Optical Fiber Technology

Transcription

1 Optical Fiber Technology 17 (2011) Contents lists available at ScienceDirect Optical Fiber Technology Laser phase noise and OFDM symbol duration effects on the performance of direct-detection based optical OFDM access network Dung Tien Pham a, Moon-Ki Hong a, Jeong-Min Joo a, Eun-Soo Nam b, Sang-Kook Han a, a Department of Electrical and Electronic Engineering, Yonsei University, 134 Shinchon-dong, Seodaemun-gu, Seoul , Republic of Korea b Electronics and Telecommunications Research Institute, 138 Gajeongno, Yuseong-gu, Daejeon , Republic of Korea article info abstract Article history: Received 19 November 2010 Revised 22 February 2011 Available online 25 March 2011 Keywords: Direct-detection Orthogonal frequency division multiplexing Optical access network Phase noise Symbol duration In this paper, we present numerical and experimental demonstration about the impacts of the laser phase noise in direct-detection based optical orthogonal frequency division multiplexing system for access-network transmission. We analyzed the system performance in the presence of the laser phase noise for various orthogonal frequency division multiplexing symbol durations. A significant phase-noise-effect difference between the access-network and long-haul transmission was verified in terms of the orthogonal frequency division multiplexing symbol duration. Ó 2011 Elsevier Inc. All rights reserved. 1. Introduction Applying orthogonal frequency division multiplexing (OFDM) as a modulation technique for the next generation optical access network is a greatly attractive subject in the area of which deals with the rapid increase of the bandwidth demands [1 3]. OFDM signal has high flexibility on various services and dynamic bandwidth allocation, which can increase the efficiency of bandwidth management. Besides, OFDM signal has high spectral efficiency; therefore, low bandwidth and low-cost optical components can be used. It is the critical point since the access networks are very cost-sensitive. For the same reason, direct-detection optical OFDM (DDO-OFDM) receivers with inherently simple and low-cost structure are more preferred than complex coherent optical OFDM (CO- OFDM) receivers. On the other hand, OFDM is highly susceptible to phase noise due to a local oscillator, in particular due to a laser diode in optical communications. It destroys orthogonality among the subcarriers; consequently, the system performance is deteriorated. This phenomenon has been reported in some papers, but all of the reports have focused on the long-haul transmission [4,5]. To the best knowledge of authors, there is no research that deals with the influence of the laser phase noise in DDO-OFDM for the range of the access network. Therefore, it is of great interest to study this subject. Corresponding author. Address: Department of Electrical and Electronic Engineering, Yonsei University, 134, Shinchon-dong, Seodaemun-gu, Seoul , Republic of Korea. Fax: address: skhan@yonsei.ac.kr (S.-K. Han). In this paper, we numerically and experimentally demonstrate the influence of the laser phase noise in DDO-OFDM system for 77-km single mode fiber (SMF) transmission, recognized as the access network. The system performance in the presence of the laser phase noise was analyzed with error vector magnitude (EVM) for the various OFDM symbol durations. Based on these analyses, a significant phase-noise-effect difference in DDO-OFDM system after the access-network range and the long-haul transmission was verified in terms of the OFDM symbol duration. 2. Numerical approach of laser phase noise It is well known that DDO-OFDM system is less sensitive to the laser phase noise compared to CO-OFDM system. In addition, it is completely insensitive to the laser phase noise in back-to-back case [6] due to the inherent beating process of the photo-detector. On the other hand, after the optical transmission, the relative delays between the center optical carrier and the subcarriers in the OFDM data sideband arises due to fiber chromatic dispersion (CD) and the phase noise is generated after photo-detector [5] System model In order to understand the laser phase-noise-effect in DDO- OFDM system, the laser phase noise and the whole system are modeled as follows. The laser phase noise /(t) is characterized by the Wiener-Lévy process with zero mean and variance 2pc t, where c is the two-sided 3-dB linewidth of Lorentzian power spectrum density (PSD) of the laser [7]. Before transmission, the /$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi: /j.yofte

2 D.T. Pham et al. / Optical Fiber Technology 17 (2011) field of the optical OFDM signal with the laser phase noise is given by X 1 2 Nsc E s ¼ e j2pf0t e j/ðtþ þ e j2pðf 0þf RF Þt e j/ðtþ c k e j2pfkt k¼ 1 2 Nscþ1 where f 0, f RF, f k are the frequencies of the center optical carrier, the intermediate frequency (IF) carrier for the up-down conversion, and the kth subcarrier of the OFDM symbol, respectively. N sc is the number of subcarriers. For simplicity, only one OFDM symbol is shown in (1). After transmission through the optical fiber link which has the chromatic dispersion parameter D, the relative delay would be concerned into the received optical OFDM signal. It is approximated as follows, E s ¼ e j2pf 0t e jð/ðtþtcþþu DðcÞÞ þ e j2pðf 0þf RF Þt X 1 2 Nsc c k e j½2pfktþ/ðtþtkþþudðfkþš k¼ 1 2 Nscþ1 where T k, T c are the time-delay differences between the kth subcarrier and the center optical carrier, and between the 3-dB Lorentzian PSD sideband of the laser and the center optical carrier, respectively. They can be expressed as follows, T k ¼ DL k2 c ðf k þ f RF Þ and T c ¼ DL k2 c c where L is the fiber length, k is the laser wavelength, and c is the light speed in vacuum. U D (f k ) is the CD-induced phase shift of the kth optical OFDM subcarrier and represented as follows, ð1þ ð2þ ð3þ ð4þ IðtÞ ¼jE s j 2 ¼ < X 1 þ 2Re e j2pf RF t 2 Nsc = c k e j ½2pfktþUDðfkÞ UDðcÞþ/ðtþTkÞ /ðtþtcþ Š : ; k¼ 1 2 Nscþ1 þ X1 2 Nsc k 1 ¼ 1 2 Nscþ1 X1 2 Nsc c k 2 c k1 e j½2pðfk1 fk2 k 2 ¼ 1 2 Nscþ1 ÞtþU D ðf k1 Þ U D ðf k2 Þþ/ðtþT k1 Þ /ðtþt k2 ÞŠ where Re(x) takes the real part of x, the superscript carries out the complex conjugation. The first term is a DC component that can be easily filtered out. The second term is the desired OFDM signal with the phase noise of the kth subcarrier, {/(t + T k ) /(t + T c )}. It has the mean of zero and the variance of 2pc(T k T c ). The third term is the second-order inter-modulation product that can be easily removed by using a radio frequency (RF) low pass filter. {U D (f k ) U D (c)}is the phase shift on the kth optical OFDM subcarrier after transmission. This term can be compensated after a single-tap equalizer by using preamble. After removing the undesired terms, the second term of the photo-current is sampled and demodulated with the fast Fourier transform (FFT). At this time, the received OFDM signal is distorted due to the phase noise. This effect can be divided into two categories, one is the phase rotation term (PRT) and the other is the interchannel interference (ICI). They were firstly reported for the longhaul application [4]. According to this analysis, the phase noise power of the kth subcarrier can be approximated as b k r 2 PRT;k þ r2 ICI;k / cl and particularly, the power of the PRT is represented as follows, ð6þ ð7þ U D ðf k Þ¼pDL k2 c ðf k þ f RF Þ 2 The photo-detector can be modeled as the square-law detector and the resultant photo-current is represented as follows, ð5þ r 2 PRT;k ¼ 2pcT T k 2pc k ¼ T samp: N SC T samp: N SC where T samp. is the sampling time. DLk 2 f k c! 2 / L2 N SC ð8þ Fig. 1. Relationship between the power of the phase noise, the ICI, and the PRT via transmission length.

3 254 D.T. Pham et al. / Optical Fiber Technology 17 (2011) Numerical analysis In simulation, the binary input was firstly packed according to the modulation format, then parallel-mapped into each subcarrier. For each subcarrier, the allocated symbol was encoded by 4-QAM modulator. The size of FFT was set to 128. Among them, 44 subcarriers were occupied with the encoded data (4-QAM). At this time, zeros were carried into the rest of the subcarriers in order to avoid aliasing products. All the subcarriers were converted to a real-valued time domain waveform, called an OFDM symbol, by using inverse FFT (IFFT). An OFDM frame was constructed as follows; four OFDM symbols were attached in front of every 40 OFDM symbols as the preamble for the timing synchronization, the channel estimation and equalization. After adding the cyclic prefix (CP) which had the ratio of 1/17 refer to the OFDM symbol, the signal was transmitted to optical channel. At the square-law photo-detector, the undesired components of the photo-current were assumed to be totally removed. The transfer function, the phase noise, and the white Gaussian noise were correspondingly added to subcarriers of the desired OFDM signal due to transmission effects. In the OFDM receiver part, according to the preamble, the time synchronization and channel equalization were done firstly then the CP was removed. After that, the signal was OFDM-demodulated by using FFT. Finally, a 4-QAM demodulator made the final decision for each subcarrier and recovered the binary data. We calculated the bit error rate (BER) performance from the measured EVM [8]. For convenience, we have rewritten the relationship between the BER and the EVM for 4-QAM format as follows, BER 4-QAM ¼ 1 2 erfc s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 1 2 EVM 2 According to this equation, the BER of 10 3 is obtained for the EVM of 32.4%. Fig. 1 shows the normalized power of the total phase noise, the ICI, and the PRT as a function of the transmission length for the certain subcarrier index (for example, the first data subcarrier). In this analysis, the laser linewidth was set to 2 MHz. We observed that the PRT power increased sharply with the transmission length. Consequently, with the long-haul transmission, the PRT might dominate the ICI and have a severer impact to system performance. In contrast, when the transmission length is relatively short, the ICI power was much larger than the PRT one; therefore, it might lead to higher system performance deterioration. The phase noise of the certain subcarrier index and the specific transmission length was tracked according to the index of the OFDM symbols (totally, 220 OFDM symbols) as shown in Fig. 2. In this analysis, the laser linewidth was set to 50 MHz. The phase noise of the first and the 22nd data subcarriers after the 77-km transmission are shown in Fig. 2a. The phase noise of the first data subcarrier was about two times larger than that of the 22nd one. It is because the frequency gap between the first data subcarrier and the center optical carrier was twofold wider than the case for the 22nd data subcarrier. The relative delay of the first data subcarrier would be double compared to that of the 22nd data subcarrier. Consequently, the phase noise variance of the former would be twofold larger than that of the latter. Fig. 2b shows the phase noise of the first data subcarrier for the 50-km and 100-km transmission. The phase noise of the 100-km case was about double larger than that of the 50-km case. It is because the double transmission length resulted in the double relative delay; as a result, it led to the double phase noise variance. All of these tendencies were originated from the CD of the fiber. Fig. 3 shows the relationship between the EVM and the FFT size according to the various laser linewidths (10 khz, 30 MHz, and ð9þ 50 MHz) and the various signal-to-noise ratios (SNR) (8.5 db and 10 db) for the 77-km transmission. The pre-defined FFT sizes were 64/128/256/512, which were corresponding to 6.8/13.6/27.2/ 54.4 ns OFDM symbol durations. The numbers of the data subcarriers were set to 22/44/88/176, respectively, in order to occupy the fixed bandwidth of 1.72 GHz. Note that in every curve, the phase noise power was supposed to be fixed with the fixed transmission length and laser linewidth. For the case of the largest FFT size, which means the longest OFDM symbol duration, the received constellation diagrams had the highest EVM values in all curves. Since the ICI dominated the PRT at the short optical transmission, the EVM of the OFDM signal was much more sensitive to ICI rather than PRT. In addition, it is well known that the narrower-subcarrier-spacing OFDM signal is more sensitive to ICI compared to the broader-subcarrier-spacing one. Therefore, the longest OFDM symbol duration, which had the narrowest subcarrier spacing, suffered the largest system performance deterioration due to ICI. It was different from the case of the long-haul transmission length where Fig. 2. Phase noise versus OFDM symbols for (a) various data subcarriers and (b) various transmission length.

4 D.T. Pham et al. / Optical Fiber Technology 17 (2011) Fig. 3. Simulated result of the relationship between the EVM and the FFT size. the longer OFDM symbol duration would yield the better system performance [4]. However, the smaller FFT size could not guarantee the better performance except for the cases of 30 MHz and 50 MHz laser linewidths at the 8.5 db SNR. For example, it was able to get better performance for the FFT size of 256 compared to that of 128 and 64 ones in the case of 10 khz laser linewidth at 10 db SNR. That is because if the FFT size was much smaller than the threshold level, the portion of the PRT power was getting larger and would be dominant in phase noise power. It is because the PRT power is known as inversely proportional to FFT size [4,9]. At this situation, the PRT effect would significantly contribute to the signal degradation. 3. Experimental results and discussion 3.1. Experimental setup The experimental setup is shown in Fig. 4 to validate the above analyses. The OFDM baseband digital signal processing (DSP) was performed offline using MATLAB and it was described in Section 2.2. The digital OFDM signal, the output of the OFDM transmitter, was downloaded into a Tektronix AWG7122B arbitrary waveform generator (AWG) which generated the desired analog baseband OFDM signal with a 10-GSample/s sampling speed and a 8-bit resolution. The inset of Fig. 4 was the RF power spectrum of the OFDM signal right after the AWG. The OFDM signal had a bandwidth of 1.72 GHz, which was allocated from 1.72 GHz to 3.44 GHz. The 44 data subcarriers corresponded to 3.44 GSymbol/s; therefore the raw data rate of 4-QAM-OFDM was 6.88 Gb/s. Since both the preamble and the CP were recognized as overhead of the OFDM signal, the effective data rate was 5.88 Gb/s. This OFDM signal was driven into a Mach Zehnder modulator (MZM) for optical conversion. The MZM was properly biased for the linear operation [10]. An external cavity laser (ECL) (10-kHz linewidth) and a commercial distributed feedback laser diode (DFB-LD) (2-MHz linewidth) were used as the laser sources which had the output wavelengths of 1550 nm and nm, respectively. The laser output power was fixed as 7.5 dbm for both cases. A polarization controller (PC) was used to co-polarize the polarization state of the input optical signal for the MZM. An erbium-doped fiber amplifier (EDFA) and a variable optical attenuator (OA) were installed after the MZM to control the input optical power of fiber not to induce undesired nonlinear distortion during the transmission due to high injection optical power. After that, this signal was transmitted through the standard SMF of 77-km. The transmitted OFDM signal was received by a photo-detector (PD) which had a frequency response up to 3.5 GHz. The post-detected OFDM signal was filtered and amplified by using a band pass filter (BPF) and a low noise amplifier (LNA) to filter out the out of band noise and amplify the received power for the proper signal recognization operated by a Tektronix DPO72004B digital phosphor oscilloscope (DPO). At this time, the recognization was realized with a 100-GSample/s sampling speed. Then, the digital OFDM signal was processed in the OFDM receiver to recover the binary data Experimental results Fig. 5 shows the transmission performance of the system implemented using the ECL for the back-to-back (B-to-B) and the 23-km SMF transmission. The insets are the constellation diagrams of the demodulated OFDM symbols after the equalization. The receiver sensitivities (BER of 10 3 ) of the B-to-B and the 23-km transmission were 15.6 and 14.9 dbm, respectively. The induced power penalty from the 23-km transmission was about 0.7 db. This power penalty was originated from the CD-induced phase-noise-effect. However, it is negligible according to the constellations. It means that the successful equalizations had been realized by the preamble. Fig. 4. Experimental setup for analyzing the laser phase noise.

5 256 D.T. Pham et al. / Optical Fiber Technology 17 (2011) Fig. 5. BER performances for the B-to-B and the 23-km SMF transmission. Insets: constellation diagrams of the OFDM signal for (a) B-to-B and (b) 23 km. Analyses of the subcarrier-dependent performance degradation based on the ECL are illustrated in Fig. 6. Fig. 6a shows the constellation diagrams of various data subcarrier indexes for the 77-km SMF transmission after the equalization. Note that the central subcarrier had the lowest frequency (the 22nd data subcarrier). We observed that the 43rd data subcarrier was more dispersive than the other data subcarriers. It is because the higher-frequency subcarrier had the lower-frequency response of experimental devices and it also had the larger phase noise variance. Fig. 6b represents the relationship between the EVM and the subcarrier index for the given transmission distances (B-to-B and 77-km). It is also able to verify that the central subcarrier index which had the lowest frequency provided the best performance. The transmission penalty increased from 3.03% to 7.75% by increasing the subcarrier frequency because of the larger phase noise variance of the higherfrequency subcarrier. These results were in a good agreement with the characteristics of phase noise in DDO-OFDM systems which were illustrated by numerical results in Fig. 2. Recently, an optical delay fiber was added just before photo-detection for temporally realigning the carrier and the subcarriers to optically pre-compensate the CD [11]. By this way, at certain transmission length and certain delay fiber length higher-frequency subcarriers can be made less vulnerable to phase-noise-effects and have better performance than lower-frequency subcarriers. Finally, we analyzed the system performance in the presence of the laser phase noise after the 77-km SMF transmission for the various OFDM symbol durations. The relationship between the EVM and the FFT size is shown according to the different laser sources Fig. 6. (a) The constellation diagrams for the subcarrier index, (b) The EVM performances according to the subcarrier index. Fig. 7. EVM versus the FFT size with two different laser sources and two different optical received powers.

6 D.T. Pham et al. / Optical Fiber Technology 17 (2011) (the ECL and the DFB-LD) and the different received optical powers ( 10 dbm and 12 dbm) in Fig. 7. We observed that the experimental result shown in Fig. 7 had the same tendency as the numerical result shown in Fig. 3. The received constellation diagrams had the highest EVM values in all curves for the FFT size of 512. That is because the transmission length was relatively short; the power of the ICI was much larger than that of the PRT, the system performance of the OFDM signal was much more sensitive to ICI rather than PRT. Moreover, it is well known that the narrower OFDM subcarrier spacing is, the severer impact ICI results into the OFDM signal. Hence, the longest OFDM symbol duration, which had the narrowest subcarrier spacing, underwent the severest performance deterioration caused by ICI. It was significantly different from the case of the DDO-OFDM long-haul transmission, such as 500 km and 1000 km, where the longer OFDM symbol duration is more attractive for the better phase noise tolerance [4]. On the other hand, the smaller FFT size could not guarantee the better performance except for the case of DFB-LD-based measurement at the 12-dBm received optical power. For example, it was able to get worse performance for the FFT size of 64 than that of 128 one in the case of ECL-based measurement at the 12-dBm received optical power. It can be explained as follows. It is well known that the PRT power is a monotonically decreasing function of the number of subcarrier [4,9]. If the FFT size, corresponding to the subcarrier number, was much smaller than the threshold level, the portion of the PRT power was getting larger and would be dominant in phase noise power. As a result, the ICI effect no longer significantly contributed to the signal degradation. The physical understanding for the inverse relationship between PRT and FFT size is still under the investigation. 4. Conclusion We have investigated, for the first time in our knowledge, numerically and experimentally the influence of the laser phase noise in DDO-OFDM system for the access-network range of 77- km SMF transmission. The system performance in the presence of the laser phase noise was analyzed for various OFDM symbol durations. It was verified that the phase noise term, which could dominate the total phase noise power and contribute to the signal degradation, was significantly different between the access network and the long-haul network based on DDO-OFDM. For the access-network transmission, the longer OFDM symbol duration may not be preferred for the better system performance because of the influence of ICI. Acknowledgments This work was supported by the IT R&D Program of MKE/KEIT (KI002037), Korea and Yonsei University Institute of TMS Information Technology, a Brain Korea 21 program, Korea. References [1] C.W. Chow, C.H. Yeh, C.H. Wang, C.L. Wu, S. Chi, C. Lin, Studies of OFDM signal for broad-band optical access networks, IEEE J. Sel. Areas Commun. 28 (2010) [2] D. Qian, T.T.-O. Kwok, N. Cvijetic, J. Hu, T. Wang, Gb/s real-time OFDM receiver for variable rate WDM-OFDMA-PON transmission, in: Proc. OFC/ NFOEC (2010) PDPD9. [3] N. Cvijetic, D. Qian, J. Hu, 100 Gb/s optical access based on optical orthogonal frequency-division multiplexing, IEEE Commun. Mag. 48 (2010) [4] W.-R. Peng, Analysis of laser phase noise effect in direct-detection optical OFDM transmission, J. Lightw. Technol. 28 (2010) [5] W.-R. Peng, J. Chen, S. Chi, On the phase noise impact in direct-detection optical OFDM transmission, IEEE Photon. Technol. Let. 22 (2010) [6] I.B. Djordjevic, PMD compensation in fiber-optic communication systems with direct detection using LDPC-coded OFDM, Opt. Exp. 15 (2007) [7] L. Tomba, On the effect of Wiener phase noise in OFDM systems, IEEE Trans. Commun. 46 (1998) [8] R.A. Shafik, M.S. Rahman, A.R. Islam, On the extended relationships among EVM, BER and SNR as performance metrics, in: Proc. ICECE, 2006, pp [9] M.S. El-Tanany, Y. Wu, L. Házy, Analytical modeling and simulation of phase noise interference in OFDM-based digital television terrestrial broadcasting systems, IEEE Trans. Broadcast. 47 (2001) [10] B.J.C. Schmidt, A.J. Lowery, J. Armstrong, Experimental demonstrations of electronic dispersion compensation for long-haul transmission using directdetection optical OFDM, J. Lightw. Technol. 26 (2008) [11] Z. Zan, L.B. Y. Du, A.J. Lowery, Experimental demonstration on the reduction of linewidth impact in a self-heterodyne optical OFDM system, in: Proc. OFC/ NFOEC, 2010, JThA8.