Java1478-fun with java, how and why spectral analysis works  (Page 5/9)

Example of time domain and frequency domain

If you were to draw a graph of the voltage impinging on the speaker coils on your stereo system over time, that would be a time series, which is a member ofthe time domain.

If you were to observe the lights dancing up and down on the front of your equalizer while the music is playing, you would be observing the sameinformation presented in the frequency domain. Typically the lights on the left represent low frequencies or bass while the lights on the right side representhigh frequencies or treble. Often there is a slider associated with each vertical group of lights that allows you to apply filters to emphasize certainparts of the frequency spectrum and to de-emphasize other parts of the frequency spectrum.

Forward and inverse transforms

There are two very similar forms of the Fourier transform. The forward transform is typically used to transform information from the time domain intothe frequency domain. The inverse transform is typically used to transform information from the frequency domain back into the time domain.

Sampled time series

The theoretical Fourier transform is defined using integral calculus as applied to continuous functions. As a practical matter, in the digital world, wealmost never deal with continuous functions. Rather, we deal with functions that have been reduced to a series of discrete numbers (or samples), which are theresult of some discrete measurement system.

(As mentioned earlier, recording the temperature in your office once each minute for twenty-four hours would produce such a discrete series ofnumbers.)

Integration and summation

In many cases, the integration operation encountered in integral calculus can be approximated in the digital world by a summation operation using discrete data. That is the case with the Fourier transform. Thus, the (simple) summation form of the Fourier transform that is applied to a discrete time series is known as the Discrete Fourier Transform , or DFT .

The FFT algorithms

The DFT is a computationally intense operation. Given certain restrictions involving the number of values in the time series and the number of frequenciesat which the spectral analysis will be performed, there is are special algorithm that can result in computational economy in performing the transform.The algorithms that are used to realize that economy are commonly referred to as Fast Fourier Transform or FFT algorithms.

DFT versus FFT

The DFT is more general than the FFT, but the FFT is much faster than the DFT. It is important to understand that these are simply two differentalgorithms for doing the same thing. Either can be used to produce the same results (but as mentioned earlier, the FFT is somewhat more restricted as to the number of time-domain and frequency-domain samples) .

Because the DFT algorithm is somewhat easier to understand than the FFT algorithm, and also more general, I will concentrate on the DFT algorithm to explain how and why theFourier transform accomplishes what it accomplishes.