Digital Recording of Extended Range Signals

Investigation of means for reducing the amount of signal-derived digital data to be stored or transmitted leads to two general lines of attack. One is to provide analog compression of the signal before digitizing. The second is to digitize the signal linearly, subsequently treating each digital number as composed of two parts, one of which identifies the number´s exponent, the other representing the value to a limited set of significant digits. Discussion of the two lines of attack in terms of discrete integers shows the similarity of concepts and points up the fundamental reasoning basic to system design. One interesting result for logarithmic analog compression is that the logarithmic base should approach unity as the available recording integers increase in number.