We propose and demonstrate a simple composite second-order (CSO) cancellation technique based on the digital signal processing (DSP) for the radio-over-fiber (RoF) transmission system implemented by using directly modulated lasers (DMLs). When the RoF transmission system is implemented by using DMLs, its performance could be limited by the CSO distortions caused by the interplay between the DML’s chirp and fiber’s chromatic dispersion. We present the theoretical analysis of these nonlinear distortions and show that they can be suppressed at the receiver by using a simple DSP. To verify the effectiveness of the proposed technique, we demonstrate the transmission of twenty-four 100-MHz filtered orthogonal frequency-division multiplexing (f-OFDM) signals in 64 quadrature amplitude modulation (QAM) format over 20 km of the standard single-mode fiber (SSMF). The results show that, by using the proposed technique, we can suppress the CSO distortion components by >10 dB and achieve the error-vector magnitude performance better than 6% even after the 20-km long SSMF transmission.
© 2017 Optical Society of America
It is anticipated that the 5th generation (5G) mobile communication systems would be able to provide a cell capacity of >10 Gb/s eventually [1, 2]. To achieve such a large capacity, the radio access technology based on massive multiple-input multiple-output (MIMO) antennas would be employed extensively at every cell site. In the cloud radio access network (C-RAN) architecture, this would require several hundred Gb/s of transmission capacity between the baseband units and remote radio heads when the mobile fronthaul network is implemented by using the conventional digital fiber-optic interface such as Common Public Radio Interface (CPRI) [3–5].
The radio-over-fiber (RoF) technology has recently regained considerable interests as an alternative transmission technology for the C-RAN-based mobile fronthaul network due to its inherent capabilities of providing high spectral efficiency and simplifying the cell sites [6–11]. For example, if we need to provide a cell capacity of 14 Gb/s by using 8 × 8 MIMO antennas, 3 sectors, and 100-MHz bandwidth of wireless signals formatted in 64-quadrature amplitude modulation (QAM) orthogonal frequency-division multiplexing (OFDM), a mobile fronthaul network using the RoF technology would require a bandwidth of only 2.4 GHz.
A directly modulated laser (DML) is considered as the most promising optical transmitter for the RoF-based mobile fronthaul network due to its excellent linearity, high output power, and cost-effectiveness [6, 7, 9, 11]. However, the use of DML in RoF transmission systems poses a couple of technical problems, including the dispersion-induced RF power fading and the nonlinear distortions caused by the interplay between the DML’s frequency chirp and fiber’s chromatic dispersion [11–13]. This is because the direct current modulation of a laser diode always accompanies frequency chirping, which, in turn, causes spectral broadening. In conjunction with the fiber’s chromatic dispersion, this spectral broadening exacerbates the dispersion-induced RF power fading. However, the deleterious effect of the RF power fading is not so serious in mobile fronthaul networks if the maximum transmission distance is <20 km and the frequencies of the radio signals driving the DML are located <3 GHz [11, 14]. On the other hand, when the RoF system is used to transport broadband signals covering multiple octaves, its performance could be limited by the dispersion-induced composite second-order (CSO) distortions [12, 13]. This is because the frequency modulation (FM) components of DML are converted to the intensity modulation (IM) components during the transmission through optical fiber due to its chromatic dispersion, which results in the CSO distortions. If we ignore the intrinsic nonlinearity of DML, the relative RF power of these dispersion-induced CSO distortions with respect to the signal’s RF power can be expressed as 15]. Thus, the CSO distortions increase rapidly with the bandwidth of RoF system and affect the high-frequency signals. For example, it has been recently shown by experiments that these CSO distortions severely degrade the error-vector magnitude (EVM) performance of the RoF signals located at >1 GHz .
In this paper, we propose and demonstrate a simple CSO cancellation technique for the high-capacity RoF transmission systems. The basic concept of the CSO cancellation was proposed in 1990s for the amplitude-modulated vestigial sideband (AM-VSB) signals used in the cable television (CATV) system [17, 18]. At that time, an analog signal processing unit, implemented by using a differentiator, a mixer, and delay lines, was utilized at the transmitter for the pre-compensation of the CSO distortions. However, it is extremely difficult to realize such an analog circuitry operating over several GHz of bandwidth with high accuracy. In addition, this pre-compensation technique requires a cumbersome calibration procedure, in which (1) the amount of the CSO distortions is first estimated, (2) a test tone is then launched into the system, and (3) the gains and delays of the analog processing unit are adjusted precisely to generate the pre-determined amount of the CSO distortions. To avoid these problems, we propose to utilize a CSO cancellation technique based on the digital signal processing (DSP) at the receiver. In contrast to the conventional analog technique, this DSP-based technique does not require such a calibration procedure and can facilitate the broadband operation. In addition, this technique can be implemented cost-effectively by using the same DSP already used at the receiver for other functions. We evaluate the performances of the proposed technique by using twenty-four 100-MHz filtered OFDM (f-OFDM) signals. These spectrally localized signals are one of the waveform candidates for 5G mobile communication systems [1, 10, 19, 20]. By using the proposed DSP-based CSO cancellation technique, we successfully demonstrate the transmission of the 64-QAM f-OFDM signals over 20-km long standard single-mode fiber (SSMF), satisfying the 8% EVM requirement of the 3rd generation partnership project (3GPP) specifications .
2. Theoretical analysis
A theoretical expression of the dispersion-induced CSO distortion component can be obtained by the time-domain analysis [12, 13]. If we ignore the intrinsic nonlinearity of DML, its optical output power can be described as 12]12]Eq. (7) can be implemented by using a DSP block at the receiver. In this case, the differential operator can be replaced with the subtraction operation when the sampling rate is sufficiently high. Thus, the CSO cancellation in the DSP can be expressed as
3. Experimental setup
We evaluated the performance of the proposed DSP-based CSO cancellation technique by using the experimental setup shown in Fig. 1. A two-port arbitrary waveform generator (AWG) was used for the generation of twenty-four 100-MHz f-OFDM signals. First, we created an OFDM signal by taking 2048-inverse fast Fourier transform of 1000 data in 64-QAM format, 200 pilots in quadrature-phase-shift-keying (QPSK) format, and 848 zeros. Thus, the subcarrier spacing was 83.3 kHz ( = 100 MHz/1200) and the symbol duration was 12 μs ( = 1/83.3 kHz). Then, a 0.8-μs cyclic prefix (CP) was appended to the OFDM symbol. The OFDM signal was sent to a finite impulse response (FIR) filter using Hamming window for the generation of the f-OFDM signal. The length and bandwidth of the FIR filter were set to be 1024 and 100.5 MHz, respectively. The odd-indexed twelve f-OFDM signals were generated from port 1 of the AWG, while the other even-indexed twelve signals from port 2. It should be noted that the even-indexed signals were half-symbol delayed with respect to the odd-indexed ones for the asynchronization between them. The twenty-four f-OFDM signals, positioned at 350 + 101 × (i-1) MHz, where i is the channel index, were finally combined by using a resistive power combiner, as shown in the inset. A butterfly-packaged DML was used in this experiment. Its optical output power and operating wavelength were 9.2 dBm and 1551 nm, respectively, at a bias current of 65 mA. The slope efficiency and FM efficiency of this DML were measured to be 0.172 W/A and 0.15 GHz/mA, respectively. After the transmission over SSMF, the signal was detected by using a PIN-TIA receiver. We set the optical power incident on the receiver to be −2 dBm regardless of the transmission distance. The responsivity and trans-impedance of this receiver were 0.8 A/W and 500 Ω, repectively. The detected signal was boosted by an electrical amplifier (gain = 18 dB) and then digitized by using a digital sampling oscilloscope (DSO) at 40 GSample/s. The CSO cancellation and demodulation of the f-OFDM signals were carried out by off-line DSP. The schematic diagram of the proposed DSP-based CSO cancellation technique is also shown in Fig. 1. The digitized received signal was first normalized, and then splitted into two paths. The upper path represents the second term in the right-hand side of Eq. (8), while the lower path is the signal tainted with the CSO distortions. Thus, we could remove the CSO distortions by subtracting the upper-path signal from the lower one. It also showed that the proposed technique could be implemented simply by using four real-number multiplications (including three constant multiplications), two real-number additions, and two memories. After the CSO cancellation, we down-converted the frequencies of the signals, sent the signal to the 1024-tap FIR mached filter, demodulated the f-OFDM signals, and finally measured the EVM performances.
Figure 2 shows the RF spectra of the received signals after setting the RF power applied to the DML to be 8 dBm (i.e., −5.8 dBm per channel). In the back-to-back condition, there were no noticeable intermodulation distortions in the RF spectrum, as shown in Fig. 2(a). Thus, we achieved a SNR better than 27 dB, which was limited by the shot and thermal noises of the receiver. However, after the 20-km long SSMF transmission, the CSO components caused by the interplay between the DML’s chirp and fiber chromatic dispersion were clearly observed outside of the signal band, as shown in Fig. 2(b). These components certainly degraded the SNR and EVM performance, as will be described later. Figure 2(c) shows the RF spectrum of the signal obtained by applying the proposed DSP-based CSO cancellation technique at the receiver. In comparison with Fig. 2(b), the CSO components were suppressed by >10 dB outside of the signal band, and, as a result, the performance of the RoF transmission system became mostly limited by the receiver noises.
Figure 3 shows the measured EVM performance. In the back-to-back condition, the EVM performance was slightly degraded as the channel frequency increased. For example, we achieved an EVM of 3.7% for channel 1 (@350 MHz), while it was measured to be 4.5% for channel 24 (@2.673 GHz). This was due to the electrical back-to-back performance: the high-frequency signal was more sensitive to the jitter performance of AWG. After the 20-km long SSMF transmission, the system performance was degraded by the dispersion-induced CSO distortions, especially for the high-index channels, as predicted in Eq. (1). As a result, these channels (i.e., channel index higher than 17) failed to satisfy the EVM requirement of the 3GPP specifications . However, when we applied the proposed CSO cancellation technique, the EVM performance was drastically improved. For example, the EVM performance of channel 24 was improved from 9.3% to 5.8%. A slight discrepancy in the performances between the back-to-back and after the 20-km long SSMF transmission was attributed to the dispersion-induced RF power fading, which caused the reduction of signal power by 1.6 dB at 2.6 GHz.
We also verified the effectiveness of the proposed DSP-based CSO cancellation technique by measuring the constellation. Figure 4 shows the constellation diagrams of channel 23 (@2.572 GHz) measured at the receiver. Comparing Figs. 4(a) and 4(b), we could clearly observe the impairments caused by the dispersion-induced CSO distortions after the 20-km long SSMF transmission. Note that we set the received power to be constant regardless of fiber lengths and the reduction of signal power caused by the dispersion-induced RF power fading was <2 dB at this channel. However, when we applied the proposed CSO cancellation technique, the contellation diagram was improved markedly even after the transmission over 20 km of SSMF.
Figure 5 shows the EVM performance of channel 23 measured as a function of the RF signal power per channel. In the back-to-back condition, the system performance was limited by the receiver noise when the signal power was low (e.g., <-15 dBm). Thus, the EVM performance was improved as the signal power increased. However, when the signal power was larger than −15 dBm, the clipping distortions of DML started to degrade the EVM. In our experiment, the best EVM performance achieved in the back-to-back condition was 4.9% (when the signal power per channel was set to be −5.8 dBm. We could not increase the signal power beyond this level due to the limited saturation power of the DML driver. After the 20-km long SSMF transmission, the EVM performance was slightly degraded by the dispersion-induced RF power fading (regardless of the RF signal power). When the RF signal power was larger than −15 dBm and the CSO cancellation technique was not applied, the dispersion-induced CSO distortions started to deteriorate the EVM and limit the performance of channel 23. Thus, the best EVM performance we could achieve after the transmission was 8.2% (at the RF signal power of −9.8 dBm), which was worse than the value in the 3GPP specifications. However, when we applied the proposed CSO cancellation technique, the EVM performance was improved drastically even when we set the RF signal power to be as high as −5.8 dBm.
For the DSP implementation of the proposed CSO cancellation technique described in Eq. (8), the differential operation can be replaced with the subtraction operation if the sampling interval is sufficiently short. Figure 6 shows the EVM performance of channel 23 measured as a function of the sampling rate when the RF signal power per channel is −5.8 dBm and the transmission distance is 20 km. We assumed that the receiver bandwidth was as wide as the signal band and the analog-to-digital converter (ADC) placed in front of the DSP satisfied the Nyquist sampling theorem. Thus, we first applied a 2.7-GHz ideal rectangular filter to the signals captured in the experiment and varied the sampling rate from 6 to 40 Gsample/s by using the sampling-rate conversion. The result showed that as the sampling rate decreased, the EVM performance became deteriorated because of the errors between the differentiation and subtraction operations. Nevertheless, as long as the sampling rate of the DSP block was higher than the Nyquist sampling rate, we could achieve the performance improvement by using the proposed CSO cancellation technique. The result also showed that the sampling rate of the DSP should be twice higher than the Nyquist sampling rate. This could be readily realized by adding a 2x upsampling at the DSP without increasing the sampling rate of the ADC.
We have proposed and demonstrated a CSO cancellation technique for the RoF transmission system implemented by using DMLs. The proposed technique can suppress the CSO distortions caused by the interplay between the DML’s chirp and fiber chromatic dispersion by using a simple DSP at the receiver. Thus, it can be readily applicable to the high-capacity RoF transmission systems utilizing broadband signals (i.e.., beyond several GHz) and is capable of the adaptive cancellation of CSO distortions without the cumbersome calibration procedure. Using the proposed technique, we successfully demonstrate the transmission of the twenty-four 64-QAM f-OFDM signals generated by using a 1.55-μm DML over 20-km long SSMF. This was equivalent to 163 Gb/s of capacity if the signals were transported by using CPRI. We believe that the proposed technique could help the practical development of the broadband RoF transmission systems implemented by using DMLs.
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (2015R1A2A1A05001868).
References and links
1. ITU-R Rec. M.2083–0, “IMT vision frame work and overall objectives of the future development of IMT for 2020 and beyond,” Sept. 2015.
2. C. Hoymann, D. Astely, M. Stattin, G. Wiksröm, J. Cheng, A. Höglund, M. Frenne, R. Blasco, J. Huschke, and F. Gunnarsson, “LTE release 14 outlook,” IEEE Commun. Mag. 54(6), 44–49 (2016). [CrossRef]
3. P. Chanclou, A. Pizzinat, F. Le Clech, T.-L. Reedeker, Y. Lagadec, F. Saliou, B. Le Guyader, L. Guillo, Q. Deniel, S. Gosselin, S. D. Le, T. Diallo, R. Brenot, F. Lelarge, L. Marazzi, P. Parolari, M. Martinelli, S. O’Dull, S. A. Gebrewold, D. Hillerkuss, J. Leuthold, G. Gavioli, and P. Galli, “Optical fiber solution for mobile fronthaul to achieve cloud radio access network,” in Proceedings of Future Network Mobile Summit (IEEE, 2013), pp. 46–56.
4. K. Tanaka and A. Agata, “Next-generation optical access networks for C-RAN,” in Proceedings of Optical Fiber Communication Conference (OFC, 2015), paper Tu2E.1. [CrossRef]
5. S. H. Bae, H. K. Shim, U. H. Hong, H. Kim, A. Agata, K. Tanaka, M. Suzuki, and Y. C. Chung, “25-Gb/s TDM optical link using EMLs for mobile fronthaul network of LTE-A system,” IEEE Photonics Technol. Lett. 27(17), 1825–1828 (2015). [CrossRef]
6. S. H. Cho, H. Park, H. S. Chung, K. H. Doo, S. Lee, and J. H. Lee, “Cost-effective next generation mobile fronthaul architecture with multi-IF carrier transmission scheme,” in Proceedings of Optical Fiber Communication Conference (OFC, 2014), paper Tu2B.6. [CrossRef]
7. B. G. Kim, H. Kim, and Y. C. Chung, “Impact of multipath interference in the performance of RoF-based mobile fronthaul network implemented by using DML,” J. Lightwave Technol. 35(2), 145–151 (2017). [CrossRef]
8. J. Zhang, M. Xu, J. Wing, F. Lu, L. Cheng, H. Cho, K. Ying, J. Yu, and G. K. Chang, “Full-duplex quasi-gapless carrier aggregation using FBMC in centralized radio-over-fiber heterogeneous networks,” J. Lightwave Technol. 35(4), 989–996 (2017). [CrossRef]
9. H. Zeng and X. Liu, “Demonstration of a real-time FPGA-based CPRI-compatible efficient mobile fronthaul transceiver supporting 53 Gb/s CPRI-equivalent data rate using 2.5-GHz-class optics,” in Proceedings of European Conference and Exhibition on Optical Communication (ECOC, 2016), paper W.1.E.1.
10. T. Pham, A. Kanno, N. Yamamoto, and T. Kawanishi, “190-Gb/s CPRI-equivalent rate fiber-wireless mobile fronthaul for simultaneous transmission of LTE-A and f-OFDM signals,” in Proceedings of European Conference and Exhibition on Optical Communication (ECOC, 2016), paper W.4.P1.SC7.71.
11. B. G. Kim, S. H. Bae, H. Kim, and Y. C. Chung, “Optical fronthaul technologies for next-generation mobile communications,” in Proceedings of International Conference on Transport Optical Network (ICTON, 2016), paper We.D2.5.
12. E. Bergmann, C. Kuo, and S. Huang, “Dispersion-induced composite second-order distortion at 1.5 μm,” IEEE Photonics Technol. Lett. 3(1), 59–61 (1991). [CrossRef]
13. D. Crosby and G. Lampard, “Dispersion-induced limit on the range of octave confined optical SCM transmission systems,” IEEE Photonics Technol. Lett. 6(8), 1043–1045 (1994). [CrossRef]
14. “White paper of next generation fronthaul interface,” China Mobile Research Institute, Alcatel-Lucent, Nokia Networks, ZTE Corporation, Broadcom Corporation, Intel China Research Center, Tech. Rep., Oct. 2015.
15. W. Way, General Technical Background: Modulation Signal Format, Coaxial Cable Systems, and Network Architecture Evolutions (Academic, 1999).
16. C. Han, M. Sung, S.-H. Cho, H. Seok Chung, S. M. Kim, and J. H. Lee, “Performance improvement of multi-IFoF-based mobile fronthaul using dispersion-induced distortion mitigation with IF optimization,” J. Lightwave Technol. 34(20), 4772–4778 (2016). [CrossRef]
17. C. Kuo and E. Bergmann, “Second-order distortion and electronic compensation in analog links containing fiber amplifiers,” J. Lightwave Technol. 10(11), 1751–1759 (1992). [CrossRef]
18. H. Lin and Y. Kao, “Nonlinear distortions and compensations of DFB laser diode in AM-VSB lightwave CATV applications,” J. Lightwave Technol. 14(11), 2567–2574 (1996). [CrossRef]
19. X. Zhang, M. Jia, L. Chen, J. Ma, and J. Qiu, “Filtered-OFDM — Enabler for flexible waveform in the 5th generation cellular networks,” in Proceedings of Global Communications Conference (IEEE, 2015), pp. 1–6. [CrossRef]
20. J. Abdoli, M. Jia, and J. Ma, “Filtered OFDM: A new waveform for future wireless systems,” in Proceedings of 16th Signal Processing Advances in Wireless Communications (SPAWC 2015), pp. 66–70.
21. 3GPP TS 36.104 version 12.5.0 Release 12, 2014.