1M Packet-Switched Router Based on the PPM Header Address for All-optical WDM Networks ABSTRACT This paper presents an all-optical 1M router architecture for simultaneous multiple-wavelength packet routing, without the need for wavelength conversion. The packet header address is based on the pulse position modulation (PPM) format, which allows the use of only a single-bitwise optical AND gate for fast packet header address correlation. The proposed scheme offers both multicast and broadcast capabilities. We’ve demonstrated a high speed packet routing at 160 Gb/s in simulation, with a low channel crosstalk (CXT) of ~ -27 dB with a channel spacing of > 0.4 THz and a demultiplexer bandwidth of 500 GHz. The output transfer function of the PPM header processing (PPM-HP) module is also investigated in this paper. Keywords Wavelength division multiplexing, packet switching, pulse position modulation, address correlation, symmetric Mach-Zehnder, optical switch. 1. INTRODUCTION The ever increasing demand for a high speed Internet access together with a high definition digital video broadcasting (video-on-demand) requires an ultrahigh speed networking from the source to the end users. Current networks based on a combination of optical fibre links, radio frequency links and twisted pair cables suffers from the speed bottleneck mainly imposed by the non-optical links. Thus the logical way forward would be to do away with all non-optical components and links as much as possible. An end-to-end all-optical packet switched network [1, 2] would offer speed, flexibility, lower power consumption and has the capability to deal with the bursty traffic. However, to be able to fully utilize this promising potential of such networks, the packet header processing (including clock recovery and packet address correlation) and routing decision would have to be performed in the optical domain rather than in the electrical domain, to avoid the speed bottleneck imposed by the slow response of existing electronic devices. In [3-5] ultra high-speed Boolean logic gates including AND, OR and XOR operating above 40 Gb/s have been reported as the key enabling technology for realising all optical routers. Currently the dominant technique for optical packet header processing is based on the sequential correlation of the incoming packet header address with all entries of a routing table within each node. This method is most viable for a small size network with a reasonable size routing table both in terms of processing speed and implementation complexity. However, for a large size network with a routing table of hundreds or thousands of entries may M. F. Chiang a , Z. Ghassemlooy a , W. P. Ng a , H. Le Minh b , and A. Abd El Aziz a a Optical Communications Research Group, School of Computing, Engineering and Information Sciences Northumbria University, Newcastle upon Tyne, UK b Department of Engineering Science, University of Oxford, UK Email: {ming-feng.chiang, fary.ghassemloy, wai-pang.ng, ahmed.shalaby}@unn.ac.uk , Hoa.le-minh@eng.ox.ac.uk Phone: +44 (0)191 227 4902, Fax: +44 (0)191 227 3684