Alternative Approaches to Fourier-Based Frequency Tracking

In the analysis of automotive vibration signals, it is often desirable to precisely track individual tonal components of a signal over time. This kind of analysis is useful in defining root cause of cyclic vibrations, allowing the engineer to localize a vibration problem to a specific subsystem. The most commonplace analysis methods are block-oriented Fourier Transform (FT) approaches. However, these methods invariably lead to time localization uncertainty and imprecise frequency resolution. Other analysis methods are available that can estimate the precise frequency of content of a signal at every time step, given certain signal assumptions. This paper studies several of these analysis methods. Narrow Band filters, Kalman filtering, Auto-Regressive signal modeling and the Phase Locked Loop are examined as methods of signal decomposition and order extraction.