A new approach to the spectral-density concept of a cyclostationary process is presented. This approach is based on the observation that a cyclostationary process can be split into a number of wide-sense stationary subprocesses. The spectral densities of these subprocesses, defined as the Fourier transform of their autocorrelation functions, are nonoverlapping functions of frequency. The sum of the spectral densities of these wide-sense stationary subprocesses yields the well-known expression for the spectral density of a cyclostationary process.

Frequency-shift-keying (FSK) signals with modulation indexm = 0.5have two significant properties. They have no discrete frequency components and nearly all the signal energy is contained within a narrow frequency region equal to\frac{3}{2}of the bit speed even without any band limiting. Unfortunately, a complete mathematical description of FSK signals in general is difficult, because FSK is a nonlinear process. A well-known exception is FSK withm = 1, because this signal can be described as the summation of two AM signals. This makes it easy to describe frequency- and time-domain properties. In this paper it is shown that by decomposing FSK signals withm = 0.5into two signals with well-understood spectral properties, it is possible to describe time- and frequency-domain properties and to calculate the effects of restricted bandwidth and linear distortion on the signal-space diagrams and eye patterns. Finally it is shown that duobinary FM can be treated as a special case of an FSK (m = 0.5) signal.