Data for: Temperature and Temporal Drift Compensation for Al2O3-gate ISFET-based pH Sensor using Machine Learning Techniques

Abstract This work presents modeling of temperature and temporal drift characteristics of Al2O3-gate Ion-Sensitive Field-Effect Transistor (ISFET) and performance enhancement of ISFET-based pH sensor by using machine learning (ML) techniques. The behavioral macromodel of ISFET is built using Simulation Program with Integrated Circuit Emphasis (SPICE), which incorporates the temperature and temporal dependent behavior of electrochemical and device parameters. The SPICE macromodel for ISFET is exported as a subcircuit block for designing the constant-voltage (CV) constant-current (CC) readout circuit. The training, validation and testing data for machine learning techniques was generated using SPICE simulations for experimental analysis. We compensate the temperature drift of the sensor by applying well-known ML techniques to its read out integrated circuit (ROIC) output. We learned the salient characteristics from the simulated data with neural networks, linear regression, support vector machines, decision trees and random forests, respectively, in order to achieve temperature drift compensated ROIC outputs for unseen data. Our results show that random forest technique achieves lowest error rates for Al2O3-gate ISFET-based pH sensor. The temperature compensated ISFET ROIC output shows excellent performance with ΔpH less than 0.025 for pH range 2–12 over a temperature range of 15 °C to 45 °C. We adopt Bayesian inference technique to capture and compensate the temporal drift in the device. The analysis shows credibility of estimated values and robustness of the Bayesian approach to temporal compensation for temperature ranging from 15 °C to 45 °C. Overall, this work provides a framework for designing robust and intelligent ISFET-based pH sensors by modeling the drift characteristics and compensating it using ML techniques.