| Frequency analysis and octave bands |
| For more or less all technical evaluations, weighting, rating etc. of sound and noise, not only the total amount of energy in a noise is important but also the frequency distribution of the sound energy. To analyze the sound in the relevant frequency range (e.g. from 20 Hz to 20000 Hz) this range must be divided in appropriate steps. Two principle methods can be used to define these frequency steps: a) a constant (absolute) bandwidth (e.g. 10 Hz, this means ranges from 10Hz ... 20 Hz; 20 Hz ...30 Hz; 30 Hz ...40 Hz ;...;19990 Hz ... 20000 Hz) or b) with constant relative bandwidth. The latter means, that the ratio between lowest and highest frequency in each interval is constant, e.g. 45 Hz...90 Hz (ratio=2); 90 Hz...180 Hz; 180 Hz ... 360 Hz; ... If the ratio as in the example is 2, we call these frequency bands for octave bands. Dividing the frequency range in bands with constant relative bandwidth corresponds best with the human perception of different frequencies also known from music. For analyzing noise, octave bands and 1/3 octave bands (ratio 3 2=1.26) are most common, but also 1/12 octave bands and 1/24 octave bands can be used. If a more detailed frequency information is required, a frequency analysis with constant absolute bandwidth (e.g. 1 Hz or 2 Hz or 10 Hz or xx Hz) can be used. The most frequently used mathematical method mostly used is called FFT (Fast Fourier Transformation).  |
