Network Architecture for a High Bandwidth WDMA/CDMA Local Area Network

We propose an optical WDMA/CDMA hybrid high bandwidth local area network using coherent homodyne detection. Local injection locking to centralized highly stable sources is used for frequency selection and coherent detection. Splitting losses in a fully connect optical local area network are significant for large systems (~1000 subscribers) necessitating the use of optical amplifiers. We examine several network configurations to optimize capacity while avoiding amplifier saturation and keeping the number of amplifiers required to a minimum. 1. A WDMA/CDMA Local Area Network Local area networks currently provide at most 100 Mb/s of aggregate data transmission, while future video-intensive applications are expected to drive requirements for bit rates to 100 Mb/s per user. Existing network protocols and electronic transmission paths cannot be adapted to support this anticipated demand. Optical transmission paths provide the necessary bandwidth, however efficient protocols that can exploit this bandwidth are still the subject of research. We propose a flexible WDMA (wavelength division multiple access) -CDMA (code division multiple access) hybrid system. The number of CDMA users at one wavelength is limited by the speed of electronic components that realize the spectral spreading (~ 10 Gb/s). For spreading codes of length 127 this corresponds to on the order of 10 active users. To increase system capacity multiple wavelengths are used, with a cluster of CDMA users operating at each wavelength. Requirements for dense packing of wavelengths are relieved by the ability of CDMA to increase system capacity at each wavelength. Several characteristics of this network set it apart from those proposed previously, including • use of bipolar codes via phase modulation • very narrow linewidth reference lasers • injection locking for frequency selection & purity • homodyne coherent detection We use bipolar codes for CDMA to take advantage of their enhanced auto-correlation and crosscorrelation properties compared to those for 0/1 codes used with on-off keying. This requires that the phase be modulated and in this paper we consider only external phase modulation. As described in [1], the phase can also be modulated by control of injection locking current, at the expense of a moderate penalty in performance. This system uses a central bank of very narrow linewidth reference wavelengths for WDMA. These reference wavelengths are distributed to each user via a fiber bus independent of the fiber network for modulated data. Users would employ these reference signals to injection lock their distributed feedback (DFB) lasers for either transmission or reception. Activating one of a bank of DFB lasers with nominal center frequency on the desired wavelength provides wavelength selection. Injection locking ensures the DFB laser adopts the frequency purity and absolute frequency reference of the reference source, hence allowing for homodyne coherent detection. 1.1 Network logical configuration Figure 1 illustrates how data is transferred from one user to another. The intended recipient has code number two on wavelength i. The first user activates the DFB laser with nominal center frequency λi, locking it to the signal from the reference bus. The output of the DFB laser is modulated externally by the data and spreading code two, and is then sent to all users via the data network. The second user decodes his message by also injection locking to λi and phase modulating by code two. Only the proper combination of wavelength and code will despread the data. This network uses W different wavelengths and a set of D different codes, each one being assigned to a different user, for a total of WD users in the network. The logical network delivers the reference signals to each user via one fiber system and all modulated data signals via an independent fiber system. This network can be implemented in any number of physical configurations. We will choose a physical configuration maximizing the number of users while maintaining 1) a robust power budget, 2) low variation in the input power of the fiber amplifiers, and 3) equal power losses for all users, i.e., all users will have roughly equal received powers to avoid the socalled near-far problem. To determine the maximum number of stations that can be supported, some definitions and assumptions are in order. Let M be the power margin (in dB), defined as M P P T R = 10log1 6 where PT is the available optical transmitter power and PR is the minimum receiver power that provides a tolerable bit-error rate (BER). Typically, M is on the order of 40 dB, representing the allowable losses in the network. 1.2 Physical configuration options The configurations considered here are a bus, tree, star and compound topologies. The bus configuration has data from each station injected into the bus via directional couplers, and broadcasts to receivers. The useful power delivered to each receiver is reduced by 1/N for an N-station network. A linear arrangement of N taps has end-to-end loss (dB) [3]: L N N = − − − 10 1 10 log( ) log α β (1) where α is the coupler power splitting ratio and β is the coupler excess loss. For typical values α = 0.1 and β = 0.8, a bus configuration is limited to a maximum of 8 users for M= 40 dB. This configuration suffers from the near-far problem, that is, the receiver signal power for each user will not be equal. The second basic configuration we consider is the minimum-depth binary-tree topology that allows delivery of uniform signal strength for all users. The excess loss for an N-station tree increases only logarithmically with N with total L N N = − 2 10 10 2 log log log