BMS sensors: picking the right type and putting it in the right place
Thermistors, PT1000s, 0–10 V and 4–20 mA, how cable length skews readings, and the sensor placement mistakes behind a lot of comfort complaints.
A BMS is only as good as what it can measure. Give it a sensor that reads 2 K high, reacts slowly or sits somewhere daft, and perfectly good control logic will look broken. Before you start rewriting a strategy or retuning loops, it’s worth being sure the sensors are telling the truth.
Temperature sensors
Temperature is far and away the most common measurement on any site, and there are two broad families.
Passive sensors change their resistance with temperature, and the controller converts that into degrees. You’ll mostly see:
- NTC thermistors. Resistance drops as temperature rises, and there are lots of different curves (10k, 20k and variants of each).
- PT1000s. 1000 Ω at 0 °C, rising by roughly 3.85 Ω per degree.
- Ni1000s and a few others, on some manufacturers’ systems.
The one rule you can’t break is that the input type set in the controller has to match the sensor exactly. A thermistor on the wrong curve doesn’t just read a fixed amount out. The error changes with temperature, so it can look spot on at 20 °C and be miles off at 60 °C. When you replace a sensor, match the type, not just the housing.
Active sensors have electronics built in and send out a standard signal, usually 0–10 V or 4–20 mA, scaled to a range like 0–50 °C. They’re handy on long cable runs, or when the controller doesn’t support a particular resistance curve.
The other sensors you’ll come across
- Humidity sensors on AHUs, server rooms and anywhere conservation matters. The capacitive elements drift, so check them against a reference every year or so.
- Air differential pressure sensors and switches for filters, fan proving and duct static pressure. Watch which port is high and which is low.
- Water differential pressure sensors for pump speed control on variable flow systems. Where they’re fitted makes a big difference (more on that below).
- CO₂ sensors for demand-controlled ventilation. Some self-calibrate on the assumption that the space empties out overnight, which isn’t true for every building.
- Occupancy sensors for lighting and room control. Coverage matters as much as sensitivity.
- Heat and water meters, which are often read over Modbus or M-Bus rather than wired as analogue signals.
Choosing a signal type
| Signal | Good for | Watch out for |
|---|---|---|
| Resistive (NTC, PT1000) | Simple, no power needed | Cable resistance adds error on long runs |
| 0–10 V | Simple and widely supported | Voltage drop on long runs, and 0 V could mean zero or a broken wire |
| 4–20 mA | Long distances; 0 mA means a fault | Needs a loop supply and correct polarity |
| Volt-free contact | Status and alarms | Normally open or normally closed? |
| Comms (BACnet, Modbus) | Lots of values over one cable | Scaling and mapping mistakes |
The big advantage of 4–20 mA is the “live zero”. A broken wire reads 0 mA, which the controller can flag as a fault instead of treating it as a genuinely low reading.
Where errors creep in
Cable resistance
With two-wire passive sensors, the cable’s own resistance gets added to the sensor’s. Here’s how much that can matter with a PT1000:
- 50 m of 0.5 mm² copper, out and back, is 100 m of conductor.
- Copper is about 0.0172 Ω·mm²/m, so the cable adds roughly 0.0172 × 100 ÷ 0.5 ≈ 3.4 Ω.
- At about 3.85 Ω per °C, that’s nearly 0.9 K of error, always reading high.
That’s why long runs use heavier cable, an active sensor, or a carefully checked offset in the controller.
Accuracy and drift
Every sensor has a stated accuracy, and some get worse with age. Humidity and CO₂ sensors are the usual offenders. Checking them against a calibrated instrument should be part of PPM, not something that only happens after a complaint.
Slow response
An immersion sensor sitting dry in a chunky brass pocket can lag the real water temperature by minutes. That lag makes loops much harder to tune, and it can hide short-lived problems altogether. A bit of thermal paste in the pocket makes a real difference.
Getting the placement right
This is where a lot of problems actually start.
Outside air sensor. In the UK, put it on a north-facing wall out of direct sun, away from extract louvres, flues and windows, and high enough that nobody fiddles with it. An outside sensor baking in the afternoon sun tells the weather compensation it’s warmer than it is, and the building goes cold.
Duct sensors. Give the air room to mix after coils and mixing boxes, because air stratifies. A sensor near the top of a duct after a mixing section can read quite differently from one near the bottom. On big ducts, or anywhere mixing is poor, use an averaging sensor with a long element that runs across the duct. And don’t mount the supply air sensor right against the face of a heating coil.
Immersion sensors. Put them where water is actually flowing, not in dead legs or capped tees. On headers, make sure they see what’s going to or coming back from the system, not just the water from one boiler.
Room sensors. About 1.5 m off the floor on an internal wall is the usual guide. Keep them away from sun, radiators, fan coil discharges, printers, coffee machines and doors. When occupants keep complaining but the BMS says the room is fine, go and look at where the sensor is before you touch the setpoint.
Differential pressure sensors. Check the tubes are the right way round: high side upstream of a filter, low side downstream. For water DP control, the sensor normally belongs out near the index circuit, the furthest or most demanding branch, not across the pump. A DP sensor across the pump throws away most of the saving a variable speed drive could make.
A quick checklist for commissioning
When you’re commissioning or taking over a site, go through each sensor and:
- confirm the sensor type matches the controller input
- compare the reading with a calibrated reference
- check the scaling on active sensors (does 10 V really mean 50 °C?)
- warm or cool it and watch the reading move
- make sure its physical location matches what the point name says
- write down any offset you apply, and why
Treat offsets as a last resort. An offset that covers up a wiring fault or the wrong sensor type will cause grief for whoever comes after you.
Once you trust the measurements, it’s worth reading up on how PID loops use them.