A Plea for Linear Units as an Alternative to Decibels and Octaves 2005-01-2549

Decibels were originally developed in the 1920s by the telephone industry (AT&T and Bell Labs). Initially the unit was the bel, derived from the name of Bell Labs and defined as the logarithm to the base 10 of the transmission loss of electrical power in telephone lines. It was also used for voice signals in telephones where the preferred unit became a tenth of a bel or decibel. The adoption of decibel for sound appears to be due principally to the dominant position of the Bell's acoustical research staff in the 1920s and 1930s. Octaves have their origin in music and were used to facilitate the use of analogue filters in partitioning the frequency scale. Partitioning into octaves divides the frequency scale into bands increasing in width by a factor of 2 with increasing frequency. To obtain greater resolution, the partitioning is often performed in 1/3 and sometimes in 1/12 octaves. However resolution remains poor particularly at higher frequencies. The ear can distinguish frequency differences that are much smaller than 1/3 or 1/12 octaves. Decibels and octave filtering originated in an era when slide rules and logarithmic tables were used for engineering computation. Nowadays digital signal processing with Fourier transforms provides much better resolution. Measurements are made in ordinary linear SI units followed by conversion into decibels and octaves. This final step is unnecessary. The data would be simpler and more informative if left in the linear SI form. Octaves and their fractions obscure frequency information. In some applications involving the measurement of vector quantities, such as sound power flow (intensity), the use of decibels and octave filtering is a definite impediment to the measurement. In this paper these questions are explored in detail. Also sound-power testing is proposed as an alternative to passby and other forms of noise testing in the auto industry. The benefits of sound-power testing are (a) it is independent of the test site and (b) it provides information identifying and quantifying the contributions of component noise sources.