IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 17, NO. 11, NOVEMBER 2005 2493 Experimental Comparison of Coherent Versus Incoherent Sources in a Four-User OCDMA System at 1.25 Gb/s Simon Ayotte, Student Member, IEEE, and Leslie A. Rusch, Senior Member, IEEE Abstract—In optical code-division multiple access (OCDMA), the optical bandwidth is accessed simultaneously by multiple users, leading to beat noise in photodetection. The choice of the optical source can have important impact on that noise. In this paper, we compare two optical sources, a broad-band er- bium-doped fiber source and a multilaser source. Experimental results for a wavelength-time OCDMA system are presented for up to four simultaneous users. Bit-error-rate curves are measured at 1.25 Gb/s with a chip rate of 10 Gchip/s. It is shown that the multilaser source outperforms the broad-band source for the design parameters of this system. Index Terms—Beat noise, code-division multiple access, multiple access interference (MAI), optical fiber communications, wave- length-time coding. I. INTRODUCTION O PTICAL code-division multiple access (OCDMA) is under consideration for applications in metro and local area networks, as well as for fiber-to-the-home. The associated asynchronous passive network and the flexibility in channel granularity allow many potential users with possibly different bit rates and qualities of service. In such a system, OCDMA channels share the fiber resources, taking full advantage of optical bandwidth and time. Wavelength-time OCDMA implements coding in two dimensions and is a good candidate to fairly distribute resources, and to increase code cardinality. In the coding process, an in- coming broad-band pulse is split in distinct chips on the time axis, each of a specific color. At reception, the matched decoder realigns the delays between the different wavelengths, recov- ering an autocorrelation peak. Interferers wavelengths not re- jected by the decoder have chip times distributed over the bit interval and are seen as multiple access interference (MAI). The literature on OCDMA focuses on developing codes with good cardinality and low MAI [1]–[4]. Although a few consider the beating between signals from multiple users [5], [6], the theoretical performance evaluations are generally made Manuscript received June 1, 2005; revised July 27, 2005. This work was sup- ported in part by Canadian Institute for Photonic Innovations (CIPI) under a TEN Grant, by the National Sciences Engineering Research Council of Canada (NSERC) under Grant CG029055, by Le Fonds Québécois de la Recherche sur la Nature et les Technologies (FQRNT), and by the Program for the Develop- ment of “La Cité de l’Optique” (PADCO) FT076892. The authors are with the Centre d’Optique Photonique et Laser (COPL), De- partment of Electrical and Computer Engineering, Université Laval, Québec, QC, G1K 7P4, Canada (e-mail: rusch@gel.ulaval.ca). Digital Object Identifier 10.1109/LPT.2005.858047 Fig. 1. Experimental setup. EDFA: erbium-doped fiber amplifier. PC: polarization controller. OSA: optical spectrum analyzer. by considering only MAI. Here, we demonstrate experimen- tally that the beating of optical sources in the receiver can be- come dominant in OCDMA systems. Bit-error-rate (BER) measurements are conducted at 1.25 Gb/s with a chip rate of 10 Gchip/s for up to four simultaneous users. We present results for two different optical sources and show that the choice of the source has an important impact on the system performance. II. SETUP The experimental setup is shown in Fig. 1. The two sources compared in this experiment are a broad-band erbium-doped fiber source and a multilaser source composed of eight semi- conductor distributed feedback lasers combined with a passive coupler. The output of the source under study is modulated with the data by an electroabsorption modulator (EAM) that is insen- sitive to polarization. For the incoherent source, the use of unpo- larized light increases the signal-to-intensity-noise ratio (SINR) by 3 dB compared to polarized light [7], [8]. For the multilaser source, the EAM eliminates the need of controlling the polar- ization of each laser before modulation. The length of the pseudorandom bit sequence for BER mea- surement is and the modulation format is return-to-zero with a duty cycle of 1/8. After modulation, the signal is am- plified, split in four, and sent to four encoders, one for the de- sired user and three interferers. Polarization controllers allow adjusting the polarization between users when the multilaser source is used. Delay lines are used to decorrelate the data of each user. The OCDMA encoded signals are recombined, am- plified, and sent to the decoder matched to the desired user. A 1041-1135/$20.00 © 2005 IEEE