IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 16, NO. 6, JUNE 2004
1597
Comparisons on PMD-Compensation Feedback Methods for Bandwidth-Rich Transmission Formats Na Young Kim, Duckey Lee, Jonghan Park, Student Member, and Namkyoo Park, Member, IEEE
Abstract—We compare the performance of polarization-mode dispersion (PMD) compensators under different feedback methods, as a function of the signal spectrum bandwidth. The return-to-zero (RZ), carrier-suppressed RZ, and chirped RZ transmission signals with various spectral bandwidths have been tested with different degree of polarization (DOP) or radio-frequency (RF) spectrum monitoring methods in the feedback circuitry of the compensator. Results show possible limitations of the DOP monitoring method in the PMD compensator particularly for the bandwidth-rich transmission formats, when compared to the conventional RF spectrum feedback method. Index Terms—Optical fiber communications, polarization-mode dispersion (PMD), polarization-mode dispersion compensation (PMDC).
I. INTRODUCTION
A
S THE channel transmission rate increase to 40 Gb/s and beyond, polarization-mode dispersion (PMD) issues became one of the most difficult barriers in expanding the transmission capacity and accelerating the channel speed of the deployed optical links [1]–[3]. In this respect, there have been numerous research activities either to mitigate the PMD effects, or to define its characteristics [4]–[6]. Still, there have been a relatively fewer number of reports on the relation between the transmitted signal and different feedback method for the PMD compensation (PMDC). The most notable investigations on these issues include a recent study by Zweck et al. on the optimum value of the fixed differential group delay (DGD) for different types of feedback signals [7], and a report by Nezam et al. demonstrating DGD monitoring range expansion techniques by using optical filtering under the degree of polarization (DOP) feedback signal [8]. Although these prior arts provide good insights for the nature and its control methods regarding the transmission signal bandwidth/feedback methods, their scopes cover only a partial set of the whole spectrum, and there still exist so far a lack for the comprehensive studies that deals with the efficiencies of feedback signals, under different modulation formats/pulsewidths in an integrated manner. In this letter, we analyzed the efficiencies of the DOP (with and without an optical filter [8]) and radio-frequency (RF) spectral line (a half and a quarter frequency of the bit rate, or their combinations [6], [9]) as feedback signals of the PMDC under different transmission signal formats, in order to Manuscript received September 4, 2003; revised March 2, 2004. The authors are with the Optical Communication Systems Laboratory, School of Electrical Engineering and Computer Sciences, Seoul National University, Seoul 151-744, Korea (e-mail:
[email protected]). Digital Object Identifier 10.1109/LPT.2004.827418
Fig. 1. Schematics of the sample transmission system and the spectrums of the different modulation formats from the transmitter.
provide a proper guideline for the implementation of PMDC. Results show that the RF spectrum in general better serves as a feedback signal than the DOP, especially for bandwidth rich transmission formats. II. SYSTEM CONFIGURATION A schematic of the transmission system analyzed in this study is shown in Fig. 1. Return-to-zero (RZ), carrier-suppressed RZ (CSRZ), and chirped RZ (CRZ) modulation formats having various pulsewidth and phase information has been applied to 40-Gb/s transmitter. An RZ signal has been generated electrically assuming a raised-cosine pulse shape; a CRZ pulse was obtained after the phase modulator using the RZ signal, and the maximum chirp value for this case was 60 GHz. A CSRZ signal also has been generated using the cascaded electrical–optical modulators [10]. Also presented in this figure are the optical power spectra for some of the tested signal pulses. To exclude other sources of impairments (nonlinearty, dispersion, etc.) from the analysis, only the PMD effect has been included in the emulation of transmission link, where a cascade of 20 various lengths of the DGD sections and polarization controllers were assumed. The importance sampling method was applied to PMD simulation for the more precise outage probability calculation [11]. The averaged DGD value and the normalized rotational rate of the principle states of polarization from this emulator were 6.1 and 4.2 ps, respectively. The optically preamplified receiver dB, NF dB) with a Gaussian optical filter (Gain of bandwidth and a second-order Butterworth electrical is the bit rate. The filter has been used in this system, where receiver electrical bandwidth was optimized for each format, for maximizing the back-to-back bit-error rate [(BER) RZ/CRZ and for CSRZ] [12]. For the transmitter, we BER. added a back-to-back power margin of 1 dB at For the PMD compensator, a three-degree-of-freedom (DOF) system composed of a polarization controller and a variable
1041-1135/04$20.00 © 2004 IEEE
1598
Fig. 2.
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 16, NO. 6, JUNE 2004
BER histogram for the CSRZ signal (50 000 realizations).
Fig. 3. BER histogram for 33% CRZ signal (50 000 realizations).
DGD has been used. To avoid the local solution, the global optimum point was obtained from the full scanning process covering all the three-dimensional DOF space. In this optimizing process, different types of feedback signals, a DOP with/without an optical filter (narrow-band symmetric type [8]), and RF spectrums with some set of transfer functions also have been tested [6], [9]. III. RESULTS We first measured the distributions of the BER before and after the PMDC with different PMD feedback signals for the cases of RZ/CRZ/CSRZ pulses at different duty cycles (33% and 66%). As an example, we illustrate here those most representative cases that clearly show the effects of the signal spectral bandwidth (Fig. 2 versus Fig. 3) on the PMDC performance, under different feedback methods (a)–(d). First, comparing figure Fig. 2(a) and 2(b), or 3(a) and 3(b), it was possible to observe the increase in the probabilities of the high BER events (which participate in the outage counts, over in this case) after the simple DOP feedback method—without an optical filter. The further increase of high-BER counts
Fig. 4. Outage probability based on 10 various forms of feedback signals.
BER. Before and after PMDC using
(after the DOP feedback compensation) for narrow pulsewidth spectral-rich transmission signals can also be seen from this analysis [compare Figs. 2(b) over (a) and 3(b) over (a)]. We attribute these phenomena to the result of narrowed DGD monitoring window (in correlation with DOP) for increased signal spectrum bandwidth [8]. More explicitly, if the link PMD value exceeds the narrowed monitoring window range, the PMDC will become to use incorrect DOP values for the compensation, which leads the compensator to the inaccurate operation. To further investigate the effect of different feedback methods in the PMDC, the DOP with narrow-band optical filter [8] and the various forms of the RF spectrum approaches have also been tested [Fig. 2(c) and (d) and Fig. 3(c) and (d)]. For the DOP monitoring method with optical filtering (for the Gaussian filter, 22 GHz/33 GHz for CSRZ, 0 GHz/52 GHz for RZ, and 36 GHz/31 GHz for CRZ were found to be optimum offset frequency/full-width at half-maximum bandwidth, respectively, after the full two-dimensional scanning process), quite an amount of BER improvement from the PMDC was observed [Fig. 2(c) over (b); yet the effect of PMDC with optically filtered DOP decreased for the transmission signals with a broad spectral bandwidth—for example, after the optical BER was reduced to filtering, the outage probability at for the CSRZ signal (bandwidth GHz), but for larger signal stayed at much higher value GHz). bandwidth (33% CRZ, BW The results from the RF spectrum feedback method also can be found in Figs. 2(d) and 3(d). The BER performances in this case showed consistently better performances when compared to the DOP approaches. To compare with the result of the optically filtered DOP approach, the outage probability was and in the case of the CSRZ signal [Fig. 2(d)] and 33% CRZ signal [Fig. 3(d)], respectively. Fig. 4 shows the outage probability as a function of the signal spectral bandwidth under different feedback methods, summarizing our results from various test environments. As can be clearly seen from the figure, the performance of the PMDC has
KIM et al.: COMPARISONS ON PMDC FEEDBACK METHODS FOR BANDWIDTH-RICH TRANSMISSION FORMATS
1599
tion points (especially for bandwidth-rich signals) owing to the optical filtering before the DOP measurement. IV. CONCLUSION We have investigated the performances of a PMD compensator as a function of the transmission signal spectrum bandwidth and the nature of the feedback signals. Results show that the DGD monitoring window expansion achieved from the optically filtered DOP measurement does not apply effectively to the PMDCs for spectrally rich transmission signals. Comparisons show that better performances of the PMDC could be realized using RF monitoring techniques, for most of the transmission signal bandwidth. This result suggests the need for considerations regarding both signal bandwidth and the types of PMDC techniques, in terms of the system complexity/cost in the implementation of the PMDC for future high-speed optical networks. REFERENCES Fig. 5. Correlations (density plot) between instantaneous DGD values in PMD emulator and PMD compensator.
strong correlations both on the source spectral bandwidth and on the choice of the feedback methods. To fully understand these behaviors, we compared the instantaneous DGD values generated in the PMD emulator against the corresponding DGD values used in the PMDC. As an example, we illustrated a sample results from the 50% RZ signal in Fig. 5. Using the DOP value as the feedback signal [Fig. 5(a)], a significant underestimation in the DGD values for the PMDC has been observed, particularly over some threshold (for this case, 60% of the bit duration). In contrast, much better correlation factors have been found between DGD values of the PMD emulator and the compensator for an optically filtered DOP [Fig. 5(b)] or the RF spectral lines [Fig. 5(c) and (d)] approaches. Still, it is worth to noting that a good correlation factor (between the link DGD and magnitude of PMD vector in PMDC) alone does not guarantee optimal compensation of PMD effects. As the instantaneous DGD value determines only the magnitude of the PMD vector, the existence of a higher order PMD will affect the direction of the PMD vector and finally the optimum operating point of the compensator. We attribute the lower performance of optically filtered DOP monitoring approach—when compared to RF spectrum method—to the suppressed higher order terms misleading the PMDC to operate at the nonoptimum compensa-
[1] P. Ciprut et al., “Second-order polarization mode dispersion: Impact on analog and digital transmissions,” J. Lightwave Technol., vol. 16, pp. 757–771, May 1998. [2] H. Kogelnik et al., Optical Fiber Telecommunications IV B. London, U.K.: Academic, 2002, ch. 15. [3] N. Y. Kim et al., “Limitation of PMD compensation due to polarization dependent loss in high speed optical transmission link,” IEEE Photon. Technol. Lett., vol. 14, pp. 104–106, Jan. 2002. [4] H. Sunnerud et al., “A comparison between different PMD compensation techniques,” J. Lightwave Technol., vol. 20, pp. 368–378, Mar. 2002. [5] M. Shtaif et al., “A compensator for the effects of high-order polarization mode dispersion in optical fibers,” IEEE Photon. Technol. Lett.,, vol. 12, pp. 434–436, Apr. 2000. [6] R. Noe et al., “Polarization mode dispersion compensation at 10, 20, and 40 Gb/s with various optical equalizers,” J. Lightwave Technol., vol. 17, pp. 1602–1616, Sept. 1999. [7] J. Zweck et al., “A comparative study of feedback controller sensitivity to all orders of PMD for a fixed DGD compensator,” in Proc. Optical Fiber Communications (OFC 2003), pp. 590–591. [8] S. M. Nezam et al., “Dependence of DOP-based DGD monitors on the optical power spectrum of equal-pulse-width data formats,” in Proc. Optical Fiber Communications (OFC 2003), pp. 112–114. [9] F. Heismann et al., “Automatic compensation of first-order polarization mode dispersion in 10 Gb/s transmission system,” in Proc. Eur. Conf. Optical Communications (ECOC’98), vol. 1, 1998, pp. 529–530. [10] A. Sano et al., “Performance evaluation of prechirped RZ and CS-RZ formats in high-speed transmission systems with dispersion management,” J. Lightwave Technol., vol. 19, pp. 1864–1871, Dec. 2001. [11] G. Biondini et al., “Importance sampling for polarization-mode dispersion,” IEEE Photon. Technol. Lett., vol. 14, pp. 310–312, Mar. 2002. [12] M. Pfennigbauer et al., “Dependence of optically preamplified receiver sensitivity on optical and electrical filter bandwidths-measurement and simulation,” IEEE Photon. Technol. Lett., vol. 14, pp. 831–833, June 2002.
