Parallel data-transmission systems using code-division multiplexing

The thesis deals with parallel data-transmission systems that operate over slowly time varying channels and use code division multiplexing. In Chapters 2 and 3 of the thesis, the transmission properties of a telephone network and an HF channel are reviewed. Brief summaries of some of the existing multiplexed systems are given in the next chapter. The latter part of the thesis describes the work done by the author as summarised below.

It is well-known that in a Code-division multiplexed (CDM) system, which operates over a channel that introduces distortion, the receiver receives many copies of each signal element with randomly varying amplitudes and delays. Thus each signal element is corrupted by interferences from time-shifted replicas of itself and from all other signal elements and their time-shifted replicas. In order to remove these interferences in any CDM system, which uses no signal in the time guard band separating two adjacent groups of elements, one requires a group of finite length sequences of ideal auto-correlation function, each of which must have nearly zero cross-correlation with other sequences of the group and their shifted replicas. The number of binary sequences in each group that satisfy the above requirement is however very small. The author has shown that this situation may be somewhat improved by the use of four phase sequences, but it is observed that the number of sequences in each group is still far below the desired value.

In order to obviate this difficulty, the author has suggested that one can look for groups of periodic sequences with the desired properties if the signal elements are periodically repeated in the time guard band. It has been shown in the thesis that even if the restriction on the auto-correlation function is relaxed, the number of sequences available in each group is restricted to n+1, where the number of elements in each sequence is 2ⁿ.

Theoretically, one would like to have as many sequences as there are elements in each of them. Therefore, it was considered useful to investigate whether a system can be designed where the receiver first acquires a knowledge of the characteristics of the channel from the received signal and later uses this knowledge in detecting the values of the data elements. Such a system is termed an adaptive system. Three different adaptive systems have been proposed in the thesis. Each of them are capable of being used with different sets of orthogonal signals. But the transmitter arrangements in these systems are different and consequently the receiver arrangements also vary. However, the processes of detection and channel estimation for all the systems are iterative. It has been shown in the thesis that the third system becomes quite attractive and can operate with a correlation detector if the group of signal elements is appropriately modified. It has also been established that each new group of signal elements is uniquely defined and the resultant distorted signals remain perfectly orthogonal to one another for any type of channel distortion. Thus the system operates with a simple correlation detector and derives the knowledge of the impulse response of the channel by a matched network. Furthermore, no training signal is required excepting at the start of transmission and the detector is held matched to the received signal during the period of detection.

All the above systems have been studied both theoretically and by computer simulation under various types of distortions and noise. The performances of the proposed systems have been compared amongst themselves and also with the existing serial time division multiplexed system.

For data-transmission over slowly time varying channels at element rates up to 2400 elements/second or perhaps 4800 elements/second, the equivalent serial system is considered to be very useful. For a higher transmission rate, the iterative processes used in the system become time consuming and require multiplication of equipment.

The CDM system, which uses 'Preferred Orthogonal Sequences' and works with two parity check elements, has the important advantage that the detection of the element values and the estimation of the sampled impulse response of the channel can be achieved simply by correlation techniques. Thus, for higher transmission rates up to 100,000 elements/second this system may have potential practical value. It is interesting to note that this system has a consistent advantage of 2 to 3 db in tolerance to additive noise over the best of the existing systems.