Signal acquisition challenges in mobile systems

In recent decades, the advent of mobile computing has changed human lives by providing information that was not available in the past. The mobile computing platform opens a new door to the connected world in which various forms of hand-held and wearable systems are ubiquitous. A single mobile device plays multiple roles and shapes human lives towards a better future. In these systems, sensor-based data acquisition plays an essential role in generating and providing useful information. The increased number of sensors is embedded in a single device in order to process various signal modalities. In practice, more than 30 data converters are required in designing a mobile system in which the data-converting blocks become among the most power-hungry components in battery-operated systems. Due to the increased variety of sensors, mobile systems are meant to face several obstacles. For example, the increased number of sensors increase system power consumption during the system operation. The increased power consumption directly affects operation time because mobile systems are powered by a limited energy source. Moreover, an increased amount of information also gives rise to bandwidth problems in communication due to the increased volume of data transmission. Also, this system design requires a larger area in a silicon die so that multiple signal paths can be placed without cross-channel interference. Therefore, the system design has presented a challenge in terms of trying to resolve the design constraints such as power consumption, bandwidth usage, storage space, and design complexity issues.

To overcome these obstacles, in this dissertation, efficient data acquisition and processing methods are investigated. Specifically, this thesis considers the problems of energy-efficient sampling and binary event detection.

This dissertation begins by presenting a new signal sampling scheme that enables higher precision signal conversion in compressed-sensing-based signal acquisition. The proposed scheme is based on the popular successive approximation register and employs a modified compressive sensing technique to increase the resolution of successive-approximation-register (SAR) analog-to-digital converter (ADC) architecture. Circuit-level architecture is discussed to implement the proposed scheme using the SAR ADC architecture. A non-uniform quantization scheme is proposed and it improves data quality after data acquisition. The proposed scheme is expected to be used for medium- or high- frequency data conversion.

Secondly, the possibility of using fewer ADCs than channels is studied by leveraging sparse-signal representation and blind-source-separation (BSS) techniques. In particular, this dissertation examines the problem of using a single ADC or quantizer system for digitizing multi-channel inputs. Mixing and de-mixing strategies are extensively studied for sampling frequency-sparse signals and the proposed multi-channel architecture can be easily implemented using today's analog/mixed-signal circuits.

The third part of this dissertation investigates a binary hypothesis testing problem. In mobile devices such as smartphones and tablet PCs, a major portion of energy is consumed in user interfaces (LCD display and touch input processing). For accurate detection and better user interface, energy-efficient sensing and detection schemes are necessary to manage multiple sensor inputs. A highly efficient detection scheme is presented that can detect binary events reliably with a fraction of the energy consumption required in the conventional energy detection.