|
|
In TDMA systems users share the same frequency band by accessing the channel in non-overlapping time intervals in a round robin fashion [Falconer, Adachi, and Gudmundson, 1995]. Since the signals do not overlap they are clearly orthogonal and the signal of interest is easily extracted by switching the receiver on only during the transmission of the desired signal. Hence, the receiver “filters” are simply windows instead of the bandpass filters required in FDMA. As a consequence, the guard time between transmissions can be made as small as the synchronization of the network permits. Guard times of 30-50 μs between time slots are commonly used in TDMA based systems. As a consequence, all users must be synchronized with the base station to within a fraction of the guard time which is achievable by distributing a master clock signal on one of the base station’s broadcast channels.
TDMA can be combined with time-division duplexing (TDD) or frequency-division duplexing (FDD). The former duplexing scheme is used for example in the Digital European Cordless Telephone (DECT) standard and is well suited for systems in which base-to-base and mobile-to-base propagation paths are similar, i.e., systems without extremely high base station antennas. Since both the portable and the base station transmit on the same frequency, some signal processing functions for the down-link can be implemented in the base station, as discussed above for FDMA/TDD systems.
In the cellular application the high base station antennas make FDD the more appropriate choice. In these systems, separate frequency bands are provided for up-link and down-link communication. Notice, that it is still possible and advisable to stagger the up-link and down-link transmission intervals such that they don’t overlap to avoid that the portable must transmit and receive at the same time. With FDD the up-link and down-link channel are not identical and hence signal processing functions can not be implemented in the base-station; antenna diversity and equalization have to be realized in the portable.
In comparison to a FDMA system supporting the same user data rate the transmitted data rate in a TDMA system is larger by a factor equal to the number of users sharing the frequency band. This factor is eight in the pan-European GSM systems and three in the D-AMPS system. Thus, the symbol rate is reduced by the same factor and severe intersymbol interference results, at least in the cellular environment.
To illustrate, consider the example from above where each user transmits 25 K symbols per second. Assuming eight user per frequency band leads to a symbol duration of 5 μsec. Even in the cordless application with delay spreads of up to 1 μsec, an equalizer may be useful to combat the resulting interference between adjacent symbols. In cellular systems, however, the delay spread of up to 20 μsec introduces severe intersymbol interference spanning up to 5 symbol periods. As the delay spread often exceeds the symbol duration the channel can be classified as frequency selective, emphasizing the observation that the channel affects different spectral components differently.
The intersymbol interference in cellular TDMA systems can be so severe that linear equalizers are insufficient to overcome its negative effects. Instead more powerful, non-linear decision feedback or maximum-likelihood sequence estimation equalizers must be employed [Proakis, 1991]. Furthermore, all these equalizers require some information about the channel impulse response which must be estimated from the received signal by means of an embedded training sequence. Clearly, the training sequence carries no user data and, thus, wastes valuable bandwidth.
In general, receivers for cellular TDMA systems will be fairly complex. On the positive side of the argument, however, the frequency selective nature of the channel provides some “built-in” diversity which makes transmission more robust to channel fading. The diversity stems from the fact that the multi-path components of the received signal can be resolved at a resolution roughly equal to the symbol duration and the different multi-path components can be combined by the equalizer during the demodulation of the signal. To further improve robustness to channel fading coding and interleaving, slow frequency hopping, and antenna diversity can be employed as discussed in connection with FDMA.
In both FDMA and TDMA systems, channels should not be assigned to a mobile on a permanent basis. A fixed assignment strategy would either be extremely wasteful of precious bandwidth or highly susceptible to co-channel interference. Instead channels must be assigned on demand. Clearly, this implies the existence of a separate up-link channel on which mobiles can notify the base station of their need for a traffic channel. This up-link channel is referred to as the random-access channel because of the type of strategy used to regulate access to it.
The successful procedure for establishing a call that originates from the mobile station is outlined in Figure 3. The mobile initiates the procedure by transmitting a request on the random access channel. Since this channel is shared by all users in range of the base station a random access protocol, like the ALOHA protocol, has to be employed to resolve possible collisions. Once the base station has received the mobile’s request it responds with an immediate assignment message which directs the mobile to tune to a dedicated control channel for the ensuing call setup. Upon completion of the call setup negotiation a traffic channel, i.e., a frequency in FDMA systems or a time slot in TDMA systems, is assigned by the base station and all future communication takes place on that channel. In the case of a mobile-terminating call request, the above sequence of events is preceded by a paging message alerting the mobile of the call request.
Named after the organization that created the system standards (Groupe Speciale Mobile) this pan-European digital cellular system has been deployed in Europe since the early 1990s [Hodges, 1990]. GSM uses combined TDMA and FDMA with frequency-division duplex for access. Carriers are spaced at 200 KHz and support eight TDMA time slots each. For the up-link the frequency band 890–915 MHz is allocated, while the down-link uses the band 935–960 MHz. Each time slot is of duration 577 μs which corresponds to 156.26 bit periods, including a guard time of 8.25 bit periods. Eight consecutive time slots form a GSM frame of duration 4.62 ms.
The GSM modulation is Gaussian minimum shift keying with time-bandwidth product of 0.3, i.e., the modulator bandpass has a cut-off frequency of 0.3 times the bit rate. At the bit rate of 270.8 Kbit/s, severe intersymbol interference arises in the cellular environment. To facilitate coherent detection, a 26 bit training sequence is embedded into every time slot. Time diversity is achieved by interleaving over eight frames for speech signals and 20 frames for data communication. Sophisticated error correction coding with varying levels of protection for different outputs of the speech coder is provided. Note that the round-trip delay introduced by the interleaver is on the order of 80 ms for speech signals. GSM provides slow frequency hopping as a further mechanism to improve the efficiency of the interleaver.
In cellular systems, like GSM or the north-American D-AMPS, TDMA is combined with FDMA. Different frequencies are used in neighboring cells to provide orthogonal signaling without the need for tight synchronization of base stations. Furthermore, channel assignment can then be performed in each cell individually. Within a cell one or more frequencies are shared by users in the time domain.
From an implementation standpoint TDMA systems have the advantage that common radio and signal processing equipment at the base station can be shared by users communicating on the same frequency. A somewhat more subtle advantage of TDMA systems arises from the possibility to monitor surrounding base stations and frequencies for signal quality to support mobile assisted handovers.
