Convolutional denoising for large-volume Seebeck calorimeter

Abstract

Ambient temperature fluctuations can affect the thermal noise and sensitivity of a large-volume Seebeck calorimeter. This paper proposes that a reference cell considerably smaller than the sample cell in the apparatus can effectively neutralize this noise. It is found that the thermal signals from both the reference and sample cells exhibit a convolution relationship. By deconvolving two distinct thermal pulses generated by abrupt changes in cooling fluid temperature, one from the reference cell and one from the sample cell, a corresponding response function is derived. And using this function, the noise is canceled out by subtracting the convoluted signal of the reference cell from that of the sample cell during isothermal calorimetry. Experimentally, utilizing this technique with two calorimeters, one with a 17.6-liter capacity and the other with a 27-liter capacity, has been shown to reduce noise by at least 5 % and 6 % from their initial values, respectively.