 Research
 Open
 Published:
Energyefficient radiooverfiber system for nextgeneration cloud radio access networks
EURASIP Journal on Wireless Communications and Networkingvolume 2019, Article number: 118 (2019)
Abstract
The paper proposes a novel adaptive radiooverfiber (RoF) system for nextgeneration cloud radio access network (CRAN), aiming to optimize the operation cost in terms of power consumption while maintaining required data rate. By jointly considering the nonlinear distortion from MachZehnder modulator (MZM) and high power amplifier (HPA) due to high peaktoaveragepower ratio (PAPR) in the electronic domain, we first provide a 2×2 multipleinput mulitpleoutput orthogonal frequency division multiplexing (MIMOOFDM) baseband model on electrical SNR (ESNR) for a single RoF transmission line. To take the modulation levels into consideration, we provide the optical signal to noise ratio (OSNR) analysis that jointly considers the electrical SNR (ESNR) model and the nonlinear effect of the optical transmission. This optical SNR (OSNR) analysis result is further used in the subsequent power consumption model for both the downlink and uplink of the considered RoF transmission system. Case studies via simulation and numerical experiments are conducted to verify that the proposed RoF system not only can reach the lowest power and spectrum consumptions at same time, but also consumes considerably less power than current RoF system.
Introduction
Today’s wireless traffic is dominated by IPbased multimedia services and applications, which have caused significant burdens on the radio access networks (RANs). Enlarging the system capacity is a simple remedy that could nonetheless cause poor equipment utilization and low energy efficiency at the base stations (BSs) due to high traffic fluctuation and network dynamics. The stringent demand on quality of service (QoS) further leaves the legacy RANs an awkward situation in dealing with the application scenarios envisioned in the near future.
Cloud RAN (CRAN), defined with LTEAdvance and those under the banner of 5G, aims to resolve the bandwidth thirsty of the current RANs while achieving high network responsiveness and QoS provisioning. With channel bandwidths up to 100 MHz and the downlink peak rates of 1 Gb/s, four radio channels per direction are provisioned at each BS, which consumes a minimal serial bit rate of 24 Gb/s in the backhaul and up to 100 GB/s is foreseen in the near future [1]. The deployment of densely distributed and centralizedcontrolled lowpower small cells can better handle the high traffic fluctuation via fast reconfiguration [2]. This is realized by equipping the central unit (CU) with multisite and multistandard baseband units, such that multiple wireless standards for geographically dispersed BSs with different levels of sectors and coverage can be supported. To this end, all the advanced technologies in the conventional RANs, such as enhanced MIMO, cooperative multipoint (CoMP), carrier aggregation, and strategies toward network heterogeneity, are modified from a distributed version to fit into a highly centralized control paradigm.
Nonetheless, the common public radio interface (CPRI) defined under LTE advanced can only support up to 6.144 Gb/s [3] and is merely adequate for a few 20 MHz channels, or a single 100 MHz radio channel [4], thus could easily render a bandwidth bottleneck. Featured by link transparency and lower bandwidth requirement per BS comparing to current CPRIbased digital transmission approaches, radio over fiber (RoF) systems are considered as a good counterpart for centralized wireless infrastructure of CRAN in the nextgeneration mobile communications.
This paper investigates an energyefficient RoF system as the backbone transmission technique for the CRAN [5]. Different from all the previously reported research, the proposed RoF system employs a 2×2 MIMO module and coherent optical orthogonal frequency division multiplexing (COOFDM) transmission technology. Thanks to the partially overlapped subcarriers and high tolerance to chromatic dispersion (CD) and polarization mode dispersion (PMD), the COOFDM transmission system demonstrates superb spectrum efficiency and bandwidth slicing flexibility. By provisioning radio access via COOFDMbased optical channels, the resultant RoF system is shown with great potential to achieve improved service provisioning granularity without losing data capacity compared with legacy WDMbased RoF systems.
The rest of the paper is organized as follows. Section 2 provides a comprehensive review on the enabling technologies for the proposed RoF system. Section 3 presents the proposed RoF system for nextgeneration CRAN and provides a multiinputmultioutput OFDM (MIMOOFDM) baseband model for simulating the endtoend RoF system, followed by OSNR calculation and power consumption analysis in Section 4. Section 5 shows the case study results that compares the proposed MIMOOFDM transmission system. Section 6 concludes the paper.
An overview
The section provides a highlevel overview on the proposed research regarding CRAN and RoF.
Cloud radio access network (CRAN)
Under the current distributed antenna systems (DAS) architecture, each BS serves as a layer2 switch by providing baseband processing functions and is further connected to the remote radio unit (RRUs) via a highspeed link, also referred to as fronthaul. Such a distributed network architecture has been considered short of flexibility and dynamic configurability for fluctuating traffic demand and multiple wireless standards, coverage, and frequency spectrum [6]. Further, since a cooling system has to be in each BS due to the presence of the processing unit, nontrivial power consumption is incurred in presence of a large number of cells.
CRAN mitigates the abovementioned problems by centralizing the baseband processing functions in the centralized unit, in which a processing unit pool is created in order to accommodate the baseband processing units of all the BSs under its cell site. With this, the capacity of baseband units (BBUs) pooled at the central unit can be dynamically and elastically allocated with sufficient flexibility. In the meantime, advanced coprocessing and synchronization techniques can be easily deployed due to the centralized and pooled BBUs, such as heterogeneous networks and cooperative multipoint (CoMP). On the other hand, the RRUs of each BBU are distributed across a wide area and are connected with the BBU via highspeed links, which are also referred to as fronthauls. According to the current industry practice, common public radio interface (CPRI) protocol, defined under 3GPP, is used on optical fibers to support these fronthaul links.
The CRPI fronthauls and the CRAN backbone are generally supported by WDM technology for exploring larger capacity and guaranteed QoS. Due to the highly dynamic environment, nonetheless, the WDMbased CRAN could waste a significant amount of energy and be subject to low channel utilization on optical transmissions that have to overprovision to meet the stringent QoS and bandwidth requirements. Another problem due to the use of digital fronthaul link is caused by analogdigital conversion that naturally increases the endtoend latency and hardware complexity. Thus, making the optical transmissions adaptive and software programmable according to the traffic variation is critical to the design of the nextgeneration CRANs. It has been widely reported that we can dynamically manipulate a number of configurations/parameters of an optical OFDM transmission line to achieve better energy and spectral efficiency [7], and these parameters include the modulation format, the code rate of the forward error correction (FEC), the symbol rate (per subcarrier), the number of polarizations per wavelength, and the number of OFDM subcarriers. Such adaptive transceivers are expected to efficiently allocate spectral bandwidth in the presence of highly dynamic traffic demands.
Radiooverfiber (RoF)
Instead of conventional CPRIbased digital link, the study considers the RoF technique for the highspeed fronthaul link inbetween the central unit BBU pool and each remote radio unit (RRU). An RoF system is characterized by simple and lowerpower RF transmitters as well as capability for dynamic resource allocation and mobility management [8], which demonstrates numerous merits against its counterparts. In spite of smaller radio ranges due to limited amplification function at the RF transmitter, it offers very low implementation cost and is considered more suitable for supporting multiple wideband radio channels than the CPRI [4].
Coherent optical OFDMbased MIMORoF is an advanced implementation of the RoF technique [9], thanks to the robustness of MIMO channel and the nature of multiple access in employing OFDM (or referred to as OFDMA). It is clear that transporting multiple subcarriers at the same wavelength (i.e., SCM) and possibly multiple wavelengths multiplexed (i.e., WDM) through the RoF links is more challenging than the CPRI transmission and thus takes more careful designs and control/management.
Similar to digital CPRI, a RoF link is typically subject to optical component nonlinearities and fiber chromatic dispersion, as well as difficulties in implementing remote control with a large number of RRUs. Although these limitations can be overcome by providing dedicated wavelengths to each RRU, it turns out to be a rather costly solution subject to a scalability problem when the number of RRUs is increasing. To provide fine granularity of bandwidth without losing scalability, a RoF transmitter could correspond to multiple RRUs by implementing a pointtomultipoint optical network architecture, such as passive optical networks (PONs). As shown in Fig. 1, longreach bidirectional RoF transmission system was demonstrated in [10], where the 65 km feeder extension of the single mode fiber (SMF) type is added to the PON with passive splitters for multiwavelength overlay.
Proposed RoF transmission system for nextgeneration CRAN
It is clear that the stateoftheart RoF schemes, such as in [10], aim to support a rather small distribution range (e.g., < 85 km), conservative modulation schemes (e.g., 64QAM or lower), and narrow channel bandwidth for downstream (20 MHz), which fail to serve in 5G mobile systems targeting wider channel bandwidth (100 MHz) and higher levels of modulation scheme (256QAM). To meet the requirements of future CRAN, the paper investigates an advanced RoF transmission system. As shown in Fig. 2, the proposed 2×2 MIMORoF system is featured with polarization division multiplexing (PDM) technology for increasing the number of provisioned BSs, as well as coherent optical (CO) detection [11] at the receiver for achieving high data rate and long distance transmission. One most important feature is that the transmit power, modulation scheme, and the number of subcarriers for each optical layer transmission can be adaptively determined and software programmed. With this, an intelligent configuration strategy should be in place in order to make transmission efficient and effective.
Note that the employment of coherent detection using local oscillator (LO) laser in the receiver, although signifiantly improved the transmission quality, will inevitably raise the system cost. Thus, the proposed system is suitable for provisioning largescale antenna arrays that are over several tens of kilometers from the baseband unit (BBU) pool.
The rest of the section presents the mathematical expressions of some important devices along the data path.
Baseband models for RoF transmission system
As shown in Fig. 3, a 2×2 MIMOOFDM baseband model for simulating the required ESNR of endtoend RoF transmission system is developed by jointly considering nonlinear distortion from both MachZehnder modulator (MZM) and high power amplifier (HPA) due to high PAPR.
PAPR in OFDM
Consider an OFDM transmission with L subcarriers at the frequencies {f_{l}, l=1,⋯,L}. Assigned to the subcarriers at {f_{l},1≤l≤L} are, respectively, the Mary data symbols $\{\bar {x}_{l}, \hspace {0.02in} l=1,\cdots,L\}$ or 0, which are independent and identically distributed random variables with zero mean and variance P.
Let T be the modulation interval and LT be the duration of an OFDM symbol (excluding the guard interval). The OFDMsignal’s complex envelope can be expressed as:
assuming the above is an idealized rectangular timedomain window, and the cyclicprefix extension of x(t) would not alter the PAPR.
The PAPR of the continuoustime signal x(t) is defined as:
where E[·] denotes the expectation value.
Electrical power amplifier model
A memoryless nonlinear HPA with a “soft limiter” inputoutput relationship is employed. For a complexvalue input y, the output equals Λ(y)e^{j∠y} [12], where:
The “clipping ratio” is defined as $\gamma =\frac {A}{\sqrt {P_{in}}}$ [12], where P_{in} denotes the average power of the input signal, γ=3 dB.
Optical MZM model
As a waveguidebased external modulator, MZM is a device widely employed in optical OFDM systems with chirpfree signals for achieving high data rate transmissions. As shown in Fig. 2, the optical OFDM signal is modulated based on the electrical OFDM signal by using the MZM. The modulation process is nonetheless subject to nonlinear and peaklimited transfer characteristics.
When using differential input data in a pushpull configuration, the transfer function of a singledrive MZM is given by [13]:
where E_{in}(t) and E_{out}(t) are the input and output in optical field, respectively; V(t) is the electrical OFDM signal; and V_{π} is the required voltage difference applied to a single electrode in order to generate a phase shift between two waveguides.
Expanding the MZM nonlinear transfer function into a Taylor series as:
The baseband equivalent polynomial model for the output electrical field of the MZM is given as:
where x_{k} and y_{k} are the discrete vectors of the applied voltage and the output voltage at the MZM, respectively; q is the order of nonlinearity; and α_{q} is the odd coefficient of the MZM nonlinear transfer function with an operating region 3V_{π}±V_{π} (i.e., at the null intensity bias point), whose output signal is approximated as a third order polynomial [14].
Optical SNR calculation
By assuming an ideal detection of optical OFDMMIMO system and the linewidths of the transmit/receive lasers to be zero, the study takes the relation between the SNR in the optical domain (OSNR) and the electrical SNR (ESNR) at ideal coherent receiver as follows [15]:
where B_{ref} is the reference bandwidth used for the OSNR measurement (≈12.5 GHz for 0.1nm bandwidth around 1550 nm); M is the constellation size of Mary quadrature amplitude modulation (QAM); m_{sys} is the system margin ≈12 dB [16]; and D Gb/s is the total system symbol transmission rate. And, this relationship is independent of whether using polarization multiplexing or not [15].
As a RoF downstream transmission line shown in Fig. 1, the available OSNR of a 0.1nm band at around 1550 nm at the optical receiver can be given by:
where P_{out} is output power of distributed feedback laser (DFB) laser source up to 16 dBm. L_{MZM} is optical excess loss of MZM with a typical value 7.75 dB for 4 channels multiplexing; L_{aligned} is passive splitter loss (1:16) with a typical value 14 dB [10]; L_{OBPF} is optical bandpass filter loss with a typical value 3 dB; L_{TOF} is tunable optical filter (or add/drop multiplexer) loss with a typical value 5 dB [17]; and α_{span}=17dB is the loss of 85km SSMF span (0.2 dB/km) [17]. For a typical EDFA, F_{EDFA} is the EDFA noise figure due to the ASE noise with a typical value of 6 dB; G_{EDFA} is the EDFA gain up to 20 dB. G_{Rx} is receiver sensitivity no better than − 30 dBm for the avalanche photodiode (APD) [17]. Some other elements along the datapath of Fig. 1 may affect the endtoend dB values, including circulator loss and array waveguide grating (AWG) loss, are jointly represented as L_{other}, which takes a values of 5.76 dB. Note that although we have considered all possible elements to the best of our survey, there could be more elements along the data path that affect the dB values. The proposed power consumption model, nonetheless, is generic to a RoF system and can be easily expanded according to any addition upon the system.
For the PDMbased RoF downstream transmission line in Fig. 2, the available OSNR of a 0.1 nm band at around 1550 nm at the coherent receiver can be given by:
where P_{LO} is output power of local oscillator (LO, DFB laser) at coherent receiver with value up to 16 dBm. G_{Co−Rx} can achieve a sensitivity of − 45.9 dBm in LRPON over 100km SSMF by using coherent receiver without amplification [11]. L_{PDM} is the PDM loss with a typical value 24.9 dB, including laser power splitter, MZM loss, AWG loss, polarization beam splitter, variable attenuator, and polarization beam combiner [16].
Power consumption analysis
Based on the ESNR requirement simulated by MIMOOFDM baseband model, we use an ideal ESNRtoOSNR transfer function to derive the corresponding OSNR for supporting the ESNR requirement. The relation between the required ESNR and OSNR is provided based on some assumptions. The OSNR is further derived to calculate the required output power of laser source. Then, we propose the power consumption model for the optical part of RoF transmission system.
Power consumption analysis
The general power consumption of optical transmitter P_{Tx} (in Watt) is given by:
where γ_{DC} is the power conversion efficiency of converting the +12 V DC power supply for transceiver modules with a value 93%; P_{map}≈(0.019·D)/log_{2}(M) is the power consumption for signal mapping by different modulation level M; P_{P/S}=0.02·D is the power consumption for paralleltoserial (or serialtoparallel) conversion; P_{DAC}≈(0.008·D)/log_{2}(M) is the power consumption for a single digitaltoanalog converter (DAC); P_{MZM}≈0.017·D is the power consumption for a singledrive MZM; and P_{IFFT−CP−TS}≈(0.16·D)/log_{2}(M) is the power consumption for the IFFT, CP, and TS modules for 512 OFDM subcarriers [16]. n_{DAC} and n_{MZM} are the number of required DACs and MZMs by different RoF systems, respectively.
The general power consumption of optical receiver P_{Rx} (in Watt) is given by:
where P_{LO} is the power consumption for the local oscillator (LO) at coherent receiver equals to the P_{laser}; P_{TIA}=(0.0188·D)/log_{2}(M) is the power consumption of transimpedance amplifier with automatic gain control for currenttovoltage conversion; P_{PD}≈0.0028·P_{CW}·D is the power consumption for a single photodiode (PD); P_{ADC}≈(0.0175·D)/log_{2}(M) is the power consumption for a single analogtodigital converter (ADC); and P_{Rx−DSP}≈(0.36·D)/log_{2}(M) is the power consumption for the DSP module of signal postprocessing at receiver for 100 km distance and 512 OFDM subcarriers [16]. n_{ADC} and n_{PD} are the number of required ADCs and photodiodes by different RoF systems, respectively.
The power consumption of the laser P_{laser} (in Watt) for downlink in CU is calculated by:
where P_{CW} (in Watt) derived from P_{out} (in dBm) is the laser continuous wave (CW) output power; D_{down} is downstream data rate; and E_{Laser} is the DFB laser energy consumption with a typical value as 1.5 pJ per bit for supporting up to 40 Gb/s [18]. For a ESNR value, P_{out} is calculated based on (6), (7), or (8) by taking M as variables.
The power consumption of the laser equipped with driver (instead of MZM) P_{laserDr} (in Watt) for uplink is calculated by:
where D_{up} is the upstream data rate and E_{LaserDr} is the energy consumption by the DFB laser with a typical value of 37 pJ per bit for supporting up to 40 Gb/s [18].
The power consumption of a link EDFA P_{EDFA} per wavelength is given by [16]:
where γ_{EDFA} is the conversion efficiency of EDFA power with a value 2%; P_{EDFAoh} is the power consumption of EDFA overhead with a value 0.69 W; h is a constant with a value 6.626×10^{−34}; and ν is the optical frequency constant with a value 1.93×10^{14} Hz.
Case study
Case study is conducted to demonstrate the energy efficiency of the proposed MIMORoF system. Our goal is to observe the performance of the proposed RoF system in terms of power consumption. We are particularly interested in how the system can satisfy a given transmission rate with minimum transmit power by manipulating its modulation level scheme.
Without loss of generality, a single BS is connected. We set D_{down}=30 Gbps for downstream data transmission, and D_{up}=10 Gbps for upstream data transmission. Each RAU/BS supports two sectors, each being equipped with 2 antennas of a 2×2 MIMO configuration. By taking the required ESNR in baseband as the simulation parameter in each scenario, we implemented VBLAST for improving capacity gain and STBC for maximizing spatial diversity in the MIMO processing unit. Each radio channel includes 512 subcarriers, and candidate modulation levels include QPSK (M=4), 16QAM (M=16), 64QAM (M=64), and 256QAM (M=256). For calculating OSNR, the distribution range is taken as 85 km and 100 km for the current and future CRAN scenarios, respectively. The required voltage difference V_{π} applied to the electrode of singledrive MZM is in the range of 6.5∼10 due to the upper bound of available OSNR. The BER requirement without using FEC at the receiver is 10^{−6}. The channel is assumed to be subject to AWGN and Rayleigh fading.
When V_{π} of MZM is small as shown in Fig. 4, the 256QAM modulation suffers significantly more signal distortion from RoF downlink than that from uplink, while the lower modulation levels of downlink suffer about the same distortion as that of uplink. This is due to the difference of the MIMOOFDM module at the BS and the antenna sides, respectively, where a larger ESNR is required for achieving the given data rate. Similar observation is gained when V_{π} of MZM is large as shown in Fig. 5, where the 64QAM and 256QAM modulations suffer more signal distortion from the RoF uplink than that from downlink, while the required ESNR of the lower modulations for downlink is only slightly affected.
For the current RoF system with small V_{π} as in Fig. 4, using a higher level of modulation can generally improve the power efficiency, thanks to the improved coding efficiency. However, using 256QAM breaks such a trend by consuming significantly more power since at this moment, the high SNR requirement at the receiver for the higher coding rate (256QAM) cannot be satisfied due to the constant attenuation of the transmission line. Thus, it has to resort to increasing the transmit power for better SNR. Note that the increase of transmit power and level of modulation also introduces the performance impairment by PAPR, which explains the big jump of the power consumption. This is still true with large V_{π} as in Fig. 5, where the highest power is consumed by QPSK, and the lowest power is also achieved by 16QAM or 64QAM.
For the proposed RoF system, on the other hand, it shows that the most aggressive modulation scheme 256QAM can be used for not only achieving the lowest power consumption in all cases for achieving the given data rate. This is due to the fact that the use of PDM and the PAPR reduction scheme can achieve better spectrum efficiency, such that the trend of highermodulationlowerpower still holds when 256QAM is being used. In summary, the proposed RoF system can save 27–69% power consumption compared with that of the current RoF system.
Conclusions
This paper introduced a novel adaptive radiooverfiber (RoF) system for nextgeneration CRAN in which energy consumption, capacity per wavelength, and distribution range are considered. By considering nonlinear distortion from both MZM and HPA, we first developed a 2×2 MIMOOFDM baseband model for simulating the required ESNR of the RoF system. Then, we proposed a novel MIMORoF system for the nextgeneration CRANs that can fit into the requirements of 5G mobile systems, where the OSNR analysis and its relation with ESNR was investigated. Lastly, a set of case studies was conducted aiming to observe the performance of the proposed RoF system in terms of its energy and bandwidth efficiency in presence of a given data rate. We demonstrated that the best energy efficiency and spectrum consumption can be achieved by using the most aggressive modulation format (256QAM) with considerably less power consumption than the current RoF system.
Abbreviations
 ADC:

Analogtodigital convertor
 AWG:

Array waveguide grating
 BBU:

Baseband unit BER: Bit error rate
 BS:

Base station
 CRAN:

Cloud radio access network
 CD:

Chromatic dispersion CW: Continuous wave
 CO:

Coherent optical COOFDM: Coherent optical OFDM
 CoMP:

Cooperative multipoint
 CPRI:

Common public radio interface
 CU:

Central unit
 DAC:

Digitaltoanalog convertor
 DAS:

Distributed antenna systems
 DC:

Digital convertor
 DFB:

Distributed feedback
 DSP:

Digital signal processor
 EDFA:

Erbiumdoped fiber amplifier
 ESNR:

Electrical signalnoise ratio
 FEC:

Forword error correction
 HPA:

High power amplifier
 IP:

Internet protocol
 LRPON:

Long reach PON
 LTE:

Longterm evolution
 MIMO:

Orthogonal frequency division multiplexing
 MIMOOFDM:

MIMO multipleinput and multipleoutput
 MZM:

machzehnder modulator
 OSNR:

Optical signalnoise ratio
 PAPR:

Peaktoaverageratio
 PD:

Photodiode
 PDM:

Polarization division multiplexing
 PMD:

Polarization mode dispersion
 PON:

Passive optical networks
 QAM:

Quadrature amplitude modulation
 QoS:

Quality of service
 RoF:

Radiooverfiber
 RRU:

Remote radio unit
 SCM:

Subcarrier multiplexing
 SMF:

Single mode fiber
 SSMF:

Standard single mode fiber
 STBC:

Spacetime block code
 VBLAST:

Vertical bell labs layered spacetime
 WDM:

Wavelength division multiplexing
References
 1
F. Tian, et al., A survey on CRAN security. IEEE Access. 55(1), 13372–13386 (2017).
 2
D. Bojic, E. Sasaki, N. Cvijetic, W. Ting, J. Kuno, J. Lessmann, S. Schmid, H. Ishii, S. Nakamura, Advanced wireless and optical technologies for smallcell mobile backhaul with dynamic software defined management IEEE. Commun. Mag.51(9), 86–93 (2013).
 3
D. Chitimalla, et al., 5G fronthaullatency and jitter studies of CPRI over ethernet. IEEE/OSA J. Opt. Commun. Netw.9(2), 172–182 (2017).
 4
D. Wake, et al., Radio over fiber link design for next generation wireless systems. J. Light. Technol.28(16), 2456–2464 (2010).
 5
H. Khodakarami, et al., Flexible optical networks: An energy efficiency perspective. J. Light. Technol.32(21), 3356–3367 (2014).
 6
D. Novak, Radiooverfiber technologies for emerging wireless systems. IEEE J. Quantum Electron.52(1), 1–11 (2016).
 7
B. T. Teipen, et al., Adaptive data rates for flexible transceivers in optical networks. J. Netw.7(5), 776–782 (2012).
 8
J. E. Mitchell, Integrated wireless backhaul over optical access networks. J. Light. Technol.32(20), 3373–3382 (2014).
 9
I. S. Amiri, et al., twobytwo MIMOOFDMRoF generation and transmission of double Vband signals using microring resonator system. Opt. Quant. Electron.48(1) (2015).
 10
M. Milosavljevic, et al., Demonstration of wireless backhauling over longreach PONs. J. Light. Technol.30(5), 811–817 (2012).
 11
D. Lavery, et al., in Optical fiber communication conference and exposition (OFC/NFOEC). Bidirectional 10 Gbit/s longreach WDMPON using digital coherent receivers (Los Angeles, 2011).
 12
Litsyn, S. (Cambridge University Press, Cambridge, 2007).
 13
K. P. Ho, Phasemodulated optical communication systems (Springer Science + Business Media, Inc., 2005).
 14
Y. London, D. Sadot, Nonlinear effects mitigation in coherent optical OFDM system in presence of high peak power. J. Light. Technol.29(21), 3275–3281 (2011).
 15
R. J. Essiambre, G. Kramer, P. J. Winzer, G. J. Foschini, B. Goebel, Capacity limits of optical fiber networks. J. Light. Technol.28(4), 662–701 (2010).
 16
B. S. Pillai, B. Sedighi, K. Guan, N. P. Anthapadmanabhan, W. Shieh, K. J. Hinton, R. S. Tucker, Endtoend energy modeling and analysis of longhaul coherent transmission systems. J. Light. Technol.32(18), 3093–3111 (2014).
 17
Gigabitcapable passive optical networks (GPON): Physical media dependent (PMD) Layer specification (2003). https://www.itu.int/rec/TRECG.984.2200303I/en.
 18
R. Tucker, Green optical communications??Part I: Energy limitations in transport. J. Sel. Top. Quantum Electron.17(2), 245–260 (2011).
Acknowledgements
This work was supported in part by the BK21 Plus Project (SW Human Resource Development Program for Supporting Smart Life) funded by the Ministry of Education, School of Computer Science and Engineering, Kyungpook National University, Korea, under Grant 21A20131600005, in part by the National Research Foundation of Korea grant funded by the Korean Government (2018R1D1A1B07051118), and in part by Kyungpook National University Research Fund, 2018.
Author information
Affiliations
Contributions
All of the authors participated in the whole process of this research work and made considerable contributions, while with the following respective focus: (1) BW drawed all the figures and carried out the simulation/experiment part of this study. (2) LP mainly contributed in the proposing idea, verifying the illustrative figures for the proposed RoF transmission system, and revising the writing. (3) PHH worked as an overall director in generating the major ideas, advising the system architecture, and organizing the overall paper writing. All authors read and approved the final manuscript.
Corresponding author
Correspondence to Limei Peng.
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Received
Accepted
Published
DOI
Keywords
 RadiooverFiber (RoF)
 Cloud radio access network (CRAN)
 Energy efficiency
 Peaktoaverage power ratio (PAPR)
 Nonlinear distortion