Open Access
Add to BibSonomyAbstract
Abstract
In this paper, the effects of asymmetric narrowband optical filtering are investigated in a 10-Gbit/s optical communication link using integrated electro-absorption modulated lasers (EML). We investigate the effect of EML chirp on link performance as well as the optimal filter bandwidth and wavelength detuning. We show that both the phase response and the spectral narrowing of the filter will enable a longer distance transmission by interacting with the EML transient chirp and compensating for the fiber chromatic dispersion. Experimentally, an 8.75-GHz filter is shown to improve the link distance by 40 km from 65 to 105 km, when transmitting over standard single mode fiber.
©2007 Optical Society of America
1. Introduction
The migration of current 2.5Gb/s networks to 10Gb/s is an opportunity for device manufacturers to innovate and meet the demands of system providers. Key factors to be considered for this upgrade are the flexibility, footprint and of course cost of the optical components being used, as for example the metropolitan network cannot support the cost of 10Gb/s components used for long haul communication systems. Electro-absorption modulated lasers (EMLs) are commonly used at 2.5Gb/s due to their cost effectiveness [1] and their enhanced reach over directly modulated lasers (DML). At 10Gb/s however the current reach of 1550nm EMLs is around 40km which is insufficient to meet the network demands, and this is driving research in extending the reach and improving link performance using low cost methods.
Various methods have been suggested to extend the reach of an EML, such as biasing it with a negative voltage, resulting in the output acquiring negative chirp [2]; or combining this negative bias with running the EML with low extinction ratio to achieve 10Gb/s transmission over 130km of SSMF [3]. The drawback to these approaches are the reduction in optical power caused by the negative biasing, and the poor short reach performance of the low extinction ratio signal. An alternative approach is to generate continuous-phase frequency-shift keying/amplitude-shift keying (CPFSK-ASK) signals by modulating both the bias on the EML in addition to the modulation signal [4]. This technique has demonstrated a reach extension of 25–75% although at the cost of an increase in both added drive complexity and electrical drive power. Finally wide-band fiber Bragg gratings have been used as dispersion compensators to extend the reach of a 10Gb/s EML over a wide range of fiber lengths from 30–110km [5].
Optical filters generally have an amplitude and phase response that interacts with many parts of an optical communication system, such as modulator chirp, channel spectrum, and resistance to fiber chromatic dispersion. Moreover, narrowband optical filters have been shown to extend the transmission distance in directly-modulated laser (DML) based links [6,7] or been used to generate single side band signals [8] but their effect on EML modulated signals with variable transient chirp has not been studied. For DML links, the filter utilizes the adiabatic chirp of the signal to generate duobinary like signals by converting Frequency modulated data into Amplitude modulated data and increase the extinction ratio [7]. In EML modulated data, adiabatic chirp is negligible while the transient chirp can be controlled by the device structure or by changing the driving bias [9]. Applying narrowband optical filtering to an EML is an attractive approach to optimize the reach of a 10 Gb/s transmission systems, as, it is conceivable that simple optical elements, such as filters, will be readily integratable with either the receiver or the transmitter to allow small form-factor enhanced reach 10Gb/s optical transceivers.
In this paper, the impact of detuned, narrowband filtering of a 10-Gb/s EML system is investigated both experimentally and through simulations. Simulations show the effect of EML chirp as well as the optimal filter bandwidth and wavelength detuning on the system performance. We show that both the phase and amplitude response of the filter will enable a longer distance transmission by interacting with the EML transient chirp and compensating for the fiber chromatic dispersion. Results show that the enhancement due to filtering is much more significant when using a narrowband filter than when changing EML chirp alone.
2. Detuned narrowband filtering of EMLs
The concept being discussed here is shown in Fig 1a. A tunable narrowband filter is used to filter the output from a 10Gb/s EML. Compared with directly modulated lasers, where the optical spectrum is broadened by large adiabatic chirp, lasers with integrated electro-absorption (EA) modulators show only transient chirp [9], which causes a frequency shift at the falling and rising edges of the pulses.
The time dependent frequency shift Δν(t) in systems with only transient chirp is described by the equation (1):
where ϕ is the phase, S(t) is the time dependent output optical power and α is the chirp factor [10]. For EMLs the chirp factor is generally set by the device architecture and is voltage dependent [9,11] with values ranging between -2 and 1. During use the EML chirp may be varied somewhat by changing the EMLs bias voltage at the expense of optical output power [9]. For standard single mode fiber a negative chirp factor leads to a decrease of the optical frequency at the rising edge of the pulse and can be used to overcome link degradation due to dispersion and extend the transmission length, while positive chirp reduces the tolerance of the pulse to dispersion and decreases the transmission length.
Fig. 1. (a) Using narrow band detuned filtering of an EML modulated signal (b) Transmission profile and chromatic dispersion of a typical 5G Lorentzian (i.e. FFP) filter.
The amplitude and dispersion response of a Fabry-Perot, Lorentzian, narrowband filter is shown in Fig.1(b). The interaction between the filter and the signal can be categorized into 3 distinct effects.
1) The amplitude response of the detuned filter can reduce the bandwidth of the signal spectrum, increasing its tolerance to dispersion at the expense of signal distortion if the filtering is too extreme.
2) The group delay at the edge of the filter’s passband can produce a high negative dispersion to compensate for dispersion in the link. As can be seen in Fig. 1(b) to optimize the overlap of the optical signal with the peak of the dispersion requires filter detuning. With a decrease in the filter bandwidth, absolute dispersion on either edge of the filter is increased.
3) The amplitude response of the filter can also modify the chirp of the signal by clipping the higher frequency components.
In the following sections these interactions will be investigated. First the effect of narrowband filtering on an unchirped EML will be simulated and the optimal filter bandwidth, optimal detuning frequency, sensitivity to filter stability and the relative impact of phase and amplitude of the filter will be studied. Secondly chirped EMLs will be simulated and the effect on filtering parameters and the optimal filter bandwidth as well as the changing impact of the phase and amplitude responses of the filter will be determined. Finally experimental results using a commercial EML will be presented.
3. Setup used for simulating the effect of narrowband filtering
Link simulations were performed with commercially available software [12]. The sample mode engine implementing the Time Domain Split Step Method was used with a simulation bandwidth of 0.256 THz, bit rate of 10 Gb/s. The simulation setup is shown in Fig 2, a non return to zero 10 Gb/s (NRZ) 223-1 pseudo random bit sequence was used to drive an EML which was comprised of a 1.55µm continuous wave laser cascaded with an Electro Absorption (EA) modulator. The output of the EML had an extinction ratio of 10 dB and passed through two equal spans of standard single mode fiber (SSMF) with 17 ps/nm/km of dispersion, each followed by an erbium doped fiber amplifier (EDFA) with a fixed output power of 0dBm. The length of the fiber spans was varied from 0 to 100 km to simulate links with up to 200 km of transmission distance. The fiber spans were followed by a narrowband Lorentzian Fiber Fabry-Perot filter with variable bandwidth from 3 to 50 GHz. After the filter, an EDFA with a fixed output power of -10dBm was used to negate the effects of the filter loss followed by a 0.5 nm filter used to reduce the amplified spontaneous emission (ASE) noise. Finally an optical receiver with a sensitivity of -19 dBm was used to measure the electrical noise limited power penalty using the semi-analytical BER technique. The power penalty of a link is defined as the additional power needed at the receiver to acquire a BER of 10-9 compared to the 0 km link containing no narrowband filter.
Fig. 2. Simulation setup: For filtering at the transmitter, narrow band optical filters (NBOF) of 3–50 GHz are placed and detuned after the EML transmitter, while for filtering at the receiver, NBOF is placed before the receiver. ASE filter of 50 GHz was centered around the data spectrum and placed before the receiver.
Simulations were performed with the narrowband filter at both the transmitter and the receiver end of the link with similar results.
4. Link simulations of unchirped EML
The following section presents the results for an unchirped EML, alpha=0. Figure 3(a) shows the power penalty as Lorentzian filters with various bandwidths from 3–17 GHz are detuned from the CW laser wavelength for a 120km link. The power penalty for the link with no narrowband filter and only the ASE filter, denoted as No filter for the rest of the paper, at 120 km was simulated to be 3.9dB. It can be seen from this figure that the main benefit of the filter is obtained by detuning the filter to higher frequencies (shorter wavelengths) whereas detuning to lower frequencies cause serious power penalty degradation. This is due to dispersion at the edges of the filter, and will be discussed in more detail in a later section.
From the data shown in Fig. 3(a) the optimum power penalty and detuning tolerance was measured as a function of filter bandwidth, and is shown in Fig. 3(b). The detuning tolerance is defined as the frequency range which generates less than 0.5 dB reduction of the optimum power penalty. In general one can see the trade-off between the improvement in power penalty and the detuning tolerance of the filter, resulting from narrower filters having better system performance. The optimum filter bandwidth for power penalty is 5GHz at 120km with a detuning tolerance of 2GHz, this means that the combined laser and filter wavelength stability has to be better then ± 1GHz which is difficult to achieve with today’s commercially available EMLs which typical have ± 1GHz wavelength stability.
Fig. 3. Simulation results after 120 km of transmission: (a) Power penalty vs. detuning for filters with different bandwidth (b) Optimum Power Penalty vs. filter bandwidth and the detuning tolerance (<0.5 dB penalty drop from optimum penalty) of each filter.
Figure 4(a) shows that the simulated optimal detuning to achieve the optimum power penalty has a linear relationship with filter bandwidth, with additional simulations (not plotted here) showing the optimal detuning does not change with transmission length after the first 40km range where the benefit of the filter phase due to dispersion compensation comes into effect.
Since the optimal detuning location for the filter is off-center, feedback signals may be needed to adjust for the filter or the laser wavelength drift. The easiest feedback signal would be the loss caused by the filter at optimal detuning which is illustrated in Fig. 4(b) for filters with different bandwidths. Obviously the power penalty improvement, the usable bandwidth and filter loss at optimal detuning are three key parameters important for judging the usability of this approach in an optical link.
Fig. 4. (a) Optimal filter detuning for filters with different bandwidths (b) Filter Loss at the optimal detuning location for filters with different bandwidths.
Figure 5(a) shows the power penalty as a function of link length for filters with different bandwidths. The plotted power penalty is the optimum power penalty found for each filter by detuning the filter at that transmission length. As can be seen, filters with narrow bandwidths shows significant back-to-back penalty but improve the link performance at higher transmission lengths where dispersion becomes the main limiting factor. The back-to back power penalty of using narrowband filters is shown at 0km transmission distance, this is due to the amplitude filtering and is discussed in the next section.
The transmission reach is defined as the distance causing a 2dB power penalty, and we define reach extension as the extension of the transmission length beyond that of the unfiltered link. Figure 5(b) plots the reach extension vs. the filter bandwidth. It can be seen from these two Fig.’s that by decreasing the filter bandwidth, the reach extension rises and peaks at 4 GHz. For filters smaller than 4GHz, a sudden fall in reach extension and rise in power penalty is observed. It is noteworthy to observe that even though at 120km, a 5GHz filter is the optimal filter; the 4GHz filter presents the overall highest reach extension of 65 km
Fig. 5. (a) Optimum power penalty vs. transmission length in the presence of filters with different bandwidths (b) Reach extension of the 2dB penalty point for filters with different bandwidths compared with the No Filter case.
Fig. 6. Simulated power penalty for 120km link containing (a) amplitude-only and (b) phase-only filters
To understand the contributions of the amplitude and the phase of the narrowband filter on the link power penalty improvement, two additional ‘virtual’ filters were simulated. One with an amplitude-only response and one with a phase-only response. These results are shown in Fig 6 (a) and (b) for different filter bandwidths for a 120 km link. As seen in Fig. 6(a), the improvement in power penalty due to amplitude filtering or spectral narrowing is minimal with the best improvement (0.8 dB) for the 5 GHz filter, with the 3 GHz filter introducing a significant penalty. As seen in this Fig. the power penalty with respect to filter detuning is almost symmetrical about the laser wavelength. Figure 6(b) shows that the power penalty with a phase only filter is improved more than with an amplitude only filter with the benefit increasing as the bandwidth becomes narrower. Also shown is the fact that only positive detuning has a beneficial effect on the power penalty improvement. The reason behind this is that the higher frequency edge of the filter acts to compensate for the dispersion in the link whereas the lower frequency edge increases the dispersion. Comparing Figs 6 (a) and 6 (b) to Fig 3 (a) it can be seen that the optimal filter bandwidth for the filter with both phase and amplitude response is a compromise between the increasing dispersion compensation due to the phase response of the filter, and the onset of the detrimental affect of filtering in the amplitude domain causing distortion in the spectrum of the signal once the filter bandwidth is below 5 GHz.
5. Interaction between EML chirp and narrowband filtering
Fig. 7. (a) 2dB power penalty limited reach as a function of filter bandwidth for various EML chirp factor (alpha). (b) Reach extension achieved by different filter bandwidths for EMLs with different alpha.
The chirp factor for standard EMLs varies as a function of the applied voltage, however link performance for a chirped EML is predominantly determined by it’s average transient chirp whose value may be tuned somewhat by changing the EML bias voltage [9]. In order to study the effect of detuned narrowband filtering in the presence of EML chirp, links containing EMLs with different fixed chirp factors of -1,-0.5, 0, 0.5 and 1 were simulated.
To clarify the naming convention here a negative chirp enhances the pulse tolerance to dispersion hence extending the reach of the transmission link in comparison with an unchirped pulse. This effect can be seen in Fig. 7(a) where the 2dB power penalty limited reach of a link with narrowband filters of various bandwidths and chirp alpha factors is plotted. At wide filter bandwidths the reach saturates to that of the link with no narrowband filter.
The transmission reach shown in Fig. 7(a) is dictated by two components:
1) The chirp of the EML. As can be seen in Fig 7 (a) at wide filter bandwidths the chirp of the EML has a pronounced effect on the reach of the optical link both with and without narrowband filtering, with the -0.5 chirped EML having the optimum reach for the unfiltered link.
2) The narrowband filtering. The chirp of the EML provides an offset to the effect of the narrowband filtering, and to isolate this effect the reach extension of the filtered links compared to the chirped unfiltered link is plotted in Fig. 7(b). The trend of reach as a function of filter bandwidth is similar for all the chirped EML links considered here and peaks at a narrowband filter bandwidth of 4GHz. Figure 7(b) shows that the reach extension due to narrowband filtering is longest for EMLs with positive chirp. This is due to the amplitude filtering of the detrimental chirp of the EML, conversely for EMLs with negative chirp the amplitude filtering is filtering the beneficial chirp and hence reducing the amount of reach extension.
Figure 7 (a) and (b) show that narrowband filtering is a more effective way of optimizing an EML link at 10Gb/s then varying its chirp factor. The peak reach extension for chirped EMLs range from 50 to 85km, when using the 4GHz filter compared to the range of -50 to 10 km obtained when only varying the EML chirp. Overall the optimized link contains an EML with -0.5 chirp factor and a 4GHz narrowband filter, the optimized reach for this link is 175 km.
Figure 8 shows the simulated results of the reach extension due to a phase-only, amplitude-only and phase-and-amplitude 4 GHz narrowband filter at the optimal wavelength detuning. This figure shows that as speculated before, the amplitude response of the filter interacts with the EML chirp and decreases the reach of links with negative chirp and increases the reach of links with positive chirp, whereas the phase, dispersion compensating, response of the filter benefits both.
6. Experimental setup and results
The experimental setup used to investigate narrowband optical filtering is similar to the simulation setup shown in Fig. 2. A commercial 10 Gb/s EML unit was used, driven by a non-return to zero (NRZ) 223-1 pseudo random bit sequence. After propagating through different lengths of standard single mode fiber and EDFAs for a maximum of 170 Km of fiber, the optical signal was filtered using a narrow-band optical filter. The thermally tunable filters were fiber Fabry-Perot filters with 8.75GHz and 17GHz Lorentzian bandwidth. The data was then passed through another EDFA with a fixed output power and then passed through an ASE filter with 150GHz bandwidth to reduce the ASE noise of the link. An attenuator was placed before a receiver with -19 dB sensitivity and a Clock data recovery unit after the receiver was used to allow the received optical power and bit error rate to be measured. As for the simulation the power penalty is defined as the additional optical power needed at the receiver to acquire a BER of 10-9 compared to the back-to-back, 0 km, case.
The spectrum and time resolved chirp of the output of the EML is shown in Fig. 9. The unfiltered bandwidth of the EML output was measured to be 1.5 GHz with a wavelength stability of ±1GHz. The time resolved chirp is shown in Fig. 9(b) and has a peak value of +4 GHz, to -3.5 GHz. The measured chirp corresponds to an alpha factor at V0 of -0.2 and at V1 of +0.07, where V0 and V1 correspond to the bias on the EML to produce a 0, or 1 at the output; the average alpha factor is -0.06. The EMLs specifications stated the 2 dB power penalty limited reach to be 40 km.
Fig. 9. (a) Optical spectrum with and without an optimized 8.75 GHz filter; (b) Measured EML modulated output and chirp.
The performance sensitivity to the tunable filter’s bandwidth and wavelength detuning relative to the central EML wavelength is shown in Fig. 10(a) for an 80km link. As suggested by the simulation the system performance is optimized when the filter is detuned to higher frequencies then the EML wavelength, and the narrower bandwidth filter has the better system performance improvement obtained at the cost of smaller tolerance to the filter detuning, which requires higher wavelength stability of both the laser and the filter. The 8.75GHz filter has an optimum power penalty of 1.55dB with a detuning tolerance of ±10 pm (total of 2.5 GHz) and the 17GHz filter has an optimum power penalty of 1.9 dB with a detuning tolerance of bandwidth of ±20 pm (total of 5 GHz). These results agree with the simulation results for the optimum detuning frequency and are close to the simulated detuning tolerance.
Fig. 10. (a) Detuned Power penalty vs. filter detuning for an 8.75 GHz and 17 GHz filter after 80 km of SMF fiber (b) Showing the power penalty versus link distance for an unfiltered EML and an EML filtered with a 8.75GHz bandwidth filter. Also shown are the optical eye diagrams obtained at a link distance of 125km.
Figure 10(b) shows the power penalty needed to obtain a BER of 10-9 versus link length for an unfiltered EML and an EML filtered with an 8.75GHz filter compared to the back-to-back link. The unfiltered EML can be seen to have a 2-dB power penalty reach of 65 km, and the filtered EML at optimal detuning, a reach of 105 km. Also shown in the diagram are the optical eye diagrams of the optimally filtered and unfiltered EML. These clearly show the signal improvement obtained using narrowband filtering. The filter was placed at both the receiver and the transmitter end of the link with similar results.
As seen in these Fig.’s, the experimental results qualitatively follow the simulation results with the optimal detuning of the 8.75 GHz and the 17 GHz filters very close to the simulated values. The reach extension of 40 Km for an 8.75 GHz filter, as compared with simulation results in Fig. 7(b) compares well to the simulated EML with chirp values between zero and - 0.5 chirp.
7. Conclusions
Here we discuss narrowband optical filtering and its use in extending the reach of 10 Gb/s EMLs. Simulation and experimental results show that narrowband filtering can enhance the reach of an EML mainly due to the dispersion properties of the filter. Once the filter bandwidth becomes <4 GHz the amplitude affects of the filter act as a detriment to the power penalty. Simulations show that narrowband filtering is a more effective way of optimizing link performance than varying the EMLs chirp. Experimentally we show a 10 Gb/s EML link can be extended with a 8.75GHz filter from 65 km to 105 km while retaining a tolerance to wavelength instability of 2.5 GHz. This tolerance is a key attribute for applications using EMLs as they typically display a detuning tolerance of 2 GHz. Simulations show optimal filter bandwidth would be 5 GHz.
Acknowledgments
Song Shang for help with real time chirp measurements.
References
1. M. Li, P. Wesolowski, L. Nederlof, M. Vaughn, M. Brisker, and R. Wagner, “USA metropolitan network value opportunities,” in Proceedings of Optical Fiber Communications Conference (OFC, 2003) 414–415
2. K. Morito, R. Sahara, and K. Sato, “Penalty-free 10 Gb/s NRZ transmission over 100 km of standard fiber at 1.55 mu m with a blue-chirp modulator integrated DFB laser,” IEEE Photon.Technol. Lett 8,. 431–433 (1996). [CrossRef]
3. Y. K. Park, T. V. Nguyen, and P. A. Morton, “Dispersion-penalty-free transmission over 130-km standard fiber using a 1.55- µm, 10-Gb/s integrated EA/DFB laser with low-extinction ratio and negative chirp,” IEEE Photon.Technol. Lett. 8, 1255–1257 (1996). [CrossRef]
4. H. Kim, S. K. Kim, H. Lee, S. Hwang, and Y. Oh, “A novel way to improve the dispersion-limited transmission distance of electro-absorption modulated lasers,” IEEE Photon. Technol. Let. 18, 947–949 (2006). [CrossRef]
5. B. J. Koch and J. F. Brennan, “Dispersion compensation in an optical communications system with an electroabsorption modulated laser and a fiber grating,” IEEE Photon. Technol. Let. 15, 1633–1635, (2003). [CrossRef]
6. S. Chandrasekhar, A. H. Gnauck, L.L. Buhl, X. Zheng, D. Mahgerefteh, Y. Matsui, K. McCallion, Z. F. Fan, and P. Tayebati, “Single Channel Transmission over 9280 km at 10-Gb/s using Small Form Factor Chirp Managed Laser Generating RZ AMI Modulation Format,” paper Th4.2.5 in post deadline Proceedings of European Conference on Optical Communication (ECOC, 2005).
7. L.-S. Yan, Y. Wang, B. Zhang, C. Yu, J. McGeehan, L. Paraschis, and A. E. Willner, “Reach extension in 10Gb/s directly modulated transmisión systems using asymmetric and narrowband optical filtering,” Opt. Express 13, 5106–5115 (2005). [CrossRef] [PubMed]
8. Q. Cheng, X. Jin, H. Chi, and X. Zhang, “Tunable Fiber Fabry-Perot Filter for Optical Carrier-Suppression and Single-Sideband Modulation in radio over Fiber Links,” Int. J. Infrared Milli Waves 27, 381–390 (2006). [CrossRef]
9. R. A. Salvatore, R. T. Sahara, M. A. Bock, and I. Libenzon, “Electroabsorption Modulated Laser for Long Transmission Spans,” IEEE J. Quantum Electron. 38, 464–476 (2002) [CrossRef]
10. F. Koyama and K. Iga, “Frequency Chirping in External Modulators,” J. Lightwave Technol. 6, 87–93 (1988). [CrossRef]
11. R. Chen and J. C. Cartledge, “Measurement-based model for the modulation properties of an integrated laser modulator and its application to systems with tight optical filtering,” J. Lightwave Technol. 23, 1683–1691 (2005). [CrossRef]
12. Optsim software, Sample mode.
