SummaryWavelength division multiplexing (WDM), the use of several distinct, separated wavelengths on a single fiber, is revolutionizing point-to-point optical links in the telecommunications industry. The separate wavelengths, however, are not truly independent but are coupled together by nonlinear interactions in the fiber material. Nonlinear four wave mixing (FWM), combined with low dispersion fibers and Er-doped fiber amplifiers, results in severe performance degradation that limits capacity and span lengths. We have developed a WDM wavelength shift keying technique with balanced detection that completely cancels FWM interference to first order, and in addition eliminates other broad bandwidth noise. Our modeling shows that this technique substantially improves performance relative to standard On-Off WDM encoding. Experimental testing of the technique is underway.
For a multiuser network the parameters of interest are the total number of users that can subscribe to the network, and the number of users that can be active simultaneously on the network for a given quality of service. The number of active users on a network is generally a small fraction of the total number of subscribers due to the bursty nature of computer traffic, but even so, design parameters of interest for a network are as many as 1000 subscribers and 100 active users. This means that network channels must have capacities approaching a terabit per second. Thus, the user interface, which extracts the desired data from the combined data stream on the channel, must be a passive optical system to avoid electronic speed limitations. We are investigating using wavelength encoding to support many more users than simple WDM can. Each user is assigned a unique combination of wavelengths as their code, a technique we call spectral code division multiplexing (SCDM). Obviously, many more combinations exist than single wavelengths. The key to performance is picking codes that have low cross correlations so that a correlation receiver can isolate the desired data. Such codes have been developed for radio frequency systems, where they are generally used in the time domain and called spread spectrum systems. These codes are bipolar, with (+1, -1) elements, and unfortunately cannot be used directly in the optical domain because optical phase cannot be determined directly.
We have invented (and patented) a new technique that permits these codes to be used in optical fiber systems while maintaining their desirable properties. This work was done in collaboration with Prof. Aazhang and his group. Our approach has a major advantage in that it avoids electronic speed limitations: the spreading or processing gain, and number of users, is independent of the transmission rate, and the combined, very high speed data on the network channel is processed by a passive all optical decoder. We have built a prototype two-user system, made bit error rate measurements, studied the influence of interference, and examined the performance limitations due to encoder/decoder optical properties. Currently we are working on integrating the encoder/decoder onto an integrated photonics chip.
This work has been supported by the Texas Advanced Technology Program, the U.S. Air Force Office of Scientific Research, and Rice University. For further information, please consult the publications below, or contact me at: young@ece.rice.edu
Selected Publications
Recent Presentations