Frequency discriminator detection in frequency-selective fading environments.

In recent years, millions of customers all over the world have been subscribing to mobile telephony services which are based on modern digital transmission. At the high transmission bit rates that these systems use, the mobile radio channel exhibits frequency-selective fading characteristics. Under such conditions, the received signal could experience significant intersymbol interference (ISI) from severe distortion to the waveform of the received baseband signal. Therefore, such techniques as adaptive waveform equalisation or adaptive maximum likelihood sequence estimation (MLSE) are used in modern digital mobile radio telephone systems to combat this undesirable ISI. These adaptive schemes have almost always been used in conjunction with coherent demodulation in the receivers.This study examines the application of noncoherent demodulation, in the form of frequency discriminator detection, as an alternative to coherent demodulation. The GSM (Global System for Mobile Communications) standard has been used as the basis for this investigation. It has been shown analytically that in the presence of frequency-selective fading, a propagation environment common to the GSM system, the use of frequency discriminator detection gives rise to nonlinear ISI in the demodulated signal. It has also been shown that frequency-selective fading could cause large unwanted "spikes" to appear in the demodulated signal, thus leading to a severe degradation in the bit-error-rate (BER) performance. Consequently, several waveform distortion cancellation schemes for combatting the nonlinear ISI have been formulated. The BER performances of these proposed schemes, under various propagation conditions, have been studied by computer simulation.Furthermore, it has been observed that the undesirable "spikes", that occur in the demodulated signal due to frequency-selective fading, could be suppressed by the use of inverse-limiting in conjunction with frequency discriminator detection. As a result, an effective adaptive detection scheme has been formulated, based on modelling the combination of the GMSK modulator, the mobile channel, the frequency discriminator, and any transmit and receive fitters, as a finite-state machine. The transmitted data is then detected using an MLSE. The BER performance of this proposed adaptive detection scheme has been extensively investigated by computer simulation. This has been carried out assuming various propagation conditions, including the two-ray fading channel model with equal path powers and relative delays of up to four bit periods, the maximum relative delay considered in the GSM system. Also, the effectiveness of the proposed adaptive detection scheme in combatting IS] has been investigated by computer simulation based on the six-ray GSM empirical propagation models for typical urban (TU), hilly terrain (HT) and rural area (RA) environments. The computer simulated results confirm that the voice grade performance required for the GSM system could be achieved by the proposed adaptive detection scheme in all the recommended GSM propagation models considered. Furthermore, the BER performance of the receiver remains unaffected by a carrier frequency offset of up to 2 kHz