Education · Temperature Sensors

Temperature Sensors Signals

Most of the temperature readings in a building ride on the simplest hardware in the trade: a resistor on the end of two small wires. No power, no polarity, no electronics — the element just changes resistance with temperature, and the controller turns ohms into degrees through a stored curve. That translation is the whole subject of this page. One question, start to finish: why does the controller need to know exactly which temperature sensor is on the wire — and how do you verify one with a meter? How the sensor physically lands on an input is Controller Wiring’s story, and how raw signals in general become values on a graphic is Analog Sensing’s, the chapter’s opener. This page is about the sensor itself: the curve families, the RTD alternative, and the ten-minute meter check that settles an argument with a graphic.

One Name, Several Curves

The workhorse BAS temperature sensor is an NTC thermistor — negative temperature coefficient: resistance falls as temperature rises, steeply and non-linearly. Near room temperature a common 10K element moves about 240 Ω for every °F — an enormous, easy-to-read signal. The “10K” in the name is a single defining point: 10,000 Ω at 77.0 °F. And here is the trap hiding in that name: a point is not a curve. Two sensors can both read 10 kΩ at 77 °F and disagree at every other temperature on Earth.

That is exactly the situation in the field. The two big industry curve families — 10K Type II and 10K Type III — share the defining point and nothing else. Type II is the steeper curve; Type III is shallower, so it reads lower at the cold end and higher at the hot end. At 40.0 °F a Type II sits near 26.3 kΩ while a Type III sits near 24.5 kΩ — close enough to look alike on a meter, far enough apart to matter on a graphic. They are not interchangeable, and neither one is “the” 10K curve; a controller asks you to pick because it genuinely cannot tell from the wire.

It gets one step messier: several vendors ship shunted 10K conventions — a 10K element with a fixed resistor in parallel to flatten the curve over the occupied range. The JCI convention pairs a 10K element with an 8.7 kΩ shunt and reads only about 4.7 kΩ at the 77 °F defining point; the TAC Type 5 convention pairs a Type III element with an 11 kΩ shunt and reads about 5.2 kΩ there. Both are still called “10K sensors” in conversation — and both would look badly broken to a tech expecting 10 kΩ at room temperature. The lesson generalizes: the label on the box is a name; the curve selected in the software is a decision. The controller needs the two to agree, and nothing in the wiring will force them to.

Which brings us to the quiet failure this section exists to teach. Wire a Type III sensor to an input configured for Type II and everything appears to work. The point reads a live, plausible temperature; it trends smoothly; it never alarms. It is also wrong — by nothing at the 77 °F crossover, by about 3 °F when the truth is 40.0 °F, and by more the further the temperature drifts from the crossover. Compare that with the loud faults: an open or shorted sensor pins the reading to a rail, where anyone can see it — the thermistor landing in Controller Wiring tells that story (an open reads full-scale cold, a short full-scale hot). A curve mismatch does the opposite. It reads plausibly wrong, forever — an outdoor-air sensor a few degrees off quietly skews economizer changeovers and lockouts for years, and nobody hunts a 3° lie.

10K Type II and Type III resistance curves overlaid, crossing at 77 degrees Fahrenheit A chart of resistance versus sensor temperature. The vertical axis is resistance on a logarithmic scale from 5 kilohms to 100 kilohms; the horizontal axis is temperature from 0 to 100 degrees Fahrenheit. Two falling curves run from upper left to lower right: 10K Type II, the steeper curve, and 10K Type III, the shallower one. A thin shaded band between them marks their disagreement. The curves cross at exactly one point, marked with a dot at 77 degrees Fahrenheit and 10 kilohms — the shared point that gives both curves the 10K name. A worked point at 40 degrees Fahrenheit shows the two curves reading 26.3 kilohms and 24.5 kilohms — pick the wrong curve in the software and the graphic reads about 3 degrees Fahrenheit off, plausibly, with no alarm. TWO CURVES, ONE NAME — 10K TYPE II vs TYPE III 100 kΩ 30 kΩ 10 kΩ 5 kΩ resistance (log scale) 0 20 40 60 80 100 sensor temperature (°F) TYPE II TYPE III at 40.0 °F: 26.3 kΩ vs 24.5 kΩ pick the wrong curve and the graphic reads ≈ 3 °F off — plausibly 77 °F — both curves read 10.0 kΩ (the one point the shared name promises)

10K Type II — steeper 10K Type III — shallower the divergence — a few quiet degrees

RTDs — the Other Family

The other family you will meet is the RTD — resistance temperature detector — a fine metal element, usually platinum, whose resistance rises with temperature and does it almost linearly. Where thermistor curves are vendor conventions, the common RTDs are standardized: a Pt100 reads 100.0 Ω at 32.0 °F on the international IEC 60751 curve, and a Pt1000 is the same curve scaled ten times. The odd cousin is Balco — a nickel-iron element near 1,000 Ω at 70 °F, roughly linear — common on older gear. Direction is a free identification clue at the meter: warm the element in your hand, and a thermistor’s resistance falls while an RTD’s rises.

The RTD’s catch is the size of its signal. A Pt100 moves about 0.22 Ω per °F — a thousand times less than the 10K thermistor’s ≈240 Ω per °F near room temperature. At that scale the copper field wiring stops being invisible: the leads have resistance too, and the input cannot tell lead ohms from element ohms. A 100 ft run of 18 AWG pair puts about 1.3 Ω of copper in series — on a 2-wire Pt100 that reads about 6 °F high, constantly, with nothing broken. Put the same 1.3 Ω in series with a 10K thermistor and the error is a few thousandths of a degree — which is exactly why thermistors get away with plain 2-wire landings and RTDs need help.

The classic help is the 3-wire landing: run a third conductor to the same end of the element, matched to the other two — same gauge, same length, same cable. That gives the input two things it can measure: the sensor loop (element plus two leads) and a lead-only loop (two leads through the jumper at the sensor head). Subtract the second from the first and the lead resistance cancels, leaving the element alone. The whole scheme rests on the leads being matched, which in one cable they are. Where lab-grade accuracy matters, a 4-wire landing measures the element directly and cancels the leads completely — and the blunter modern fix is the Pt1000: ten times the base resistance makes the same lead ohms ten times less significant (that 1.3 Ω run becomes about 0.6 °F), which is a big part of why newer HVAC controllers favor it.

A 3-wire RTD landing and how the input cancels lead resistance A controller input block on the left with three terminals, T1, T2, and T3, connected by three field conductors, L1, L2, and L3, to a sensor head on the right containing a Pt100 element. Each conductor carries a small resistor symbol — every lead adds a few ohms of copper. Inside the sensor head, L1 lands on the top of the element while L2 and L3 are jumpered together to the bottom of the element. A caption explains the compensation: measuring T1 to T2 reads the element plus leads L1 and L2; measuring T2 to T3 reads leads L2 and L3 alone; subtracting the two cancels the matched leads and leaves the element resistance by itself. CONTROLLER T1 T2 T3 3-wire RTD input SENSOR HEAD Pt100 element L1 L2 L3 every lead adds a few ohms of copper T1→T2 reads element + L1 + L2 · T2→T3 reads L2 + L3 — the leads alone subtract the two: matched leads cancel and the element remains

Knowing both families, the field split makes sense. Thermistors own BAS space, duct, and return sensors: they are cheap, their signal is enormous, lead resistance is a non-issue, and any sensor on the same curve is a drop-in replacement. RTDs show up where accuracy and stability pay their way: plant and process sensors, high-temperature applications past a thermistor’s comfort, and especially matched pairs — an energy (BTU) measurement lives and dies on the supply-return ΔT, and a pair of matched platinum elements holds that difference honestly for years. RTDs also drift less over the long haul, which is why the sensor whose number feeds a bill is rarely a thermistor.

Prove It With a Meter

Now the payoff — the workflow that turns all of this into a verdict. A space reads 80 °F and the occupants say it is freezing; an OAT point disagrees with the weather; two duct sensors a foot apart disagree with each other. There are exactly three suspects: the sensor, the input, and the configuration — and the graphic cannot testify against itself. The meter separates them, before anyone swaps hardware on a hunch.

The workflow is four steps. One: disconnect the sensor from its input terminals, so the meter reads the sensor alone and not the sensor in parallel with the input’s own circuit. Two: meter the ohms across the sensor’s leads. Three: turn ohms into a temperature using the R/T table for the type the sensor is supposed to be — the Thermistor / RTD Calculator’s R/T tables are exactly this lookup, both directions, for every common curve on this page (and if the type itself is in doubt, its Identify tab ranks every curve against measured points). Four: compare that temperature against a reference thermometer sitting where the sensor sits. A working 10K sensor in a 70 °F room meters near 11.9 kΩ on a Type II curve — numbers in that neighborhood are the everyday sanity check that the sensor, at least, is telling the truth.

The verdict comes from what disagrees. If the metered ohms match the reference temperature through the right table but the graphic still reads wrong, the sensor is honest — move the suspicion inside: the input, or more often the configuration, starting with which curve the point has selected. If the meter reads open or near-zero, the sensor or its run has failed — localize it by metering again at the sensor head: healthy ohms at the head with an open back at the panel convicts the run and its terminations; an open at the head convicts the sensor itself. And if the ohms translate to a temperature that sits a consistent few degrees off the reference, you are back in the first section of this page — suspect a curve mismatch before you condemn hardware, because checking the software’s curve selection is free and replacing a good sensor with an identical good sensor fixes nothing.

Two boundaries on what the meter can prove, both deliberate. A sensor can pass every check above and still steer the system wrong because it is measuring the wrong air — placement, stratification, and the averaging elements that big ducts need are their own subject for another page. And when a healthy sensor disagrees with the reference by a small, honest margin, trimming it with a calibration offset is its own discipline with its own traps — also a page of its own. What this page hands you is the core of the argument: the wire carries ohms and nothing else. Only the curve turns ohms into a temperature, the controller can only be told which curve — and the meter, a table, and a reference thermometer are how you audit that the label, the wire, and the configuration are all telling the same story.

← Back to Education