An Architecture for Real-Time Arbitrary and Variable Sampling Rate Conversion With Application to the Processing of Harmonic Signals

The paper presents a new solution for sampling rate conversion and processing of harmonic signals with known but possibly varying fundamental frequency. This problem is commonly found in particle accelerators, for tracking the beam signals whose revolution frequency varies during the acceleration ramp. It is also common among many other fields such as speech and music processing, removal of mechanical noises, filtering of biomedical recordings, active crack imaging, etc. The key element in the proposed solution is a new architecture for a Farrow-based resampler, in which the resampling ratio can take any value and can be modified continuously to follow the signal fundamental frequency. The combination of two complementary resamplers creates a processing region where signal synchronous processing is performed. The resampler architecture is optimized for modern FPGA features. It decouples the processing and sampling clocks, and uses a single processing (hardware) clock whose frequency remains fixed. The functional model was migrated to Xilinx System Generator and the overall performance is evaluated with an application that filters a periodic signal whose frequency follows a known linear ramp in the presence of additive white noise.