There is a multitude of process parameters nowadays that need to be measured in the industrial environment (temperature, pressure, humidity, force etc.). Out of these, undoubtedly the most common one is temperature, as it influences most manufacturing parameters. It is no wonder then that many solutions have been developed over time to measure it. There are a few general categories any industrial temperature sensor will fall into: thermocouples, RTDs (Resistance Temperature Detectors), thermistors and integrated silicon sensors. There is no “best sensor” rather they all have pros and cons which need to be individually evaluated for each application. The RTDs are the most expensive, but they also provide best accuracy and best resolution for the measurement. This, however, only if appropriate analogue circuitry will be used (which of course will add cost to the already high price of the sensor itself). The appropriate analogue circuitry constitutes the subject of this article.
RTDs are regarded as the best quality temperature sensors (when it is worth paying for them). They provide accurate and stable measurements over time, and, most important, they provide a linear resistance-temperature characteristic. Below, you may see the resistance-temperature characteristic of the most common RTD, the PT100, which gives 100Ohms at a temperature of 0 Celsius degrees.
RTDs also represent a continuously expanding technology, better materials being researched and used, further improving the characteristics of the sensors.
The purpose of the analog circuitry buffering the sensor is to transform the resistance variation of the sensor in a variation of voltage which can easily be converted in digital values by an ADC. Although there are several methods to do this, the most common one is to build an analogue precision constant current source that will force a known and constant current through the RTD. The variation of the voltage will linearly depend on the variation of the sensor resistance, and thus on the temperature.
Great care should be taken care for the excitation current to be as low as possible. At the end of the day, the RTD is a normal passive device, which dissipates power as heat, so a higher current through the sensor will determine higher self-heating and thus will introduce errors in the measurement results. A good practice is to keep the excitation current below 1mA, but the drawback of such a small current is that it translates the temperature variation to a quite narrow voltage interval. If this is the case, a higher resolution ADC will be required in order to obtain a satisfactory resolution of the final result (of the measured temperature). For, instance, the resistance variation of PT100 between 0 and 100 Celsius is 38.5Ohm, and a 1mA constant current source would translate this in a 38.5mA voltage interval (between 100mV and 138.5mV) – hardly a wide interval for the plain 10-bit ADC usually provided on chip by the microcontrollers.
If using the proper devices tough, excellent accuracy and resolution can be obtain from an RTD. The table above only shows the resistance-temperature variation between 0 and 200 Celsius. The RTD is quite linear in that interval, but even if wider ranges are required, simple first order up to third order mathematic formulas may be used to estimate the temperature based on the measured resistance. Even if this introduces some strain on the software algorithms, it must be weighed if this is acceptable against a maximum +/-4.3 Celsius at the highest end of the measurement range (800 Celsius). The graph below indicates the measurement error (with appropriate analogue circuitry) against the measured temperature (note than in the most common measurement interval, between 0 and 100 Celsius, the measurement error is minimum):
For this kind of performance to be achieved, it is best for the RTD to be excited with a constant and stable current. There are several ways to build a constant current source, of which one is shown in the schematic below:
The way it works is rather simple: because the way an opamp generally functions, U1A will always have the same voltage at its two inputs. The voltage at its negative input is determined by the R1/R2 voltage divider to be at about 4.5V, thus, the same voltage will be found at its positive input. This will determine a 0.5V voltage drop on the R3 resistor, which generates about 1mA of current through it. The same current will be the collector current for Q1, irrespective of the value of the RTD sensor. Q1 is used to buffer the sensor itself with a high impedance of the current source (for an ideal voltage source, the output impedance is minimum, but for an ideal current source, the output impedance is maximum). Care should be taken in the way the components for this current source are selected. The opamp should be low offset and all the resistors should be 0.1% tolerance. Any deviation in the values of these components will adversely and considerably affect the generated current, thus the voltage drop on the RTD to be measured.
Another major error source that might influence the measurement is the wire resistance between the sensor itself and the measurement circuit. The situations you may be confronted with in industry might be various. The sensor might be located in the most various environments, from freezers to steam pools for wood treatment. Many of these environments are simply to rough for electronic devices to operate in, therefore the device containing the measurement circuit has to be located in a different position, usually in a building which is tens or even hundreds of meters away from the location of the sensor. Thus, the copper wires connection the sensor to the electronics become very long, and their resistance becomes significant against the small resistance values of the sensor itself.
In order to compensate not only the absolute resistance of the wire, but also the variation of this resistance across the temperature, there are methods to use 3-wire or 4-wire sensors. The 4-wire configuration means that each terminal of the sensor is connected to the electronic circuit with two parallel wires. The 3-wire configuration means that one terminal of the sensor is connected to the electronic circuit with two parallel wires, while the other terminal is connected with a single wire. Either of these configurations allows for the wire lengths to be compensated, taking the assumption that all the wires between the sensor and the circuit have the same length and the same resistance.
The example below depicts the 3-wire configuration. The RTD sensor is on the left, and the measurement circuit is on the right. The resistances modeling the three wires are Rw1=Rw2=Rw3. We will consider the voltage drop on these wires to be Vw1, Vw2 and Vw3. Due to the fact that the 1mA current from the constant current supply is flowing through both Rw1 and Rw3, it will give an equal voltage drop over these two wires: Vw1=Vw3.
The current flowing through Rw2 is really negligible, due to the fact that it is equal to the input current of the opamp, which is in the range of nanoampers, given an appropriate device is chosen.
Writing the equation for the output voltage of the amplifier, we will get:
But R6=R5 so we get:
Combining these with the equation above and taking into account that Vw1=Vw3 we obtain:
So the voltage at the output of the amplifier will be equal to the voltage drop on the sensor itself, no matter what the resistance of the wires is (as long as the wires are the same length and the same resistance).
In order to ensure a clear signal reaches the ADC, the employment of a low-pass filter is sometimes a good addition to the circuit, between the wire resistance compensation circuit and the ADC itself. The filter can range from the simple RC or LC low-pass filter to more complicated topologies like the Sallen-Key filter specially designed for high gain and which includes an additional operational amplifier (again, think of the trade-offs, since it is more expensive).
Thus, a generic block diagram of the entire analogue circuitry required to accurately measure an RTD can be the one below:
As mentioned before, due to the small variation of the voltage signal at the output of the analogue circuitry, a higher resolution ADC is usually required (at least 12 bit).
Choosing an ADC is a significant task in itself, and there are other tutorials and articles out there to help you on this (for instance: ADCS7476 12-Bit A/D Converter or 16 Top Kickass 16 bit ADConverter).
The microcontroller is really not that important for the application. The criteria based on which it should be selected should be driven mostly by any potential algorithms in which the digital value of the temperature would be involved in, or by other requirements of the application (driving a display etc).