PON-FTTX Architecture and Bandwidth Analysis for Future Broadband Communications

Abstract

Huge traffic and high bandwidth requirement of 5G and beyond networks call for holistic planning to establish seamless and cost-efficient transmission. Current and future passive optical networks (PON) will undoubtedly play an active role in actualizing a high-speed and cost-efficient networks through integration with 5G radio access networks (RAN) architecture. In doing this, fast speed modulation at each connection in the 5G xhaul architectures is required to cope with the strict latency and bandwidth requirements at each section. In this chapter, PON evolution up to the current and future PONs is reviewed to study different modulation approaches, their limitations, and complexities. We further reviewed different PON architectures and proposed usage possibilities for 5G and beyond networks.

1. Introduction

With no doubt, the future of broadband communications can be securely placed upon fiber networks considering its greener infrastructure, the highly efficient and unbeatable speed with the possibility to connect everything to everything.

We have gone beyond the era of just fiber to the home (FTTH) as we focus now on the connection of homes, industries, 5/6G wireless base stations, businesses, smart cities, and several other massive high-speed technologies on a single-fiber infrastructure [1]. Unarguably, fronthaul, middlehaul, and backhaul jointly referred to as any-haul in 5/6G optical wireless communications (OWC) cannot sufficiently guarantee their expected output in terms of speed, latency, and spectral efficiency without improved channel architectures and that is where fiber comes in. Painstaking efforts are, however, ongoing for optimized and proper usage of fiber infrastructure in terms of the transmitter, channel, and receiver redesigning to meet the requirements of future broadband communications [2].

One major area of upgrade and infrastructure investment is the passive optical networks (PONs) which will serve as an integrator for different traffic [3]. PON is an enabler for the delivery of high-speed gigabit broadband services to end users. The current advancement in PON is the introduction of 25G high-speed PON (HSP) [1] and the plan is on top gear in the development of 50G HSP. Higher-speed PON like 100 Gb/s have also gained attention, which is considered to be the future of PON. Nevertheless, as the data-rate is increasing to meet the bandwidth demands of new and future applications, there are significant challenges associated with bandwidth limitation of opto/electronic devices and channel impairment, such as chromatic dispersion (CD) and other transmission impairments [4].

25/50G HSP has been agreed to use an O-band wavelength to eliminate CD. Future PON, which may be an upgrade of flexible 10G per wavelength time and wavelength division multiplexing (TWDM-PON) to a higher rate in C and L bands, for instance, 100G per wavelength will eventually trigger high CD. Without doubt, 6G communication fronthaul may need such a high data-rate to cope with the required speed and latency. Hence, there may be the need to use an advanced modulation approach such as self-coherent or full-coherent scheme as it will be highly challenging to achieve such transmission with intensity modulation direct detection (IM-DD).

In all these, cost, energy efficiency, and footprint are important factors that must be properly planned as 5G/6G strictly depends on sustainability and simplicity. In this chapter, PON evolution in terms of capacity improvement of each released standard is presented. Furthermore, the simplicity and complexity of different modulation technologies with some results to showcase performance are also presented, followed by a proposed architecture of inclusion of PON in 5G and beyond mobile broadband architecture. We also study in more detail the possible modulation technologies and possible PON architectures that are suitable for 5G and beyond fronthaul transmission.

2. PONs and evolution beyond 50G

With typical PON technology, broadband data can be shared between optical line terminal (OLT) located at the central office (CO) and several optical network units (ONU) located at the end users, with several other transmission coordination suites such as management protocols, convergence layer and physical medium dependent layer [5, 6].

The connection between the OLT and ONU can be guaranteed through optical distribution networks (ODN), a fast-speed, high-bandwidth optical link that operates as a passive connection without electromagnetic interference [7].

PON connections are usually referred to as Fiber to the X (FTTX) where X could be home, industries, buildings, and other end users [8]. It can also be referred to as the last mile of fiber communications which serves as a connection from the CO to end users. The major goal of PON technology is to guarantee high speed data transmission and to reduce the cost of service provision since PON is passive and does not require any power installations for its operations [9].

PONs have evolved for over two decades with the first ITU-T standard (G.983) published in 1998 for asynchronous transfer mode PON (ATM-PON or APON). APON has a downstream (DS) bitrate of 155.52 or 622.08 Mb/s and an upstream (US) of 155.52 Mb/s, respectively. The next standard released by ITU-T is called GPON with an improved bitrate of 2.5/1.2 Gb/s DS and US bandwidth. GPON networks have spanned millions of subscribers as the most widely used PON networks up to date having a maximum splitting ratio and a maximum reach of 1:64 and 20 km [10].

Furthermore, GPON was earlier upgraded from its nominal line rates to an XG-PON with asymmetric bitrates of 10/2.5 Gb/s. Further progress led to the abandoning of XG-PON and an upgrade to a full symmetric 10/10 Gb/s XGS-PON. Both XG-PON and XGS-PON technologies operate on the same frequency windows, having DS on 1577 nm and US on 1270 nm wavelengths, respectively [4, 11].

In 2015, the next generation passive optical networks (NG-PON2) were launched offering data-rate of 10 Gb/s per channel and an aggregate of 40 Gb/s, 4-wavelength multiplex transmission in DS connection [8]. While previous PONs mainly focus on broadband services and residential users, NG- PON2 technology is based on TWDM, a hybrid combination of time-division multiplexing (TDM) and wavelength-division multiplexing (WDM) with four or eight optional multiplexed wavelength for both DS and US communications. The bitrate could be symmetric with 10 Gb/s or 2.5 Gb/s per wavelength or asymmetric with 10/2.5 Gbps DS/US communications. NG-PON2 has a maximum splitting ratio of 1:256 and maximum signal reach of 40 km [12, 13].

Due to the continuous demand for high-speed transmission by heterogeneous applications, such as new radio and advanced video services on the current 5G networks, IEEE, Full Service Access Network (FSAN) and the Institute of Telecommunication Union, Telecommunication Standardization Sector (ITU-T), who coordinate technologies standardization, have proposed new access network standards beyond 10 Gb/s per wavelength, such as 50G Ethernet PON (50G-EPON) from IEEE and HSP from ITU-T, respectively [14]. Although HSP is still under discussion and investigation, there is a possibility of 50/25 or 25/25 Gb/s in DS and US directions.

The timeline of legacy PONs and the proposed future PONs with the aggregate DS/US data-rate evolution is shown in Figure 1. Future standards are also being forecast to either be 100G or 200G per wavelength as the case may be. These could be fixed or flexible PON technologies.

Figure 1.

PON evolution and standard from ITU-T and IEEE showing timelines, architectures, and bitrates.

2.1 Fixed PON

It is a technology whereby data-rate is uniformly provided to all connected ONUs based on the requirement of the most distant ONU to the OLT. The major disadvantage of this technology is bandwidth and power wastage since even the closer ONUs receive the same provision as the distant ones [15].

2.2 Flexible PON

PON can be made flexible in terms of data-rate provision by the OLT and in terms of ONU accessibility to resources in the OLT. By this approach, a single OLT can provide flexible data-rates to different ONUs which can be based on different modulation technologies [16]. Using software-defined coherent approach in [15], different phase modulation coherent technologies are multiplexed and made available to ONUs based on demands conveying data-rate up to 100 Gb/s on time division multiple access (TDMA) architecture. Another approach to flexible PON can be seen in TWDM-PON, which gives users a time slot on any of the architecture wavelengths [17]. Flexible PON creates an avenue for cost and energy minimization through sleep mode operation by switching off the unused channel while operating in full mode during burst hours [17]. With this approach, data-rate can be provisioned as 50–100G flexible rate PON for future networks which is allocated based on user demands [16].

2.3 Intensity modulation-direct detection (IM-DD) and challenges for future PONs

Intensity modulation is an approach that uses the amplitude property of the optical signal as a carrier to transmit information in the fiber. It is the oldest and simplest form of modulation in optical communications. To date, PON and other short-reach access networks are based on IM-DD non-return-to-zero (NRZ) signaling [18].

Signal transmission through PON usually requires a light source such as a DFB laser and an optical modulator. Although, some lasers acting as pulse lasers can act as a direct modulated laser (DML). An electrical information signal can be transmitted over an optical signal by modulating any of the light properties (intensity, frequency, or phase) before the signal is launched into an optical fiber. With DML, the electrical signal can be directly injected into the laser cavity to modulate any of the optical properties through stimulated emission. High-pulse broadening and induced chirp limit the use of DML for high-speed transmission above 10 Gb/s.

Another approach known as external modulation laser (EML) that uses a continuous wave (CW) laser and an external modulator such as an electro-absorption modulator (EAM), Mach Zehnder modulator (MZM) and ring resonator (RR), can guarantee improved transmission over DML but with extra complexity.

The simplest EML demonstration is with EAM which we have tested experimentally in our laboratory for 10/25 Gb/s PONs. EAM is a semiconductor device that can be used to modulate the intensity of a light beam by changing its absorption spectrum via applied electric voltage. In principle, EAM is designed through the Bulk process using the Frank–Keldish approach or through multi-quantum well (MQW) using quantum confined stacked effect (QCSE) [2]. QCSE is the dominant effect leading to electro-absorption of MQWs. It decreases the band gap energy of quantum wells which leads to a decrease in the entire absorption spectrum toward lower photon energies and this increases the overlap of electrons and hole functions. Investigation shows that MQW-EAM is preferred over Bulk-EAM for achieving a high extinction ratio (ER) as a result of the QCSE operation [8].

We have demonstrated both by simulation and experimentally, the use of EAM for NRZ 10/25 Gb/s transmission for single channel XGS-PON and HSP.

For the 10 Gb/s demonstration, a discrete indium phosphide (InP)-based EAM photonic chip of size 50 μm from Fraunhofer Heinrich–Hertz–Institut (HHI) is used. The setup for the experimental demonstration is shown in Figure 2. Here, an external DFB laser with 8.19 dBm output power at 1577 nm launches an optical signal into the EAM chip while a 10 GHz FPGA is used for 10 Gb/s NRZ electrical signal generation in addition to bias generated through a DC source. The electrical signal with DC is used to modulate an intensity property of the optical signal from the DFB as shown in the setup. The modulated signal is sent through a single-mode fiber (SMF), decoded with a PIN photodetector (PD), and passed through BERT for error analysis. Eye diagrams for back-to-back (B2B) and 20 km transmission are presented in Figure 3 while the BER curve is shown in Figure 4.

Figure 2.

Experimental setup of dynamic measurement with electro-absorption modulator (EAM) PIC for 10 Gb/s signal.

Figure 3.

Eye diagrams at the output of a PIN photodiode for (a) B2B and (b) 20 km, SMF transmission of 10 Gb/s downstream XGSPON.

Figure 4.

BER against optical power showing downstream performance of a 10 Gb/s XGS-PON for B2B and 20 km.

With the increase in demands for speed and high-traffic networks as signal bitrate moves above 10 Gb/s, dispersion sustainability, high power budget, signal reach, bandwidth requirement, receiver sensitivity, and high insertion loss (IL) are some issues to deal with in meeting 5G and beyond demands.

Current 25/50 Gb/s HSP has been standardized on O-band to eliminate the effect of CD which relaxes the implication of CD on high bitrate signal transmission. The current major challenges are the cost and complexity of developing opto/electronic components to meet the current bandwidth demands, especially the front-haul mobile access a with latency of about 150 μs.

To analyze symmetric 25 Gb/s O-band transmission, an experimental lab test was carried out. For the DS direction, a commercial 25G- EPON-small form-factor pluggable-28 (25G-EPON SFP28) was used as the OLT. The simple test setup is shown in Figure 5(a) consisting the OLT, an SMF, an attenuator, a power meter for optical power measurement and an ONU as the receiver. The OLT is an EML, and it is based on IEEE802.3ca which works both as GPON and XGS-PON with a maximum transmission rate of 25.781 Gb/s on 1358 nm wavelength. For the ONU, an LTF7603-BH+ Hisense commercial 25GS-PON SFP28 was used at 1328 nm wavelength and having data-rate of 25.781 Gb/s.

Figure 5.

(a) 25G-PON downstream experimental demonstration. (b) 25 Gb/s eye diagram of B2B downstream transmission using SFP28 OLT. (c) Downstream scenario BER curve for B2B and 20 km transmission.

Based on the setup in Figures 5(a), 25 GHz NRZ electrical signal is generated from a pattern generator and modulates the SFP28 through the test board. Transmission from an OLT to ONU was tested on B2B and 20 km SMF while an attenuator was used to vary the optical power to the receiver. The decoded signal by the ONU is sent to an oscilloscope with the B2B eye diagram as shown in Figure 5(b). The transmission BER curves for B2B and 20 km are also presented in Figure 5(c). The impact of dispersion at the wavelength (1358 nm) is approximately 3.4 nm/nm/km, which has no significant effect on the quality of the received signal, confirming the approximation of the B2B and 20 km curves in Figure 5(c). According to the expected value of sensitivity, a BER of 10⁻⁹ was achieved with −21.4 dBm, ensuring an improved performance than the value quoted in 25G-PON standard. By using the BER of 10⁻³ as a reference, a sensitivity of −26.6 dBm for B2B and − 26.1 dBm after 20 km of fiber was obtained.

In the US direction, the LTF7603-BH+ SFP28 at 1286 nm DML transmits 24.8832 Gb/s data-rate which was generated from a pattern generator to modulate the DML. The setup is presented in Figure 6(a).

Figure 6.

(a) 25G-PON upstream experimental demonstration. (b) 25 Gb/s eye diagram of B2B upstream transmission using SFP28 ONU. (c) Upstream scenario BER curve for B2B and 20 km transmissions.

At the receiver side, in this case, the OLT is used having an APD-based receiver capable of decoding a 25 Gb/s signal. Transmission here through SMF is the same as that of DS scenario. The decoded signal is sent to an oscilloscope and the eye diagram is presented in Figure 6(b). B2B and 20 km transmissions for the US direction are presented in Figure 6(c).

For a sensitivity of −22.4 dBm, BER values of 10⁻⁵ and 10⁻⁷ were obtained for B2B and for 20 km, respectively. Based on the results, it can be seen that better performance was achieved when at 20 km, which is related to the dispersion since the wavelength used for transmission is in the negative dispersion range (approximately −3.8 ps/nm/km). As a result, instead of the fiber having a negative impact and causing signal degradation or interference between symbols, it compensates for this effect at this transmission distance.

Other forms of intensity modulation to alleviate the bandwidth limitation of NRZ can be achieved through a multi-level approach such as pulse amplitude modulation level-4 and level-8 (PAM-4/PAM-8) that transmit 2 and 3 bits per symbol. PAM-4 and PAM-8 present an attractive alternative to NRZ in terms of bandwidth management. While PAM-4 doubles NRZ spectral efficiency (SE), PAM-8 could achieve a double SE of PAM-4 with two and three bits of logical information per clock period in PAM-4 and PAM-8. Nevertheless, these technologies also suffer from a low signal-to-noise ratio (SNR) due to reduction in amplitude of the eye signal compared to NRZ [19, 20]. Besides this, there is a need for a power hungry and expensive digital-to-analog converter (DAC) for signal encoding in PAM-4 and PAM-8 schemes with increased complexity as the level increases. This introduces an extra challenge as PON and 5G are highly cost-sensitive.

2.3.1 Electrical digital pre/post compensation IM-DD

An electrical pre/post-compensation approach can be used to mitigate major challenges of transmitting high bitrate signals with IM-DD [3, 21, 22, 23]. Although these introduce an extra cost and complexity to PON technology but, reduce most of the channel impairment challenges [21]. Hence, the cost and complexity are trade-off with performance.

For pre-compensation, simple electrical equalized digital signal processing (DSP) can be introduced before optical signal modulation. This could be in the form of signal pre-emphasis to compensate for opto/electronic bandwidth limitation or CD pre-compensation for channel impairments mitigation. Common DSP for pre-compensation utilizes a feed-forward equalizer (FFE) finite impulse response (FIR) filter placed before an optical modulator at the OLT. The complexity of the DSP depends on the nature of the channel, bitrate, and opto/electronic components used for the design. For instance, in [4, 23], a very heavy DSP algorithm coupled with highly expensive in-phase and quadrature-MZM (IQ-MZM) was used for 100 Gb/s per wavelength transmitter design. This, of course, increases complexity and cost for PON and may not be an acceptable option for 5G and beyond, at least for now.

For the post-compensation, a highly impaired optical signal decoded by a PD at the receiver can be equalized electrically to reshape the signal to its original form. Some of the common post-equalizations used for this purpose are FFE, infinite impulse response (IIR), decision feedback equalizer (DFE), continuous time linear equalization (CTLE), maximum likelihood sequence estimation (MLSE), and several other low-heavy DSP equalizations [4, 22, 23, 24].

Considering the cost and complexity of introducing DSP in PON, planning of the best location to implement the DSP is important. For instance, placing a heavy and expensive DSP at the OLT can share the cost among several connected ONUs. This could be more advisable than implementing a DSP at each of the ONUs. In this case, as seen in [4, 23], simple equalization with less cost can be implemented only at the receiver. An illustration of DSP-based OLT with options of low/high-cost opto/electronic receivers is presented in Figure 7. Here, expensive DSP can be introduced at the OLT while ONUs of varying complexity and cost can be implemented as illustrated in the diagram.

Figure 7.

50G-HSP with pre-equalized DSP at OLT and different receiver’s specifications redrawn from [11, 18].

2.4 Complexity and cost leveraging optical-domain-based-DAC PON

Multi-level intensity modulations (PAM-4 and PAM-8) require electrical DAC for signal encoding which introduces an extra cost to implement. One simple alternative to solve this problem is using optical-domain-based-DAC [25, 26, 27]. optical-domain-based-DAC multiplexes two or more binary signals with different amplitudes in the optical domain through either hybrid modulation [25, 28], dual drive MZM approach with power divider [26], or a polarization multiplexed 2-external modulator cascaded in parallel [27]. Optical-domain-based-DAC approaches are not only possible for intensity modulation or binary signal multiplexing but researchers in [29, 30] also demonstrated the possibility of multiplexing two or more PAM-4 into an array of nPAM-4 using cascaded MZM. They went further to demonstrate the generation of quadrature amplitude modulation (QAM) in a coherent (COH) scheme by multiplexing several QAM signals using optical-domain-based-DAC, eliminating the expensive electrical DAC in the conventional approach.

A simple simulation-based implementation of the approach for multilevel intensity modulation using two cascaded electro-absorption modulators and a DFB laser at 1550 nm is presented in Figure 8(a) while the eye diagram resulting from the modulation can be seen in Figure 8(b) for 25 GBaud signal transmission. With this approach, two binary NRZ signals are successfully multiplexed using optical-domain-based-DAC into a PAM-4 optical signal as shown in the eye diagram.

Figure 8.

(a) All optical DAC modulation scheme for PAM-4 signal generation (b) eye diagram 25 GBaud PAM- 4 signal generated using OpticalDAC scheme. CW = continuous wave laser, MSB/LSB = most/least significant bit, L = EAM-length, ER = extinction ratio.

2.5 Self-coherent PON

Scaling the capacity of short-reach optical links relies on using multiple wavelengths or multiple fibers carrying conventional noncoherent modulation formats such as on-off keying (OOK) and, more recently, higher-order PAM [31]. These modulation formats utilize the intensity property of light for information encoding which becomes more challenging as the demand for symbol rate increases.

In addition, intensity modulation schemes have little tolerance to linear propagation effects, such as CD and polarization-mode dispersion and therefore, further scaling of the system capacity to meet future demands may require the use of conventional full-coherent receivers (COH).

COH are widely used in long-haul communications which requires optimization and redesigning to meet the cost and complexity requirements of short-reach applications [32, 33]. COH is based on the intra-dyne scheme, which requires two optical hybrids and four pairs of balanced photodiodes for dual-polarization transmission systems, making its overall cost unacceptably high for short-reach applications. Short-reach applications are highly cost-sensitive, single PD-based DD transceivers will continue to be relevant due to its low cost and complexity [34] but they impose an irreversible loss of phase information upon the square-law detection.

Self-coherent (SCOH) transceivers based on a single sideband (SSB) scheme can be utilized to recover the missing phase information through the carrier-signal beating terms in DD optical communication systems [35, 36]. Nevertheless, the performance of SSB systems is severely degraded because of the signal-signal beating noise (SSBN) generated upon square-law detection. This is because of the interference caused by SSBN on the wanted baseband signal to carrier beating terms within the signal’s bandwidth which significantly degrades the receiver’s sensitivity [37].

To reduce the impact of SSBN, a minimum phase signal-based Kramers-Kronig (KK) receiver can be employed [38, 39]. The minimum phase condition of the signal implies that log magnitude and phase are related by the Hilbert transform, and this requirement can be fulfilled by adding a constant direct current (DC) value in the SSB complex signal [40]. However, the nonlinear operations (logarithmic and exponential) in the Kramers–Kronig algorithm demand the DSP to be operated much faster than the Nyquist sampling rate to accommodate spectral broadening [41].

Some approximations have been proposed to reduce the sampling rate, but these tend to require higher tone powers which result in a receiver sensitivity penalty [42, 43]. A possible solution to reduce the sampling rate is to find a way of getting around the execution of nonlinear operations. For instance, an upsampling-free KK method uses a Taylor expansion approximation of the associated nonlinear operations which require a higher carrier-to-signal power ratio (CSPR) [41]. To address the aforementioned high sampling rate and higher tone power requirements of the conventional KK and upsampling-free KK methods, respectively, a DC-Value method exploring the SSB and DC-Value properties of the minimum phase signal was proposed [44]. The DC-Value method has the potential to provide an upsampling-free phase reconstruction process at low tone power operation [44]. In the following, the transceiver architecture of the SCOH setup is described in detail.

2.5.1 SCOH transceiver architecture

A general architecture for the minimum phase signal-based SCOH transceiver is proposed to satisfy a minimum phase condition upon detection. The transceiver setup can be used for both signal reconstruction techniques namely, KK and DC-Value method. The required minimum phase condition can be achieved by an SSB signal transmission with a high enough CSPR value. We have proposed two different techniques to generate an SSB signal with the desired CSPR value.

First, a digital method is proposed where the carrier is inserted into the complex baseband signal in the transmitter DSP stage. The optical method is the second approach, where the carrier is inserted in the optical domain. In the following, we present a brief discussion about the general architecture of the SCOH transceiver in consideration of the digital and optical methods.

Transceiver architecture of SCOH based on the digital method is presented in Figure 9. The complex baseband signal with bandwidth B can be generated at the transmitter side by employing advanced modulation formats such as QPSK and QAM. Following that, a virtual carrier at a frequency of $f_o=\frac{B}{2}$ is added with a high enough power (notice that $f_o=-\frac{B}{2}$ is equally possible). In addition, the signal is passed through the DAC and then optically modulated using an IQ Mach-Zehnder modulator (IQ-MZM). This IQ-MZM generates an SSB optical signal which is launched to an optical fiber after an optical amplification. The incoming signal is first amplified at the receiver end using an optical pre-amplifier and detected using a signal PD. The signal is then captured using a real-time oscilloscope and processed offline using a receiver DSP. In the receiver DSP, first, the full optical field of the signal is reconstructed using a signal reconstruction algorithm (such as the KK method, upsampling-free KK method, or DC-Value method) and then applied to the standard coherent receiver post-DSP for the symbol detection.

Figure 9.

The general architecture of the minimum phase signal-based SCOH transceiver with a digital method to generate an optical SSB signal. TX: Transmitter, RX: Receiver, ODN: Optical distribution networks.

The transceiver architecture for the optical method is shown in Figure 10. In this case, the complex baseband signal is first modulated using an IQ-MZM like in a standard coherent transmitter. Further, a continuous wave (CW) tone at frequency $f_c+\frac{B}{2}$ is generated from the laser source using a frequency shifter. The power of a CW tone can be varied to achieve the desired CSPR value. The output of an IQ-MZM and frequency shifter output is added using an adder (i.e., an optical coupler) that generates an SSB signal with the desired CSPR value. The rest of part channel and receiver are the same as we discussed in the digital method.

Figure 10.

The general architecture of the minimum phase signal-based SCOH transceiver with an optical method to generate an optical SSB signal.

2.5.2 Experimental results

The experimental results for the digital method are shown in Figure 9. To reduce the cost, the booster amplifier is not used during the experiment. Since no booster amplifier is employed in the link, the launch power is adjusted by carefully increasing the modulation depth (by increasing the arbitrary waveform generator (AWG) peak-to-peak output voltage) of the IQ-MZM. For 30 Gbaud QPSK transmissions, the AWG output is varied from 200 mV_pp to 675 mV_pp with a step size of 25 mV_pp. The graph of log₁₀ BER with respect to modulation depth and CSPR after 80 km SSMF (2 spans of 40 km each) is presented in Figure 11. This analysis helps in identifying an optimum value of the CSPR and AWG peak-to-peak output. The results shown in Figure 11 indicate that the system performance gets deteriorated at both extreme points of CSPR and modulation depth. If CSPR is too low, then system performance is limited by SSBN perturbation. On the contrary, too high CSPR tends to induce receiver sensitivity penalty. Likewise, lowering the modulation depth results in low-launched power which degrades the optical signal-to-noise ratio (OSNR) at the output of the pre-amplifier. Also, the pre-amplifier saturates in case of very low optical is received at the receiver end, and limits the system performance. On the other hand, too high modulation depth causes an IQ-MZM to operate in a highly nonlinear regime which produces nonliterary in the system.

Figure 11.

After transmission of 80 km SSMF, performance analysis of 30 Gbaud QPSK as a function of CSPR and modulation depth employing DC-value method (a) 3D view (b) 2D view. Redrawn from [45].

Furthermore, as shown in Figure 11, the optimum operating point lies around AWG = 550 mV_pp, with a CSPR of 10 dB. Further increasing the AWG output voltage (greater than 550 mV_pp) causes the modulation nonlinearities to act as a dominant source of noise and degrade the system performance [38].

Results of the performances of the three different signal reconstruction methods, namely KK, upsampling-free KK [41], and the proposed DC-Value [44] (with 5 iterations) at the optimum AWG operating point of 550 mV_pp and 2 samples per symbol (SPS) DSP are presented in Figure 12. These results show that the proposed DC-Value method requires 2.7 dB less CSPR to achieve the same accuracy as an upsampling-free KK method and provides 13% BER improvement. In addition, when the modulation depth is increased above 550 mV_pp, the modulation nonlinearities become a dominant source of noise and degrade the system performance. Also, in Figure 13, results of the system performance for AWG operating at 650 mV_pp are presented with an indication that the overall system performance is degraded by modulation nonlinearities for all three signals.

Figure 12.

System performance as a function of CSPR at an optimum AWG operating point of 550 mVpp employing three signal reconstruction methods. (FEC limit @ 2.4 × 10–2 with 20% overhead). Redrawn from [45].

Figure 13.

System performance as a function of CSPR for an increased modulation depth to 650 mVpp employing three signal reconstruction methods. (FEC limit @ $2.4\times with20\%overhead)$. Redrawn from [45].

2.6 Full-coherent PON

Coherent optics is an advanced approach that offers an improved signal transmission over the conventional IM-DD. Through this approach, the amplitude, frequency, polarization, and phase of the optical signal can be simultaneously modulated and transmitted across the fiber for higher spectral efficiency. The complete modulated signal is decoded at the receiver through integrated coherent receiver and DSP [46].

The need for coherent optics arises due to limitations of traditional IM-DD when data-rate increases above 50 Gbps ITU channel, leading to spectral broadening and inter-symbol interference (ISI). With a coherent approach, several bits can be encoded into a symbol and several symbols can be transmitted on the same fiber. Some advantages of coherent transmission are the generation of a high-power budget with a high splitting ratio through the introduction of a local oscillator (LO) at the receiver called coherent receiver. Electrical filter is also used to increase wavelength selectivity.

The major constraint of coherent optics is its complexity and cost of installation which makes coherent PON far from reality in the current and near future PON. For instance, the need for LO in coherent optics and several number of photodiodes at the receiver will require high cost. Besides this, high multi-level modulation requires high speed analog and digital converter (ADCs) as well as digital to analog converter (DACs) which are highly expensive [8].

Nevertheless, several works have contributed to high-speed coherent-PON proposing its possibility for future PON. The work in [47, 48] presented a contribution to coherent-PON in preparation for 6G networks. The researchers in [47], proposed a 100 Gb/s to 1 Tb/s coherent-PON using WDM architecture and DSP receiver. The purpose of the work is to allow over-provision of the required bandwidth of 5G and beyond networks at the fronthaul section which requires very high bandwidth as we will discuss later in this chapter. Likewise, the review of high-speed coherent PON was made in [48] to study the possibility of improving power consumption and footprint which are some major challenge of introducing coherent optics in PON. Some possible suggestion as mentioned in the chapter is the miniaturization of DSP in order to reduce footprint and allow its application in small form-factor aperture (SFP). As coherent PON is not the direct objective of this chapter, only its objectives and limitations are briefly mentioned. Further study is required for the likeliness of its usage in future PON.

2.7 Possible architecture and their challenges for future PON

In the current phase of modern innovations and explosive bandwidth demand in broadband communications, it is generally believed and accepted that PON will play a highly significant role in next generation mobile/wireless communications in any-haul scenarios [49]. Continuous architectural forecast and planning is therefore imperative to put PON in the global acceptable position for 5/6G communications. This will enable fiber and wireless (5G/6G) convergence for enabling green transmission and moderate cost [49].

Architecture plays a significant role in overall PON system performance in terms of spectral efficiency, cost, complexity, speed, and system impairment management. All these factors must be adequately considered while choosing the right architecture for 5G/6G any-haul scenarios.

Most of the past PON releases (APON, GPON, XGS-PON) and the current HSP are based on TDM-PON. TDM-PON is a single wavelength architecture with centralized bandwidth allocations to end users based on their demands. At the DS region, OLT broadcasts its bandwidth to all ONUs with media access control (MAC) addresses of each of the users. The ONU, however, while receiving the broadcast, select its right data packet based on the MAC address in its memory. US transmission is based on time allocation and therefore, each ONUs utilizes full US bandwidth at its given time slot [50]. Due to the simplicity and wide acceptability of TDM-PON, there is high chance that it will be used for 5G and future networks fronthaul. Nevertheless, issues with latency are another challenge that we will address in the subsequent section.

The major problem of TDM-PON is the limited resource availability to end users which can be solved with wavelength division multiplexing PON (WDM-PON).

WDM-PON, on the other hand, multiplexes several wavelengths at the OLT which are then assigned to each ONU in the DS direction. WDM-PON can offer unlimited bandwidth to all users through dedicated wavelength allocation per user [51] which solves the common issue with TDM-PON. There are challenges of high-power consumption and tunability in the ONUs section of WDM-PON which could be an issue for 5G front-haul system.

TWDM-PON is similar to WDM-PON having a minimum of four wave- lengths/channels but each channel on TWDM-PON can utilize any of the wavelengths in the architecture [13]. It is a hybrid combination of TDM and WDM PON whereby a typical TDM-PON is embedded within each of the wavelength of a WDM-PON [50]. The major challenges of TWDM-PON are the tunability issue at the ONUs, burst mode receiver complexity at the OLT side, four wave mixing degradation, high-optical power budget and ONUs high-power consumption.

The last possible architecture is the point-to-point PON (PTP-PON), which uses a dedicated optical cable directly from the OLT to each ONUs. The cost of implementation is the major disadvantage of PTP-PON.

Table 1 gives a comparative summary of the advantages and disadvantages of each of the PON architectures mentioned above and the table serves as a guide for 5G/6G any-haul system plan.

PON Architecture	TDM-PON	WDM-PON	P2P-PON	TWDM-PON
Power Budget	moderate	low	low	moderate
CD	high	high	high	high
ONT required power	moderate	moderate	moderate	moderate
Burst mode Tx at OLT	moderate	—	—	high
ONT tenability	—	moderate	moderate	high

Table 1.

Challenges of different PON architectures in planning for future PON.

3. Broadband any-haul PON for 5G

The global architecture of mobile 5G networks will link global core networks to user equipment with in-between connections comprising of back-haul, mid-haul and front-haul respectively. Each of the connectors is envisioned to be of fiber links and most probably PON-FTTH [52]. The bandwidth requirement at each link, however, varies and this is a major area of concern in 5G network planning. The major interest here is not on the radio access architecture but majorly on the optical bandwidth specifications to meet the system requirements of each optical link in the global architecture.

3.1 General architecture

An architectural view of bandwidth requirements at each segment of mobile broadband communications from the 5G core to the user equipment is presented in Figure 14. The latency of backhaul, midhaul, and fronthaul of 5G are 10 ms, 1–2 ms and 50–200 μs, respectively. It is more challenging to meet the fronthaul latency and bandwidth requirements and at such, a very high speed more than 50 Gb/s data-rate would be required and at least a 100 Gb/s could be suitable for future 6G fronthaul transmission. In this category, a more sophisticated modulation approach such as coherent 100G-PON is more appropriate but this introduces extra cost and energy consumption. For the midhaul, it would be more suitable to use a 25G-PON for mobile connection since its latency is more relax when compared with fronthaul section.

Figure 14.

Possible PON architecture connecting core cloud to RU which is then accessed by the user equipment’s (UE) over PON channel and possible modulation technology to meet the bandwidth and latency requirements at each section.

3.1.1 Fronthaul

The connection between the distributed unit (DU) and radio units (RUs) is called fronthaul in 5G-RAN broadband communication. The DU and RU are traditionally called baseband unit (BBU) and remote radio head (RRH) in 4G networks [53]. RUs are several antennas that operate on multiple input and multiple output (MIMO) and user equipment (UE) can access signal through a wireless connection to the RUs as illustrated in Figure 14. Fronthaul as the last mile connection is highly challenging because of high sensitivity requirement to meet the throughput, latency, and reliability demands of advanced 5G applications which makes fronthaul to be the most important architecture in 5G technologies [54, 55, 56]. For instance, the latency in this connection must not be more than 150 μs to meet channel conditions both for user experience and for RAN signaling processed at the DU such as hybrid automatic repeat request (HARQ). Hence, it requires very high speed modulation, for instance, above 50 Gb/s bitrate. A 100 Gb/s specified in IEEE 802.3 and mentioned in [57] for fronthaul internet aggregation could be very suitable for 5G fronthaul connection. This could be possible only on the O-band around 1300 nm region due to CD impairment if we are to use IM-DD except we use self-coherent or coherent modulation as suggested in [58] in order to stay in C or L band which will also require advanced DSP. It is also important to note here that transmission over fiber introduces some latency due to forward error correction (FEC) in the order of milliseconds in addition to several switching, queuing and buffering processes that are carried out during the round trip time signal processing. This adds extra complexity and challenges to cope with high-bitrate transmission that requires FEC [57].

In terms of PON architecture, a suggestion of flexible PON on TWDM would be more appropriate having several flexibility and advantages over fixed TDM-PON.

Nevertheless, it is a trade-off between cost/complexity and efficiency. With TWDM-PON, advanced modulation with 50 Gb/s or more per wavelength may be possible in C or L band with additional DSP for channel impairment management. An illustration of this suggestion is also indicated in Figure 14.

Since 6G is already under discussion and forecasts to be launched in 2030, the required fronthaul bandwidth and latency should be 10 times better than 5G. According to [59], the pick data-rate expected is about 1 Tb/s with latency as low as 10 𝜇s and SE 10–100 times better than 5G. It is also envisaged that the 6G will operate on very high frequency, above 20 GHz unlike the sub-6 GHz frequency used  by 5G. In this frequency, the true advantages of millimeter wave (mm-wave) will be established. Technologies such as advanced software defined networks (SDN), advanced network function virtualization (NFV) with full-scale artificial intelligence beyond what is currently implemented in 5G will be introduced in 6G [60].

3.1.2 Midhaul

In 5G, midhaul is the connection between DU and CU. Its required bandwidth is close to that of backhaul with the exception of tighter latency than the latency of backhaul. The combination of fronthaul and midhaul in 5G establishes the 5G networks as a technology of choice with varieties of traffics on a single trunk. If DU and CU are together, there would be no need of midhaul and traffic from DU would be directly connected to core networks through backhaul channel [61]. According to Figure 14, midhaul latency must not exceed 2 ms which relaxes its latency while compared to fronhaul counterpart with stricter latency of 150 μs. Also, in terms of distance span, it is envisaged to be around 20–40 km. For modulation bandwidth, a HSP of about 25 Gb/s should be sufficient for smooth transmission within the midhaul links [53, 58].

3.1.3 Backhaul

The wired/wireless connection between the CU and the cloud core is referred to as backhaul in 5G networks [60]. This connection is very essential to 5G networks, and it requires high bandwidth and large capacity to manage millions of devices that are accessing the core networks from remote connections to the CU. Several traffic types are converged into a single channel infrastructure which reduces the cost of managing several connections to the core but it is highly challenging. Although, debate has been ongoing on either to connect 5G backhaul on wired or wireless. Wired connection through fiber seems the best option due to its large bandwidth and low latency. The major disadvantages of fiber backhaul are the cost and complexity of installation most especially in smart cities scenarios. Nevertheless, fiber as an option can be sufficiently set to provide over 25 Gb/s speed with less than 10 ms latency using a coherent modulation technology that could span for several kilometers as illustrated in Figure 14.

The work in [47] presents a typical illustration of coherent PON which could serve as technology for 5G backhauling. In the work, WDM approach is used with 8-channel coherent WDM-PON with 100 Gb/s bit rate per channel totaling 800 Gb/s. Although, the cost of DSP is the major challenge in this type of modulation, Nevertheless, it provides sufficient bandwidth and it can cover long-distance requirements for 5G backhauling. As described in Figure 14, considering the huge traffic that are accessing 5G core networks, DWDM-PON with a sufficient bit rate will be most suitable to leverage the traffic. The DWDM also has its setbacks in terms of cost and complexity.

4. Conclusions

In this chapter, the possibilities of using PON for 5G architectural plan were reviewed. Considering the high bandwidth possibility and cost efficiency of PON, bandwidth and latency requirements of 5G networks can adequately be met using PON. However, different modulation technologies are used in PON with the legacy PONs based on intensity modulation. However, as the bit rate increases, advanced modulation approaches such as semi-coherent and full coherent modulation are employed such as 100G-PON for complex and long-reach transmissions. This chapter provides the possibilities to implement 5G connections using any of these modulation approaches and further reviews different PON architectures for managing several traffics and increasing system spectral efficiency.

Acknowledgments

This work is supported by the European Regional Development Fund (FEDER), through the Competitiveness and Internationalization Operational Program (COMPETE 2020) of the Portugal 2020 framework [Project POWER with Nr. 070365 (POCI-01-0247-FEDER-070365)].

It is also supported in part by the Fundação para a Ciência e a Tecnologia (FCT) Fellowship under Grant: DOI 10.54499/2021.03815.CEECIND/CP1653/CT0005 (https://doi.org/10.54499/2021.03815.CEECIND/CP1653/CT0005). It is also supported in part by FEDER, through the CENTRO 2020 programme, project ORCIP under Grant CENTRO-01-0145-FEDER-022141 and MSCA RISE programme through project DIOR under Grant Agreement 10100828.