Errors Abound When Switching Temperature Sensors
This last part in a series on signal switching explains how to achieve the best measurement performance when switching thermocouple, RTD, or thermistor inputs.
Maurizio Basso, National Instruments, Italy; Michel Haddad, National Instruments, USA -- Test & Measurement World, 10/1/2000
 |
| PXI-based systems often need to switch multiple inputs from a variety of thermistors. |
Data acquisition systems often have to accept temperature signal inputs from either thermocouples, resistance temperature detectors (RTDs), or thermistors. For switching these inputs you have the choice of using mechanical or solid-state switches. All combinations of sensor and switch introduce certain levels of measurement error so you need to consider what sensor and switch arrangement best suits your application. If you follow some simple guidelines, given later in this article, you’ll be able to optimise the switching arrangement and minimise the effects of those errors.
Mechanical switches, or electromechanical relays, use an energising coil to open or close metal contacts. Solid-state switches are either solid-state relays with isolated contacts, or FET switches with non-isolated contacts. Solid-state relays use optical isolation to control a MOSFET switch capable of switching high voltages and currents. In FET switches, digital control directly connects to the gate of the MOSFET, which limits switching voltages to approximately 10 V.
 |
| Figure 1. Rm and Vm are the main sources of error when switching sensors with mechanical relays or solid-state switches. |
In order to decide which switch type performs best in each situation, you need to understand how each switch type departs from the ideal of a perfect open or closed circuit. Figure 1 shows the complete equivalent circuit for an electromechanical relay. For solid-state relays or FET switches, an LED, or a digital control circuit replaces the coil.
You can generally assume parameters such as resistance and capacitance within switches are negligible for many data acquisition inputs, but with thermocouple or RTD inputs the real value of these parameters can add measurement errors.
Resistances, Rp (about 1000 MW), and capacitances, Cp (a few pF), indicate the value of the parasitic components between the signal terminals (IN and OUT) and the switch control terminals (for example, relay coil). These values explain the current leakage and the capacitive coupling between the signal and the control terminals. Parasitic resistances and capacitances between the signal terminals IN and OUT and the switch control terminals explain current leakage and noise coupling into the signal path.
When a switch closes, Rm is the resistance in series with the signal path. Rm can vary significantly between mechanical relays, where Rm is a few milliohms, and solid state switches where Rm can be several hundreds ohms.
In the case of an electromechanical relay, Vm represents the thermal EMF voltage that the switch contact generates and which appears in series with a signal. Thermal EMFs result when two different metals make contact and coexist in a temperature gradient. Even if contacts are of the same material, they can still generate a small EMF because their composition always differs slightly. In addition, temperature gradients can also exist between the two sides of a contact inside the same switch enclosure.
For solid-state relays and FET switches, Vm is mostly due to voltage offset in the semiconductor material that makes up the switch. Vm can range from a few microvolts for mechanical relays to several millivolts for solid-state relays.
Alternatively, when a switch is open, although the IN and OUT terminals should be perfectly isolated, Rk (less than 1000 MW) and Ck (a few pF) still connects the input and the output terminals.
If you think that these parasitic elements remain constant and you can therefore compensate for them when calibrating your system, bear in mind that some of these values change with time and with changing physical environmental parameters, such as humidity and temperature. For example, the value of Rm will vary with relay operation because of contact oxidation.
Table 1 summarises the main differences between mechanical relays and their solid-state counterparts that will affect temperature sensor switching.
| Table 1. Comparison of the main features of mechanical and solid-state switches that affect sensor switching. | |||
| Mechanical | FET Switches | Solid State | |
| Rm | < 60 mW | > 20 W | kWs |
| Vm | microvolts | millivolts | millivolts |
| Switching Speed | up to 300 cycles/s | 100 ns to 10 µs | <1 ms |
| Life | 5 to 1000 million operations | infinite | infinite |
Thermocouple Is Common Sensor
Sixty percent of temperature measurement applications use thermocouple sensors, which generate a voltage proportional to measurement temperature. The voltage varies from –80 to +80 mV and changes by tens of µV/°C. With this kind of sensor, you must take into account the parasitic voltage Vm because its value can be significant when compared to the signal. Therefore, when selecting your front-end switching system, pay special attention to the offset voltage specifications.
Practical Tips for Switching Thermocouples
• Use two-wire differential switching. The thermal EMFs of the two switch contacts tend to cancel out, and ground-loop effects are minimised.
• Minimise the number of switches in series with the signal path, and so reduce the overall effect of their errors.
• Minimise thermal differentials between the switch terminals. For example, make sure there is not a source of heat — such as a power resistor — close to one of the switch terminals.
• Use low thermal EMF and low-offset relays.
• Measure a zero reference. To do this properly, you need to determine the offset, thermal or semiconductor induced, of every channel in your system. You can consider this “short-test” to be part of your system calibration. As previously mentioned, these offsets vary with ambient temperature and humidity, so you may have to perform this calibration before every measurement or operate in a stable environment.
• Use latching relays. The main source of heat that causes thermal EMFs in an electromechanical relay is the coil. Latching relay coils only energise for a very short period and virtually eliminate the heat source. You’ll negate this benefit if you operate the coil too frequently.
• When using one-wire switching, you can reduce thermal EMFs by arranging two one-wire contacts in series, back-to-back, in the same relay package. The thermal EMFs generated across each set of contacts tend to cancel each other.
Switching RTDs or Thermistors
RTDs and thermistors are temperature sensors whose resistance changes with temperature. A common type of RTD, the Pt100, has a nominal value of 100 W at 0°C with a resistance sensitivity around 0.4 W/°C. For thermistors, popular nominal values at ambient temperature are 2252 W, 5000 W, and, 10,000 W. The 5000 W thermistor has about 200 W/°C sensitivity at room temperature.
The parasitic resistance, Rm, is most important when switching resistance sensors. This error will be particularly significant if the resistance measured is small and your system uses solid-state relays. In fact, the closed channel resistance for some solid-state relays can easily exceed 1 kW, which could seriously affect your measurement results if you are using two-wire resistance measurements.
Practical Tips for Switching RTDs
Because an RTD resistance is fairly small, especially when compared with the Rm resistance of a solid-state relay, pay particular attention when switching this type of sensor.
• Use a four-wire resistance measurement technique, which will eliminate the effects of lead resistances and Rm in series with the signal path. (Refer to the NI Developer Zone at www.ni.com/zone for more details of four-wire and two-wire resistance measurements.)
• If you cannot use a four-wire measurement, then use mechanical switches with their very low on-resistance. Note that your cabling will also introduce lead resistance errors that add to your basic electromechanical ON resistance.
• If you cannot use a four-wire measurement, but must use solid-state relays (for example, for speed), then you’ll have to calibrate the switching system channel-by-channel to take into account the on-resistance of the switches. In this case, short circuit the input at the RTD terminals and measure the resistance offset of the system. Many system instruments or test programs, such as LabVIEW or Measurement Studio, provide you with the ability to implement a mathematical null function to easily subtract this value from the RTD measured value for a correct measurement. Switch resistance may vary during warm-up, or with environmental changes, so for the best measurement accuracy use null before every reading.
Because thermistor sensors have nominal values of several kWs changing at 200 W/°C, switching errors are less troublesome. However, there are still a couple of practical tips.
Practical Tips for Switching Thermistors
• Use a four-wire measurement with solid-state relays. Although thermistor sensor’s resistances are higher than RTD values, a solid-state relay’s resistance will still produce significant errors.
• Use a two-wire measurement with mechanical switches because the adverse effect of the cabling and electromechanical switches on the measurements will be very small. T&ME
Maurizio Basso is the Computer-Based Instruments Product Marketing Manager with National Instruments. Michel Haddad is R&D Section Group Manager with National Instruments.
