/filters → calibration/relocation.
Quick Links
Key Takeaways
A zone temperature sensor is only as honest as its placement, wiring curve, and network freshness.
Never argue from a single number on a graphic, use a reference probe and a short, controlled test to build proof.
Trend temperature + supply temp + airflow/valve + occupancy together; lies fall apart when you correlate.
The Coffee Machine Case
Two weeks into a “mysterious warm zone,” I found the sensor mounted six inches from a new countertop espresso maker. Every morning at 8:15 a.m., the machine breathed heat across the sensor, the controller dutifully clamped cooling, and by 9:30 the space roasted. Graphics showed a steady “72°F.” The operator trusted the number; the number wasn’t trustworthy. We moved the sensor, set a short trend, and the “problem” disappeared.
Zone temps rarely lie on purpose. They lie because physics wasn’t respected, wiring wasn’t verified, or software was trusted more than the world in front of you. This article gives you a repeatable, courtroom-level way to prove what’s true and fix it, fast.
Quick Win: 15-Minute Truth Test
You can end 80% of debates in one visit:
Baseline (5 minutes): Place a known-accurate reference probe two inches from the sensor. Let it stabilize.
Nudge (2–3 minutes): Warm the sensor gently with your palm from 1–2 inches away (no breathing on it). Watch both readings.
Recovery (5–7 minutes): Remove your hand. Observe how fast each reading returns to baseline and if they land at the same value.
What good looks like: Both signals rise together, peak similarly, and recover to within ±0.5–1.0°F of each other in a similar time.
What a lie looks like: One signal flat-lines, overshoots, lags excessively, or recovers to a different baseline. That’s your smoking gun.
Playbook: Catching Lying Zone Temperature Sensors
1) Choose Your Referee (Establish Truth)
Use a calibrated air probe (not IR).
Stabilize 3–5 minutes before calling a number.
If the space is large or stratified, use two probes (desk height and 6 ft high).
Outcome: You now have a number you trust more than the graphic.
2) Check Physics First (Placement & Environment)
Bad mounts: Behind TVs/monitors, next to coffee machines, near exterior doors, under supply diffusers, on sun-soaked walls.
Drafts: Displacement diffusers or sneaky door undercuts can over-cool sensors.
Solar gain: Morning or afternoon beams can bias readings by 2–5°F.
Stratification: Tall spaces can see 2–8°F vertical deltas.
Fix: Relocate to representative breathing zone (4–5 ft from floor, away from heat sources and diffusers). Consider aspirated enclosures in low-air-movement areas.
3) Verify the Sensor Type, Curve, and Wiring
Thermistors are not interchangeable. 10k Type II ≠ 10k Type III ≠ 20k.
Confirm curve in the controller. Use a simulator to dial a known ohms value and verify displayed °F/°C.
Check for polarity on active sensors (4–20 mA / 0–10 V), correct scaling, and shield grounding (one end only).
Long runs add resistance; splices corrode; terminals loosen.
Fix: Set the correct curve, rescale inputs, re-terminate, replace corroded runs, or isolate shields. Document the curve in the BAS point description.
4) Inspect the Object, Flags, and Filters
In BACnet, look at Present_Value, Reliability, Out_Of_Service, Status_Flags (fault/overridden).
Overly aggressive smoothing/averaging can lag the signal; too little can cause hunting.
Hidden offsets added “to make people happy” stack up over years.
Fix: Clear overrides and offsets; set sensible filters (e.g., 10–30 s). If multiple sensors vote, document the math (min/max/average/median) in the logic.
5) Prove Freshness (Polling, COV, & Stale Values)
MS/TP polling intervals, COV increments, and busy trunks can freeze a displayed value that looks steady.
Check timestamp/last change; trend the raw input at the controller, not just the front-end alias.
A flat line during real changes = stale.
Fix: Tighten COV increments, fix trunk issues, or increase poll rate for critical points. Make sure the UI shows timestamps.
6) Run the 15-Minute Truth Test (and Save the Trend)
Baseline → gentle nudge → recovery.
Trend Zone Temp, Supply Air Temp, Airflow/Valve Position, Occupancy/Mode, and the Reference Probe.
Expect a coherent story: when zone temp rises, cooling should respond; when it recovers, outputs should relax.
Fix: If the story breaks, you’ve isolated which actor is lying (sensor, network, or control logic).
7) Calibrate and Document (Only After Physics & Scaling)
If the sensor is consistently off by a small, stable amount, use calibration offset at the controller (not at the UI).
Note offset, date, and reason in the point description.
If drift recurs, replace the sensor, offsets are not long-term band-aids.
8) Solve the Real Problem (Relocate/Shield/Redesign)
Relocate the sensor, add a guard to stop tampering, or switch to averaging in big rooms.
For low-airflow niches, consider aspirated sensors (small fan draws air across the element).
If people gather heat around the sensor (conference rooms), move it to representative flow or use multiple sensors with voting logic.
Troubleshooting (Symptom → Cause → Fix)
Reads 4–6°F too warm every morning→ Solar gain or a heat source nearby (appliances, AV, people hotspot).→ Move sensor, add shading, or average multiple locations.
Flat line at an even number→ Stale BACnet value or forced override.→ Check Reliability/Status_Flags, trunk health, and poll/COV; clear overrides.
Slow to react, always behind the room→ Excessive smoothing/averaging or heavy wall mass behind sensor.→ Reduce filter time constant; add standoff plate; consider aspirated housing.
Wild sawtooth swings→ Drafts or short-cycling from an over-aggressive PID/PI loop reading a noisy signal.→ Shield from drafts, tune loop (deadband/filters), fix diffuser throw.
Good with probe, bad in BAS→ Wrong curve, scaling, or unit conversion (°C vs °F) at the input.→ Correct the input type; validate with a simulator.
Good at controller, wrong at UI→ Front-end aliasing or wrong engineering units/format string.→ Fix the UI mapping; show timestamps on graphics.
Occupant says “it’s cold,” BAS says “72°F”→ Local airflow hitting occupants; mean radiant temperature (cold window) biasing comfort.→ Adjust diffuser, add radiant compensation or relocate sensor away from draft.
FAQ
How accurate should a zone temperature sensor be?
±0.5–1.0°F (±0.3–0.6°C) is typically acceptable for comfort control when correctly placed and filtered.
Where should I mount it?
Breathing zone: 4–5 ft AFF, away from direct sun, diffusers, exterior walls, and heat sources. For large rooms, use averaging or voting logic.
Why not use an IR gun?
IR reads surface temperature, not air. Air is what the sensor and people “feel.”
Type II vs Type III thermistor, does it matter?
Yes. Curves are different even if both are 10k. Pick the right curve in the controller or your reading will be wrong.
How long do I wait for stabilization?
3–5 minutes for quick checks; longer (10–15) in low-airflow or high-mass spaces.
When should I use multiple sensors?
Big rooms, high ceilings, conference rooms, or spaces with strong solar or occupant gradients. Average or use median/voting logic.
Is an offset “okay”?
Only after you’ve nailed placement, curve, and polling. Document it. Repeat offsets = replace the sensor.
Field Checklist
Calibrated reference air probe (date verified)
15-minute truth test run and saved trend
Sensor placement clear of sun, diffusers, appliances, and drafts
Correct curve/scaling verified (with simulator if needed)
BACnet flags and timestamps clean; value is fresh
Filters/averaging tuned (not laggy, not twitchy)
Fix documented: relocation/calibration/logic tweak + point descriptions updated
Results & ROI
Fewer callbacks: A single relocation + calibration can eliminate chronic comfort calls in a zone (hours saved per month).
Energy savings: A biased +3°F warm reading can under-cool a space, triggering space heaters and comfort complaints; a biased cool reading over-conditions. Expect 2–5% HVAC energy swing from persistent bias in many office zones.
Trust: Operators stop debating screenshots when you bring a probe, a clean trend, and a 15-minute experiment. Confidence closes tickets faster.
Next Read
Wrap-Up
Zone temperature sensors aren’t villains, they’re mirrors. If the mirror is in a bad spot, has the wrong curve, or isn’t being refreshed, the reflection is useless. Bring a trustworthy probe, run the 15-minute truth test, trend the right supporting points, and fix in this order: physics → wiring/curve → flags → polling → filters → calibration/relocation. That sequence ends arguments and callbacks.
Comments