Novel architectures and multiplexing strategies for optical CDMA.

The work presented in this thesis explores avenues to efficiently use code-division as the multiple-access technique in optical communication systems. Coding in the time domain is first considered. Novel architectures that allow the use of bipolar codes with simple direct detection receivers are first presented, for the case of coherent optical sources. We show that a proper combination of electro-optic elements and all optical processing elements leads to encoder and decoder complexities much reduced compared to the all optical solutions, without sacrificing the chip rate. This is further enhanced when we propose the use of Kronecker sequences whose inner sequences are chosen to be perfect autocorrelation sequences. In that case, even programmability of the delay lines is not required. We then modify the proposed architecture so that they can be used with incoherent optical sources. Although the sources can only be modulated in power, we show that still bipolar codes can be used and maintain their properties. A detailed performance analysis of the proposed architectures is made, including multiple-access interference, shot noise and thermal noise. Two types of modulation: amplitude-shift-keying (ASK) and orthogonal coding (OC) are considered. Results presented give the probability of bit error as a function of the number of simultaneous users, for many code lengths values. Performances obtained using different combinations of inner and outer sequences are compared, and requirements in terms of number of delay lines and bandwidth of electro-optic elements are identified. The effect of partial cross-correlation due to non optimized optical filtering is studied by means of an upper bound on the probability of bit error. A new analytical technique is derived to analyze optical CDMA systems using coherent optical sources with direct detection. A new and highly efficient coding technique based on amplitude spectral encoding of incoherent optical sources such as light-emitting diodes is then proposed. This is one of the major contributions of this thesis. Architectures for the encoder and decoder are given, and suitable codes are identified. The proposed system is insensitive to near-far effects and has a spreading gain independent of the bit rate. This property makes this coding technique a suitable one for networks serving users with different bit rates requirements. Detailed evaluation of the probability of error is done, and practical considerations such as the spectral shape of the sources, finite extinction ratio of the amplitude mask as well as its programmability speed are quantified. We show that more than 100 simultaneous users can be accommodated with a probability of error of 10$\sp{-9}$, for a code length as small as 127. Finally, the use of this coding strategy as the routing technique for Asynchronous Transfer Mode (ATM) switches is discussed