Optimized Circuitry For Sensor Interfaces: In Cmos And In Brains

There is a need to optimize processing circuitry for sensor interfaces in both modern electronics and biology. Several analogs typically exist between sensing systems in electronics and biology: they have transducers to convert one type of energy or signal into one that can be processed by circuitry composed of transistors or neurons, they utilize an amplification stage to boost the sensed signal and suppress unwanted background signals, and then they have some means to store the signal or translate it into useful information. Optimization usually refers to power consumption, or, more specifically, optimizing the energy cost for processing a given amount of information. Circuit optimization requires selecting an appropriate architecture, bandwidth or speed, and output precision for a given task. In this dissertation I present several examples of circuitry in modern electronics and biology optimized for sensor interfaces. Chapter 1 serves as an introduction to the dissertation. I discuss the similarities of electrical and biological sensor systems, how circuitry works to reduce unnecessary bandwidth and dynamic range of sensed signals for low power processing, and give an essential background to energy efficient CMOS amplifier design. In Chapter 2 I present an orthogonal current reuse amplifier; a topology that circumvents the fundamental noise-power tradeoff in amplifiers by reusing bias current across independent amplifiers. This technique effectively increases am- plifier gm /ID linearly with every additional amplifier at a small cost in headroom voltage. Chapter 3 discusses the dynamics of gamma band oscillations recorded from olfactory bulb slices recorded with microelectrode arrays. Persistent gamma oscillations are induced in slice using methods previously reported for hippocampal slices and are shown to have multiple regions of coherent oscillatory activity across slice. Chapter 4 presents microelectrode arrays developed in CMOS that scale to large electrode counts (+1000), have high spatiotemporal resolution (20kHz at 50[MICRO SIGN]m pitch), and have integrated photosensors for correlating recorded electrical activity with optical stimuli. In Chapter 5 I present a high-speed imager (> 1kfps) for calibrating MEMS intertial sensors in real-time. The imager utilizes polar symmetry to directly extract angular rotation information far more efficiently than standard, cartesianbased imagers