This paper experimentally investigates the effectiveness of electronic dispersion compensation (EDC) for signals limited by self phase modulation (SPM) and various dispersion levels. The sources considered are low-cost conventional directly modulated lasers (DMLs), fabricated for operation at 2.5 Gb/s but modulated at 10 Gb/s. Performance improvement is achieved by means of electronic feed-forward and decision-feedback equalization (FFE/DFE) at the receiver end. Experimental studies consider both transient and adiabatic chirp dominated DMLs sources. The improvement is evaluated in terms of required optical signal-to-noise ratio (ROSNR) for bit-error-rate (BER) values of 10-3 versus launch power over uncompensated links of standard single mode fiber (SSMF).
©2009 Optical Society of America
Next generation access and consequently metro networks are targeting in the development or deployment of cost effective solutions able to offer high bandwidth connectivity to the end user, wider network coverage and accommodation of a large number of users over a shared infrastructure. However, as certain system parameters (e.g. number of users and transmission distances) increase, the introduced impairments (i.e. additional losses and increased chromatic dispersion) degrade significantly the system’s performance. As a consequence, the upgradability (in bit rate or/and coverage) of these systems is limited.
In the recent past, electronic equalization has been proven to be a powerful and cost effective tool for the correction (to some extend) of certain types of impairments, like chromatic dispersion, polarization-mode dispersion, electronically induced spectral limitations and optical filtering effects [1-6] that have linear characteristics in the optical domain. In particular, for extended reach access and metro networks, the dispersion and bandwidth limitation effects are very significant since uncompensated links and low cost components with reduced performance quality characteristics are common in these networks. Based on the aforementioned concept, the authors have recently demonstrated that low-cost 2.5 Gb/s rated directly modulated laser (DML) sources can be operated at 10 Gb/s by the combination of offset optical filtering with post receiver decision feedback equalization (DFE) .
In order to optimize the performance of the proposed scheme and observe its operating limits, the current work presented in this manuscript focuses on the identification of the optimum launch power levels allowed in the fiber links. Actually, the current work is a study on the capability of the equalizer to correct or not the combined effect of self-phase modulation (SPM) (due to the high launch power in the fiber) and chromatic dispersion (due to fiber propagation) of a chirped optical data signal (produced by the low-rate DML source operated at 10 Gb/s).
Relevant studies on the issue of SPM mitigation with the use of equalization have been performed only for externally modulated laser (EML) transmitters. A preliminary work was presented in  where electronic dispersion compensation (EDC) was proven effective for SPM but ineffective with both SPM and strong cross-phase modulation. However in that work , a maximum likelihood sequence estimation (MLSE) type of equalizer was used, whereas in the study presented here the equalizer type is FFE and DFE. Although MLSE offers better performance compared to FFE/DFE (depending on its memory), the FFE/DFE technique has lower complexity and power consumption than MLSE. Recently, the effectiveness of EDC for the mitigation of SPM for EML systems was reported in  and it was found that the advantage of equalization depends on the amount of uncompensated dispersion. To the best of our knowledge, there are no studies focusing on the mitigation of SPM with the use of adaptive electronic processing at the receiver end for DML-based transmitters (for transient and adiabatic chirp dominated DML) targeting application for extended access or metro networks with uncompensated links.
2. Experimental set up and parameters
The experimental set up for all measurements is shown in Fig. 1. The transmitter consisted of a common DFB laser emitting at 1542.14 nm and rated for 2.5 Gb/s operation. The laser was driven directly at 9.952 Gb/s by the pulse pattern generator (PPG) with a non-return-to zero (NRZ) pseudorandom bit sequence (PRBS) of length 223-1. Depending on the current threshold and the peak-to-peak voltage of the input driving signal, the laser source was operated either as adiabatic or transient chirp dominated DML. Based on chirp measurements that we have performed in other experiments , for the same laser source and under the same driving conditions, the adiabatic and transient chirp behaviour was apparent. The adiabatic chirp dominated laser had a 2.5 GHz estimated peak-to peak chirp (with an adiabatic chirp component around 1.5-1.7 GHz, and a transient chirp component around 0.8-1 GHz) and relaxation oscillation frequency around 6-7 GHz. The transient chirp dominated DML had a chirp value of 8 GHz and relaxation oscillation frequency around 4-5 GHz. Transient chirp dominated lasers achieve higher dynamic ER values (5.5 dB) than adiabatic chirp dominated lasers (2.5 dB), at the expense though of significant amplitude overshoot. The chirp values for the two laser types provided above are important since their interaction of the DML chirp with the fiber dispersion and SPM effect is the dominant source of distortion in the experimental studies presented here.
The SPM effect was introduced in the examined system by varying the optical power launched into the fiber span via a variable optical attenuator (VOA). This effect was examined for both types of laser sources and various fiber lengths of standard type of fiber (SMF-28TM). No optical dispersion compensation was used in the link. Before the receiver, an optical signal to noise ratio (OSNR) emulator was used, consisting of a VOA and an erbium doped fiber amplifier (EDFA) (with a noise figure of ~5 dB), in order to alter the level of OSNR for the measurements. Following this, a JDSU-TB9TM tuneable optical bandpass filter (BPF) with 3 dB bandwidth at 28-GHz and low group delay was inserted before the photodiode.
After the receiver, an integrated electronic dispersion compensation (EDC) circuit was used. The EDC circuit had an FFE part and a DFE which were independently controlled by 5 and 2 taps respectively. The tap spacing in the FFE/DFE module was half the bit period and the tap weights were adjusted within the normalized range of [-1 1] (which refers to a [0, 3.3] voltage levels with a logical zero tap value being approximately 1.65 V). The equalizer had the opportunity to work on FFE mode with 5 taps or on DFE mode with 5 taps in the FFE part and 2 taps in the DFE part. The values of the taps were adjusted through a controller and in terms of optimum BER for each of the measured cases.
The basic experimental procedure involved varying the channel launch power level injected into the fiber span and measuring the required OSNR (ROSNR) value at the receiver for a bit error rate (BER) value of 10-3. Measurements were obtained for different lengths of fiber with and without the use of equalization in order to investigate the role of DFE/FFE in extending the allowable launch power when signals that are created by directly modulated lasers are limited by SPM. It should be noted that the clock and data recovery (CDR) circuitry in the equalizer module operates at around 9.952 GHz. Therefore, it was not possible to use a data rate of 10.664 Gb/s for the inclusion of FEC overhead. However in the results that will be presented in the next section, it is very important to mention that the difference in the performance benefit of the equalizer for the mitigation of SPM will be negligible at 9.952 Gb/s compared to the case of 10.664 Gb/s. In this manuscript, the main focus is to study the benefit of the equalizer and the difference in the performance improvement with respect to the case without equalization. Of course, it is expected that the absolute numbers of OSNR in all cases (either if equalization is used or not) will change slightly since a higher rate will affect the shape of the modulated signal coming out of the DML.
3. Results and discussions
Measurements on ROSNR (for 10-3 BER) versus different signal launch power levels into the link are presented in Fig. 2 and Fig. 4 for the cases of transient and adiabatic chirp dominated DMLs respectively. In these figures, the lower right graph depicts the OSNR improvement that is achieved with DFE for different launch power levels, in comparison to the case without equalization. Moreover, in both cases, measurements have been obtained for three different lengths of fiber links (15 km, 30 km and 50 km) and comparisons are performed among the cases without the use of equalization, with FFE (5) and DFE (5,2). The ROSNR for the back-to-back case is 22.5 dB and 21.7 dB for transient and adiabatic chirp dominated DML respectively, when no equalization is used. The RSONR values for a BER of 10-9 drop to 16.3 dB (for transient DML) and 20.3 dB (for adiabatic DML) with the use of DFE.
For the case of transient chirp dominated DML (see Fig. 2), it is observed that the performance improvement for 15 km and 30 km with the use of DFE ranges between 4 and 5.7 dB. This is almost constant for launch power levels up to 12 dBm (in both transmission distances) and slightly increases for higher power levels. However, for longer distances (50 km) the ROSNR for 10-3 BER varies significantly with respect to the launch power level, when no equalization is used, showing an optimum at around 12 dBm of input power. This is caused by the fact that the transmitted distorted chirped pulses interact with the spectral broadening of the pulse, due to SPM, and as a result the signal distortions smooth out at high launch power conditions, resulting in improved signal quality. In this case, the use of DFE offers significant OSNR improvement for a large range of launch power levels; (with DFE an improvement of 11 dB is observed for 17 dBm of input power). Therefore the ROSNR for 10-3 BER is almost stable (around 18 dB) independently of the launch power levels.
Regarding the adiabatic chirp dominated DML, and when no equalization is used, a similar trend is observed (see Fig. 4) in terms of ROSNR increase with respect to launch power level for different transmission distances. The ROSNR is further increased for longer link lengths due to the interaction of SPM with the fiber dispersion. Similar to the case of transient chirp dominated DMLs, the use of DFE offers a significant improvement in terms of ROSNR, which increases as the launch power level and transmission distance increases and is evident after a launch power level of around 13 dBm; (up to 17dB ROSNR improvement is observed for launch power of 17 dBm and transmission over 50 km of SSMF). With the use of DFE, the ROSNR variations with respect to the launch power level are no more than 1.3 dB for all the studied link lengths. Furthermore, it is shown in Fig. 2 and Fig. 4 that FFE (5), shows similar performance improvement trends as DFE (5,2) but offers moderate improvement. FFE has a simpler structure than DFE and may be equally used in cases where maximum performance improvement is not a strict issue.
The corresponding eye diagrams after the receiver and before the equalizer for the two cases of transient and adiabatic chirp dominated DML are presented in Fig. 3 and Fig. 5 respectively. From these eye diagrams, it is evident that as the launch power level increases, the eye opening decreases due to the SPM nonlinear impairment. However, it was shown that even the highly degraded eye diagrams can be corrected with the use of DFE.
According to the aforementioned results it is observed that the transient chirp dominated DML shows better performance than adiabatic chirp dominated DML at low BER (10-3) and when no equalization is used. This is mainly due to the higher initial ER  that transient chirp dominated DMLs have. It is noteworthy to mention that the spectral position of the receiver’s optical filter with respect to the signal’s central frequency plays an important role in the measured performance. The filtering effect has been studied in detail in  identifying its optimum position with equalization. In order to have a more practical approach and focus only on the performance improvement that the equalizer offers against chromatic dispersion and SPM (i.e. avoiding the filtering effect), the optical filter position was specified for each fiber length in order to achieve optimum performance without equalization and it remained the same when equalization was used. According to the studies performed in  it was shown that the performance of adiabatic chirp dominated lasers in the case where equalization is used can be further improved by tuning the optical filter at the receiver. This improvement can be up to 3 dB of ROSNR for a BER of 10-3 for low values of launch power.
Finally, it is noteworthy to mention that the ROSNR for BER 10-3 was selected in order to present comparative results with and without equalizer and therefore identify the effectiveness of equalization in the mitigation of SPM. As it has been presented in , it is not possible to achieve a BER value of 10-9 for any transmission distance, without the use of equalization. One may refer to , where the ROSNR for different BER levels and different distances is examined and discussed, when equalization is used with the same type of transmitters.
This paper highlights the effectiveness of DFE/FFE equalization for increasing the SPM tolerance in uncompensated systems that utilize low cost DML-based transmitters. In the studies reported in this manuscript, common low-cost 2.5 Gb/s rated DMLs sources were used but modulated at 10 Gb/s. Moreover, DML sources with different chirp characteristics were considered operated in transient and adiabatic chirp dominated mode. For both types of DMLs (transient and adiabatic) it was found that DFE/FFE equalization can significantly increase the system performance when the signal is degraded by SPM and chromatic dispersion impairments. This improvement increases further in cases where the launch power level is very high (greater that 12 dBm) and therefore the SPM effect is more pronounced.
The important finding of this study in terms of system design is that the introduction of equalization at the receiver end allows significantly high transmission power levels (>12dBm) to be launched in the fiber links at the expense of a small only degradation (1-2dB) compared to normal input power levels (i.e. up to 6 dBm) that are commonly used in systems. Therefore, the design tolerances in terms of launch power level are relaxed and consequently the system power budget can be further increased with almost no additional performance degradation. This is particularly important for solutions targeting the extension of passive optical networks in bit rate, number of users and reach. In this case, higher launch power levels allows both the length extension of the uncompensated links in access network and the increase in the number of splitting ports in the distribution coupler/splitter. Furthermore, it should be noted that the specific DML used in this study was not optimized for operation at 10 Gb/s. The system performance can be further improved if DMLs with higher operating bandwidth (and therefore fabrication cost) are used in combination with electronic equalization at the receiver end.
This work has been supported by the European Commission through the 7th ICT-Framework Program under the project SARDANA (217122) and the Network of Excellence BONE-project.
References and links
1. G. S. Kanter, A. K. Samal, and A. Gandhi, “Electronic dispersion compensation for extended reach,” in Proc. Optical Fiber Communication Conf. (OFC 2004), Washington, DC, Paper TuG1, (2004).
2. T. Nielsen and S. Chandrasekhar, “OFC 2004 workshop on optical and electronic mitigation of impairments,” J. Lightwave Technol. 23, 131–142 (2005). [CrossRef]
3. H. Griesser, J.-P. Elbers, C. Fuerst, H. Wernz, and C. Glingener, “Increasing the dispersion tolerance of 10 Gb/s duobinary modulation by electrical distortion equalization,” in Proc. Eur. Conf. Optical Communications (ECOC 2004), Stockholm, Sweden, Paper We.4.P.106, (2004).
4. H. Haunstein and R. Urbansky, “Application of electronic equalization and error correction in lightwave systems,” in Proc. Eur. Conf. Optical Communications (ECOC 2004), Stockholm, Sweden, Paper Th.1.5.1, (2004).
5. J. D. Downie, M. Sauer, and J. Hurley, “Experimental measurements of uncompensated reach increase from MLSE-EDC with regard to measurement BER and modulation format,” Opt. Express 14, 11520–11527 (2006). [CrossRef] [PubMed]
6. M. Rubsamen, P. J. Winzer, and R.-J. Essiambre, “MLSE receivers for narrowband optical filtering,” in Proc. Optical Fiber Communication Conf. (OFC 2006), Washington, DC, Paper OWB6, (2006).
7. S. Chandrasekhar and A. H. Gnauck, “Performance of MLSE receiver in a dispersion-managed multispan experiment at 10.7 Gb/s under nonlinear transmission,” IEEE Photon. Technol. Lett. 18, 2448–2450 (2006). [CrossRef]
8. J. D. Downie, J. Hurley, and M. Sauer, “Behavior of MLSE-EDC With Self-Phase Modulation Limitations and Various Dispersion Levels in 10.7-Gb/s NRZ and Duobinary Signals,” IEEE Photon. Technol. Lett. 19, 1017–1019 (2007). [CrossRef]
9. I. Papagiannakis, D. Klonidis, A.N. Birbas, J. Kikidis, and I. Tomkos, “Performance improvement of low-cost 2.5 Gb/s rated DML sources operated at 10 Gb/s,” IEEE, Photon. Technol. Lett. 20, 1983–1985 (2008). [CrossRef]
10. I. Papagiannakis, D. Klonidis, A.N. Birbas, J. Kikidis, and I. Tomkos, “Transmission performance improvement studies for low-cost 2.5 Gb/s rated DML sources operated at 10 Gb/s,” in Proc. Eur. Conf. Optical Communications (ECOC 2008), Brussels, Belgium, Paper Tu.1.D.5, (2008).