Spectral Analysis and Correlation

Spectral analysis is the process of breaking down a signal into its components at various frequencies, and in the context of acoustics there are two very different ways of doing this, depending on whether the result is desired on a linear frequency scale with constant resolution (in Hz) or on a logarithmic frequency scale with constant percentage resolution. The fundamental connection between the time domain and the frequency domain, the Fourier transform, is most easily interpreted in terms of linear time and frequency scales, at least in the practical version now used to calculate it, the FFT (fast Fourier transform). However, expressing a spectrum on a linear scale automatically restricts its frequency range, since the upper frequency decade occupies 90% of the scale and the upper two decades 99% of the scale.