Abstract
Wearable devices with multi-modal sensing are needed to provide a well-rounded report on the user’s health and environment. Studies have shown that air pollutants like ozone, VOCs etc., when present in unhealthy concentrations, cause both immediate and long-term respiratory and cardiovascular effects, and exacerbate underlying comorbidities. Therefore, a single wearable device for monitoring personal exposure to environmental parameters of concern, and co-related physiological signs is necessary. Also, in order to enable uninterrupted sensing, self-powered operation achieved by energy harvesting technologies is a viable solution. However, the readily available power density from these ambient sources is severely limited (~20µW/cm2), and does not match the high-power consumption of existing multi-modal analog front-ends. Therefore, it is critical to realize low-power AFEs, while maintaining good sensing performance metrics. Further, in addition to the above-mentioned wearable personal device, IoT sensor nodes supporting voltage, current, resistive, and capacitive signals are also necessary to realize a wide range of applications such as safe hospital room conditions, smart environment, smart homes etc. However, majority of existing multi-modal front-ends either support a sub-set of these signals or deliver application-specific performance metrics. This requires a design effort to account for every small change in the sensing requirements. Therefore, IoT sensor nodes can benefit from a universal V/I/R/C analog front-end with added performance metrics’ reconfigurability. This dissertation addresses the design challenges with respect to wearable devices for joint physiological and environmental monitoring in two research themes. In the first theme, the AFE power consumption for ambient gas and respiration rate sensing is optimized while ensuring good performance metrics. While the first theme optimizes AFE’s power consumption, now the active sensors and their peripheral devices dominate the system power. The second theme in this dissertation mitigates this by introducing design techniques to reduce PPG and gas modalities' LED power. Also, a context-aware event-driven method is presented to improve the quality of information and system availability by monitoring only the relevant signs, and reducing unnecessary data generation. Lastly, this dissertation explores a universal analog front-end design to support low-power voltage, current, resistive, and capacitive sensing with on-demand reconfigurability to adjust the AFE’s performance according to application needs.
