Open Access
Add to BibSonomy
Abstract
A code reservation technique is proposed to reduce the peak-to-average power ratio (PAPR) of digital code-division multiplexing (CDM) based channel aggregation for mobile fronthaul. We numerically investigate the relationship between the PAPR and the number of aggregated channels during the CDM based channel aggregation, and experimentally verify the transmission performance with the code reservation technique. The PAPR of aggregated signal, which is composed of 48x20MHz Long Term Evolution (LTE) signal mapped with 64 quadrature amplitude modulation (QAM), is reduced with the help of code reservation technique. The PAPR reduction enables larger optical modulation index (OMI) per channel at the linear operation region of a directly modulated laser (DML), leading to the optical signal-to-noise ratio (OSNR) improvement. After the transmission over 10km standard single mode fiber (SSMF), the receiver sensitivity can be improved by 4dB owing to the PAPR reduction.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Centralized radio access network (C-RAN) is one of the most promising technologies for next generation wireless network, which can improve the network capacity, transmission performance, power and cost efficiency, by using massive multiple-input-multiple-output (M-MIMO) and coordinated multi-point technique [1–3]. The mobile fronthaul connecting the baseband unit (BBU) and radio remote units (RRUs) is a crucial part of C-RAN. It uses fiber as the transmission medium, in order to increase the transmission capacity immensely. The fronthaul interface has been defined by common public radio interface (CPRI) [4]. Generally, it achieves the digitization of the wireless signal from the core network, but it requires extremely large bandwidth due to its binary feature. For example, for one Long Term Evolution (LTE) sector with a carrier bandwidth of 20MHz, CPRI needs over 1.2Gbit/s bit-rate [1]. In upcoming 5G era, such CPRI interface is stressful to satisfy the enormous capacity requirement, especially under the scenario of enhanced mobile broadband (eMBB). In order to solve the bandwidth inefficiency of current mobile fronthaul, both frequency division multiplexing (FDM) and code division multiplexing (CDM) based channel aggregation have been proposed with the assistance of advanced digital signal processing (DSP) technique. Recently, the FDM-based channel aggregation using the DSP has been comprehensively investigated [5,6]. It uses fast Fourier transformation (FFT) to achieve the aggregation of each channel in the frequency domain, and inverse fast Fourier transformation (IFFT) at the receiver to separate the aggregated signal for each users by the use of the Hermitian symmetry to obtain real value signal. One of the disadvantages for the FDM-based channel aggregation is its high computational complexity, due to lots of multiplication operations arising in the FFT/IFFT algorithm. Recently, we have put forward a CDM-based channel aggregation, using the orthogonal Walsh code to keep each channel independent in the code domain [7]. Compared with the FDM-based solution, the CDM-based channel aggregation only introduces the addition operation, leading to substantial reduction of computational complexity. Meanwhile, it is compatible with current wireless communication system, because CDM has several advantages of high security, anti-interference, and low probability of intercept [8,9].
Orthogonal frequency division multiplexing (OFDM) as one of the most attractive modulation techniques has been widely used in the 3G and 4G wireless communication systems. As a multicarrier modulation technique, OFDM has the ability to resist the frequency selective channel and achieve high spectrum utilization. However, it suffers high peak-to-average power ratio (PAPR), because of the superposition of multiple subcarriers. By taking into account of the channel aggregation, the PAPR will become even worse, which is harmful to the devices in the mobile fronthaul, such as directly modulated laser (DML) [10]. When the signal amplitude exceeds the linear operation region of DML, clipping phenomenon will occur and consequently degrade the transmission performance. To reduce the PAPR of OFDM signal, many solutions have been proposed, including partial transmission sequence (PTS), selective mapping (SLM), nonlinear companding transforms, tone reservation, block-scaling, and discrete Fourier transform spread [11–14]. In particular, as for the FDM-based channel aggregation, several techniques have been put forward to reduce the PAPR, such as tone reservation technique [15], phase pre-distortion scheme [16], joint interleaving technique [17], and multi-dimensional crest factor reduction [18]. However, there is no specific solution to address the problem for the CDM-based channel aggregation on mobile fronthaul, to the best of our knowledge.
In this paper, we propose a PAPR reduction technique for the CDM-based channel aggregation in mobile fronthaul by using the code reservation. We theoretically investigate the PAPR characteristics with respect to the number of aggregated channel, and experimentally verify the capability of code reservation technique for the PAPR reduction of CDM-based channel aggregation. After the transmission over 10km standard single mode fiber (SSMF), 4dB improvement of receiver sensitivity is achieved.
2. Operation principle
2.1 Digital CDM based channel aggregation
Figure 1(a) shows the schematic of mobile fronthaul with the channel aggregation. By taking the downstream transmission into account, the LTE signals from core network undergoes the baseband processing and the channel aggregation. Then the signals are modulated on the optical carrier in the BBU [19]. At the RRU part, after the optical-to-electrical conversion, the de-aggregation, and frequency up-conversion operation, the wireless signals are recovered and transmitted into the air. Figure 1 (b) shows the operation principle of CDM-based channel aggregation/de-aggregation, where M denotes the number of channels and N is the order of Walsh codes. Generally, N ≥ 2M needs to be satisfied. Complex signal of each channel is divided into the real and the imaginary part, and each part is assigned with a specific code sequence [20,21]. Then each symbol is repeated with a period of N, and the code sequence which is comprised of ± 1 is used to determine the sign of symbol. Finally, all signals are summed up to generate the aggregated signal. Clearly, the computation complexity of CDM-based channel aggregation can be substantially reduced with only sign selection and addition calculation in comparison with the FDM-based channel aggregation [7], which can satisfy the delay requirement of 5G system. At the RRU part, the signal is distinguished by the code orthogonality. As for the FDM-based channel aggregation technique, each channel owns a specific frequency range, while for CDM-based counterpart, signal from different channels occupies the same range of frequency spectrum.
Fig. 1 (a) Schematic of mobile fronthaul with the channel aggregation, (b) operation principle of digital CDM-based channel aggregation/de-aggregation.
2.2 PAPR characteristics of digital CDM-based channel aggregation
Since the OFDM signal is the sum of multiple subcarriers [14], the range of amplitude will enlarge correspondingly with the growing number of subcarriers. When the OFDM signal with high PAPR is aggregated, the PAPR of the aggregated signal becomes even severe. Therefore, we start to investigate the relationship between the PAPR value and the number of aggregated channels. Figure 2 shows the complementary cumulative distribution function (CCDF) of PAPR with respect to the channel numbers for the CDM-based channel aggregation. The CCDF is a parameter to describe the PAPR characteristic, which denotes the probability when PAPR exceeds a specific value. With the growing channel number to be aggregated, the high PAPR occurs more frequently. In particular, when 60 channels are aggregated, the PAPR reaches up to 16.3 dB when the CCDF is equal to 10−3. Normally, both the high power microwave amplifier (HPMA) and DML have a limited linear operation range. Those devices will bring about the peak-clipping effect under the condition of high-PAPR signal input [22], and the generated clipping noise will degrade the transmission performance. Therefore, it is necessary to reduce the PAPR value, when the CDM-based channel aggregation is utilized in mobile fronthaul.
2.3 Code reservation technique
We propose an algorithm to reduce the PAPR of CDM-based channel aggregation for mobile fronthaul, which we call the code reservation technique. Compared with the deliberate digital clipping technique, which leads to significant signal distortion and performance penalty, the proposed technique can avoid those disadvantages. Code reservation technique utilizes the peak-cancellation signal distributed among the reserved code sequence, in order to reduce the PAPR. Figure 3(a) shows the operation principle of the technique.
Fig. 3 (a) Operation principle of code reservation technique, (b) signal with the code reservation technique, (c) signal without the code reservation technique, (d) CCDF of PAPR with respect to the clipping ratio, (e) CCDF of PAPR with respect to the iteration time.
If we assume that N is the order of Walsh code, and M is the number of channels to be aggregated. As for the CDM-based channel aggregation, 2M code sequences are necessary, as shown in Eq. (1)
where S is the aggregated signal, is the n-th symbol of channel k, T is the symbol duration, and is the sequence of Walsh code. Then the signal is clipped at the peak-clipping level, as shown in Eq. (2)where A is the peak-clipping level. The clipping noise is distributed among all the code sequences from 1 to N. Then we extract the clipping noise distributed among the code sequences from 2M + 1 to N by doing the orthogonal decomposition, as shown in Eq. (3). Please note that here we make full use of the redundant code sequences in order to obtain the outstanding PAPR mitigation. If the number of reserved code sequence is reduced, more channels can be used to aggregate the wireless signals, leading to the capacity improvement of mobile fronthaul. However, the performance of PAPR mitigation is consequently degraded. Such operation is almost the same to the de-aggregation process at the receiver-side DSP.where is the clipping noise distributed among on the unused Walsh sequences. After converting it into the spread signal, we can obtain the peak-cancellation signal , as shown in Eq. (4).Then we combine the peak-cancellation signal with the input signal, as shown in Eq. (5), in order to reduce the corresponding PAPR,
Since the two parts of Eq. (5) are orthogonal with each other in the code domain, no interference occurs among LTE channels. Therefore the code reservation technique can reduce the PAPR without the distortion of wireless signal. Moreover, the performance of PAPR reduction can be further improved in case we do the iteration from Eq. (2) to Eq. (5). Each iteration of proposed algorithm includes one clipping operation with the help of comparator and one CDM aggregation/de-aggregation operation. As shown in Figs. 3(b) and 3(c), we compare the normalized amplitude of temporal waveform for the signal with/without the code reservation enabled PAPR reduction. The peak value of signal is reduced and the amplitude of signal after the code reservation technique is more evenly distributed. Furthermore, the peak-clipping level and the number of iterations are two important parameters to be optimized. We investigate the effect of both two parameters on the performance of code reservation technique enabled PAPR reduction. In Fig. 3(d), we show the effect of clipping ratio which is defined as the peak-clipping level to the average amplitude of the original signal. It can be seen that, with the growing clipping ratio, the PAPR increases gradually when the value of CCDF is between 100 and 10−1. We can conclude that small clipping ratio is more effective for our proposed PAPR reduction technique. When the clipping ratio is further reduced to less than 7dB, the corresponding PAPR mitigation becomes saturated even under the condition of more iteration numbers. As a result, the clipping ratio of 7dB is set in our next investigation. Figure 3(e) shows the CCDF of PAPR with respect to the iteration times. When the number of iteration increases, the PAPR continues to decrease, but the speed of convergence becomes slower and marginal improvement becomes less. By taking the delay into account, we set the maximum iteration times at 10 in our investigation.
3. Experimental setup and results
Figure 4 shows the experimental setup of the CDM-based channel aggregation. 48x20MHz LTE signals with 64-QAM format are generated via the offline-DSP. For each channel, 1200 subcarriers with 2048 IFFT points are used. The 48 LTE signals can be organized to a representative macro-cell configuration which has 8 × 8 MIMO with a channel aggregation of two 20MHz signals and 3 directional sectors [5]. Signals from different channels are aggregated by the CDM-based aggregation with 128-order Walsh code sequences generated by the Hadamard matric. The bandwidth of aggregated signal is 2GHz(30.72 MHz/2 × 128) [6,7], as shown in Fig. 4. Meanwhile, there are 32 unused code sequences. Then the code reservation technique is used to reduce the PAPR. Next the signal with low PAPR is introduced to arbitrary waveform generator (AWG, Tektronix AWG 7122C) with a sampling rate of 4GSa/s. To achieve cost-effective mobile fronthaul, a DML at 1550nm is used to generate corresponding optical signal, and the HPMA is used to adjust the amplitude of electric signal before the electronic-to-optical conversion. After the 10km SSMF transmission, a photodetector (PD) with 3dB bandwidth of 10GHz is used to perform optical-to-electrical conversion. After that, the signal is captured by a 12.5GSa/s real-time sampling oscilloscope (OSC, Tektronix DPO 73304D) for the offline DSP, consisting of synchronization, channel estimation, channel de-aggregation, and OFDM demodulation. The error vector magnitude (EVM) value is measured in order to evaluate the transmission performance of each channels. Figure 5 shows the power-to-bias current (P-I) curve of used DML, whose threshold current is around 15mA. The maximum bias current is limited to 100mA to prevent the DML damage. As shown in Fig. 5, the DML is well performed even with high input current. Since practical low-cost DMLs may not have such a large dynamic range, and in order to investigate the influence of nonlinearity introduced by the DML, the bias current in our experiment is set to only 45mA to generate 5mW output optical power. The loss of 10km SSMF together with two connectors is 3dB, and the launch optical power of digital CDM-based channel aggregation signal is 1dBm.
Firstly, we investigate the performance of the CDM-based channel aggregation signal with the code reservation technique under the condition of optical back-to-back (B2B) transmission, when the received optical power (ROP) is −7dBm. For the ease of comparison, the aggregated signal without the PAPR reduction is presented under the same experimental condition. Figure 6 shows the mean EVM of 48 channels as a function of optical modulation index (OMI) per channel, because each channel has similar performance in the CDM-based channel aggregation system. The OMI/ch is defined as
where is the input voltage, and are the bias current and threshold current, respectively, and is internal resistance of DML. The measurement of OMI/ch is to obtain an optimal analog signal modulation for subsequent fiber optical transmission experiment. When the OMI/ch is small, signals with/without the code reservation technique both have a poor performance, due to the low optical signal-to-noise ratio (OSNR). When the OMI/ch increases, the signal is mainly degraded by the nonlinearity of DML, and consequently the peak-clipping starts to degrade the transmission performance. As shown in Fig. 6, signal with the PAPR reduction enables larger OMI/ch, when the EVM is equal to 8%, which is specified by LTE-A for 64-QAM. It means that higher OSNR can be achieved. Meanwhile, the minimum EVM of the signal with the code reservation technique is lower than that without the use of code reservation technique. The minimum EVMs with and without the code reservation technique are 3.3% and 4%, respectively. Therefore, we can conclude that the code reservation technique for the CDM-based channel aggregation can improve the dynamic range of the DML without extra signal distortion, leading to the use of low-cost DML for the mobile fronthaul.Secondly, we measure the EVMs of all 48 channels with and without the code reservation after the 10km SSMF transmission and compare them with optical B2B case. The OMI/ch is 29% and the ROP is −7dBm. As shown in Fig. 7(a), all channels have an EVM below the threshold of 8%, and the signals with and without the code reservation technique have a mean EVM value of 3.3% and 4.9%, respectively. After the 10km SSMF transmission, the signal without the code reservation has a mean EVM value of 7.4%, but not all channels can satisfy the EVM threshold. However, the signal using the code reservation has mean EVM value of 5.5%, and EVM of all channels is below 8%. Figures 7(b)-7(e) are the corresponding constellation diagrams of the received signal with and without the PAPR reduction under conditions of optical B2B and 10km SSMF transmission.
Fig. 7 (a) EVMs of all 48 channels under conditions of optical B2B and 10km SSMF transmission, (b-c) constellation diagrams of signal with/without the code reservation under the condition of optical B2B transmission, (d-e) constellation diagrams of signal with/without the code reservation under the condition of 10km SSMF transmission.
Finally, to further quantify the transmission performance of proposed code reservation technique, we measure the EVM value with respect to the ROP under conditions of both optical B2B and 10km SSMF transmission. Figure 8 shows the mean EVM of 48 channels as a function of ROP under different transmission scenarios, where the OMI/ch is 29%. For optical B2B transmission, the receiver sensitivity at 8% EVM threshold has around 1dB improvement by using the proposed code reservation technique. For the case of 10km SSMF transmission, the EVM is degraded due to the chromatic dispersion-induced penalty in comparison with that of optical B2B transmission. Please note that, the wavelength channel can be transmitted in the O-band at 1310nm in order to avoid the fiber dispersion induced penalty. Furthermore, the minimum required ROP is improved from −20dBm to −24dBm owing to the PAPR reduction, indicating 4dB improvement of receiver sensitivity after the 10km SSMF transmission.
Fig. 8 Mean EVM of 48 channels versus the ROP under condition of optical B2B and 10km SSMF transmission.
4. Conclusion
We propose a code reservation technique to reduce the PAPR for the digital CDM-based channel aggregation in mobile fronthaul, and experimentally verify its transmission performance. The clipping noise is mitigated and a larger OMI/ch can be achieved. We experimentally verify the transmission performance of mobile fronthaul, when the 48 × 20MHz LTE signals are aggregated into one single wavelength channel and the redundant code sequences are used for the peak-cancellation signal. For the optical B2B transmission, the mean EVM can be improved from 4.9% to 3.3% using the proposed technique. 4dB improvement of receiver sensitivity is achieved after the 10km SSMF transmission, indicating of that both more power budget of mobile fronthaul and flexible wireless channel aggregation can be obtained.
Funding
National Natural Science Foundation of China (NSFC) (61701359); National Key Research and Development Program of China (2016YFE0121300); Wuhan Basic Applied Research Project (2017010201010101).
References and links
1. A. Pizzinat, P. Chanclou, F. Saliou, and T. Diallo, “Things you should know about fronthaul,” J. Lightwave Technol. 33(5), 1077–1083 (2015). [CrossRef]
2. X. Liu, F. Effenberger, N. Chand, L. Zhou, and H. Lin, “Efficient mobile fronthaul transmission of multiple LTE-A signals with 36.86-Gb/s CPRI-equivalent data rate using a directly-modulated laser and fiber dispersion mitigation,” in Asia Communications Photonics Conf. (ACP) (2014), paper AF4B.5. [CrossRef]
3. V. Jungnickel, K. Manolakis, W. Zirwas, B. Panzner, V. Braun, M. Lossow, M. Sternad, R. Apelfröjd, and T. Svensson, “The role of small cells, coordinated multipoint, and massive MIMO in 5G,” IEEE Commun. Mag. 52(5), 44–51 (2014). [CrossRef]
4. Common Public Radio Interface, “CPRI Specification V7.0,” Stand. Doc. Specif. 0, 128 (2015).
5. X. Liu, H. Zeng, N. Chand, and F. Effenberger, “Efficient mobile fronthaul via DSP-based channel aggregation,” J. Lightwave Technol. 34(6), 1556–1564 (2016). [CrossRef]
6. X. Liu, H. Zeng, N. Chand, and F. Effenberger, “Experimental demonstration of high-throughput low-latency mobile fronthaul supporting 48 20-MHz LTE signals with 59-Gb/s CPRI-equivalent rate and 2-us processing latency,” in European Conference and Exhibition on Optical Communications (ECOC) (2015), Paper We.4.4.3. [CrossRef]
7. H. Li, Q. Yang, S. Fu, M. Luo, X. Li, Z. He, P. Jiang, Y. Liu, and S. Yu, “Digital code-division multiplexing channel aggregation for mobile fronthaul architecture with low complexity,” IEEE Photon. J. 10(2), 7902710 (2018).
8. R. L. Pickholtz, D. L. Schilling, and L. B. Milstein, “Theory of spread-spectrum communications-a tutorial,” IEEE Trans. Commun. 30(5), 855–884 (1982). [CrossRef]
9. S. Kaiser, “OFDM code-division multiplexing in fading channels,” IEEE Trans. Commun. 50(8), 1266–1273 (2002). [CrossRef]
10. B. G. Kim, H. Kim, and Y. C. Chung, “Impact of multipath interference on the performance of RoF-based mobile fronthaul network implemented by using DML,” J. Lightwave Technol. 35(2), 145–151 (2017). [CrossRef]
11. S. S. K. C. Bulusu, M. Crussiere, J.-F. Helard, R. Mounzer, Y. Nasser, O. Rousset, and A. Untersee, “Quasi-optimal tone reservation PAPR reduction algorithm for next generation broadcasting systems : a performance/complexity/latency tradeoff with testbed implementation,” IEEE Trans. Broadcast. PP(99), 1–7 (2018).
12. M. S. Hossain and T. Shimamura, “Spectrum efficient DSI-based OFDM PAPR reduction by subcarrier group modulation,” IEEE Trans. Broadcast. PP(99), 1–7 (2017).
13. T. Jiang and Y. Wu, “An overview: peak-to-average power ratio reduction techniques for OFDM signals,” IEEE Trans. Broadcast 54(2), 257–268 (2008). [CrossRef]
14. S. H. Han and J. H. Lee, “An overview of peak-to-average power ratio reduction techiques for multicarrier transmission,” IEEE Wirel. Commun. 12(2), 56–65 (2005). [CrossRef]
15. M. Sung, C. Han, S.-H. Cho, H. S. Chung, and J. H. Lee, “Improvement of the transmission performance in multi-IF-over-fiber mobile fronthaul by using tone-reservation technique,” Opt. Express 23(23), 29615–29624 (2015). [CrossRef] [PubMed]
16. F. Lu, M. Xu, L. Cheng, J. Wang, S. Shen, J. Zhang, and G. K. Chang, “Sub-band pre-distortion for PAPR reduction in spectral efficient 5G mobile fronthaul,” IEEE Photon. Technol. Lett. 29(1), 122–125 (2017). [CrossRef]
17. A. K. Ajami, H. A. Artail, and M. M. Mansour, “PAPR reduction in LTE-Advanced carrier aggregation using low-complexity joint interleaving technique,” in IEEE Wireless Communications and Networking Conference (WCNC) (2015), pp. 675–680. [CrossRef]
18. H. Chen, J. Li, C. Yin, K. Xu, Y. Dai, and F. Yin, “Multi-dimensional crest factor reduction and digital predistortion for multi-band radio-over-fiber links,” Opt. Express 22(17), 20982–20993 (2014). [CrossRef] [PubMed]
19. X. Liu, H. Zeng, and F. Effenberger, “Bandwidth-efficient synchronous transmission of I/Q waveforms and control words via frequency-division multiplexing for mobile fronthaul,” in IEEE Global Commun. Conf. (GLOBECOM) (2015), Paper SAC 21. [CrossRef]
20. E. H. Dinan and B. Jabbari, “Spreading codes for direct sequence CDMA and wideband CDMA cellular networks,” IEEE Commun. Mag. 36(9), 48–54 (1998). [CrossRef]
21. N. Wada and K. I. Kitayama, “A 10 Gb/s optical code division multiplexing using 8-chip optical bipolar code and coherent detection,” J. Lightwave Technol. 17(10), 1758–1765 (1999). [CrossRef]
22. W. O. Popoola, Z. Ghassemlooy, and B. G. Stewart, “Pilot-assisted PAPR reduction technique for optical OFDM communication systems,” J. Lightwave Technol. 32(7), 1374–1382 (2014). [CrossRef]
