There are several approaches to the Monte Carlo simulation of a time-varying multipath radio channel with a given Doppler spectrum. This paper concentrates on a particular method in which the channel, having a discrete tapped-delay-line structure, is simulated by generating a time-varying Doppler spread channel tap for each resolvable delay path. Each channel tap is created by the addition of multiple sinusoidal echoes, where the frequency of an echo is assigned randomly based on the channel Doppler spectrum. A model of this type has three key parameters: the sampling rate, the number of sinusoidal echoes and the total number of samples used in the channel realisation. Quantitative criteria are developed which provide guidelines for selecting values of parameters in the simulator design. These criteria are illustrated by examining the performance of two simulated channels, one having the Jakes Doppler spectrum the other a Gaussian Doppler spectrum. The performance is measured in terms of a normalised root-mean-squared-error between the specified and the achieved Doppler spectra. We conclude that to keep the NRMSE of the achieved Doppler spectrum to within 10%, 200 sinusoids are sufficient, provided that a suitable sampling frequency is selected.