ABSTRACT
Currently the most commonly used representation is a 16-bit system; 16 bits allow 216 ¼ 65,536 separate levels to be represented in binary form. Often a digital audio processing system will use 16-bit DACs followed by 20-bit or more ‘internal processing’ (Figure 2.1.9). Representing the digital audio more accurately than its ori-
ginal conversion allows for many stages of audio processing, gain, filtering, equalisation, etc., each adding rounding errors3
(rounding noise) to the signal. The use of these extra ‘internal’ bits means rounding errors will always be smaller than the least significant bit (lsb) of the original conversion. This process is analogous to performing calculations on a calculator which is accurate to a large number of digits, then finally rounding the result back to the accuracy of the original data.
