Basics of Time-Based Signal Processing

Over time CMOS technologies are optimized with shrinking feature sizes to enable the implementation of high-performance integrated circuits (ICs) with packed densities and numerous configurable features. Recent trends insist on fully integrating analog, mixed signal, and digital design on the same IC. Digital circuit designers have leveraged the benefits of technology trends with enhanced switching speed and increased transistor density. However, despite supply scaling in sub-100 nm CMOS technologies, there has not been any significant improvement in total power consumption, and the power density has almost remained unaffected. On the other hand, shrinking voltage headroom, increasing noise, and drop in intrinsic gain and dynamic range with low-feature sizes have not been very helpful for analog and mixed-signal designers. They mostly opt for circuits with large areas and high power dissipations to meet performance requirements. There has been an increased interest in finding alternatives to traditional voltage-based signal processing as it is mainly dependent on power-hungry area inefficient operational-transconductance amplifiers OTAs. A viable emerging alternative is time-based signal processing which has the potential to achieve analog-like circuit functionalities with highly digital-like circuits in scaled technologies [88]. The basic idea has been to modulate analog voltage- or current-based information with square-wave signals such that signal amplitudes are represented by transition edges. Although this technique existed earlier, there has been increased interest and effort to explore various time-based circuit topologies in the last decade. As illustrated in Fig. 3.1, time-based architectures exploited reduced feature sizes, unlike voltage-based circuits with faster transition times, which eventually translates to higher time resolution or, in other words, increased quantization capability. Typically, the time-based signal processing as in practice is accomplished by either time-based arithmetic units [89–91] using pulse width and time difference as signals or ring oscillators, using frequency outputs as signals. This chapter discusses the fundamentals of time-based signal processing. It briefly explains the working principles of various building blocks, focusing on circuits using time-based arithmetic units.