Passive switched-capacitor based filter design, optimization, and calibration for sensing applications

Abstract

This thesis discusses the limitations of wideband software defined radios (SDR). Spectrum sensing is identified as an important aspect required of SDR based cognitive radios. Several architectures and implementations are reviewed. An RF sampler followed by analog signal processing is identified as a critical block enabling low-power wideband digitization. Passive switched-capacitor charge-domain processing is introduced. Its implementation simplicity and lack of active power consumption are enticing. The effect of kT/C noise is analyzed in detail. Linear and nonlinear computation errors are modeled, and circuit techniques for their reduction are developed. Simulations are used to optimize power and computational dynamic range. For spectrum sensing, the design of CRAFT (Charge Re-use Analog Fourier Transform): an RF front-end channelizer for wideband software defined radios based on a 16-point analog-domain FFT is described. The design relies on charge re-use to process a 5GS/s input with an average output SNDR of 47dB, and consumes only 12.2pJ/FFT conversion (3.8mW). These numbers represent orders of magnitude improvement on the work reported previously in literature. The thesis also discusses the system-level modeling and mitigation of circuit non-idealities in CRAFT. These design principles enable this implementation to achieve a large dynamic range even at high speeds. Additionally, these techniques can be easily extended to improve the performance of other passive switched-capacitor designs.