high rate communications and networking

  • Faculty D. Erasme, P. Gallion, C. Gosset, Y. Jaouën, C. Ware

Signal processing for optical time division multiplexing transmission

Clock recovery at transmission end or in routing nodes is an essential and yet challenging functionality in the case of high-bit-rate digital signals. A phase locked-loop can take advantage of ultrafast nonlinear optical devices (SOA or PPLN) as phase comparator. It allows clock recovery of RZ signals—and NRZ in some cases—as well as OTDM demultiplexing by recovering the “sub-clocks”. Collaborating with the Technical University of Denmark and the National Institute for Materials Sciences of Japan, we demonstrated sub-clock recovery and full 1/64 OTDM demultiplexing at 640 Gbit/s.. This was the second-ever demonstration of clock recovery at that high a bit rate, and the first involving a PPLN device, which was presented among record-setting postdeadline papers in OFC'2008. This activity, in the framework of e-Photon/ONe+ and EURO-FOS was rewarded by an invited paper in the Journal of Lightwave Technology and the “Letter of the Month” of Electronics Letters. Bit rate was pushed up to 0.87 Tbit/s, including phase modulation. Now that this level of performance and versatility has been shown, this activity is now reduced in favor of network-oriented functionalities.

Coherent optical communication combined with Digital Signal Processing

Progress in digital signal processing and optical integration have enabled a new generation of optical transmission systems using complex modulation formats, coherent detection and digital algorithms to compensate for transmission impairments. Within ANR-TCHATER, we proposed more efficient coding/decoding forward error correcting (FEC) solutions suitable for high bit rates. We investigated the unique properties of space-time codes for optical transmission systems. For the first time, we have shown that space-time coding could efficiently mitigate Polarization Dependent Loss impairments (PDL). We have shown also that performances are very different from those in wireless transmission and explained why. In collaboration with the Karlsruhe Institute of Technology, we proposed the first experimental implementation of Polarization-Time code (PT) for optical communications. The performance of Silver, Golden and Alamouti PT codes for PDL mitigation were compared in reference to the uncoded case. A very-high-baud transmission platform is currently in development, including a 100 Gb/s transmitter/receiver and a 400 km recirculating loop. Its versatility and upgradability allows investigating different aspects of digital optical communications: Tx/Rx characterization, propagation impairments, new detection schemes, digital processing and coding techniques dedicated to the optical channel. As higher modulation formats are more sensitive to signal distortions, accurate estimators and more robust equalizers are still required for QAM formats. New adaptive blind and decision-directed equalizer based on Pseudo-Newton gradient-descent algorithm well matched to QAM and offering a better convergence speed with little extra computational load has been successfully introduced. The channel having very slow time variation (with respect to data rate), we proposed a block-wise implementation of the blind-time CMA equalizers. Moreover, a new Carrier Frequency Oscillator (CFO) estimator well adapted to QAM modulation yields remarkable performance and allows penalty-less operation.

Cross-layer networks, network architecture and packet switching

Today's conventional OSI-layer-model networks face the critical challenge of unsustainable energy consumption growth. Solving it will require a drastic redesign, new network architectures optimised globally, beyond the artificial barriers imposed by network layers. Such cross-layer networks are already being developed to support mobile devices; in the case of fixed networks, optics has the potential to shine. Switching and routing, which is currently performed mostly in the electrical domain, entailing costly optical-to-electrical conversions is an emerging energy-cost driver. We have worked on the practical implementation of all-optical packet switching nodes at two occasions. An experimental proof of principle for a global all-optical node architecture based on serial-to-parallel conversion for header recognition concluded the first investigation. Further work initiated with SUP'Com Tunis and supervised conjointly whose objective used optical coding (OCDMA) to label the packets and performed all-optical routing using fiber Bragg grating coders/decoders and an optical flip-flop for routing. Experiments were performed in NTUA (Athens) thanks to the EURO-FOS NoE. Another study led us estimate to the performance an optical ring WDM network architecture by providing the parameters of the statistical distribution χ². However, contention is a major issue in all-optical switching: without practical buffering techniques, packets bound for the same destination must sometimes be dropped. Hybrid switching nodes aim to solve this problem by supplementing an optical switching matrix with an electronic memory. Our work on this architecture started with C. Ware's sabbatical in Columbia University (New York); part of a Ph.D. was dedicated to a performance analysis of the hybrid node (in collaboration with other LTCI groups COMNUM, RMS, and with Mitsubishi El.).

Complex signal analyses through pulsed coherent optical detection and optical sampling

Optical linear sampling uses the coherent optical mixing of the signal to be received, and statistically characterized, with a pulsed local oscillator (LO) providing simultaneously a temporal gating and a mixing gain. Oversampling, using several samples during symbol duration, takes benefit of improved modulation schemes and radiofrequency-based digital signal processing. Bandwidth related issues are solved afterward by using interleaved optical time bin and many fold detection. Synchronized linear optical sampling has been introduced, in real-time configuration, to overcome post detection electronics and analog-to-digital converter limitations as the bit rate increases. We have derived a general theory for the optical transfer function and the signal-to-noise ratio of pulsed coherent detection, the results of which can be used in different applications. On different grounds, a prototype based on asynchronous linear optical sampling was developed, especially for high frequency measurement (> 100GHz). We have derived and implemented digital blind signal processing techniques for digital clock recovery to display periodic signals and eye diagrams, for optical carrier recovery to display constellations, and for time resolved techniques to study frequency dynamic of optoelectronic devices. This work is partially included in the ANR-OCELOT project.

Related experimental means and novel characterisation concepts

Since 2008, we have defined a platform based programme, called “PENSER 100GIGA”, to implement both direct detection and coherent high-bit-rate optical techniques for the study of communication systems and related optoelectronic advanced devices. These new platforms allow the generation, detection and characterization of vectorial optical signals at 100 Gb/s. “PENSER 100GIGA” is ready to be upgraded up to 400 Gb/s, with a final target at 1 Tb/s. This investment project evaluated 500k€ has won a SESAME financing from region Ile-de-France (2010-2014). Within this programme we are developing a pulsed optical source, that now stands as a new field of investigation. It is a 1.5 µm widely tunable repetition rate, from 1 MHz to 10 GHz, femtosecond pulsed optical source, with frequency noise characteristics compatible with asynchronous linear optical sampling. The pulse duration allows THz bandwidth measurement

Image return top of page