This paper presents the measured performance of a 64 kbit/s switched-diversity FSK receiver subjected to simulated frequency-selective Rayleigh fading. The single receiver input is switched between two, three, or four uncorrelated Rayleigh-fading signals whenever the instantaneous receiver output falls below a threshold. The optimum level of this threshold relative to the mean signal level is relatively insensitive to frequency selectivity and vehicle speed. A nearly optimum threshold may be determined using an AGC amplifier and fixed comparator. Switched diversity is a powerful tool for combating frequency selectivity and Rayleigh fading. Two-branch switched diversity can achieve 10-2BER, with 6 dB lower SNR than that needed without diversity. In the absence of frequency selectivity, two-branch switched diversity can perform within 3 dB SNR of maximal-ratio diversity and within 1 dB of selection diversity. Switched diversity can achieve BER's lower than the irreducible single-channel BER produced by frequency selectivity. Four-branch switched diversity typically requires 4 dB lower SNR to achieve a given BER than does two-branch switched diversity.

We present the analysis and design of an efficient hybrid digital/analog multichannel hardware simulator for frequencyselective Rayleigh-fading mobile radio paths. The simulator has eight RF outputs and is linear and reciprocal. All outputs exhibit mutually independent Rayleigh fading over a 45 dB range and have identical time delay spreads. Analog RF techniques are combined with a digital signal processor (DSP) to generate random path modulation. Frequency selectivity from time delay spread is created using several RF path delay elements. Simulated vehicle velocity is continuously variable from 6 to 60 mph. The simulator is thus a flexible tool for the study of mobile radio transmission techniques.

This paper presents measured error-rate performance of binary FSK with discriminator detection in simulated Rayleigh fading with time delay spread. Measured relations between error rate and carrier-to-noise ratio are shown for a range of time delay spreads. We have verified that the effect of time delay spread on error rate can be described by only one parameter, the ratio of the rms delay spread to the transmitted baud time. Intersymbol interference from time delay spread causes an irreducible error rate for FSK, even for infinite carrier-to-noise ratio. For 64 kbit/s binary FSK, the 2.6μs rms delay spread typical of the urban environment produces an irreducible error rate of 3 percent. The measured irreducible error rate for binary FSK with discriminator detection is proportional to the fourth power of the ratio of rms delay spread to baud time. This functional dependence is identical to that predicted analytically for FSK with incoherent matched-filter detection.

This paper presents distributions of the intervals between fades measured on a hardware-simulated Rayleigh-fading mobile radio channel. For intervals between deep fades, these distributions take on a Poisson form. Measured results are in good agreement with a simple analytical model. Distributions for the intervals between shallower fades exhibit preferred durations caused by peaks in the Doppler fading spectrum.

This paper presents measured fade duration and levelcrossing statistics of a hardware-simulated mobile radio channel. These measurements were taken over a wide range of fade depths. The measurements agree well with existing analytical results and extend them to show the distribution of fade durations for shallow fades. Fine detail in the fade-duration distributions for shallow fades is produced by peaks in the fading spectrum.