Event-Based Data Acquisition And Digital Signal Processing In Continuous Time

We review techniques in which uniform sampling is abandoned, with samples generated and processed instead only when there is an input event, thus causing dynamic energy use only when demanded by the information in the signal. Methods to implement event-based A/D converters and digital signal processors (DSPs) in this context are reviewed. It is shown that, compared with traditional techniques, the techniques reviewed here lead to circuits that completely avoid aliasing, respond immediately to input changes, result in lower in-band quantization error, and exhibit dynamic power dissipation that decreases when the input activity decreases. Experimental results from recent test chips, operating at kilohertz (kHz) to gigahertz (GHz) signal frequencies, are reviewed.