RTD Excitation Current and Measurement Circuit
A platinum resistance thermometer is the default choice when temperature has to be measured accurately and repeatably over a wide range. Its resistance is nearly linear with temperature, it is stable over years, and the accuracy of the measurement then depends almost entirely on how the excitation current and the wiring are handled.
Why the Excitation Current Matters
The sensor gives a resistance, and the circuit converts that to a voltage by passing a known current through it. Errors in the current appear directly as temperature errors, so a hundred parts per million change in current is a hundred parts per million change in the reading, which for platinum is about twenty five millithousandths of a degree per degree.
That places a demand on the current source that a simple resistor from the supply cannot meet. A reference and an operational amplifier, or a dedicated current source with a matched internal reference, is the usual answer, and the temperature coefficient of the current source sets the drift of the whole instrument.
The current also sets the resolution. A hundred microamps through a hundred ohm sensor at zero degrees gives ten millivolts, and a change of one degree changes it by about forty microvolts. That is the number the converter and the amplifier have to resolve.
Choosing the Current
A current of one milliamp is the classic value for a wire wound or thin film platinum sensor in a metal sheath. It gives a useful voltage without dissipating too much power in the element, and most sensors are specified for a current in that region.
A smaller current reduces self heating but also reduces the signal, and the noise of the following amplifier then becomes a larger fraction of the reading. The choice is a trade between thermal error and electrical noise, and the datasheet often gives the measuring current the accuracy figures were obtained at.
Where the sensor is a small thin film element on a substrate that cannot dissipate heat, the current is reduced to a tenth of a milliamp or less and the amplifier is chosen for a low noise at a low source impedance. The trade is real and should be made explicitly rather than by copying a circuit.
Two, Three and Four Wire Connections
A two wire connection places both lead resistances in series with the sensor. A copper lead of a few tenths of an ohm against a hundred ohm sensor is a few tenths of a percent, which is already a degree of error, and the copper also changes with the ambient temperature.
A three wire connection puts the lead resistance into two arms of the measurement so that it cancels, provided the two leads are identical. It is the common compromise in industrial installations and works well with a matched cable.
A four wire connection separates the current path from the voltage sense path entirely, so the lead resistance has no effect at all. It is the standard in the laboratory and in any installation where the cable is long or the ambient varies along it. The extra pair of conductors costs less than the error it removes.
The Kelvin Connection in Practice
A Kelvin connection means bringing the sense conductors to the sensor terminals separately from the current conductors, so that the voltage is measured at the element rather than at the end of the cable. The junction of the two pairs is the definition of the measurement point.
In practice the sense pair should be a twisted pair routed with the current pair, and the junction should be at the sensor terminals, not at the instrument. If the sense pair joins at the instrument, the measurement includes the lead resistance and the arrangement has gained nothing.
The same principle applies to a shunt resistor or a current sense element, and the general layout treatment is the same. The requirements for a low resistance measurement node are set out in our guide to mixed signal board design.
Self Heating and Sensor Construction
Self heating is the power dissipated in the element divided by its dissipation constant. A sheathed sensor in a moving liquid has a dissipation constant in the tens of milliwatts per degree, while the same element in still air is far worse, and the self heating error can differ by a factor of ten between the two installations.
The error appears as a reading that is high by a fixed amount once the sensor has settled, and it takes minutes to appear because of the thermal mass. A reading that drifts upward for the first ten minutes and then settles is showing self heating rather than a circuit fault.
The construction of the sensor also sets its response time and its immunity to vibration. A thin film element on a ceramic substrate responds quickly and is fragile, while a wire wound element in a ceramic former inside a stainless sheath responds slowly and survives rough handling.
Reference Resistor and Ratiometric Measurement
An accurate measurement usually compares the sensor with a reference resistor in series with it, driven by the same current. The ratio of the two voltages gives the resistance ratio directly, and the absolute value of the current cancels if the converter is ratiometric.
The reference resistor has to be stable with temperature, so a metal foil or a precision wire wound part is used. Its tolerance sets the scale error of the instrument, and it is the one component worth specifying to a few parts per million.
Placing the reference resistor close to the sensor terminals rather than near the electronics reduces the effect of the current source drift, because both voltages are measured across resistors in the same loop. The current only has to be constant over the duration of one measurement, not over months.
ADC Interface and Filtering
A delta sigma converter with a ratiometric reference and a programmable gain suits this measurement well. It can resolve the small voltage change and it rejects the mains frequency if its output rate is chosen to be a multiple of the mains period.
An external amplifier is usually unnecessary with a modern converter, and adding one introduces its own offset and drift. Where the sensor current is very small, a low noise instrumentation amplifier ahead of the converter is still worthwhile.
Filtering should be limited to what the application needs. A temperature measurement that reports once a second does not need a wide bandwidth, and a digital filter with a long average is easy to implement and effective against random noise.
Cable Length and Shielding
A long cable adds capacitance across the sensor and picks up interference from anything running beside it. A twisted pair for the current loop and a separate twisted pair for the sense connection keeps the loop area small, and a screen earthed at the instrument end only avoids a ground current in the screen.
Where the cable runs into a different building or past a large drive, an isolated measurement front end removes the ground potential difference entirely. The wider rules for keeping a low level analogue node free of coupled noise are covered in our guide to mixed signal board design.
Verification and Faults
Verify the channel in a stirred bath against a reference thermometer at three temperatures spread across the range. A constant offset suggests a reference resistor error, while a gain error across the range suggests an incorrect nominal resistance.
A reading that is high and rises slowly points to self heating. A reading that changes when the cable is moved points to a bad connection or to a two wire connection being used where the installation needs four wires.
An intermittent reading in an industrial installation is often a moisture problem in the terminal head rather than an electronics fault. Recording the raw resistance as well as the temperature makes that distinction, and the assembly and release points that prevent the connection problems in the first place are collected in our PCB design release checklist.
FAQ
What excitation current should I use for an RTD? One milliamp is the classic value for a sheathed platinum sensor. Reduce it for a small thin film element where self heating is a concern.
Is a three wire connection good enough? For a short matched cable in a stable ambient, yes. Use four wires for a long cable or where the ambient varies along its length.
Why does my RTD reading creep upward? Self heating is the most likely cause. Check the excitation current and the thermal path between the sensor and the medium.
