We can measure vibration of a machine and generate a time waveform (time domain). Fast Fourier Transform (FFT) can then be performed to produce a spectrum (frequency domain). One thing to consider is FFT is performed on a block of samples, which are called the time record. One assumption made in the FFT calculation is that the time record is continuous. This means the signal before and after the time record are identical.
Let us analyze the time waveforms shown in the Figures. When we perform the FFT on a block of data, the FFT calculation “assumes” that the data continuous before and after the block of data. If we are analyzing a pure sine wave, and there is an integer number of cycles in the time record, this assumption is correct. However, it is seldom true that the time record starts and ends at zero. If the FFT is performed and the signal is discontinuous (start and end of time record do not have the same value), we may see boardening of peaks within the spectrum. This phenomenon is called leakage.
To solve this issue, the shape of the time record can be changed so that the values at the start or end of the record are the same. This is known as windowing data. One thing to note is windowing does not change the frequency of content, but it affects the shape of the spectral peaks and the amplitude levels. It is clear that when windowing is performed on time records, there is no longer any sudden change in amplitude at the start and end of the record, thus there is no leakage.
There are a number of window functions, each with a different shape and each having a different affect on the resultant spectrum.
– Hanning window is most commonly used in vibration analysis of rotating machinery. The start and end of the time record are forced to zero amplitude. It results in good frequency accuracy, but It affects the amplitude.
– Flat Top window has better amplitude accuracy, but the frequency accuracy is poor.
– Hamming window is similar to Hanning window except the start and end do not go to zero amplitude.
A block of time record that goes to zero at the start, but does not go to zero at the end. This phenomenon is called leakage.