680
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 16, NO. 2, FEBRUARY 2004
Filtering Effects in a Spectrum-Sliced WDM System Using SOA-Based Noise Reduction Anoma D. McCoy, Student Member, IEEE, Benn C. Thomsen, Morten Ibsen, and D. J. Richardson
Abstract—We present an experimental investigation into the effects of receiver filtering on the intensity noise in a spectrum-sliced incoherent light system incorporating semiconductor optical amplifier (SOA) based noise reduction. Spectral filtering of the SOA output degrades the signal quality, reducing the benefit offered by the SOA. However, narrow filters are required to reduce the crosstalk in high channel density systems. We characterize this tradeoff and find the optimum receiver bandwidth for varying channel spacing configurations in a spectrum-sliced wavelength-division multiplexed (WDM) system. Index Terms—Optical noise, semiconductor optical amplifier (SOA), spectrum slicing, wavelength division multiplexing (WDM). Fig. 1. Three-channel spectrum sliced WDM system incorporating saturated SOAs for intensity noise reduction. The receiver filter is tuned to channel 1.
I. INTRODUCTION
S
PECTRUM-slicing using incoherent sources has been proposed as an attractive and viable solution for cost-sensitive WDM access networks. Though an economical alternative to laser technology, incoherent sources such as light-emitting diodes and erbium-doped fiber amplifiers (EDFA) are limited in performance by the excess intensity noise inherent in such thermal-like sources [1]. For a given channel width, this intensity noise poses an upper limit on the achievable signal-to-noise ratio and the signal quality can only be improved by reducing the bit rate or increasing the channel bandwidth. A number of intensity noise suppression techniques have been proposed to overcome these limitations and increase achievable system capacity [2]–[4]. One such approach uses the nonlinear gain compression of a saturated semiconductor optical amplifier (SOA) to suppress the intensity noise of the input thermal light [5]. This technique has the added benefit in that, ideally the SOA can also be used for signal modulation and amplification [6]. Reducing the intensity noise enables finer channel granularity and thus higher spectral efficiency. However, the spectral broadening produced by the saturated SOA is more pronounced when narrow, steep input spectral slices are used, and filtering this broadened output at the receiver degrades the signal quality [7]. Therefore, there is a tradeoff between crosstalk and intensity noise that must be considered in the design of high channel density spectrum-sliced systems incorporating SOA-based noise reduction. In this letter, we experimentally characterize this tradeoff and find the optimum channel bandwidth in a 3 2.5 Gb/s spectrum-sliced system.
Manuscript received August 18, 2003; revised September 20, 2003. The authors are with the Optoelectronics Research Centre, University of Southampton, SO17 1BJ Southampton, U.K. (e-mail:
[email protected]). Digital Object Identifier 10.1109/LPT.2003.821077
II. EXPERIMENT Prior to the three-channel system measurements, the effects of receiver filtering were characterized for a single-channel scenario. These single channel measurements demonstrate the impact of filtering on source intensity noise in the absence of crosstalk. In the three-channel configuration, both the crosstalk and the intensity noise are influenced by the receiver/demultiplexer filter bandwidth. Note that with good receiver filter extinction, the crosstalk will be dominated by the two adjacent channels. Thus, in this study we assume that the three-channel performance is sufficient to represent a higher channel count WDM system. The experimental setup for the three-channel system is shown in Fig. 1. The amplified spontaneous emission (ASE) from an EDFA is polarized and spectrally sliced into three channels using 0.24 nm (3-dB bandwidth) fiber Bragg gratings (FBGs). The center wavelength of the subject channel (channel 1) was fixed at 1551.3 nm while the crosstalk channels (channels 2 and 3) were tuned to different wavelengths in order to achieve the desired channel spacings. An EDFA is used after each grating in order to ensure sufficient input power to saturate the SOA. The three channels are then individually modulated with nonreturn-to-zero (NRZ) data at 2.5 Gb/s using LiNbO modulators. The measurements presented here are for a pseudorandom binary sequence (PRBS), however, no penalty was observed with a PRBS. At the receiver, channel 1 is demultiplexed using a bandwidth-tunable FBG (the 0.24-nm bandwidth was obtained by replacing the tunable FBG with a low dispersion grating identical to the input spectral slice). The dispersion introduced by the receiver filter has a negligible impact on system performance at 2.5 Gb/s. A high-speed sampling scope is used to detect the filtered signal and measure the system as a function of receiver filter bandwidth.
1041-1135/04$20.00 © 2004 IEEE
Authorized licensed use limited to: UNIVERSITY OF SOUTHAMPTON. Downloaded on June 24, 2009 at 07:17 from IEEE Xplore. Restrictions apply.
MCCOY et al.: FILTERING EFFECTS IN A SPECTRUM-SLICED WDM SYSTEM
681
2
Q-Factor
4
6
-9
BER of 10
8
BER of 10
No filter 1.3nm filter 0.7nm filter 0.24nm filter No SOA, no filter
10
-15
12 -15
-10
-5
0
Received Power (dBm)
Receiver input spectra with and without SOA-based noise reduction.
III. RESULTS AND DISCUSSION The noise suppression offered by the saturated SOA is the result of small signal gain compression and intrachannel four-wave mixing (IC-FWM) [5], [8]. The spectral broadening due to the IC-FWM has been observed previously and can be significant depending on the input bandwidth and spectral shape [7]. This is illustrated in Fig. 2, which shows the receiver input spectrum for the 0.8-nm (100-GHz) channel spacing configuration, with and without using SOA-based noise reduction. The single channel measurements as a function of received power and receiver filter bandwidth are shown in Fig. 3. The corresponding system output spectra are given in Fig. 4, where the inset shows the receiver filter transfer functions. For receiver filter widths greater than 1.3 nm, the system is detector noise limited at power levels 2 dBm while for narrower receiver bandwidths this limit shifts to lower powers due to the increase in intensity noise. The intensity noise floor of the unfiltered SOA output occurs at . However, using a steep 1.3-nm receiver filter increased this noise floor to . Although the signal degradation is noticeable here, the absolute measurement is low enough to be unnoticed in a routine bit error rate (BER) measurement. However, the narrower receiver filters (0.7 and 0.24 nm) introduce significant system penalty. Note from Fig. 4 that due to the flat filter passband, the 3-dB bandwidth of the system output spectrum does not change appreciably as the receiver filter bandwidth is varied. While is a measure of system performance and is influenced by modulation parameters, RIN is a better measure of the signal quality and directly quantifies the level of intensity noise. Thus, we used RIN measurements to characterize the signal degradation caused by spectral filtering at the receiver. For the purposes of this study we also compare the measured RIN of our system to that calculated numerically for the equivalent thermal light spectrum. Using the analysis presented in [9, eq. (7)], we are able to predict the RIN for an arbitrary thermal light spectrum. The measured RIN for the system without noise suppression (no SOA) is 104.5 dB/Hz, which as expected, closely
Q
20
0
No filtering 1.3nm filter 0.7nm filter 0.24nm filter
10dB/div
For the single channel measurements, channels 2 and 3 were disabled and the system performance was characterized in terms of and relative intensity noise (RIN). For the RIN measurements, the modulator was bypassed to produce a continuous-wave (CW) signal, and the high-speed scope was replaced by a 125-MHz high-sensitivity photodetector and an electrical spectrum analyzer.
Q
Fig. 3. Single channel measurements as a function of power and receiver filtering. decreases with decreasing filter bandwidth (3-dB bandwidth given).
Normalized Spectra (10dB/div)
Fig. 2.
-20
Wavelength (0.5nm/div)
-40
-60
-80 1550.0
1550.5
1551.0
1551.5
1552.0
1552.5
1553.0
Wavelength (nm)
Fig. 4. Detector input spectrum (system output) for single channel only. Inset shows receiver filter transfer functions.
matches the calculated value of 104.8 dB/Hz. However, when SOA-based noise reduction is incorporated, the measured RIN improves significantly (23-dB improvement with no receiver filtering) from that predicted by the thermal light model, indicating a deviation from the Gaussian statistics of incoherent light [10]. The measured RIN and the corresponding thermal light model predictions are shown in Fig. 5. These results clearly show the impact on the noise suppression due to receiver filtering. The observed signal degradation can be understood qualitatively by considering the nonlinear interactions that occur within the saturated SOA [10]. The small signal gain compression is approximately uniform across the amplifier bandwidth [11], and produces an increased correlation between the frequency components of the input light. A further increase in spectral correlation is caused by the IC-FWM that occurs within the device. These yield reduced fluctuations in the output intensity which consists of the superposition of the powers of the various spectral components. Filtering of this signal modifies the spectral profile, reducing the correlation and thus counteracting the noise suppression benefits offered by the SOA. Having examined the effects of receiver filtering on the intensity noise alone, we next characterized the tradeoff between crosstalk and intensity noise for the three-channel system by measuring as a function of receiver filter bandwidth. The system performance is given in Fig. 6 for channel spacings of 0.6, 0.8, and 1.0 nm. Note that the receiver filter we used had good out-of-band extinction (typically 30 dB), thus minimizing the crosstalk from outside of the receiver filter
Authorized licensed use limited to: UNIVERSITY OF SOUTHAMPTON. Downloaded on June 24, 2009 at 07:17 from IEEE Xplore. Restrictions apply.
682
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 16, NO. 2, FEBRUARY 2004
-100
RIN (dB/Hz)
-110
-120
-130
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Receiver filter 3dB bandwidth (nm)
Fig. 5. Single channel RIN as a function of receiver filter bandwidth. The measured system output RIN with SOA noise suppression ( ) and the calculated RIN for the equivalent thermal light spectrum ( ) are shown. Measurements are at 100 MHz and a detector input power of 14 dBm.
0
6
0.6nm 0.8nm
Q-Factor
using SOA-based noise reduction. Spectral broadening produced by the saturated SOA is prominent with steep narrow input spectral slices. Filtering of this broadened output degrades the signal quality, reducing the noise suppression benefits of the SOA. However, filtering is required at the receiver to minimize crosstalk, thus imposing a tradeoff between channel crosstalk and intensity noise. We experimentally determined the optimal receiver bandwidth in a three-channel spectrum-sliced system for three different channel spacing configurations. It was found that the optimum receiver bandwidth is wider than the chosen channel spacings due to the substantial signal degradation caused by spectral filtering of the saturated SOA output. This study clearly demonstrates the need to consider the effects of channel and receiver filter shape/width in the design of high channel count spectrum-sliced systems employing SOA-based noise reduction.
8
ACKNOWLEDGMENT 1.0nm
The authors would like to thank Mr. R. Mokhtar for his help in preparing the tunable-bandwidth grating, and Dr. J.-H. Lee and Mr. T. Kogure for helpful discussions.
10
single channel 12 0.6
0.8
1.0
1.2
1.4
1.6
Receiver filter 3dB bandwidth (nm)
Q
Fig. 6. Three-channel system measurement at 0 dBm for varying filter widths and channel spacings of 0.6, 0.8, and 1.0 nm. Single-channel measurements are shown for comparison.
passband. As is expected, the broader receiver filters allow more crosstalk from the adjacent channels. Also shown in the measurement as a function of graph is the corresponding filter bandwidth for the single channel only scenario, which represents the intensity noise limit for the system at the varying receiver filter bandwidths. From Fig. 6, the optimum channel bandwidth for the 0.6-, 0.8-, and 1.0-nm channel spacings are 0.7, 1.0, and 1.4 nm, respectively. The system is intensity noise limited for bandwidths less than the optimum value, after which it becomes crosstalk limited. In each case, the optimum receiver/demux filter bandwidth is greater than the channel spacing, which is in contrast to the system without SOA noise reduction, where the optimal receiver filter bandwidth is approximately half the channel spacing [12]. In systems using SOA-based noise suppression, receiver filtering results in a more dramatic increase in intensity noise in comparison with the thermal light case. This results in the optimum receiver bandwidth being shifted toward a higher level of adjacent channel overlap. We would like to point out that our experimental setup was not optimized for low cost, as our main objective was to investigate the filtering effects on the noise suppression of the SOA. A more cost-effective system configuration is proposed in [3]. IV. CONCLUSION We have experimentally characterized the filtering effects on the system performance in a spectrum-sliced WDM system
REFERENCES [1] P. R. Morkel, R. I. Laming, and D. N. Payne, “Noise characteristics of high-power doped-fiber superluminescent sources,” Electron. Lett., vol. 26, pp. 96–98, 1990. [2] A. J. Keating and D. D. Sampson, “Reduction of excess intensity noise in spectrum-sliced incoherent light for WDM applications,” J. Lightwave Technol., vol. 15, pp. 53–61, Jan. 1997. [3] F. Koyama, T. Yamatoya, and K. Iga, “Highly gain-saturated GaInAsP/InP SOA modulator for incoherent spectrum-sliced light source,” in Conf. Indium Phosphide and Related Materials, 2000, pp. 439–442. [4] J. H. Han, J. W. Ko, J. S. Lee, and S. Y. Shin, “0.1-nm narrow bandwidth transmission of a 2.5 Gb/s spectrum-sliced incoherent light channel using an all-optical bandwidth expansion technique at the receiver,” IEEE Photon. Technol. Lett., vol. 10, pp. 1501–1503, Oct. 1998. [5] M. Zhao, G. Morthier, and R. Baets, “Analysis and optimization of intensity noise reduction in spectrum-sliced WDM systems using a saturated semiconductor optical amplifier,” IEEE Photon. Technol. Lett., vol. 14, pp. 390–392, 2002. [6] P. Healey et al., “Spectral slicing WDM-PON using wavelength-seeded reflective SOAs,” Electron. Lett., vol. 37, pp. 1181–1182, 2001. [7] A. D. McCoy, B. Thomsen, M. Ibsen, and D. J. Richardson, “Experimental study on receiver filtering effects in a spectrum-sliced incoherent light WDM system using SOA based noise reduction,” in Proc. European Conf. Optical Communication (ECOC 2003), pp. III-866–867. [8] S.-J. Kim, J.-H. Han, J.-S. Lee, and C.-S. Park, “Intensity noise suppression in spectrum-sliced incoherent light communication systems using a gain saturated semiconductor optical amplifier,” IEEE Photon. Technol. Lett., vol. 11, pp. 1042–1044, Aug. 1999. [9] H. S. Kim, R. P. H. Haaksman, T. P. Newson, and D. J. Richardson, “Noise properties and phase resolution of interferometer systems interrogated by narrowband fiber ASE sources,” J. Lightwave Technol., vol. 17, pp. 2327–2335, Nov. 1999. [10] M. Shtaif and G. Eisenstein, “Experimental study of the statistical properties of nonlinearly amplified signals in semiconductor optical amplifiers,” IEEE Photon. Technol. Lett., vol. 9, pp. 904–906, July 1997. [11] L. Kazovsky, S. Benedetto, and A. Willner, Optical Fiber Communication Systems. Norwood, MA: Artech House, 1996. [12] K. Akimoto, J. Kani, M. Teshima, and K. Iwatsuki, “Spectrum-sliced 25-GHz spaced, 155 Mbps 32 channel WDM access,” in Proc. Conf. Lasers and Electro Optics (CLEO 2001).
2
Authorized licensed use limited to: UNIVERSITY OF SOUTHAMPTON. Downloaded on June 24, 2009 at 07:17 from IEEE Xplore. Restrictions apply.
