A Hybrid EOG System Featuring Flexible a-IGZO TFTs and a 180 nm CMOS IC

Abstract

Flexible wearable sensors are becoming increasingly popular due to their capability to monitor physiological signals while ensuring high user comfort. Wearable sensors built using only flexible electronics result in inferior system performance due to the inherent limitations of flexible technologies. Flexible hybrid electronics (FHEs) combine the strengths of flexible technologies (low cost, larger area coverage, and user comfort) with the high performance and energy efficiency of silicon-integrated circuits (Si ICs). Here, an FHE system is described that addresses the challenge of system functions partitioning between flexible and Si subsystems. The presented design approach is applied in this work to a multichannel electro-oculography (EOG) system, but can be expanded to various FHE application domains. The proof of concept fully integrated eight-channel FHE EOG system exploits a front-end in a-IGZO TFT technology, interfaced with a custom Si IC that supports the front-end and provides conversion of the acquired signals to the digital domain. In this way, the low cost, flexible form factor, and the large area character of the flexible thin-film transistor (TFT) technology are leveraged, while exploiting at the same time the accuracy, computational capability, and energy efficiency of the Si IC. We propose a hybrid high-pass filter (hHPF) as an interface between the electrodes and the front-end, whereby TFTs and capacitors on foil are controlled by the Si IC. The hybrid approach exploits the capabilities of flexible electronics to bring functionality close to the electrodes and its potential to cope with a high number of sensor channels, with the better precision of CMOS, enabling a fully integrated filter without discrete components. The TFT front-end converts the input signal voltages to a pulsewidth-modulated (PWM) representation, which is then quantized in the Si IC, to provide a digital output. The Si IC enables conversion to the digital domain with much higher efficiency compared to state-of-the-art TFT solutions. The digital output is further used in an algorithm for feature extraction and classification to detect the desired signals. The experimental electrical characterization of the system demonstrates an input-referred noise of $6.68~\mu $ $\mathrm{V}_{rms}$ in a 1–30 Hz band, with a chopping frequency of 1 kHz. The input impedance offered by the integrated hHPF is 6.88 G $\Omega $ at 0.089 Hz (with multiplexing disabled). The total power consumption per channel is $84~\mu $ W. In vivo experiments have also been performed against a gold standard (a commercial bio-potential amplifier) to confirm a faithful representation of the eye activities. The signals acquired during in vivo experiments are successfully classified using a support vector machine (SVM) classifier, with an average accuracy of 87.6% $\pm ~2.9$ %.