Optical filtering has been used to extend the reach of directly modulated laser in 10Gb/s WDM systems via two separate mechanisms: narrowing the broadened spectrum, and converting frequency modulation into useful amplitude modulation. We investigate in detail, the impact of asymmetric and narrowband optical filtering at the transmitter or receiver. Experimental demonstrations for both shorter distance and long-haul like transmission using optical filtering are performed. The transmission reach is nearly doubled from <25-km to >45-km without dispersion compensation. 1400-km error-free transmission (Q>15.6-dB) is further achieved over dispersion-managed link for a directly modulated DFB laser within an 8×10-Gb/s WDM system
©2005 Optical Society of America
Direct modulation schemes have recently attracted increased attention in 10 Gb/s WDM transport systems . The simplicity of direct modulation has always offered most cost-effective transmitters compared to external modulation techniques using continuous-wave (CW) laser diodes followed by electro-optic (EO) modulators, or integrated external-modulated lasers (EML) with electro-absorption (EA) modulators. Direct modulation system performance, however, has been limited by the intrinsic chirp of directly modulated lasers (DMLs), and induced spectrum broadening . More specifically, at 10-Gb/s, transmission distances are limited to tens of kilometers over single-mode fiber (SMF) without dispersion compensation.
In order to enhance the achievable distance, a number of approaches have been investigated at different data rates, from 2.5 to 40-Gb/s, over different distances, and up to several hundred kilometers [3–6]. Such approaches have included (i) modification of the DML via incorporating other components (e.g., gratings), or reshaping the corresponding driving current [7–9] (ii) mid-span spectral inversion (over 200 km) , (iii) electrical equalization in the receiver (20-km) , (iv) a 0.2-nm filter at high driving current (38.5-km) , (v) the deployment of negative dispersion fiber (100~320 km) [13–14] (vi) dispersion-supported-transmission (DST) (~250 km) , (vii) interferometric noise reduction through intrabit frequency evolution or optical interferometer configuration [16–17]. However, most these approaches have typically required significant changes in embedded fiber links, or replacement of deployed transmitters or receivers. In this sense, methods that enhances system performance with a simply modification is highly desirable. Electronic equalization has been considered as one such promising approach, but with rather limited yet performance enhancements . In this paper, we investigate the value of such a method based on optical filtering of only one endpoint of a WDM directly modulated link.
In externally modulated systems, optical filtering has been widely used in high performance communication systems to narrow the signal spectrum and increase the spectral efficiency up to 160% [19–20]. Likewise, in directly modulated systems, depending on the filter characteristics, optical filtering can improve the performance via two major mechanisms: narrowing the broadened spectrum and converting chirp-induced frequency modulation (FM) to useful amplitude modulation (AM) for transmission [21–23].
In this paper, we experimentally investigate the impact of asymmetric optical filtering in a 10-Gb/s directly modulated system using a commercial DML with a wavelength ~1.55-µm. By using an optical filter with a Gaussian amplitude profile and a 3-dB bandwidth of 0.3-nm, experimental results show nearly double the reach from <25 km to >45 km without dispersion compensation. Furthermore, we demonstrate error-free transmission (Q >15.6 dB) at 10-Gb/s up to 1,400 km when using the DML. We use dispersion management along the link and incorporate asymmetric optical filtering at the receiver to achieve our result. The DML is inserted into an 8×10-Gb/s WDM system. The maximum transmission distances for the DML are ~1400 km, 1100 km and 580 km for a link with residual dispersion values of -0.54 ps/nm/km, 0 ps/nm/km and +0.60 ps/nm/km, respectively.
2. Asymmetric narrowband filtering
In directly modulated systems, chirp is caused by the change in the refractive index induced by carrier injection during modulation, as well as the power-dependent photon intensity distribution along the laser cavity . The time dependent frequency deviation Δν(t) is
where ϕ is the phase, P(t) is the output optical power, α is the linewidth enhancement factor and κ is the adiabatic chirp coefficient. The first part corresponds to the dynamic chirp caused by pulse transitions, while the second part κP(t) is the adiabatic chirp term related to the frequency difference between “1”s and “0”s. Time evolution of wavelength change or frequency chirping may broaden the optical spectrum and degrade signal quality quickly along the fiber link. Moreover, the broadened spectrum is asymmetric with the complex field as 
Here P 0(t) is determined by the input modulation sequence, and the phase ϕ(t) is expressed as an integral of the instantaneous frequency ν(t). Thus the obtained spectrum density can be perceived as a convolution of the data AM spectrum and a related chirp induced FM term.
When an optical filter is applied, as shown in Fig. 1, two mechanisms come into play:
(i). Compared to external modulation, there are many unwanted frequency components in the broadened spectrum that results from direct modulation. While strong filtering has been used in external modulation systems to increase the spectral efficiency, an optical filter with a bandwidth narrower than the broadened spectrum in directly modulated systems may remove these frequency components, resulting in a “cleaner” signal that may transmit a longer distance. The chirp-induced spectrum broadening is not symmetric, thus the filter should not be placed at the center of the optical spectrum, i.e., asymmetric filtering is needed.
In addition, optical filtering can increase the extinction ratio by cutting off some low driving current (during modulation) induced frequency components, therefore the “0”s in data sequences will be further reduced to the zero level although the “1”s remain the same. Better extinction ratio can improve the transmission performance due to higher immunity to the optical noise, especially in long-haul transmission systems.
(ii). Fiber grating filters have been shown to improve the frequency response of a DML , as well as reduce the intensity noise of the DML , and have also been applied to reduce the chirp of semiconductor optical amplifiers (SOAs) . The change of relative intensity noise (RIN) of the laser can be expressed as 
where T′=dT/d ω|ω0 is the slope of the grating transmission versus frequency at the center frequency ω0, Ω0 is the laser resonant frequency. A grating whose transmission T increases linearly with ω will compensate an increased intensity with a decreased transmission. Numerical analysis shows that more than 5-dB reduction of RIN is possible using the appropriate grating filter. In terms of frequency response, the optical field envelope exiting the laser as 
where m is the AM modulation index, and β is the phase modulation index, Ω is the modulation frequency, and θFM is the phase by which the laser FM leads the AM. For a DML,
From Eq. (4) and the corresponding Fourier transform Ẽin ω, the output optical field from the fiber grating combination can be taken as
where t(ω) is the complex field transmission function of the grating, β” is the fiber dispersion parameter, and L is the fiber length. It was found that : (a) The AM system response strongly depends on the detuning of the optical frequency from the grating center frequency, and (b) In the presence of a certain distance of transmission fiber, larger and flatter system response is expected, as well as higher frequency at which the first dip occurs.
In addition, when the grating filter has pre-designed chirp or dispersion, it will interact with the laser chirp (or even the dispersion along the fiber link) and convert the frequency modulation into useful amplitude modulation and add to the transmitted signal, resulting in better performance. Note since most of the chirp or dispersion of an optical filter is introduced by the nonlinear phase response on the sloping edge, the filter should also be detuned from the center wavelength of the carrier, which is a second reason for asymmetric filtering. For example, a DML with positive chirp may require the center frequency or wavelength to be located at the negative edge of the filter.
3. Short-distance transmission without dispersion compensation
In the experiments, we choose an optical filter with the Gaussian transmission profile shown in Fig. 2(a). The filter is a thin film based filter with a 3-dB bandwidth of ~0.25 nm (~30-GHz). The filter has a certain amount of chromatic dispersion across the passband on the order of tens of ps/nm. The frequency response curve of the DML used in our experiments is measured and shown in Fig. 1(b), the driving current is set to ~40 mA.
First we investigate transmission over SMF without dispersion compensation. The experimental setup is shown in Fig. 3. The DML is directly modulated at 9.95328-Gb/s (223-1 PRBS) with an amplitude of 2 V. The output power from the DML is ~5 dBm, with an extinction ratio of ~9 dB when driven with ~40 mA at a ~-0.9 V bias. Different lengths of SMF comprise the transmission link. For the case of optical filtering at the transmitter, the narrow-band optical filter (NBOF) is used after the DML; for the case of filtering at the receiver, the filter is placed after the transmission fiber. The receiver consists of an optical attenuator, a pre-amplifier (EDFA2) and another wideband optical filter OF (>1-nm). The attenuator is used to keep the input power into the pre-amplifier constant, thus the optical signal-to-noise-ratio (OSNR) will not affect the Q-measurement results. OF is used to reduce the ASE noise from optical amplifiers. The input power into the PIN receiver is fixed at ~-3 dBm. Q measurements are performed at the receiver using decision threshold adjustment. In the back-to-back condition, we find that through optical filtering, the Q factor can be improved by more than 2 dB at different bias voltages, as shown in Fig. 4(a). In addition, typical eye diagrams are inserted in the setup for comparison. This improvement is mainly due to the extinction ratio improvement using optical filter since the optical filter can cut off low current induced frequencies as discussed in the theory section. Figure 4(b) and (c) show the eye diagrams for comparison.
By changing the link length, we compare the performance improvement using optical filtering. Figure 5(a) shows the experimental results in terms of the transmission distance versus Q factor using optical filtering at the transmitter (after the DML). Two bias conditions are compared here: -0.9 V and -1.4 V. The maximum transmission distances (with a Q-factor of ~15.6 dB used as the threshold for maximum transmission distance) are almost doubled using filtering under these two bias voltages (i.e. from ~20-km to >40-km with -0.9V bias and from ~25-km to >45-km with -1.4V bias). In addition, we can see that without filtering, although the Q factor at -1.4V bias is lower than the one at -0.9V for the back-to-back condition, the -1.4V bias case results in a longer transmission distance than the one at -0.9V: the -0.9V bias case suffers a much more rapid Q-factor degradation as the transmission distances increases. This is the case both with and without optical filtering. To show the similarities between filtering at the transmitter and receiver, we compare the two in Fig. 5(b) with an eye diagram (filtering at the transmitter case) after transmission through 45-km of SMF. Although the achievable transmission distance is very close to each other, the trends of Q-variation at different distances are different between two cases, partially due to the nonlinear chirp effect when filtering at the transmitter (i.e., the Q-factor increases first and then decreases), the FM to AM conversion improves the performance through a certain distance of fiber. In order to further evaluate this effect, we use a well-accepted software package to perform simulations and compare to our experimental results, especially for the case of filtering at the transmitter. As shown in Fig. 5(c), a Gaussian-profile filter is used. The chirp parameter of the DML is set to be 4. We detune filters with different bandwidths until we reach an optimum position and see that the transmission distance after filtering is significantly improved compared to the case without filtering (although the best Q-factor is decreased). Within a certain distance (typically ~40–50 kilometers), the performance in terms of Q-factor remains constant, resulting in a much wider transmission window for the DML. Since the sloping edge of the Gaussian filter is still not sharp enough, we apply a 3rd- order super-Gaussian filter into the simulation for comparison, as shown in Fig. 5(d). Note that super-Gaussian filter has already been used in high capacity transmission systems to increase the spectral efficiency . The sharper slope induced by the super-Gaussian filter provides a stronger narrowing effect, thus slightly better performance may be achieved. Similar to our experimental results, the Q-factor also increases a little first and then decreases although the Q values are less than the experimental results since a high chirp parameter (4.0) is used in the simulation.
More importantly, the position of the filter on the chirped spectrum has a significant effect on the Q factor, i.e., the filter slope should match the chirp of the laser, as shown in Fig. 6, where we can see that asymmetric filtering (detuning from the spectrum peak by ~0.15-nm) can provide better performance, while centering the filter at the point of maximum output power is not an optimum position, as shown by the two eye diagrams inserted. This effect is also simulated as shown in Fig. 6(b). We investigate the sensitivity of system performance to the filter position, as practical implementation of the proposed solution call for high robustness to the position of the narrowband optical filter, which would not require an expensive stabilization manufactuting process. Figure 6(b) shows that the sensitivity of the Q to the detuning of a Gaussian filter for a transmission distance of 35 kilometers is a relatively flat for a frequency window of more than 4 GHz, even for the narrowest Gaussian filter. This requirement can be met by most of the electrically controlled filters.
These transmission results, without dispersion compensation, confirm the effectiveness of applying asymmetric narrowband optical filtering into directly modulated systems, and motivate us to find the feasibility of this approach in a “long-haul” transmission system, where both optical amplification and dispersion compensation are extensively employed.
4. Long-haul transmission
The performance of direct modulation in “long-haul” transmission is equally attractive. In order to test this, we use a recirculating fiber-loop testbed to evaluate the long-distance transmission performance of DMLs. Figure 7 shows the experimental setup of the recirculating loop testbed. The same DML with a wavelength of ~1554.7 nm is used. We place the DML into an 8-channel WDM system with 100-GHz (0.8 nm) spacing between channels. The other seven channels are externally modulated using LiNbO3 electro-optic modulators, and range from ~1552.2 to ~1557.8 nm. All channels are modulated at 9.95328 Gb/s with 223-1 PRBS, and are then decorrelated using a spool of fiber. The dispersion-managed recirculating loop consists of ~80-km SMF and ~12-km DCF (dispersion of -1348 ps/nm at 1550 nm). The input power per channel into the SMF and DCF is fixed at 1.0 dBm and -4.0 dBm, respectively. We insert different short spools of SMF (0, ~2.5, and ~5.5-km) into the loop to vary the residual dispersion value - 0, ~2.5 and ~5.5-km SMF correspond to ~-0.54 ps/nm/km, ~0 ps/nm/km and ~+0.60 ps/nm/km residual dispersion (Dres), respectively. A long-period-grating (LPG) is used to avoid noise accumulation and perform gain equalization along the link and an attenuator is used to balance the optical power without changing the OSNR. The OSNR after ~1400-km transmission is ~21 dB. The receiver consists of a pre-amplifier, the same filter as used in short distance transmission, and a PIN photodiode. Similarly, due to the intrinsic positive chirp of the DML, the carrier center is located at the negative sloping edge of the filter.
We measure the Q-factor for various transmission distances and determine the effects of asymmetric (sideband) or symmetric (center) filtering. Figure 8 shows our results after ~940-km transmission. Figure 8(a) shows the output optical spectrum and the eye diagram when using center filtering (no offset from the carrier), while Fig. 8(b) shows the output optical spectrum and the eye diagram when using asymmetric filtering at the receiver. In our experiment, center filtering cannot maintain error free transmission after only ~300-km transmission even in the negative residual dispersion case.
The overall transmission performance in terms of Q factor is shown in Fig. 9. As a typical one, Fig. 9(a) shows the Q-factor values for all the eight channels after ~1100-km transmission with ~0-ps/nm/km residual dispersion. The difference between the Q factor of the best externally modulated channel and the DML channel is ~0.8-dB, and all channels are error-free (Q > 15.6 dB). We believe that the major cause of Q degradation is OSNR degradation. Figure 9(b) shows the Q-factor for the DML channel vs. transmission distance for varying values of residual dispersion - maximum error-free transmission distances are ~1400 km, ~1100 km, and ~580 km for residual link dispersion values of -0.54 ps/nm/km, 0 ps/nm/km, and +0.60 ps/nm/km, respectively. Similar to the simulation results and short-distance experimental demonstration using the filter at the receiver, the degradation of Q follows one direction (i.e. no fluctuations) towards the worst case. In addition, due to the positive chirp of our DML, the negative residual dispersion (-0.54 ps/nm/km) results in the best transmission performance.
In conclusion, we investigated the impact of asymmetric narrowband optical filtering at the transmitter or receiver in a 10-Gb/s directly modulated WDM system, significantly improving system performance. Optical filtering can narrow the broadened spectrum resulting from direct modulation. Moreover, it enables a additional chirp management mechanism by FM to AM conversion based on the interaction between the nonlinear phase response of the sloping edge of the optical filter, the laser chirp, and the link chromatic dispersion. Experimental results show the effectiveness of this approach in doubling the reach from <25 km to >45 km without dispersion compensation. In addition, error-free transmission (Q> 15.6 dB) at 10-Gb/s up to 1,400 km using a single DML within an 8×10-Gb/s WDM system is also demonstrated. We note that the current approach of a narrowband optical filter integrated with 10 Gb/s DML holds promise for cost reduced WDM transport particularly in multi-service metropolitan networks.
The authors acknowledge support from Cisco Systems Inc.
References and Links
1. H. J. Thiele, L. E. Nelson, and S. K. Das, “Capacity-enhanced coarse WDM transmission using 10 Gbit/s sources and DWDM overlay,” Electron. Lett. 39, 1264–1266 (2003). [CrossRef]
2. A. Zadok, H. Shalom, M. Tur, W. D. Cornwell, and I. Andonovic, “Spectral shift and broadening of DFB lasers under direct modulation,” IEEE Photon. Technol. Lett. 10, 1709–1711 (1998). [CrossRef]
3. K. Sato, “Semiconductor light sources for 40-Gb/s transmission systems,” J. Lightwave Technol. 20, 2035–2043 (2002). [CrossRef]
4. A. Wonfor, R. V. Renty, I. H. White, J. K. White, A. E. Kelly, and C. Tombling, “Uncooled 40 Gb/s transmission over 40 km single mode fiber using multi-level modulation of a highly linear laser,” in Proc. OFC, MF60, Atlanta, GA, 2003.
5. K. Soto, S. Kuwahara, A. Hirano, M. Yoneyama, and Y. Miyamoto, “4×40 Gbit/s dense WDM transmission over 40-km SMF using directly modulated DFB lasers,” in Proc. ECOC, We1.5.7, Stockholm, Sweden, 2004.
6. B. Wedding and W. Poehlmann, “43 Gbit/s transmission over 40.5 km SMF without optical amplifier using a directly modulated laser diode,” in Proc. ECOC, We2.6.6, Stockholm, Sweden, 2004.
7. L. Illing and M. B. Kennel, “Shaping current waveforms for direct modulation of semiconductor lasers,” IEEE J. Quantum Electron. 40, 445–452 (2004). [CrossRef]
8. K. Nakahara, T. Tsuchiya, T. Kitatani, K. Shinoda, T. Kikawa, F. Hamano, S. Fujisaki, T Taniguchi, E. Nomoto, M. Sawada, and T. Yuasa, “12.5-Gb/s Direct modulation up to 115°C in 1.3-µm InGaAlAs-MQW RWG DFB lasers with notch-free grating structure,” J. Lightwave Technol. 22, 159–165 (2004). [CrossRef]
9. D. Mahgerefteh, A. M. Benzoni, P. S. Westbrook, K. S. Feder, P. I. Reyes, P. Steinvurzel, B. J. Eggleton, R. G. Ernst, L. A. Reith, and D. M. Gill, “DMRZ: a directly modulated 10-Gb/s RZ source for ultralong-haul WDM systems,” IEEE Photon. Technol. Lett. 14, 546–548 (2002). [CrossRef]
10. M. C. Tatham, X. Cu, L. D. Westbrook, G. Sherlock, and D. M. Spirit, “Transmission of 10 Gbit/s directly modulated DFB signals over 200-km standard fiber using mid-span spectral inversion,” Electron. Lett. 30, 1335–1336 (1994). [CrossRef]
11. 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]
12. P. A. Morton, G. E. Shtengel, L. D. Tzeng, R. D. Yadvish, T. Tanbun-Ek, and R. A. Morgan, “38.5 km error free transmission at 10 Gbit/s in standard fiber using a low chirp, spectrally filtered, directly modulated 1.55 µm DFB laser,” Electron Lett. 33, 310–311 (1997). [CrossRef]
13. I. Tomkos, B. Hallock, I. Roudas, R. Hesse, A. Boskovic, J. Nakano, and R. Vodhanel, “10-Gb/s transmission of 1.55-µm directly modulated signal over 100 km of negative dispersion fiber,” IEEE Photon. Technol. Lett. 13, 735–737 (2001). [CrossRef]
14. H. S. Chung, Y. G. Jang, and Y. C. Chung, “Directly modulated 10-Gb/s signal transmission over 320 km of negative dispersion fiber for regional metro network,” IEEE Photon. Technol. Lett. 15, 1306–1308 (2003). [CrossRef]
15. B. Wedding, B. Franz, and B. Junginger, “10-Gb/s optical transmission up to 253 km via standard single-mode fiber using the method of dispersion-supported transmission,” J. Lightwave Technol. 12, 1720–1727 (1994). [CrossRef]
16. C. S. Wong and H. K. Tsang, “Improvement of directly modulated diode-laser pulse using an optical delay interferometer,” IEEE Photon. Technol. Lett. 16, 632–634 (2004). [CrossRef]
17. G. Yabre, “Improved direct-modulation characteristics of a semiconductor laser by FM/IM conversion through an interferometer,” J. Lightwave Technol. 14, 2135–2140 (1996). [CrossRef]
18. T. Nielsen and S. Chandresekhar, “OFC 2004 Workshop on Optical and Electronic Mitigation of Impairments”, J. Lightwave Technol. 23, 131–142 (2005). [CrossRef]
19. A. Agarwal, S. Banerjee, D. F. Grosz, A. P. Kung, D. N. Maywar, A. Gurevich, and T. H. Wood, “Ultra-high-capacity long-haul 40-Gb/s WDM transmission with 0.8-b/s/Hz spectral efficiency by means of strong optical filtering,” IEEE Photon. Technol. Lett. 15, 470–472 (2003). [CrossRef]
20. N. Yoshikane and I. Morita, “160% spectrally-efficient 5.12-Tb/s (64×85.4 Gb/s RZ DQPSK) transmission without polarization demultiplexing,” in Proc. ECOC, Postdeadline Th4.4.3, Stockholm, Sweden, 2004.
21. M. McAdama, E. Peral, D. Proenzano, W. K. Marshall, and A. Yariv, “Improved laser modulation response by frequency modulation to amplitude modulation conversion in transmission through a fiber grating,” Appl. Phys. Lett. , 71, 879–881 (1997). [CrossRef]
22. H.-Y. Yu, D. Mahgerefteh, P. S. Cho, and J. Goldhar, “Improved transmission of chirped signals from semiconductor optical devices by pulse reshaping using a fiber Bragg grating filter,” J. Lightwave Technol. 17, 898–903 (1999). [CrossRef]
23. M. McAdama, D. Proenzano, E. Peral, W. K. Marshall, and A. Yariv, “Effect of transmission through fiber gratings on semiconductor laser intensity noise,” Appl. Phys. Lett. 71, 3341–3343 (1997). [CrossRef]
24. T. L. Koch and R. A. Linke, “Effect of nonlinear gain reduction on semiconductor laser wavelength chirping,” Appl. Phys. Lett. 48, 613–615 (1986). [CrossRef]