Limitations of the Gaussian model for radar signals

We show that the modeling of radar signals by a Gaussian random process has serious limitations in representing the various cases of radar signals. We show that, although the Rayleigh density function is a good model for the envelope of the target signal, the signal can not usually be modeled as a Gaussian process, i.e., the samples are not jointly Gaussian in spite of the use of this assumption by several authors. The case of coherent signals with unknown Doppler is also examined