1. Introduction

Future dynamic optical networks will offer increased flexibility and utilization of overall transmission capacity. Optical performance monitoring (OPM) is an indispensable tool for the efficient operation and management of such dynamic optical networks [1]. Chromatic dispersion is a major transmission impairment affecting the overall performance of high-speed fiber-optic networks and hence must be effectively compensated. Particularly, in dynamic networks CD compensations need to be adaptive in nature since each individual wavelength-division-multiplexed (WDM) channel may accumulate different amounts of CD by traversing different lengths of fibers due to network reconfigurablity enabled by optical add-drop multiplexers (OADM) [2]. In addition, CD is also subjected to change with temperature and other physical effects [3]. Therefore, it is imperative to have an efficient in-line CD monitoring technique which could provide essential information for adaptive CD compensation.

Several classes of techniques have been proposed in recent years that are capable of monitoring in-line CD for dynamic optical networks [4]. Unfortunately, most of these techniques are data rate and modulation format dependent. Since the envisaged future optical networks will probably incorporate multiple modulation formats as well as different data rates, there is a considerable interest in the development of monitoring techniques which could accommodate mixed modulation formats at different data rates. Such beneficial feature may significantly reduce the monitoring costs. These requirements have led to a considerable interest in the OPM techniques utilizing statistical properties of the received signal [5]. Two types of techniques, namely asynchronous amplitude histograms (AAH) and asynchronous delay-tap sampling (DTS), are especially popular in this context. Using the technique of AAH, Kozicki et al. [6] and Li et al. [7] have successfully demonstrated CD monitoring in phase-modulated systems. Although, AAH based-techniques have shown the potential for CD monitoring in several modulation formats and at different data rates, the results are significantly affected by OSNR changes as well as PMD effects [6]. Another inherent practical problem in dealing with AAH is that the distributions for different signal amplitudes are severely overlapped due to the absence of timing information in asynchronous sampling. Therefore, the separation of individual distributions for parameters extraction is a challenging task. The situation becomes even more complex in case of DQPSK and higher-order modulation formats. Since the signal amplitude in the overlapped region of the pulses (due to pulse spreading caused by CD) will split into multiple levels (due to constructive and destructive interferences depending upon the phases of the adjacent pulses), it will give rise to more peaks in the overall amplitude histogram of the signal and make distributions separation even more difficult. Asynchronous DTS by providing the distributions of signal amplitude slopes instead of signal amplitude itself (i.e. in case of AAH) substantially reduces distributions overlap problems [8]. However, the CD monitoring results are still degraded by both OSNR variations and PMD effects [9,10]. Another major disadvantage of asynchronous DTS is that the delay value between the taps is a function of symbol rate and hence needs to be precisely adjusted for various data rates. This would require a high precision tunable electrical delay line with large tuning range depending upon the range of data rates being monitored. It has been shown in [9] that a slight mis-adjustment in the tap-delay value may lead to very large monitoring errors. On the other hand, Yu et al. [11] proposed a CD monitoring technique by detecting the relative group delay between the upper and lower VSB signals (induced by CD) through clock recovery and phase-sensitive detection. Nonetheless, clock recovery is complicated and data rate dependent. Consequently, it is not suitable for CD monitoring in networks with multiple data rates.

In this contribution, we propose a simple CD monitoring technique that combines several advantageous features of the aforementioned techniques while avoiding their drawbacks. This technique estimates the relative group delay between the two VSB signals by measuring the differences in the sampled amplitude levels of two sideband signals which are sampled simultaneously but asynchronously. Since the information is obtained from the signals amplitudes without the need for clock recovery, this technique can work at various data rates and is also applicable to multiple RZ modulation formats such as on-off keying (OOK), differential phase-shift keying (DPSK) and differential quadrature phase-shift keying (DQPSK). Unlike methods using AAH and DTS, the averaging of results at various pulse locations in the proposed technique allows the cancellation of noise effects, making it less prone to OSNR changes. Furthermore, in contrast with AAH based-techniques, no separation of overlapped distributions for parameters extraction is needed, which is often quite challenging and may contribute to monitoring inaccuracies. In terms of implementation complexity, the proposed scheme also offers certain advantages. (1) Unlike DTS based-techniques, no adjustment of tap-delay is needed, thus enabling the use of same hardware for multiple data rates; (2) Unlike [11], our technique does not require sophisticated and expensive clock recovery. The CD monitoring ranges of our proposed technique for 10/12.5/20 Gsym/s RZ-OOK, RZ-DPSK and RZ-DQPSK systems are comparable with the techniques presented elsewhere in the literature [6,7,9,10,12,13] with good monitoring sensitivities.

2. Operating principle

The operating principle of proposed CD monitoring technique is shown in Figs. 1 and 2 . Since the two sidebands of the modulated optical signal (carrying the same data information) occupy different wavelength ranges as shown in Fig. 1(a), there will be a relative group delay due to CD between the two sideband signals. Hence, if two filters with their centre frequencies ± Δf from the carrier frequency f _c are used to filter out the two sidebands as shown in Fig. 2, the resulting signals after photodetection will have a relative delay τ with respect to each other as shown in Fig. 1(b) [11]. In our proposed technique, the two sideband signals are sampled simultaneously and due to the relative delay between the VSB signals, the sampled signal amplitudes (x _i, y _i) will be different in general as shown in Fig. 1(b). Since the two signals amplitudes are sampled asynchronously, the sampling time T _sampling between the two sample pairs is not necessarily related to the symbol period T _symbol and can be much larger. Now, if the sample pairs (x _i, y _i) are plotted against each other it will result in a scatter plot shown in Fig. 1(c)–1(f). In the absence of CD, τ = 0 and therefore same signals amplitudes will be sampled and hence the sample pairs (x _i, y _i) will be located along the diagonal D as shown in Fig. 1(c). However, in the presence of CD, τ ≠ 0 and x _i≠ y _i in general. Consequently, the sample pairs (x _i, y _i) will deviate away from the diagonal D and the amount of deviation is related to the CD of the link as shown in Fig. 1(d)–1(f). To monitor CD from the scatter plot, we can calculate the shortest distance d _i of the sample pair (x _i, y _i) to the diagonal D which is given by [14]

 

Fig. 1 (a) Upper and lower VSB filtering and (b) relative group delay between the two sideband signals after photodetection. Plot of sample pairs (x _i, y _i) for a 10 Gbps RZ-DPSK signal with (c) CD = 0 ps/nm; (d) CD = 200 ps/nm; (e) CD = 500 ps/nm and (f) CD = 530 ps/nm. The colour bars show the number of occurrences of sample pairs.

Download Full Size | PDF

 

Fig. 2 Experimental and simulation setup for CD monitoring using sideband optical filtering and subsequent asynchronous amplitude sampling for 10/12.5/20 Gsym/s RZ-OOK, RZ-DPSK and RZ-DQPSK systems.

Download Full Size | PDF

3. Experimental setup and results

Experiments and numerical simulations are performed to demonstrate the validity of the proposed monitoring technique. The experimental setup for the proposed monitoring technique is shown in Fig. 2. The 10/12.5 Gbps RZ-OOK and RZ-DPSK signals with 50% duty cycles are generated and transmitted over a single mode fiber (SMF). An Erbium-doped fiber amplifier (EDFA) is used to add ASE noise to the signal and a variable optical attenuator (VOA) is used to change the OSNR in the range between 20 and 40 dB (0.1 nm noise bandwidth) in order to investigate the OSNR dependency of the proposed technique. The accumulated CD of the link is varied in small steps from 0 ps/nm to + 600 ps/nm by using a CD emulator (comprising of different lengths of fibers). At the monitor a coupler is used to tap part of the optical signal for monitoring. In our experiments, we used a fixed power level (−6 dBm) for monitoring. A DWDM demultiplexer with a channel spacing of 100 GHz and a 3 dB bandwidth of approximately 88 GHz for each channel is used to filter out the two VSB signals. The use of demultiplexer channels as filters ensures symmetrical transfer functions for the two sideband filters. Alternatively, a 3 dB coupler and two identical tunable optical filters can also be used to realize VSB filtering. The signal carrier frequency is located at equal frequency differences from the centre frequencies of the two demultiplexer channels. The transfer functions of the two demultiplexer channels and the spectra of the resulting VSB signals are shown in Fig. 3(a) , 3(b), and 3(c) respectively. The two VSB signals are then detected independently using two photodetectors. The two sideband signals having a relative delay (as shown in Fig. 3(d) and 3(e)) are then simultaneously and asynchronously sampled to collect 100,000 sample pairs which are then used to calibrate F _CD for CD monitoring.Experimental and simulation results for the proposed monitoring technique are shown in Fig. 4 . For 10 Gsym/s, it is clear from Fig. 4(a) and 4(b) that F _CD is sensitive to accumulated CD in the range of 0 to + 600 ps/nm (0 to + 565 ps/nm), 0 to + 525 ps/nm (0 to + 500 ps/nm) and 0 to + 550 ps/nm for RZ-OOK, RZ-DPSK and RZ-DQPSK systems respectively where the numbers in the brackets indicate the ranges observed in experiments. The small discrepancies between the simulation and experimental results are contributed by several undesirable factors such as non-identical response of the two photodetectors and the drift in carrier frequency with time. From our experimental observations, the frequency drift in the range of a few GHz does not cause serious monitoring inaccuracies especially at higher data rates. Some of these factors have been appropriately addressed in the post-processing for e.g. the amplitudes of the two sideband signals are equalized by measuring the non-identical responsivities of the respective photodetectors. To demonstrate the applicability of the proposed technique to different data rates, results for 12.5 Gsym/s RZ-OOK and RZ-DPSK signals are shown in Figure 4(c) and 4(d) which demonstrate a monitoring range of 0 to + 375 ps/nm (0 to + 335 ps/nm) and 0 to + 325 ps/nm (0 to + 315 ps/nm) for RZ-OOK and RZ-DPSK systems respectively. The decrease in monitoring range with an increase in symbol rate is attributed to the decrease in symbol period of the signal. Finally, due to hardware limitation, the validity of the proposed monitoring technique for 20 Gsym/s RZ-OOK, RZ-DPSK and RZ-DQPSK systems is demonstrated through numerical simulations using commercial software Virtual Photonics Inc. (VPI) [15]. A monitoring range of 0 to + 175 ps/nm, 0 to + 140 ps/nm and 0 to + 130 ps/nm is obtained for RZ-OOK, RZ-DQPSK and RZ-DPSK systems respectively as shown in Fig. 4(e). It may also be noticed from Fig. 4 that F _CD changes quasi-linearly with accumulated CD over most of the monitoring range and the saturation mainly occurs near the end thus enabling good sensitivity over the measurement range. From the above results, it is evident that the proposed technique is capable of monitoring accumulated CD for multiple modulation formats and different data rates with reasonable monitoring range and sensitivity.

 

Fig. 3 (a) Measured transfer functions of two demultiplexer channels used for VSB filtering. Optical spectra of received and two VSB-filtered signals for (b) 10 Gbps RZ-DPSK and (c) 12.5 Gbps RZ-OOK systems. CD induced shift for upper and lower VSB signals for (d) 10 Gbps RZ-DPSK and (e) 12.5 Gbps RZ-OOK systems.

Download Full Size | PDF

 

Fig. 4 Dispersion parameter F _CD vs. accumulated CD for (a) 10 Gsym/s (simulation) (b) 10 Gsym/s (experimental) (c) 12.5 Gsym/s (simulation) (d) 12.5 Gsym/s (experimental) and (e) 20 Gsym/s (simulation) using various modulation formats.

Download Full Size | PDF

To analyze the resilience of proposed CD monitoring technique against OSNR variations (due to the averaging of results at various pulse locations), we performed CD monitoring experiments for 10 Gsym/s RZ-OOK and RZ-DPSK systems for OSNR values in the range of 20−40 dB (0.1 nm noise bandwidth). The results are demonstrated in Fig. 5 which clearly show that F _CD is not perturbed by the OSNR variations. Thus, the calculation of mean distance F _CD using (2) effectively averages out the noise contributions as anticipated. Note that the CD monitoring using AAH as well as DTS in conjunction with Hough transform exhibit OSNR dependencies [6,9,10]. Therefore, the proposed technique certainly offers advantage in this regard.

 

Fig. 5 Effect of OSNR variations on the CD monitoring for 10 Gsym/s (a) RZ-DPSK and (b) RZ-OOK systems. The noise resolution bandwidth is 0.1 nm.

Download Full Size | PDF

Finally we investigated the PMD dependence of proposed technique through numerical simulations for 20/40 Gbps RZ-DQPSK systems. The results are shown in Fig. 6 . It is clear from the figure that the CD estimation errors remain relatively small for DGD values till 10 ps and 5 ps for 20 Gbps and 40 Gbps RZ-DQPSK systems respectively while they start increasing afterwards. This is due to the fact that PMD causes deterioration of the pulse shape for both sideband signals resulting in variation of F _CD with DGD. Dependencies on PMD have also been reported in AAH and DTS based CD monitoring techniques [6,9]. Considering typical PMD coefficient value of 0.1 ps/km^1/2, DGD tolerances of 10 ps and 5 ps reflected through our simulations for 20 Gbps and 40 Gbps systems translate into fiber lengths of 10x10³ km and 2.5x10³ km respectively, which are above typical link lengths for reliable data transmission at such data rates. Therefore, we believe that for typical optical networks lengths the performance of proposed monitoring technique will not be significantly affected by PMD.

 

Fig. 6 Effect of DGD on the CD monitoring for (a) 20 Gbps and (b) 40 Gbps RZ-DQPSK systems.

Download Full Size | PDF

4. Discussion

The comparison of proposed technique with some of the existing CD monitoring techniques is summarized in Table 1 . Like the methods using AAH and DTS, the proposed technique successfully demonstrates CD monitoring for various modulation formats at different data rates, without requiring any hardware modification. Such cost-effective feature is not exhibited by the technique utilizing sideband filtering with clock phase-shift detection because the clock recovery circuitry is data rate dependent, expensive and complicated. On the other hand, the tap-delay in DTS based-techniques needs very precise adjustment. The tuning range of tap-delay must also be quite large depending upon the range of data rates being monitored thus adding to the hardware complexity. Our technique avoids this complexity at the cost of an additional asynchronous amplitude sampling port which is data rate as well as modulation format independent. Despite being simple in nature, our technique outperforms AAH and DTS based schemes in decoupling the effect of OSNR. The introduction of averaging feature helps in getting rid of deleterious noise effects thus making CD monitoring resilient to OSNR changes. The processing of samples from two VSB signals for the calculation of F _CD is alsoquite straight forward as it does not rely on extracting parameters from the overlapped distributions unlike methods using AAH. Finally, the samples acquired at the two asynchronous sampling ports in our technique may also potentially be exploited for simultaneous monitoring of other impairments for e.g. OSNR, like in AAH and DTS based-techniques. An additional advantage in this case is that the amplitude samples are acquired in parallel at the two ports which may halve the data acquisition time and henceforth reduce monitoring time. The monitoring ranges demonstrated by the proposed technique are comparable with most of the existing methods. Techniques based on AAH [6,7], DTS [9,10] and clock-tone based methods [12,13] typically exhibit monitoring ranges of 0 to + 600 ps/nm for 10 Gsym/s signals. The technique based on clock recovery and phase-shift detection [11] demonstrates a broader measurement range of ± 70 ps/nm for 40 Gsym/s RZ signal (considering the square dependence of CD on symbol rate, this corresponds to a monitoring range of ± 1120 ps/nm for 10 Gsym/s systems). The monitoring range of the proposed technique can potentially be extended by introducing dispersion offset fibers inside the CD monitoring module. Measuring the dispersion parameters F _CD for the individual branches (incorporating different amounts of dispersion offsets) and by defining appropriate mapping rules, the range can be broadened. Since the proposed technique relies on the detection of optical signal intensity and lacks any phase information, the monitoring curves are expected to be symmetrical for positive and negative dispersions as in the methods using AAH and DTS. If the information about the dispersion sign is needed then this can be obtained by introducing an offset fiber of known CD in the monitoring module. Determining the corresponding increase or decrease in the dispersion parameter F _CD will directly provide information about the sign of accumulated CD being monitored. Finally, the nonlinear effects in long-haul fiber-optic transmission systems may also slightly affect the performance of the proposed technique especially for low accumulated CD values.

Table 1. Comparison of proposed technique with other CD monitoring techniques

View Table

5. Conclusions

In this paper, we proposed and experimentally demonstrated a simple and cost-effective technique for CD monitoring in 10/12.5/20 Gsym/s RZ amplitude and phase-modulated systems through asynchronous sampling and subsequent processing of two VSB signals. This technique enables CD monitoring for several RZ modulation formats at different data rates with good monitoring ranges and sensitivities with simple hardware and signal processing. The OSNR dependence of CD monitoring is minimized through averaging of results at various pulse locations and the DGD tolerances of the proposed monitoring technique are also investigated.