UCLA

The communication systems community has been working towards integrated Software-Defined Radios (SDRs) and Cognitive Radios (CRs) that can reduce cost and enhance connectivity. In light of the technology bottleneck at the analog-to-digital converter (ADC) and the in-applicability of off-chip filters, the integrated analog front-end is entrusted with the task of sharp, linear, and programmable signal selection required for SDRs and CRs. Traditional analog filtering techniques, however, incur a high penalty in power consumption, area, and linearity to provide the required sharpness and programmability. Similarly, recent efforts that use compressive sensing to acquire wideband spectra have also faced a bottleneck in the complexity and re-configurability of the analog measurement front-end.

Towards enabling SDRs and CRs, this dissertation proposes a new perspective on the design of anti-alias filters that defies the traditional trade-off between cost, linearity, and programmability. The technique, termed Filtering by Aliasing (FA), anticipates the aliasing operation at the sampler instead of avoiding it. The pre-sampling circuitry is modulated, using the high-speed switching techniques popular in state-of-the-art receivers, to provide significantly enhanced filtering responses at the sampling instances. The dissertation describes how the FA technique, by varying the resistor of a single-pole passive RC filter for example, provides programmable anti-alias filtering comparable to a 7th-order Butterworth filter.

On the compressive sensing front, this dissertation proposes a new approach to the acquisition of sparse spectra using Random Filtering by Aliasing (RFA). RFA acknowledges the existence of noise in realistic spectra and accordingly simplifies the analog measurement stage, moving most of the complexity to the low-cost, highly reconfigurable digital domain. RFA achieves significantly better resolution, lower cost, and better programmability than existing schemes.