A new thermopile-based setup

Thermopile temperature sensors are made of a thin photoabsorbent membrane, characterized by a small thermal mass and a low reflection coefficient, that is suspended on an array of serially-interconnected miniaturized thermocouples.

The membrane exchanges heat, as thermal radiation, with the surroundings until the achievement of the thermal equilibrium. The thermocouples have one junction in contact with the membrane and the other in contact with a heat sink at the base of the sensor. A variation of the membrane temperature, respect to the sensor case, causes a voltage signal (by Seebeck effect), which is amplified by the serial connection of the junctions. The output signal of the sensor is proportional to the incident heat-flux and consequently to the temperature of the emitting body. The calibration of the sensor is usually performed through a black-body emitting simulator. The response times of these sensors, typically of the order of milliseconds, depends mainly on the thermal mass of the membrane.

The membrane and the array of thermocouples are usually locked in a hermetically-sealed package, with a silicon window, which allows the transmission of the IR thermal radiation. The electronic noise, mainly of Johnson type, is generally low. This permits to reach a very good sensitivity, if an accurate calibration procedure is performed and the temperature of the sensor case is kept constant. However, the emissivity coefficient of the emitting bodies and the geometrical configuration of the measurement can introduce relevant errors in the temperature measurement. Indeed, emissivity coefficient values lower than 0.6 leads to a different thermal equilibrium state, due to the balance of the absorbed and emitted thermal radiation from the sensor, which can be approximated by a black body. Moreover, the geometrical configuration of the measurement has to be carefully considered to obtain a correct estimation of the temperature. Radiation emitted from the background or from other bodies can be absorbed by the membrane if the angle of view of the sensor is not completely covered by the studied material or if the sample is semitrasparent or very reflective [161]. Nevertheless, the small dimensions, the mechanical stability, the good sensitivity and the very low cost (of the order of tens of euro) make these temperature sensors very useful.

Our new experimental setup is based on a commercial thermopile (ZTP-135SR of General Electric Company), which is composed of a 0.7x0.7 mm² photo-absorbing membrane placed on the top of an array of 60 junctions. A silicon window (diameter 2.4 mm) allows the transmission of the thermal radiation with a wavelength between 6 µm and 16 µm and defines a view angle of 90°. The time constant of the sensor has been experimentally estimated by placing an optical

chopper between the sensor window and a copper plate kept at a constant temperature, which acts as a source of constant IR radiation (Figure 5.1.a). In this way, the sensor receives an IR square-wave. The time constant is considered as the time needed for the sensor signal, at every half-period, to reach 63% of its maximum value (Figure 4.1.b). The obtained time constant is 27 ms. It has to be noticed that the minimum time required for achieving the maximum signal value is about 85 ms and hence the operative frequency limit, in case of simulation of thermomagnetic cycles, should be about 6 Hz.

Figure 5. 1 (a) Sketch of the experimental setup used to measure the response time of the

thermopile. An optical chopper is inserted between the thermopile and a copper plate (on the left), kept at a constant temperature. (b) Response curve of the thermopile exposed to a square wave, obtained chopping the IR radiation emitted from the copper plate.

The experimental setup, designed for adiabatic temperature change measurements, is schematically showed in Figure 5.2. The structural part of the instrument is composed of a long plate of copper, water-cooled, which acts as a multipurpose optical bench, on which one can align the temperature sensor and the sample holder. The plate is inserted in a cylindrical vacuum chamber (diameter: 22 mm, maximum vacuum: 10^-5 mbar), which can be placed in between the poles of a low-inductive electromagnet that is able to generate a maximum magnetic field of 2.4 T. The sensor and the sample are placed in a copper box (8x9x7 mm³ – see Figures 5.2.b and 5.2.c), closed on the top with a brass plate thermally connected to the box by a thermoconductive paste (Arctic Silver Ceramique, 𝑘 = 7 Wm^-1K^-1). The sensor is inserted in a face of the box and thermally connected with it in order to keep both at the same temperature. The sample is suspended at the centre of the box with a frame of nylon wires. This solution allows to reduce the thermal contact between the sample and the environment and, so, to minimize the heat lost because of conductive dissipation.

The sample is placed 1 mm away from the sensor so that a minimum surface of

4.5x4.5 mm² is enough to fully cover the view angle of the sensor. This helps to reduce the error due to background radiation. The longer axis of the sample is positioned parallel to the magnetic field to minimize the effects of shape anisotropy.

Figure 5. 2 Sketch of the experimental setup. (a) Vacuum chamber with the optical copper bench. (b) Copper box with the sensor (on the right side), the sample and the copper finger (on the left side). (c) Top view of the sketch b). (d) Diagram of the electronic control system.

A thermoelectric module, located under the box and in contact with the water-cooled plate, and a Pt100 sensor allow to control and stabilize the box temperature (𝑇𝑏) between 260 K and 350 K (see Figure 5.2). The temperature stabilization is made through an ITC503 Temperature Controller (Oxford Instruments). A thermoconductive paste ensures a good thermal contact between the box and the module.

The magnetic field change is obtained by switching on and off the electromagnet, which takes about 1 s to reach its maximum field (𝐻𝑚𝑎𝑥). The time constant of the exponential rise (63% of 𝐻_𝑚𝑎𝑥) is lower than 0.3 s. The field change is controlled through a DAQ board (National Instruments BNC-2120) and synchronized with the signal acquisition by a dedicated control program developed in Matlab.

The output signal of the sensor is amplified and filtered with a notch filter (𝑓₀= 50 Hz) and a low-band filter (𝑓_𝑡= 200 Hz), then acquired through the DAQ at a sampling frequency of 10 kHz (Figure 5.2.c). It was verified that the magnetic field does not perturb the thermopile response. A short spike of electric noise appears, however, during the field change. This noise, independent of the sensor signal, is subtracted from the measurements.

The calibration of the sensor is performed for each sample, placing in contact with the back side of the sample, by a thermoconductive paste, a thermally controlled copper finger (thermal bath) with a Pt100 sensor and a resistive heater (on the left of the copper box in the sketch of Figure 5.2.b). An ITC503 temperature controller (Oxford Instruments) is used to stabilize the temperature of the finger, independently of 𝑇𝑏 (which is kept constant), within 0.01 K. The calibrations are obtained by sweeping the temperature (on heating and on cooling) with a rate lower than 0.5 Kmin^-1 to minimize the thermal gradient between copper finger and sample. The short temperature range (about 5 K above 𝑇_𝑏) of every calibration makes possible to fit the obtained 𝑉(𝑇) curve with a linear function 𝑇(𝑉) = 𝑎𝑉 + 𝑏. Small variations of the coefficients have been observed on repeating the calibration. This contributes with an error of 0.1 K to the measured temperature. To reduce the error due to background IR radiation we repeated the calibration at every 𝑇_𝑏. During the measurement of the adiabatic temperature variation, the copper finger is removed from the box and the sample remains thermally insulated from the environment.

This calibration procedure allows to reduce the errors in the temperature measurements due to different surface emissivity, which is peculiar of every measured specimen and depends on the emissivity coefficient, the surface area and state and the emitting angle. An alternative solution to overcome this drawback could be to cover the surface with a thin layer of black paint in order to maximize and standardize the emissivity for every sample to be measured.

This allows using the same calibration when measuring different materials [139]. In this way however, we measure the temperature change of the black paint and not the direct effect of the sample surface [180,181].