Comprehensive experimental measurement data are presented comparing the performance of an optical receiver with MLSE-EDC technology against a standard receiver for signals with uncompensated chromatic dispersion. Signals with the NRZ and duobinary modulation formats are investigated. We find that the MLSE-EDC technology provides greater uncompensated reach advantage for both formats as the allowable measured BER is increased, demonstrating the EDC technology has greatest effect and application in conjunction with strong forward error correction. We also measure signal quality vs. dispersion with constant OSNR and draw similar conclusions along with further insights into the application space of the EDC technology for both modulation formats.
©2006 Optical Society of America
Electronic dispersion compensation (EDC) has been given significant attention and interest recently because of its potential to correct for signal distortion in the electrical domain after detection in a photoreceiver. Types of distortion commonly amenable to some degree of correction are chromatic dispersion, polarization mode dispersion (PMD), and electrical filtering [1–10]. The ability to electronically compensate for some of these impairments may allow looser component tolerances and/or system design rules, thereby promoting system cost savings and new network configurations. In particular, uncompensated metro networks may benefit from significantly longer uncompensated reach lengths through the application of receiver-based EDC, allowing longer links with fewer regeneration sites and the increased use of reconfigurable optical add/drop multiplexers (ROADMs).
One particularly promising version of EDC is the Maximum Likelihood Sequence Estimation (MLSE) approach. This nonlinear equalization approach may have performance advantages in general over other algorithms such as feed-forward equalization (FFE) and decision feedback equalization (DFE) for the correction of chromatic dispersion distortions . MLSE is a digital signal processing approach that operates on a sequence of bits rather than on a single bit at a time. It attempts to find the sequence of digital data that is statistically most likely to have generated the detected optical signal. In this work, we experimentally evaluate the performance of a receiver with MLSE-EDC technology with respect to signal distortion caused by chromatic dispersion, and specifically examine the behavior as a function of the underlying signal quality as manifested by the measured bit error ratio (BER). We test the non-return-to-zero (NRZ) modulation format as well as the duobinary format, which offers an inherently greater tolerance to dispersion [11–13]. We find that the receiver with MLSE-EDC offers a greater advantage over a standard receiver for higher BER values, making it especially suitable for use in systems that also employ forward error correction (FEC). The trend is the same for both modulation formats, although the absolute and percentage-wise reach advantage achieved is greater for NRZ. However, the MLSE-EDC receiver also shows some significantly different behavior for the NRZ and duobinary formats, and we determine link distances and system configurations for which either a standard receiver or an MLSE-EDC receiver is advantaged in uncompensated systems with both signal types.
2. Experimental configuration
The experimental system set-up used for all measurements is shown in Fig. 1. A single wavelength at 1531 nm was modulated with either NRZ or duobinary format with a Mach-Zehnder modulator and launched into up to 3 spans of standard single-mode fiber. The spans were generally of length up to approximately 100 km, but were of variable lengths depending on the measurement. Variable optical attenuators (VOAs) controlled the channel power into fiber amplifiers as well as the amount of amplified spontaneous emission (ASE) combined with the signal after propagation through the fiber spans. The channel launch power into each span was kept at about 0 dBm to minimize nonlinear impairments. The channel OSNR (measured with 0.1 nm resolution) at the end of the link was measured and the signal was detected with a photoreceiver. An optical filter with bandwidth 0.6 nm was used to filter out ASE noise prior to detection. The data and clock signals were output to a bit error ratio tester (BERT) for BER measurement. The bit rate was 10.3125 Gb/s and the pseudorandom bit sequence (PRBS) transmitted was of length 231 -1. The bit rate tested was constrained by equipment issues unrelated to the receivers or the EDC technology, but is still close to FEC rates for 10 Gb/s data rates.
Two different commercially available photoreceivers were used in the system experiments for comparison. One receiver (Rx) was a standard photoreceiver with a PIN photodetector, trans-impedance amplifier, and associated clock and data recovery circuitry. The second Rx was from the same manufacturer, but had MLSE-EDC circuitry in the back-end electronics. The MLSE Rx digital equalizer comprises a 3 bit A/D converter operating at up to 25 Gsamples/s and a four-state (memory m=2) Viterbi decoder. Given the general equivalence of the PIN photodetectors used, the two receivers allowed reasonably fair evaluation of the MLSE-EDC technology implemented in the second receiver.
3. Experimental results
In the first set of experiments, the measurement quantity was the required OSNR value of the signal at the receiver in order to achieve a specific BER value. In general, the BER values tested were 10-9, 10-6, 10-4, and 10-3. This was done for various transmission distances over standard single-mode fiber, from 0 km out to the maximum distance measurable. The first modulation format studied was NRZ, and the results comparing the performance of the standard receiver to the MLSE-EDC receiver are shown in Fig. 2. The results clearly show a back-to-back penalty suffered by the MLSE-EDC Rx that is strongly dependent on the measurement BER. For example, for BER=10-9, the back-to-back penalty is over 4 dB. This penalty monotonically decreases with increasing measurement BER, and is less than 1 dB for BER=10-3. We suspect this penalty is likely due to the limited resolution of the A/D converter, and is not necessarily fundamental to the MLSE algorithm. This reasoning may be consistent with the decreasing nature of the penalty as the measurement BER increases. Another observation from the data in Fig. 2 is that the reach advantage offered by the MLSE-EDC is also a function of the measurement BER. In fact, the reach advantage increases with increasing measurement BER, such that the maximum advantage of the MLSE-EDC receiver over the standard receiver occurs for BER=10-3, both in terms of percentage and actual distance advantage.
We next look at the corresponding data for duobinary signals, measured and evaluated in the same manner. The data for the duobinary format signals is shown in Fig. 3. We immediately note that one striking result was the inability to measure a BER value of 10-9 or lower at any distance with the MLSE-EDC receiver. The MLSE-EDC receiver simply did not produce a signal with a BER this low, much less error-free, for any OSNR value up to ~35 dB. On the other hand, the back-to-back penalty of the MLSE-EDC receiver was found to be smaller for the duobinary signals than for NRZ, and as for NRZ, this penalty decreased with increasing measurement BER such that it was basically zero for BER=10-3. However, there was a penalty for the MLSE-EDC receiver for non-zero transmission distances out to roughly 200 km for all measurement BER values. It was only after ~200 km that the MLSE-EDC receiver’s performance and reach advantage over the standard receiver was manifested. As for NRZ, the reach advantage of the MLSE-EDC receiver with duobinary signals generally increased with increasing measurement BER, although the percentage and absolute increases found for this format were significantly smaller than for NRZ. This is consistent with findings from previous research [6, 7] although our experimental results presented here show a somewhat smaller advantage from the MLSE-EDC receiver for duobinary than some previous results. However, the experimental findings depend strongly on the characteristics of the standard receiver, MLSE-EDC receiver, and Mach-Zehnder modulator used in the transmitter.
The data in Figs. 2 and 3 can be summarized in terms of the reach increase afforded by the MLSE-EDC receiver in comparison to the standard receiver as shown in Fig. 4. In this figure, we present the reach increase obtained for both NRZ and duobinary modulation formats as a function of the measurement BER value. The reach increase is defined at an OSNR penalty level of 5 dB in comparison to the back-to-back value of the standard receiver. The advantage is expressed in terms of both percentage increase and absolute distance increase. As discussed above, both formats experience a larger advantage from the MLSE-EDC receiver for higher measurement BER values.
We next compare the NRZ and duobinary formats directly in the format of required OSNR to achieve a BER value of 10-3 as a function of uncompensated transmission distance. These results are presented for both the standard receiver and the MLSE-EDC receiver in Fig. 5. For both receivers, the duobinary format has an initial back-to-back penalty of 2-3 dB, but then offers significantly longer uncompensated reach because of its well-known higher dispersion tolerance. The reach advantage of duobinary over NRZ is about 100% (~120 km) for the standard receiver but is reduced to approximately 45% (~90 km) for the MLSE-EDC receiver, measured at the 5 dB OSNR penalty level.
Finally, it is interesting and instructive to compare the performance of the standard and MLSE-EDC receivers for these modulation formats in a different way. In another set of experiments, we kept the OSNR of the signals constant, and measured the signal BER as a function of uncompensated transmission distance. The signal Q value was then calculated from the BER data. The OSNR used for all measurements was the maximum possible OSNR achievable at the longest distance tested for a given modulation format. This sort of comparison may represent the expected performance of real systems somewhat more closely and can suggest clearly the range of transmission distances for which the MLSE-EDC technology is essential and necessary.
The first format evaluated in this measurement approach is NRZ and the results from these experiments are shown in Fig. 6 as signal Q value vs. transmission link length. The OSNR of the signal for all measurements with both receivers was about 29 dB. We find that the MLSE-EDC receiver provides superior signal quality and performance for all distances and enables transmission out to ~250 km with a Q value of about 11.5 dB or greater. It also shows that the advantage of the MLSE-EDC in comparison to the standard receiver generally increases with transmission distance, or equivalently with increasing BER. This behavior qualitatively agrees with the results found earlier and summarized in Fig. 4.
Similar experiments were conducted with the duobinary signal and the observed results are shown in Fig. 7. The signal OSNR value for all measurements was about 26 dB. The results are significantly different from the NRZ data and show that the standard receiver provides better performance for all distances out to approximately 240 km. The signal was error-free with the standard receiver for distances between 40 km and 225 km, while the MLSE-EDC receiver provided nominally flat signal quality for all distances out to ~250 km with Q values over that range of 14-15 dB. It was only for transmission beyond 240 km that the MLSE-EDC receiver provided better performance but it did give more than 2 dB improvement in the Q value at 270 km.
Lastly, we compare the NRZ format with MLSE-EDC receiver against the duobinary format with the standard receiver. This is of some interest as a means to understand which system approach may provide the greatest benefit to uncompensated transmission in metro/regional networks. In this case, the signal OSNR for all measurements was about 26 dB, and Fig. 8 shows the results for measured Q value as a function of transmission distance. We observe that the duobinary signal with standard receiver has better signal quality for all distances out to the maximum length measured of ~270 km. The Q value at this distance is ~11.5 dB which corresponds to the FEC threshold for standard Reed-Solomon FEC encoding. However, we also see that at this threshold Q value, the actual difference in achievable distance is only on the order of 10–15 km. Therefore, if the system were designed to operate at this minimum Q value, then either approach of duobinary with standard receiver or NRZ with MLSE-EDC receiver would produce approximately the same maximum network link length. In this case, the relative overall transmitter and receiver cost for the two approaches may be the primary factor determining the appropriate system choice.
4. Summary and conclusion
We have experimentally investigated the performance of an MLSE-EDC receiver against a standard receiver for NRZ and duobinary modulation formats with a focus on the behavior as a function of measured BER. We found that for both formats, the MLSE-EDC provides a greater advantage for higher BER values with a general monotonically increasing relationship. The percentage advantage of the EDC is significantly higher for the NRZ signal than the duobinary signal, and the same is true for the absolute reach increase afforded by the EDC technology. We also made measurements for increasing transmission distances with constant signal OSNR and observed that the MLSE-EDC receiver provides significantly improved signal quality for the NRZ signal at all distances, at least all distances with a measurable BER. On the other hand, we found that for the duobinary signal, the standard receiver provided the highest signal quality up to ~240 km. It was only beyond this transmission distance that the MLSE-EDC receiver proved beneficial for the duobinary signal. However, it is also interesting to note that the measured signal quality in terms of BER or Q factor is nearly flat for all transmission link lengths out to approximately 250 km using the duobinary signal and the MLSE-EDC receiver. Finally, a direct comparison of NRZ with MSLE-EDC receiver vs. duobinary with standard receiver shows a fairly clear advantage to the latter system approach for uncompensated metro/regional network links lengths up to the maximum achievable of ~270 km. However, practically speaking there appears to be only a small difference in the maximum reach between the two systems when evaluated at the standard FEC Q threshold of about 11.5 dB.
References and links
1. G. S. Kanter, A. K. Samal, and A. Gandhi, “Electronic dispersion compensation for extended reach,” Optical Fiber Communication Conference (OFC 2004) (Optical Society of America, Washington, D.C., 2004), paper TuG1.
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 Equalisation,” European Conference on Optical Communications (ECOC 2004), Stockholm, Sweden, paper We4.P.106, (2004).
4. M. D. Feuer, S,-Y. Huang, S. L. Woodward, O. Coskun, and M. Boroditsky, “Electronic dispersion compensation for a 10-Gb/s link using a directly modulated laser,” IEEE Photon. Technol. Lett. 15, 1788–1790 (2003). [CrossRef]
5. A. Farbert, S. Langenbach, N. Stojanovic, C. Dorschky, T. Kupfer, C. Schulien, J.-P. Elbers, H. Wernz, H. Griesser, and C. Glingener, “Performance of a 10.7 Gb/s Receiver with digital equaliser using maximum likelihood sequence estimation,” European Conference on Optical Communications (ECOC 2004), Stockholm, Sweden, paper Th4.1.5, (2004).
6. H. Haunstein and R. Urbansky, “Application of Electronic Equalization and Error Correction in Lightwave Systems,” European Conference on Optical Communications (ECOC 2004), Stockholm, Sweden, paper Th.1.5.1, (2004).
7. J.-P. Elbers, H. Wernz, H. Griesser, C. Glingener, A. Faerbert, S. Langenbach, N. Stojanovic, C. Dorschky, T. Kupfer, and C. Schulien, “Measurement of the dispersion tolerance of Optical Duobinary with an MLSE-Receiver at 10.7 Gb/s,” Optical Fiber Communication Conference and Exhibition and The National Fiber Optic Engineers Conference on CD-ROM) (Optical Society of America, Washington, D.C., 2005), paper OThJ4. [PubMed]
8. C. Xia and W. Rosenkranz, “Performance enhancement for Duobinary Modulation through Nonlinear Electrical Equalization,” European Conference on Optical Communications (ECOC 2005), Glasgow, Scotland, paper Tu4.2.31, (2005).
9. A. Faerbert, “Application of Digital Equalization in Optical Transmission Systems,” Optical Fiber Communication Conference and Exhibition and The National Fiber Optic Engineers Conference on CDROM) (Optical Society of America, Washington, D.C., 2006), paper OTuE5. [CrossRef]
10. V. Curri, R. Gaudino, A. Napoli, and A. Nespola, “Advantages of using the Electronic Equalization together with innovative modulation formats in dispersion-limited systems,” 2004 IEEE LEOS Annual Meeting Conference Proceedings, paper ThB1, (2004).
11. A. Price and N. Le Mercier, “Reduced bandwidth optical digital intensity modulation with improved chromatic dispersion tolerance,” Electron. Lett. 31, 58–59 (1995). [CrossRef]
12. S. Kuwano, K. Yonenaga, and K. Iwashita, “10 Gbit/s repeaterless transmission experiment of Optical Duobinary Modulated Signal,” Electron. Lett. 31, 1359–1361 (1995). [CrossRef]
13. D. Penninckx, M. Chbat, L. Pierre, and J.-P. Thiery, “The Phase-Shaped Binary Transmission (PSBT): A new technique to transmit far beyond the chromatic dispersion limit,” IEEE Photon. Technol. Lett. 9, 259–261 (1997). [CrossRef]