Dr. Dobb's Journal October 1998

Digital Audio Basics

Digital audio consists of a sequence of samples recorded at a specific rate. Each sample represents the voltage of an audio signal at a particular instant in time. Unless you are interested in changing the pitch of the recorded signal, samples must be played back at exactly the same rate at which they were recorded. Each sample is generally represented as either a 16-bit or eight-bit number. Typically 16-bit samples are standard two's complement integers. But eight-bit samples are often in an unusual format called "offset 128" or "excess 128." In this format, the smallest negative voltage is represented by 0, a zero voltage is 128, and the largest positive voltage is 255. In other words, you add 128 to a two's complement number and treat the result as an unsigned integer. I do not know the origin of this format, but I'm sure it must have saved a NAND gate somewhere back in 1978. In a stereo signal, the samples are typically interleaved. As with any digitization, the higher the sample rate and the more bits in the sample, the greater the fidelity of the recording. Sample rates are usually specified in hertz, though you should think of it as samples per second, not cycles per second. Table 1 lists some common sample rates.

Table 1: Common sample rates. *On the original Macintosh, the sound sample clock was based on the horizontal retrace signal of the video circuitry: 60.15 video frames per second x (342 visible horizontal lines+28 "invisible" lines for vertical retrace)=22255.5 samples per second.

-- R.C.

Back to Article