WN3- 1 F FWM Noise Reduction in Bursty Metro Transmission Hyunchin Kim, Lubo Tancevski, An Ge and Gerard0 Castanon Alcatel USA, 1000 Coit Rd., M/S: 446-310, Plano, TX 75075-5802 Tel: 972-996-2849 Fax: 972-996-5902 email: Hyunchin.Kim@usa.alcatel.com Abstract: We analyze the occurrence of FWM noise in bursty metro network transmission of Ethernet packets directly on the WDM layer. We obtain more than 5 dB reduction in FWM noise for traffic loads less than 0.7 both for uniform and self-similar traffic. Gigabit Ethernet is increasingly gaining in importance as a protocol for the future WDM metropolitan networks due to its low cost and relatively straightforward implementation - indeed efforts are under way to implement 10 Gb/s Ethernet interfaces. The question of efficient framing mechanism for Gigabit Ethernet, however, remains still open with the same underlying expectation that it should be straightforward and inexpensive. In this manuscript we analyze the performance of a framing protocol which does not use idle patterns to fill up gaps between packets - it rather only envelops the packets. As such, it results in power fluctuations on the line which are statistical in nature and are determined by the nature of the traffic model. The consequence of this is that non-linear effects which are power dependent also become statistical in nature. We analyze the performance with regards to one particular effect, the four-wave mixing FWM noise. Because of the statistical nature of the transmission, not all the channels are present all the time, meaning that there will be reduction in the generated FWM noise. We derive an analytical reduction factor, compare it to the reduction obtained through simulations and note an excellent agreement. We further note that the reduction is not dependent on the burstiness of the traffic. \ J Sampling in time a) 0 -1 n U. n n w -2 w 0 4 -3 -4 -45 -40 -35 -30 -25 -20 -15 -10 FWM power (dBm) b) Fig. 1: Simulation model: a) graphical representation of the traffic and the sampling procedure; b) histogram of recorded FWM powers for traffic load 0.5, self-similar model with Hurst H=0.9. The simulation model is depicted in Fig. la). Packets are transmitted at every wavelength independently and the space between packets is not filled with idle patterns - rather, there are power gaps on the line. In this scenario, the power on the line becomes a random variable and therefore the generated FWM noise is a random variable itself. In order to obtain a statistics of the occurrence of the noise, the power is sampled at regular intervals and the noise evaluated at each of those intervals using eq.(8) in [l].