Spectral loudness summation: shortcomings of current standards.

Abstract

There are two competing national standards for the calculation of loudness of steady sounds, DIN 45631 and ANSI S3.4. Their different concepts of critical bands lead to different predictions for broadband sounds. As that discrepancy is neither constant nor linear but highly frequency-dependent, the present study investigates spectral loudness summation in three frequency regions, at various levels, and using two different methods. The results show that both algorithms overestimate loudness; however, DIN 45631 comes closer to the subjective evaluations and often falls within their interquartile range. The overestimation by the standards is particularly large in the frequency range from 2 to 5 kHz.