Synthesis and study of evolutionary optimised sensor linearisation with translinear & FPGA circuits.

Linear response characteristics of sensors are affected by ageing and changes in material properties. However, existing methods for improving these characteristics are disadvantaged by recurrent full-scale errors and insufficient speed. Accordingly, this study aims to reduce temperature sensor non-linearity through implementation of evolutionary optimised non-linear functions via a translinear-based application-specific integrated circuit. Moreover, the translinear-based analogue computation method is compared to a proposed field programmable gate array (FPGA)-based digital computation methodology to develop an alternative method for linearisation at sensor nodes (edge computing). The optimal values of the linearisation parameters for selected thermocouples are determined using the covariance matrix adaptation evolutionary strategy. Specifically, these optimal values are input in order to realise the non-linear functions. Experimental results reveal that FPGA offers a lower full-scale error than that of the translinear-based computation, although the translinear method can accomplish the required linearisation with fewer components (e.g. Data Converters), less delay and less power. Further, analogue computation through the proposed single chip-based method can utilise the latest processes, e.g. gallium nitride or pseudomorphic high electron mobility, for an enhanced error reduction. The comparative result shows that the translinear method with recommended processes can correct the non-linearity of sensors in all aspects. [ABSTRACT FROM AUTHOR]