Time-domain graphs of various modes of pulse modulation. At A, pulse amplitude modulation; at B, pulse width modulation; at C, pulse interval modulation; at D, pulse code modulation.
Pulse code modulation, such as is shown at above figure D, is one form of analog-to-digital (A/D) conversion. A voice signal, or any continuously variable signal, can be digitized, or converted into a train of pulses whose amplitudes can achieve only certain defined levels.
In A/D conversion, the number of states is always a power of 2, so that it can be represented as a binary-number code. Fidelity improves as the exponent increases. The number of states is called the sampling resolution, or simply the resolution. A resolution of 23 = 8 (as shown in above figure D) is good enough for voice transmission, and is the standard for commercial digital voice circuits. A resolution of 24 = 16 is adequate for high-fidelity (hi-fi) music reproduction.
The efficiency with which a signal can be digitized depends on the frequency at which sampling is done. In general, the sampling rate must be at least twice the highest data frequency. For an audio signal with components as high as 3 kHz, the minimum sampling rate for effective digitization is 6 kHz; the commercial voice standard is 8 kHz. For hi-fi digital transmission, the standard sampling rate is 44.1 kHz, a little more than twice the frequency of the highest audible sound approximately 20 kHz).