3-wire vs 4-wire RTD: what actually changes, and what your Allen-Bradley module does with the fourth wire

A contractor once handed me a box of 4-wire Pt100 probes for a retrofit and asked which terminal the fourth lead went to. The card in the rack was a 1746-NR8, and it has three terminals per channel. Not four. Three.

That is the part almost nobody tells you when they explain the difference between 2-wire, 3-wire and 4-wire RTDs. The theory is real and it matters. But on every Allen-Bradley RTD input card I have worked with, the fourth wire has nowhere to land, and the card treats your expensive 4-wire probe as a 3-wire sensor whether you like it or not. Below is what each wire is actually doing, the numbers that decide whether 2-wire costs you anything, and how the four common Rockwell RTD cards handle it, straight from their manuals.

What the extra wires are buying you

An RTD is a resistor. That is the whole basis of the measurement, and it is also the whole problem: the copper running from your panel out to the probe is a resistor too, and the instrument cannot tell the two apart on its own.

Put a number on it. A Pt100 sits at 100 ohms at 0 C, and the 385 curve means the resistance moves about 0.385 ohms for every degree. So every ohm of copper in the loop reads as roughly 2.6 degrees of temperature that is not there. Not a rounding error. On a long run in a hot cable tray it can be tens of degrees.

Lead resistance translated into error on a Pt100 (385 curve, roughly 0.385 ohms per degree C):

Total lead resistance in the loop	Apparent error on a Pt100	Same lead resistance on a Pt1000
0.5 ohms	about 1.3 C	about 0.13 C
1 ohm	about 2.6 C	about 0.26 C
2 ohms	about 5.2 C	about 0.5 C
5 ohms	about 13 C	about 1.3 C
10 ohms	about 26 C	about 2.6 C
25 ohms	about 65 C	about 6.5 C

Those numbers are arithmetic off the temperature coefficient, not a spec from a datasheet, and they always read high because added resistance looks like added heat.

So, the three configurations:

A 2-wire RTD gives the card no way to separate cable from element. Whatever the copper adds goes into the reading and stays there.

Add a third conductor back to that same end of the element, and the card can measure one leg's resistance directly and subtract it out. The catch is in the assumption: it assumes the other leg is identical. Rockwell says it plainly in the 1746-NR8 manual, publication 1746-UM003, that the module reads the resistance of one wire and takes the other to be equal, and that a real difference between the two shows up as error.

Four wires drops that assumption entirely. One pair pushes the excitation current, a separate pair measures the voltage right at the element, and because almost no current flows in the measuring pair its resistance stops mattering. That is the lab-grade answer.

And on three of the four cards below, it is also the answer you cannot use.

All of these cards top out at three wires

This is not a quirk of one old module. It runs across generations.

The 1746-NR8 manual states it in the overview: the module is designed to accept input from RTD sensors with up to 3 wires, and when you use a 4-wire sensor one of the two compensation wires is not used and the sensor gets treated like a 3-wire. It even tells you not to overthink which one you strand, because it does not matter which sense wire is left open.

Same story on the MicroLogix side. Publication 1762-UM003 describes the 1762-IR4 as configurable for 2-wire or 3-wire RTD, says channels are compatible with 4-wire sensors but the fourth sense wire is not used, and repeats it in the wiring chapter: leave one sensor wire open.

ControlLogix does not change the answer, only the wording. The 1756-IR6I wiring examples in publication 1756-UM009 show a 3-wire RTD, then show the 4-wire example with a single note under it: wiring is the same as the 3-wire RTD with one wire left open.

The newest of the four is the most interesting. In publication 5069-UM005, the table of analog devices the 5069-IY4 supports lists a 2-wire RTD device and a 3-wire RTD device. A 4-wire RTD is not in that table at all, and the phrase does not appear anywhere in the manual. For actual terminal-level wiring the analog manual sends you to the specifications document, publication 5069-TD001, so I would confirm there before ordering probes rather than assume from silence.

The practical upshot is the same for all four: if a spec sheet somewhere in your project says 4-wire, and the input card is one of these, you are getting 3-wire accuracy. Buy the probes that fit the card, not the ones that fit the ideal.

The 1746-NR8 terminal block: three terminals per channel, named RTD, Sense and Return. There is no fourth.

The terminals get renamed on every platform

Nothing electrical changes between these cards. The labels change constantly, which is exactly why a drawing copied from an older panel lands people on the wrong screw.

Module	Platform	Channels	Terminal names per channel
1746-NR8	SLC 500	8	RTD x, Sense x, Return x
1762-IR4	MicroLogix 1200 expansion	4	EXC, Sense, RTN, plus an NC terminal
1756-IR6I	ControlLogix	6	IN-x/A, IN-x/B, RTN-x/C
5069-IY4	Compact 5000	4	See publication 5069-TD001

On the 1756-IR6I there is one more line worth reading twice, because it is a jumper and jumpers get forgotten. For two-wire resistor work, including calibration, the manual says IN-x/B and RTN-x/C have to be shorted together. Same manual also tells you not to put more than two wires on any single terminal.

The 1756-IR6I. Six isolated channels on a 20-pin RTB, with terminals called IN/A, IN/B and RTN/C instead of RTD, Sense and Return.

2-wire is where the damage actually happens

The 3-wire versus 4-wire argument is mostly academic on this hardware. The 2-wire decision is not, and there are hard numbers attached to it.

Publication 1746-UM003 puts a ceiling on what 3-wire compensation can even do: the module compensates up to an overall cable impedance of about 25 ohms, and it wants lead resistance kept under that figure. Then it gets stricter. To hold the accuracy, it says the resistance difference between the cable lead wires has to be equal to or less than 0.01 ohms. That is a tight tolerance, and it is the real argument for buying decent instrument cable instead of using whatever spare pair is in the tray.

Rockwell also tells you outright not to run 2-wire on long cable, and if you are stuck with it, to fight the resistance with copper: use a heavier gauge, AWG 16 instead of AWG 24, and cable with lower resistance per foot. The 1746-NR8 terminal block takes one AWG 14 conductor. The 1762-IR4 block takes two.

The 1762-IR4 will flag your working 2-wire sensor as open

This one costs people an afternoon. On the 1762-IR4, running a channel in 2-wire configuration sets the open-circuit status bit for that channel. Not because anything is broken. Publication 1762-UM003 explains it: 2-wire does not let the module compensate for lead resistance, so the module keeps putting the uncompensated value in the input data file and sets the OCx bit in word 4 for any enabled channel using a 2-wire configuration, as a flag that the data carries lead wire error.

Read that as a technician at two in the morning and it looks exactly like a broken sensor. Reading is present, temperature is plausible, and the open-circuit bit is on. Nothing is wrong. The card is telling you the reading is uncompensated, which you already knew when you landed two wires instead of three.

Which is a good reason to just run three. The third conductor costs almost nothing on a new pull, and it takes a permanent status bit off your diagnostics.

The 1762-IR4. Four channels, and a 2-wire sensor on any of them permanently reports open circuit by design.

The settings that will fool you after the wiring is right

Wiring is only half of a temperature reading. Every one of these cards has at least one configuration setting that produces a wrong number from perfectly good copper.

Integer mode turns the 1756-IR6I into an ohmmeter

Publication 1756-UM009 is blunt about it: integer mode does not support temperature conversion on the temperature-measuring modules, and if you choose integer mode the 1756-IR6I is strictly an ohms module. Sensor type linearization to a temperature value only happens in floating point mode.

So the failure mode is a channel sitting there reporting something near 110 when the process is at room temperature, and somebody hunting the wiring for a fault that is really a radio button in the module properties. 110 is not a broken reading. It is 110 ohms, which is what a Pt100 does at about 25 C.

Excitation current is not the same choice on every card

The 1746-NR8 gives you 0.25 mA or 1.0 mA. The 1762-IR4 gives you 0.5 mA or 1.0 mA. Lower current means less self-heating in the element and a worse signal-to-noise ratio, and the two cards do not offer the same low end, so a configuration cannot simply be carried across.

On the 1746-NR8 that choice also quietly moves your temperature range, and this is the trap. Take a 1000 ohm platinum 385 element, the sensible pick if you are stuck with long leads, since the same lead resistance costs you a tenth of the error. Run it at 0.25 mA and the range in publication 1746-UM003 is -200 to +850 C. Run the same element at 1.0 mA and the range collapses to -200 to +50 C. Fifty. A 500 ohm platinum element does the same thing, dropping from +850 to +390.

There is a combination the card refuses outright, too. Pick a 10 ohm copper RTD and set the excitation to 0.25 mA and the 1746-NR8 raises a configuration error, because that pairing is not allowed.

Lead resistance measurement doubles your scan time

The 1746-NR8 lets you measure lead resistance never, once every five minutes, or on every single acquisition. Set it to always and you pay for it. At the 28 Hz filter a channel scan runs 125 ms, or 250 ms with lead resistance. Eight channels, 28 Hz filter, lead resistance on every acquisition, and the module update time is 2000 ms by Rockwell's own arithmetic.

Two seconds is fine for a tank. It is not fine feeding a PID loop that somebody tuned expecting fresher data. Periodic measurement is the setting I reach for on anything that is actually controlling.

On the 5069-IY4 the sensor type picks your range for you

Newer platform, different behavior. In publication 5069-UM005 the RTD input range on a 5069-IY4 follows the sensor type you select, and Studio 5000 sets it automatically: a 100 ohm Pt 385 lands on 1 to 500 ohms, a 1000 ohm Pt 3916 on 8 to 4000 ohms. Pick the wrong sensor type and you have quietly picked the wrong range as well.

It also has a feature the legacy cards do not, and it is genuinely useful on copper elements. The 10 ohm copper offset lets you dial out a known error in units of 0.01 ohms, so a copper RTD reading 9.74 ohms at 25 C gets an offset of -0.26 entered on the configuration tab, or -26 in the tag. Temperature units go beyond C and F on this one as well, to Kelvin, Rankine and a custom scale.

The 5069-IY4 handles current, voltage, RTD and thermocouple on four channels, and its supported device table lists 2-wire and 3-wire RTDs only.

Sorting out which wire is which at the panel

Say a probe shows up with three leads and no tag, or the colors do not match anything in the drawing. You do not need the datasheet for this, and you do not need to guess. A meter answers it in about a minute, because the geometry of a 3-wire element only allows one answer.

Two of the three wires are tied to one side of the sensing element. Measure between all three pairs. One pair reads close to zero, a fraction of an ohm, and that pair is your compensation pair, both landing on the element's common end. Each of those two wires reads the element itself against the third wire, right around 100 ohms for a Pt100 near freezing and closer to 110 at normal room temperature.

From there the terminal assignment follows the card. On a 1746-NR8, the compensation pair goes to RTD and Sense, since Sense's whole job is reading the lead resistance on that same excitation leg. The lone wire goes to Return. Take the reading with the module powered down and the leads off the terminals, or the excitation current will sit on top of your measurement.

While the meter is out, it is worth knowing what the card considers broken, because the thresholds are specific rather than intuitive. The 1746-NR8 declares an open circuit when the excitation current falls under half of what you selected, and a short circuit when the compensated element resistance comes out below 3 ohms. The short-circuit test only applies to RTD input types, not to the direct resistance ranges, since those legitimately start at zero.

Near zero ohms across two leads means you have found the compensation pair. Either of those against the third wire reads the element itself.

One of these four is no longer current hardware

If you are speccing new, this changes the list. The 1756-IR6I now lives in an appendix of publication 1756-UM009, and Rockwell labels the section as reference information regarding the discontinued 6-channel 1756-IR6I, 1756-IT6I and 1756-IT6I2 modules. The revision I am working from is 1756-UM009G, dated March 2025, and it points at publication 1756-RM011 for migrating from the 6-channel isolated analog modules to the newer 8-channel ones.

That does not make a working IR6I in a running rack a problem. It makes buying one new a problem, and it means a rack with several of them is a bigger conversation than a card swap. Check the individual catalog number on Rockwell's lifecycle status tool before you build a bill of materials on any of it, because lifecycle state is assigned per catalog number and moves over time.

Questions I get asked about this

How does a 3-wire RTD work?

The third lead doubles up on one terminal of the element, and that is what lets the card read a single lead's resistance by itself and take it back out of the total. The whole thing rests on that second lead matching the one that got measured. Publication 1746-UM003 spells that out and warns that any real mismatch between the legs turns into measurement error.

Can I use a 3-wire RTD as a 2-wire?

Physically yes, and you will lose the compensation. On a 1762-IR4 you will also earn a permanent open-circuit status bit on that channel, which is the module telling you the data is uncompensated. If the third wire is already run to the panel, there is no reason to leave it off.

What is the color code for RTD wires?

Colors vary by standard and by manufacturer, and none of the Rockwell module manuals specify them, because the module cares about terminal function rather than insulation color. Check the sensor's own datasheet, or skip the question entirely and identify the compensation pair with a meter as described above.

Does cable length actually affect the reading?

On 2-wire, directly and permanently: every ohm in the loop is roughly 2.6 degrees on a Pt100. On 3-wire it is compensated up to a point, and on a 1746-NR8 that point is about 25 ohms of overall cable impedance, with the two legs needing to match within 0.01 ohms for the compensation to hold accuracy. Long runs are where a Pt1000 element earns its price, since it divides that lead error by roughly ten.

Why is my 1756-IR6I reading ohms instead of temperature?

It is in integer mode. Publication 1756-UM009 states that in integer mode the module is strictly an ohms module, and turning a sensor type into degrees needs floating point instead.

In stock, tested, and backed by a 2-year warranty

IQElectro keeps the Allen-Bradley RTD and analog input cards on the shelf across all four of these platforms: 1746-NR8 for SLC 500, 1762-IR4 for MicroLogix expansion racks, 1756-IR6I for ControlLogix, and 5069-IY4 for Compact 5000, along with the surrounding analog and temperature modules. The IR6I is worth a mention on its own, since it is the one you can no longer order new. Everything ships tested with a two-year warranty. Send the catalog number off the card you are replacing and we will confirm what is on the shelf first.