IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 24, DECEMBER 15, 2007 2003 Joint Effect of MLSE and Receiver Filters Optimization on Dispersion Robustness of IMDD, DPSK, DQPSK, and Duobinary Modulation G. Bosco, Member, IEEE, V. Curri, Member, IEEE, E. Roffè, and P. Poggiolini, Member, IEEE Abstract—We analyze the effectiveness of a 32-state maximum- likelihood sequence-estimation (MLSE) receiver on chromatic dis- persion robustness of optical transmission based on several binary modulation formats: intensity modulation direct detection, differ- ential phase-shift keying, and duobinary line coding. Multilevel differential quadrature phase-shift keying modulation is also an- alyzed using a four-state 2-bit/symbol joint MLSE processor. For all modulation formats, receiver filters are optimized together with the use of the MLSE technique. Index Terms—Chromatic dispersion (CD), maximum-likelihood sequence-estimation (MLSE), receiver filters. I. INTRODUCTION I T IS well-known that the use of maximum-likelihood sequence-estimation (MLSE) receivers may strengthen the robustness to chromatic dispersion (CD) of optical communi- cations systems. The use and benefits of MLSE for standard intensity modulated direct detection (IMDD) modulation for- mats have been extensively investigated [1]–[4], also studying the advantages given by the increasing of algorithm complexity through the number of states [5]. Interesting theoretical and simulative investigations have been presented on a branch metric alternative to the Gaussian one, showing that the use of the square root of electric signal as decision signal, in conjunc- tion with the Gaussian metric, can give relevant advantages for the robustness to CD of IMDD modulation. At the same time, several studies have demonstrated that, in most optical communication scenarios, the use of alternative modulation formats can give advantages in terms of system per- formance. Duobinary line coding (DB) and two- and four-level differential phase-shift keying (DPSK) and differential quadra- ture phase-shift keying (DQPSK) seem to be the best candidate to substitute IMDD in optical communications. Experimental results on the application of MLSE receivers to DB and DPSK have been presented in [8]–[10], demonstrating how this tech- nique can give some advantages in CD robustness, but the po- tentialities of MLSE for DB, DPSK, and DQPSK need further investigations. Another technique that may improve performance for all the formats is the optimization of receiver (Rx) optical and electrical filters [10]. Manuscript received July 18, 2007; revised September 3, 2007. This work was supported by the FP6 NOE E-Photon/ONe+, by WP-VD-T, and by WP-JP-E. The simulator OptSim was supplied by RSoft Design Group, Inc. The authors are with Dipartimento di Elettronica, Politecnico di Torino, 10129 Torino, Italy (e-mail: curri@polito.it). Digital Object Identifier 10.1109/LPT.2007.909673 Fig. 1. Considered system setup for IMDD, DB, and DPSK. The purpose of this letter is to present the potentialities of MLSE applied together with Rx filter optimization for DB, DPSK, and DQPSK by comparing their performance to that of IMDD. We carried out a simulative analysis and first analyzed the advantages on systems based on typical Rx filters. We then optimized the Rx filters showing, for the first time to our knowledge, the advantages of simultaneous use of MLSE and optimal filters, using a 32-state MLSE Rx. II. SYSTEM SETUP We considered the system setup shown in Fig. 1, operating at a bit rate equal to 10.7 Gb/s. For the IMDD modulation, the Mach–Zehnder modulator (MZM) is driven between and and the driver is bandwidth limited to BW GHz. For DB modulation, the MZM is driven between and and BW GHz. For DPSK modulation the MZM is driven between and and BW GHz. DQPSK modulation is obtained transmit- ting two (in-phase and quadrature) DPSK data flows at half the bit rate. For DB, DPSK, and DQPSK modulations, the pseudo- random bit sequence (PRBS) is properly precoded. The PRBS degree is 16 and the number of simulated bits is equal to . The fiber is linear and purely dispersive, with a dispersion value ps nm/km typical of G.652 standard single mode fibers at 1550 nm. Amplified spontaneous emission (ASE) noise loading is performed after the fiber to obtain the desired optical signal-to-noise ratio (OSNR), which is measured over a noise bandwidth equal to the bit rate. The Rx photodetector and elec- trical circuitry are assumed noiseless. For IMDD and DB mod- ulations, the receiver is composed by a second-order supergaus- sian optical filter with bandwidth , followed by an ideal pho- todetector and a fifth-order postdetection filter with bandwidth . The same filters are used for DPSK, with the optical filter followed by an asymmetric Mach–Zehnder interferometer (with 1 bit delay) and a couple of balanced photodetectors. The values of filter bandwidths are displayed in Table I. The MLSE processor consists of an ideal A/D converter whose samples are sent to a parallel bank of 64 branch metric computation stages. The extracted metric is sent to a 32-state 1041-1135/$25.00 © 2007 IEEE