RTD temperature sensor calibration: three jobs that share one name

Every few jobs I end up in the same argument. Somebody's commissioning agent writes "calibrate all temperature sensors" on a punch list, the controls contractor answers that nobody calibrates an RTD, and both of them are partly right. RTD temperature sensor calibration is not one procedure. Three separate jobs share the name, and each one checks a different piece of the loop.

Pick the wrong one and you can burn a day with a dry block and still read two degrees high. Or enter an offset that quietly hides a sensor on its way out. Here is what each job actually checks, the numbers that decide whether anything is wrong to begin with, and the order I work them in.

Look up what the sensor is allowed to be off by first

A good share of the RTD temperature sensor calibration calls I take are not calibration problems. Someone held a handheld thermometer next to a Pt100, found half a degree of disagreement, and opened a ticket. At most process temperatures half a degree is inside what a Class B sensor is allowed to be, brand new, in spec, straight out of the bag.

IEC 60751 does not publish tolerance as a single number. Each class is a formula, and the allowed band widens as the temperature climbs. That is the part people skip when they quote a class off a datasheet.

Allowed sensor error at each process temperature, worked out from the IEC 60751 tolerance formulas. These figures are the sensor alone, before the module, the cable or any drift is counted.

Each class carries its own valid range, and wire wound and thin film do not get the same one. Class AA covers -50 to +250 °C wire wound and only 0 to +150 °C film. Class A runs -100 to +450 °C wire wound, -30 to +300 °C film. B and C both reach -196 to +600 °C wire wound, with film at -50 to +500 °C for B and -50 to +600 °C for C. So a Class AA figure quoted at 400 °C is meaningless, because the class does not exist up there.

The standard also puts a wiring rule on the tight classes. Anything better than Class B has to be 3-wire or 4-wire, and 4-wire is what it recommends. A 2-wire Class A probe is a contradiction on paper before anybody lands a conductor.

Datasheets muddy this with two naming systems. The bare platinum element is marked W0.1, W0.15, W0.3 or W0.6 when it is wire wound, F0.1 through F0.6 when it is thin film. That number is the first constant in the same tolerance formula. The assembled probe is what carries AA, A, B or C. Same arithmetic, and for wire wound types the loose element is rated across a wider span than the finished thermometer built around it.

If what you need is the expected resistance rather than the allowed error, the 2022 edition moved the printed resistance table to informative and made the formula itself the specification. Above zero it is Rt = R0 (1 + At + Bt²), with A = 3.9083 x 10-3 and B = -5.775 x 10-7. Below zero a third term joins in, C = -4.183 x 10-12, applied as C(t - 100)t³. Run 100 °C through that on a Pt100 and you get 138.51 ohms. That value and 100.00 ohms at the ice point are what a meter check is really being compared against.

The three things people call RTD temperature sensor calibration

They cover different parts of the chain, they cost different amounts of time, and only one of them can tell you the sensor is going bad.

The three jobs that get filed under RTD temperature sensor calibration, and how much of the loop each one actually covers.

Calibrating the sensor

This is the only version of RTD temperature sensor calibration that can find a bad sensor. The element goes into a known temperature alongside a reference probe. You record what it reads at each point and compare that against the resistance the curve says it should have given you.

You cannot adjust a Pt100. There is no trim screw on a platinum element. What comes out of sensor calibration is a table of corrections, and those corrections then have to land somewhere: coefficients in a smart transmitter, or an offset in the controller. Skipping that last step is how a sensor gets "calibrated" and goes on reading exactly as wrong as it did before.

It is worth the trouble because the sensor is where the error lives. Fluke's calibration guidance puts the sensor at more than 75 percent of the total error coming out of a temperature transmitter, and that matches what fails in the field. The element is the part sitting in the process, thermally cycled, vibrated, and now and then run past its rating.

Calibrating the module or the transmitter

This half of RTD temperature sensor calibration never touches the sensor. Precision resistors go in where the sensor was and the electronics get trimmed against known values.

On a 1756-IR6I the procedure runs out of the module properties in the software, and you put the module in Program mode or inhibit its connection first, while it stays in Run mode. It wants two references: 1 ohm for the low point, 487 ohms for the high point, applied to every channel being calibrated. Rockwell recommends doing all channels in one pass instead of one at a time. When it finishes, the new gain and offset and the date get stored on the module.

The SLC 500 RTD module works differently, and it is worth knowing that before you go hunting for an adjustment that does not exist. The 2-pin connector on that board is factory setup only. What you get instead is autocalibration: the module runs its own A/D conversions on zero and full scale and corrects itself. Leave any channel's calibration disable bit at zero and it does that every five minutes. It is not converting input data while it runs, and the cycle costs 510 ms plus 125 ms for each unique combination of input type and excitation current on the card. Rockwell's own advice is to trigger one after anything that swings the cabinet temperature, a door left open for an hour being the obvious case.

Neither version sees a drifted sensor or a damaged cable. It proves the electronics are honest about the ohms presented at the terminals. Nothing beyond that.

The single-point offset, which covers the module and the cable together

This is the one almost nobody runs, and it solves more real problems than either of the others. It is the only job that measures the actual cable in the actual tray.

Rockwell documents it for the SLC 500 RTD module as an optional procedure, and the claim attached is a specific one. Better than ±0.2 °C for the module and cable combination, as long as the RTD ends up working within ±50 °C of the temperature you calibrated at. The sequence runs like this.

1.    Cycle power to the chassis.

2.    Pick a calibration temperature within ±10 °C of the control point.

3.    Work out the exact resistance for that temperature, to ±0.01 ohms, from a published resistance table.

4.    Pull the RTD and land a fixed precision resistor of that value in its place. Rockwell asks for a 2 ppm temperature coefficient part.

5.    Let the module read that resistor through the real cable and report a temperature.

6.    Subtract the calibration temperature from what the module reported. The difference is your offset.

7.    Reconnect the RTD.

8.    Subtract the offset from the measured temperature in ladder.

Say the control point sits at 77 °C. You look up 129.75 ohms for that temperature, wire a resistor of that value in at the sensor end of the run, and the module comes back with 78.4 °C. Your offset is 1.4 °C, and it gets subtracted in logic from then on.

The single-point offset: the only part of RTD temperature sensor calibration that includes the field cable you actually installed.

The useful detail is what invalidates it. Replacing the RTD does not, because the offset describes the module and the cable rather than the sensor. Disturbing or degrading the cable does, which is a decent argument for running this after the tray work is finished instead of during construction.

Ohms are not degrees, and the module spec gives you ohms

Every RTD temperature sensor calibration figure printed on a module datasheet is in ohms or in percent of range. The temperature error that comes out the far end depends on where you are sitting on the curve, because the curve is not a straight line.

Rockwell works the example on the 1756-IR6I. Calibrated at operating temperature, with that temperature reasonably stable, accuracy is better than 0.1 percent of full range for the first year. On the 1 to 487 ohm input range the error gets computed across 507 ohms of span rather than across the sensor's usable range, so worst case is 0.507 ohms. Convert that at 0 °C with a Platinum 385 sensor and it lands at -1.25 to +1.2 °C. The same 0.507 ohms at 200 °C works out to roughly ±1.4 °C.

Same module, same error in ohms, different answer in degrees. Anyone stacking a sensor tolerance on top of a module spec without doing that conversion first is adding two numbers that are not in the same units.

Ice water, a dry block, or a box of resistors

Four ways to source a temperature for RTD temperature sensor calibration, and they do not prove the same things.

What each temperature source actually verifies, and what it leaves untested.

The decade box deserves its own warning, because substituting one for the sensor is the most common thing that gets written into a report as RTD temperature sensor calibration and is not. It exercises the transmitter and the input, and leaves out the component carrying most of the error.

The ice bath is real, but it is a single point and it is not quite the number people assume. The international temperature scale fixes the triple point of water at 0.01 °C exactly, which is what a reference cell reproduces. A bucket of crushed ice and distilled water gets close to that, and how close depends on how well it was packed and how clean the water was. Fine as a go/no-go check on a loop running near freezing. Not a certificate.

Read the symptom before you schedule the calibrator

Most of the temperature complaints I get handed are not drift at all, and RTD temperature sensor calibration is the wrong tool for almost all of them.

Symptoms that look like a calibration problem and usually are not.

Two of those are worth another sentence. A reading that moves when someone shifts the cable bundle is not a calibration problem at any level. The excitation through an RTD is small, the signal is a resistance, and an unshielded run next to switching gear is going to wander. And the 2-wire versus 3-wire question comes up constantly whenever drives are involved. The third conductor compensates lead resistance. It does nothing about electrical noise, and buying new sensors to fix a drive problem is money spent at the wrong end of the loop.

So how often

Nobody can hand you an RTD temperature sensor calibration interval off the shelf. IEC 60751 specifies what the sensor has to be when it is manufactured, not how often you check it afterward. What sets the number in practice is the contract, the regulator if you answer to one, and how hard the sensor's life is. Drift comes from thermal cycling and process exposure, not off the calendar.

The equipment-level triggers are firmer. Rockwell wants an autocalibration after anything that moves cabinet temperature substantially. The single-point offset only needs redoing if the cable is disturbed or degraded. And if you have historian data, a slow one-way creep on a loop that used to sit flat tells you more than any interval somebody typed into a spec.

Questions I get asked about this

Can RTD temperature sensor calibration be done on the sensor itself?

You can calibrate it in the sense of measuring its error against a reference and writing that error down. Adjusting it is another matter, because a platinum element has no trim anywhere on it. The correction has to be entered downstream, in the transmitter coefficients or in the controller, or the measurement goes on being exactly as wrong as it was.

How do I check an RTD with a multimeter?

Disconnect one end so you are reading the element instead of the whole loop, then measure resistance and compare against the curve. On a Pt100 that is 100.00 ohms at 0 °C and 138.51 ohms at 100 °C. Two cautions. On a 2-wire connection you are measuring the leads too, and on a long run that is degrees, not hundredths. And a meter pushes current through the element to read it. The standard holds measuring current to whatever keeps self-heating under 25 percent of the class tolerance, which usually means no more than 1 mA on a 100 ohm wire wound element. Treat a meter reading as a go/no-go test, not as RTD temperature sensor calibration.

What equipment do I need for RTD temperature sensor calibration?

A temperature source that actually heats and cools, a reference probe with a current certificate, and a way to read resistance accurately. A dry block with an external reference covers most field work. A stirred bath does better on uniformity when the sensor can leave the plant. Precision resistors and a decade box cover the module and transmitter side, and only that side.

Does a Class A sensor buy me anything on a standard input card?

Sometimes very little. Work the numbers: sensor tolerance from the class formula, plus module error converted from ohms into degrees at your operating temperature. If the card is contributing north of a degree and the class difference is a quarter of one, you paid for accuracy the card then throws away.

When the card is the part that has to go

Once RTD temperature sensor calibration points at the input card rather than the sensor, the question turns into what is on the shelf. IQElectro keeps the Allen-Bradley RTD input modules in stock across the platforms these loops actually live on: 1756-IR6I for ControlLogix, 1769-IR6 for CompactLogix, 1746-NR8 and 1746-NR4 for SLC 500, 1762-IR4 for MicroLogix expansion racks, and 1794-IR8 on Flex I/O. Everything ships tested with a two-year warranty. Send the catalog number off the card you are pulling and we will confirm what is on the shelf before you order.