Open Access
Add to BibSonomyAbstract
This is the second of a two-part paper intended to provide technical insight and rationales behind the recently approved ITU-T G.989.2 Recommendation: the physical media dependent layer specification of the 40-gigabit-capable passive optical networks (NG-PON2). While Part 1 of the paper discusses topics related to the optical link design, Part 2 focuses on wavelength control, technology feasibility, management and control channel design, and potential future standardization directions of such a multi-wavelength PON system. As the NG-PON2 system will continue to evolve, technology extensions are also discussed to provide guidance for future research.
© 2015 Optical Society of America
I. Introduction
In this second part of the paper reviewing physical media dependent (PMD) layer aspects of the NG-PON2 standards [1], we examine wavelength control, technology feasibility, management and control channel design, and potential future directions for the NG-PON2 PMD standard.
Two technologies are specified in the NG-PON2 standards, namely, a hybrid time and wavelength division multiplexed (TWDM) PON and a point-to-point wavelength overlay (PtP WDM). One of the challenges to implement both of these technologies for a NG-PON2 system is to realize the tunable transmitter (Tx) at the optical network unit (ONU) at the low-cost points necessary for residential fiber-to-the-home services. In Section II, the requirements on the tunable Tx wavelength characteristics are reviewed along with potential methods to meet them. In Section III, the main subsystems are discussed and the envisaged technologies to realize them are described. In some cases, small modifications of existing technologies are all that is required, but, for the more challenging functions and options, significant steps in technology are needed.
In Section IV, we describe new management and control channels to enable NG-PON2 ONUs to be managed when the conventional approaches, based on in-band methods, are not possible for various service and operational reasons.
Even though the ITU-T G.989.2 Recommendation was only recently completed, a range of possible enhancement options is already being considered. In Section V, potential future directions of the NG-PON2 PMD standard are foreseen with particular emphasis on enhanced capacity.
II. Wavelength Control
In this section, we discuss three key aspects pertinent to control the multiple wavelengths in an NG-PON2 system, namely, wavelength stability, wavelength accuracy, and wavelength locking. The related wavelength control parameters, specified in G.989.2, are also introduced.
A. Wavelength Stability
The wavelength stability of an NG-PON2 system is qualified through a parameter called maximum spectral excursion (MSE). MSE is specified for two main reasons: 1) to prevent optical power in one wavelength channel from leaking into an adjacent wavelength channel causing cross-talk induced degradation, and 2) to ensure that the transmitter operates within the desired wavelength multiplexer (WM) channel passband to achieve the specified link performance. The spectral excursion of a Tx in a stationary wavelength channel state is defined as the absolute difference between the nominal central frequency of the wavelength channel and the point of the Tx spectrum furthest from the nominal central frequency.
For TWDM downstream (DS), the channel spacing (CS) is 100 GHz and the required MSE is . In this case, it is relatively simple for a tunable filter to follow the maximum received power through a local feedback loop; thus tight wavelength calibration of the ONU Rx filter is not necessary.
For the upstream (US), the CS is not fixed and can range from 50 to 200 GHz, with the MSE specified for three specific values of CS, as shown in Table I. These MSE values were selected by taking into account the need to avoid overly stringent requirements on wavelength calibration (for calibrated lasers), power variation over wavelength, and tuning granularity.
TABLE I. MSE Values for the NG-PON2 US Direction (the MSE for Intermediate Values of CS is Interpolated)
Transmitter optical power wavelength dependency is specified in G.989.2 to be and for 2.5 and 10 Gb/s upstream US transmission, respectively. It was necessary to define these limits because optical power wavelength dependency causes a tuning error, when a wavelength locking mechanism based on a received optical power measurement is used (see Subsection II.C). This tuning error, along with the MSE specification, restricts the allowed Tx spectral excursion. A maximum tuning granularity of CS/20 was then chosen to avoid exceeding the MSE whenever a wavelength correction was attempted.
To arrive at the MSE values in Table I, the loss and isolation characteristics of both thin-film filters (TFFs) and arrayed waveguide gratings (AWGs) were considered, as well as Tx spectral widths, calibration accuracies, tuning granularity, and acceptable power penalties. While a full analysis is out of scope for this paper, some general considerations can be discussed.
For the example of , it is possible to stay within MSE using a laser with a spectral width of 15 GHz, calibrated to , and tuning granularity of 5 GHz (i.e., CS/20). When combined with a WM (with of adjacent channel isolation and a 0.5 dB passband width of 20 GHz), cross-talk penalty and insertion loss increase due to laser spectral excursion being kept below 0.1 and 0.5 dB, respectively.
Note that short-term spectral excursion caused by the ONU laser switching on/off at the burst boundaries must comply with the MSE limit. See proposed solutions to counteract this effect in [2–4].
B. Wavelength Accuracy
The calibration accuracy of an ONU’s tunable Tx, i.e., its capability of transmitting with a spectral excursion within specified limits, has an impact on its cost. Three levels of ONU tunable Tx calibration accuracy are defined in the NG-PON2 standards: sufficient calibration, loose calibration, and no calibration. Loosely calibrated and uncalibrated ONUs cannot be guaranteed to transmit within the MSE limit by themselves and strictly require additional mechanisms, as described in the following subsection.
C. Wavelength Locking
The intrinsic wavelength accuracy and stability of an ONU’s tunable optical devices may be relaxed to reduce cost, but this must be compensated for by adopting alternative wavelength locking mechanisms.
In the DS direction, as anticipated in Subsection II.A, the ONU receiver (Rx) can autonomously find the best wavelength tuning point by looking for the highest received optical power and/or the lowest error rate on its operation channel.
In the US direction, ONU Tx wavelength locking may be more complicated and rely on OLT feedback. In G.989.2, the wavelength locking method assumed in specifying the optical parameters is one in which the ONU Tx wavelength is “dithered” to obtain lock. To facilitate this, two assumptions are made: 1) an OLT Rx can perform a sufficiently accurate power, but not wavelength, measurement, and 2) the ONU Tx launch power does not vary significantly for small wavelength changes (as addressed in Subsection II.A). One possible implementation of the dithering scheme is to realize a closed wavelength control loop between the ONU and the OLT to reach and maintain the optimal ONU Tx tuning point. In this case, the filter shape of the OLT WM plays a fundamental role as the wavelength variation of the ONU Tx is transformed into a measurable power variation at the OLT Rx, as shown in Fig. 1.
Under OLT control, the ONU transmits at different times at two slightly differing wavelengths ( and ) around the central wavelength . If is perfectly coincident with the OLT WM channel center and if the WM channel passband has a symmetrical shape, then the OLT Rx would detect no power change between the different transmission times, in which case no corrective action is necessary.
For the cases shown in Fig. 2, the power change would reveal both the magnitude and sign of the wavelength misalignment. As a result, the OLT would instruct the ONU to apply an appropriate tuning correction.
Fig. 2. Dithering result for a mistuned ONU. (a) Negative correction required. (b) Positive correction required.
As outlined in Appendix VIII of G.989.2, to support the closed loop method described here, the WM may need to conform to additional constraints with respect to filter bandwidth, monotonicity, and passband ripple. Two factors must be guaranteed: 1) a mistuning of the ONU Tx is correctly detected by the OLT received power measurement, before the MSE limit is exceeded, and 2) power variations due to particular WM filter shapes (in addition to the previously addressed wavelength dependency of the ONU Tx) do not interfere negatively with the tuning mechanism. Note that, in the case of a high bit error rate (BER) prior to the application of forward error correction (FEC), the wavelength locking mechanism may be more conveniently based on a BER measurement, instead of on the power measurement that was assumed in the previous discussion.
III. Technology Feasibility—New and Challenging Subsystems
The fundamental decision regarding the choice of a multi-wavelength system with tunable direct-detection ONU transceivers for the NG-PON2 systems was made after a thorough analysis of potential alternatives. Compared to previous generations of standards-based PON systems, the decision has opened an additional area of flexibility and configuration options through the wavelength domain. This led to the requirement for new DWDM-capable components with cost-efficient implementations. In the following subsections, we focus on three key implementation choices that address different WDM aspects, namely, 1) tunable transceivers, 2) directly versus externally modulated lasers for ONUs, and 3) wavelength multiplexers.
A. Tunable Transceivers
NG-PON2 ONUs need to be equipped with tunable transceivers capable of CS of as low as 50 GHz. This leads to requirements and specifications concerning the tuning capability, including the operating wavelength band, spectral excursion, and tuning granularity. The requirement for tuning capability, of both Tx and Rx, increases the overall complexity of the transceivers. In combination with other requirements, such as support of a high loss budget for the optical distribution network (ODN) or high bit-rates, this would demand a sophisticated transceiver design. Tunable transceivers, especially for high ODN optical path loss (OPL) classes and bit-rates of 10 Gb/s, are therefore challenging for low-cost implementation.
Tunability in the ONU receiver can be achieved by placing tunable filters in front of the respective avalanche photodiode (APD). Ideally, this is done by integrating tunable filters into the Rx optical subassembly. Several tunable filter technologies have been shown. For example, 10 Gb/s tunable Rx, based on thermally tuned thin-film Fabry–Perot filters covering up to 800 GHz of spectrum, has been demonstrated [5]. This Rx has achieved a sensitivity of at a pre-FEC BER of (cf. commercially available APDs [6] with sensitivity at BER). Note that PtP WDM transceivers must work without FEC, as there are applications requiring low latency, e.g., mobile front-haul.
Narrow-band laser tuning can be achieved with, e.g., thermally tuned DFB lasers (up to tuning range) or electronically tuned (three-section) DBR lasers (up to ) [7]. Their implementation as small form factor pluggable plus (SFP+) modules is highly desirable, especially in pay-as-you-grow deployment scenarios.
Transmitters used for commercially available tunable SFP+ modules are mostly based on DBR laser designs with multiple tuning sections and potentially using Vernier effects to achieve full-band tuning [7]. In addition, these Txs use external Mach–Zehnder modulators to achieve low optical path penalty (OPP, see Part 1, Section V) at the required reach.
For SFP+ modules, a maximum power consumption of 1.5 W is specified in the multi-source agreement [8]. In order to achieve sufficiently low power consumption of the thermo-electric cooler (TEC, one of the main power consumers in an SFP+), the launch power of these modules is limited to around 0 dBm. Today, most tunable SFP+ modules fall into the range of 1.5–1.8 W for power consumption. Since a tunable filter must be added for NG-PON2 transceivers (possibly consuming another ), it is doubtful that the 1.5 W target will be achieved for launch power in the near future. Somewhat higher power consumption is tolerable within the SFP+ form factor, but this must be reflected in the respective host equipment (ONUs, OLT). Therefore, tunable SFP+ modules can be regarded as a starting point for reduced-cost tunable transceivers.
Some configurations in NG-PON2 may require additional optical amplification. For TWDM, the 10 Gb/s E1 OPL class in unamplified Type A configurations (see Part 1, Section III for more details) is challenging. Therefore, an amplified Type B configuration may be a cost-efficient alternative.
Similarly, for the highest PtP WDM bit-rate class, it is very challenging to achieve the N1 OPL class without optical amplification in the DS and without FEC. The G.989.2 specifications require ONU and OLT sensitivity of (at the R/S reference point) and (at the S/R-CG reference point), and ONU and OLT launch power of (at R/S) and (at S/R-CG), respectively. These reference points are defined in Part 1 of this paper. At the S/R-CG reference point, these specifications can be met using optical amplification in the OLT. For the ONU, this leads to the necessity of Rx with sensitivity of at least without FEC, and minimum launch power of the ONU Tx of . This sensitivity and launch power can be achieved by APDs [9] and cooled DFB or DBR lasers, respectively. However, both parameters need improvement for today’s commercial tunable SFP+ modules.
Optical amplification can be implemented at the OLT with a booster or preamplifier in the transmit and receive directions, respectively. Practically, this could be with per-channel semiconductor optical amplifiers (SOAs) integrated into the respective Tx or Rx, or with dedicated SOAs or erbium-doped fiber amplifiers operating on the respective channel group.
SOAs integrated with either Tx or Rx have been investigated for a long time [10]. In [11], a performance comparison of different booster amplifier implementations for 10 Gb/s Tx for PON applications is described. SOAs integrated with Rx are aimed at sensitivity increases, relaxed electrical amplifier requirements, or supporting bit-rates over 10 Gb/s [12].
B. Externally Modulated Versus Directly Modulated Lasers for an ONU Transceiver
Maintaining the ONU Tx wavelength within the allowed MSE for the US path presents several challenges, especially for 10 Gb/s.
The externally modulated laser (EML) Tx is well adapted to high data rate transmission over fiber distances up to 100 km due to its low frequency chirp. Figure 3 shows an optical spectrum of a 10 Gb/s EML Tx. The signal spectrum is about 14 GHz wide at the 15 dB point. If 100 GHz CS is chosen for the US channels, the allowed MSE is ±20 GHz. The accuracy to which the spectral peak of this Tx must be controlled, to stay within the MSE, is therefore ±13 GHz. Such an EML Tx is generally considered in the industry as feasible at acceptable cost with known technology and techniques.
The measured optical spectrum of a 10 Gb/s DML Tx is shown in Fig. 4. Due to modulation induced frequency shift, the DML spectrum has two distinct peaks corresponding to the “0” and “1” logic levels, respectively. The power difference between the two peaks is related to the ER of the Tx, in this case about 7 dB (close to the 10 Gb/s TWDM US requirement of 6 dB). The frequency chirp in this example is about 25 GHz, which is high for a typical DML laser as will be explained below. The spectral width of 35 GHz at the 15 dB point is 15 GHz wider than the EML due to frequncy chirp. Thus, the accuracy to which the spectral peak of this Tx must be controlled is . Such tight wavelength control is generally viewed as not feasible without integrating a costly wavelength locker in the Tx. In Fig. 4, it may also be seen that the “1s” peak and “0s” peak must straddle the filter center to fit within the MSE zone.
Fig. 4. Gaussian-shaped filter and optical spectrum of a 10 Gb/s DML Tx measured with a resolution bandwidth of 2 GHz and tuned to fit within MSE.
From the transmission perspective, although a DML is a good candidate for 2.5 Gb/s US transmission, it might not be so for 10 Gb/s transmission due to the chirp induced OPP. However, there is a method [13] to mitigate both the chirp and tight wavelength control difficulties in order to take advantage of the DML Tx. In an NG-PON2 system, a WM filter is used at the OLT prior to the Rx; see more details in Subsection III.C. With appropriate design, the filter will attenuate the power of the zero bits, thus simultaneously reducing the dispersion penalty, and enhancing the ER at the OLT Rx. Note that the WM filter shape is not defined in G.989.2 and is left to the implementer. In Fig. 4, we give the example of a common Gaussian-shaped filter passband, with the optical spectrum of the DML Tx overlaid. This DML is purposely designed with larger than normal chirp to benefit from the filtering method [13] described below.
Figure 5 shows the impact on the OLT Rx sensitivity from tuning a DML across the filter center wavelength. Note that here the Rx sensitivity is measured at BER and the power is measured directly at the Rx input, not at the S/R-CG where the OLT Rx sensitivity is defined in G.989.2. The actual OLT Rx sensitivity must be significantly better than what is specified at S/R-CG to account for power penalties, losses from the WM, connectors, and the diplex filter.
As seen in Fig. 4, tuning the laser to fit within the MSE zone requires a positive shift of 12 GHz for the laser peak (the “1s” peak). From Fig. 5, this leads to a degradation of the OLT Rx sensitivity from about (without filter, horizontal dashed line) to . If instead, the laser peak is shifted left to between and of the filter center, the sensitivity is enhanced to . With this placement, the DML has an optical frequency tolerance of , significantly larger than the original tolerance of . However, the potential cross-talk of the “0s” power into the left adjacent channel must be managed using a WM filter with adequate isolation.
In summary, with the MSE requirement applied to the “1s” peak rather than the “0s” peak, both DML and EML Tx may be used effectively for US transmission. The DML is more dependent on the WM filter shape and tuning accuracy. Whether the DML can now be considered a good candidate for 10 Gb/s transmission depends on the relative weight given to higher transmitted power (a DML advantage) versus required tuning accuracy (a DML disadvantage).
C. Wavelength Multiplexers
As the NG-PON2 system employs WDM technology, a WM is used at the OLT to combine and split the wavelengths for DS and US signals, respectively. The component technologies for the WMs could be AWGs or TFFs. WMs using TFFs are composed of a number of individual simple filter elements to create the desired WM properties. AWGs have a cyclic property of multiple passband wavelengths in a single port, with the free spectral range (FSR) designating the wavelength spacing between these passbands. In general, AWGs are widely deployed for a large number of wavelength channels () with a narrow CS of 100 GHz or less. This is because a large number of input and output ports are easily fabricated using silica-based planar lightwave technology.
On the other hand, a conventional TFF-based WM module offers a small number of channels (four is typical) because it consists of several TFFs and other free-space optics. However, TFFs have several advantages over AWGs, such as low insertion loss, low adjacent channel cross-talk (high isolation), and low cost when a small number of channels is needed. When TFFs are used for a large number of channels, these advantages gradually disappear. The crossover point where AWGs become more advantageous than TFFs is technology and vendor dependent.
In the rest of this section, we discuss three key aspects of WMs: cyclic transmission characteristics of AWGs, OLT configuration options using TFFs and AWGs, and adjacent channel cross-talk analysis of AWGs.
1) Cyclic Characteristics of AWGs:
AWGs have a cyclic property of multiple passband wavelengths in a single port. This cyclic property enables a specific implementation of TWDM that allows a wide variation in US wavelength [14]. In the context of this paper, the term “cyclic AWG” is used to designate an AWG with unique wavelength routing properties as shown in Fig. 6.
Fig. 6. Cyclic AWG US wavelength routing. The four colors represent four different OLT channel terminations.
The cyclic AWG depicted has 100 GHz CS with enough channels to occupy the entire US band from 1524 to 1544 nm, corresponding to the wide band option. No matter where the ONU laser is tuned within the US band, it will be less than 50 GHz from the center of a channel that routes to one of four OLT channel terminations (CTs). The routing of wavelengths to CTs is shown by different colors, and the routing repeats every 400 GHz.
With the cyclic AWG, it is possible to use an ONU laser that is thermally tuned with only a heating element. While the nominal wavelength of a laser may vary over most of the US band, if the heater can change the laser die temperature by , then the laser can tune to any desired CT. The main disadvantage of the approach is that, due to ambient temperature changes, the laser will have to retune (jump 400 GHz) to remain connected to the desired CT. If the tuning time is long, as it generally is with a thermally tuned laser, then the service outage from periodic retuning may not be acceptable. This approach does hold the promise of a low-cost tunable laser without the need for a TEC, although there may be other concerns related to limited launch power and increased ASE noise passing through the cyclic AWG.
2) OLT Configuration Options:
Here we review three use cases of AWGs and TFFs in an NG-PON2 OLT, as shown in Fig. 7. Configuration I is composed of a WM and transceivers that have a single wavelength pair of US and DS signals. Configuration II contains a WDM filter, WMs, and transceivers, which have dual optical interface ports. Configuration III consists of a WDM filter, WMs, a four-channel Tx, and a four-channel burst-mode Rx.
One of the advantages of Configuration I is that the transceivers have a single optical input/output interface port, which is a common feature in existing PON systems. Another advantage is that a single WM device leads to a simple filter configuration. However, if an AWG is used as the WM device, it offers off-grid CS (e.g., 105 GHz) for US channels in the C-band, while it has on-grid spacing (e.g., 100 GHz) in the L-band because of its FSR characteristics. TFFs cannot be adopted for Configuration I.
In Configurations II and III, both the TFFs and AWGs are applicable as WMs. Dual-port transceivers, which are common in general telecom markets, can be adopted, while optical connection among mux/demux and transceivers would need duplex patchcords. Configuration III is attractive because a small footprint can be expected if four-channel device integration technologies employed in high-speed Ethernet transceivers (CFP or QSFP pluggable modules) can be applied. However, Configuration III may not be able to support pay-as-you-grow deployment scenarios. In summary, specific OLT configuration depends on the types of WMs and transceivers to be employed and the network operator use case.
3) Adjacent Channel Cross-talk Analysis of AWGs:
Since G.989.2 specifies CS for US signals of 50 GHz minimum and 200 GHz maximum, the power penalty caused by inter-channel cross-talk may not be negligible. Here we provide an example study of the relationship between CS and adjacent channel cross-talk, denoted in the case of AWGs. Figure 8 shows the calculated dependency on CS for three different 3 dB passband values of , , and . The purple line at shows an example target value for 0.5 dB inter-channel cross-talk penalty, as described in Appendix VIII of G.989.2.
Fig. 8. Example study of the relationship between the channel spacing and adjacent channel cross-talk of AWGs with various passbands.
In the simulation, a semi-flat-top profile for the AWG passband was assumed. A passband width margin was set to be , which allows for an acceptable yield of AWGs based on the fabrication tolerance. As shown in Fig. 8, better can be obtained as the CS increases for the same passband. For the same CS, becomes smaller as the passband narrows. The results thus indicate that a smaller passband and a wider CS lead to a sufficient . Note that the results depend on specific parameters of AWGs. For example, if a smaller passband width margin of would be assumed, a better could be expected as shown with a green dotted line ().
IV. Management and Control
The fundamental difference between TWDM and PtP WDM systems in terms of management and control is that TWDM is an end-to-end specification of a complete system in layers 1 and 2. It extends the protocol specifications for prior PON generations, including management and control functions of the TWDM link. Meanwhile, PtP WDM is specified to transport client services with minimal or no payload modification; hence management and control features of the PtP WDM link must be added in a new way. An auxiliary management and control channel (AMCC) is thus defined to provide such functions.
The physical layer aspects of the AMCC were not included in the first published version of G.989.2. The intention of this section is to provide readers with advanced insight into the ongoing AMCC work.
A. PtP WDM AMCC
In PtP WDM, an AMCC is added to each individual wavelength channel in both transmission directions because the payload is passed transparently instead of being terminated by the PON. Transparency is needed, e.g., for mobile front-haul transport, to avoid any added latency. In such cases, no signaling channel would be available without an AMCC. Without a signaling channel, automated ONU tuning and monitoring would not be possible.
For PtP WDM, two physical layer methods of the AMCC are discussed: transparent and transcoding. The difference between the two methods is the way the AMCC content is transported over the physical channel, while the content of the AMCC remains independent from the transmission method.
To transport a payload bitstream transparently, without terminating any part of its frame structure, the AMCC has to be added to the payload at the same wavelength with minimum interference. This method is called transparent AMCC. In the case in which the payload data encoding is converted into a different line code, the AMCC data can be transmitted by code transformation. This method is called transcoding AMCC. Figure 9 compares these two methods. In both methods, the same management and control information is transmitted, and the interfaces to the higher-layer control systems are identical. Both methods differ in the PMD adaptation and PMD layers.
Fig. 9. Relationship of the AMCC TC layer with other system functions for the transparent (left) and transcoded (right) options.
Two implementations of the transparent AMCC are currently under discussion. These are based on baseband overmodulation and low-frequency pilot tones, respectively.
With baseband overmodulation, the framed AMCC data are amplitude-modulated on top of the payload bitstream as a baseband signal. It can be detected with a simple threshold detector after low-pass filtering of the receive signal. Currently, a low modulation index of is discussed to limit penalties on the payload data signals to . Furthermore, overmodulated AMCC bit-rates of 150 kb/s are under discussion and have been demonstrated.
The second implementation of the transparent AMCC is based on the use of a low-frequency pilot tone. At the Tx side, the client signal is mixed with the management and control channel in a low-frequency domain so as to offer transparent AMCC with minor interference between the AMCC and the payload data. The modulation depth of the management and control channel will be very shallow: it will be optimized to not harm the quality of the client signal. At the Rx side, the management and control signal is demultiplexed after the detection of the optical signal.
Transcoding AMCC adds management and FEC information into additional code words of the payload data-encoding scheme.
At the Tx side, the PtP WDM payload data in Code #1 and management information are combined into a new Code #2. FEC can be employed as a part of the transcoding to improve the transmission gain. The Rx decodes Code #2 to obtain the PtP WDM management information, and the payload data are converted back to Code #1. At both the Tx and Rx sides, there are no requirements for modulation and extra signal processing other than transcoding. The same chip sets, physical interfaces, and signal carrier of the PtP WDM payload data channel can be reused.
B. TWDM AMCC
The approach used in TDM-based PON systems for discovery and ranging of a newly connected ONU is to create a small traffic interruption (quiet window) in the US frame in which the ONU can send identification information, e.g., its serial number, without interfering with active ONUs.
In a TWDM PON, if the ONU Tx is not sufficiently calibrated in wavelength, transmission of the US discovery burst could occur in any US channel. It is therefore necessary to open the quiet window simultaneously on all channels. If this is not acceptable, e.g., because the channels belong to separate systems or independent network operators or if the service on one channel should not be interrupted by ONUs being added to another, a new approach must be followed.
One approach is the adoption of a low-level low-frequency (LL-LF) AMCC on which the ONU can communicate US during the activation phase. The AMCC signal must be sufficiently weak to minimize the penalty caused by the interference onto other active ONUs, while at the same time it must be of an adequate level to permit its correct detection by the OLT Rx. In [15], it is shown that both conditions are not easily achieved unless additional measures are taken.
To facilitate the AMCC signal detection, the noise generated by active ONUs inside the passband of the AMCC Rx can be reduced either in the frequency or in the time domain. In the first case, proper coding or filtering of the transmitted data is necessary, which could, however, add an additional penalty. In the second case, the bandwidth allocations to the active ONUs can be withheld for a short time interval (time gap in Fig. 10) to allow for a (data-generated) noise-free detection of the AMCC signal [15]. During such “time gaps,” samples of the AMCC signal can be acquired until acquisition of a full activation message is completed. It should be noted that these time gaps can be generated asynchronously on different wavelength channels, so that independent operation of different channels is possible.
V. Future Directions of NG-PON2 PMD Standard
Since the approval of G.989.2 in December 2014, the ITU-T has started development of an amendment to specify low-loss power budget classes. The general agreement is to specify at least two low-loss power budget classes for PtP WDM with WR-ODN. The basic class would support a 20 km reach. Another class would meet a higher requirement of the 40 km reach. This work is actively ongoing in the ITU-T Q2/SG15 working group.
In the more distant future, as the underlying technology is improving, the enhancement of NG-PON2 technology may be driven by the access application, or by completely separate markets such as optical transport or data-center communication. The driver for these enhancements might be to increase capacity or to reduce cost, or both. At present, there are three obvious directions for the NG-PON2 system.
The first future direction is to scale the number of wavelengths. Transport systems have long used the increase in wavelength count as a path towards more capacity, and the NG-PON2 system works in much the same way. There are two issues that wavelength scaling must address: spectrum reallocation and tuning capability. The current spectrum of NG-PON2 systems is the leftover spectrum from the previous three PON systems (G-PON, RF-overlay, and XG-PON1). Some previous system spectrum must be reclaimed to use for NG-PON2 expansion; however, which system will be turned down? It would make sense that the oldest systems (G-PON and RF-overlay) would be selected. However, these have spectrum that is far away from the defined NG-PON2 wavelength bands. Contiguous spectrum growth would require the turning down of RF-overlay and XG-PON1. Whatever spectrum is recovered, the system specifications must then address the availability of tunable components to cover that. One of the design principles of NG-PON2 systems is to keep the tuning range small, so that many technologies can address it. If we increase capacity by growing the spectrum, then a new generation of tunable devices will be required. In that sense one can say that wavelength scaling presents several tough trade-offs to be made.
The second future direction is to scale the line rate. When key decisions were made during NG-PON2 standardization, 10 Gb/s was the most economical high-speed link rate, and the next higher rate (40 Gb/s) was very expensive. However, the introduction of 100 Gb/s Ethernet and particularly the series of PHYs has raised the availability of 25 Gb/s components. There is an ongoing project in IEEE 802.3 to define a 25 Gb/s Ethernet interface. From a technical perspective, 25 Gb/s is considerably easier than 40 Gb/s, both from electronic signal integrity and fiber chromatic dispersion views. Thus, 25 Gb/s might be the next data rate to be supported by the NG-PON2 system.
The third direction is to consider alternative line codes. NRZ has served optical access since its inception, but with the advance of data rates, other codes become more suited. If the NG-PON2 PHY remains intensity modulated direct detection, then the codes to be considered will be simple. The two leading candidates are duobinary and PAM-4. Both of these have the salient feature that they require about half the analog bandwidth as the comparable NRZ link. This opens the door for supporting a higher data rate on the existing generation of 10 Gb/s PHYs. Importantly, these kinds of codes are being developed in industry for use in data-center communications. This might be the easiest path to 25 Gb/s.
At the present time, it is not entirely clear which of these directions will be developed first. The operational and technical difficulties of rearranging the wavelength plan might tend to keep that option in the background, at least in the short term. The combination of a higher line rate using a new line code looks quite attractive, and so this might be the next enhancement of the NG-PON2 system, to reach 100 Gb/s total capacity. However, operator requirements for this capability will need to be established. In particular, would customers need to access just 10 Gb/s as their peak information rate, or would some want to intermittently access the entire 100 Gb/s? If the latter were true, then some form of bonding of the wavelength channels would be required. This channel bonding itself would be a new feature, so far completely unexplored in optical access systems but commonly used elsewhere.
There may be other improvements required in the future. For example, a higher power budget, hence splitting factor and reach, that would allow network operators to consolidate central offices. However, this will probably be the subject of a complete new generation of PON systems.
VI. Summary and Conclusion
In this second part of the NG-PON2 paper, we highlighted topics relevant to the system design aspect, such as wavelength control, technology feasibility, and management and control functions.
The specification of MSE is important for controlling ONU Tx wavelengths within an allowed spectral interval to minimize cross-talk and optimize link performance.
The wavelength calibration accuracy of ONU tunable Tx has a direct impact on cost. The NG-PON2 PMD standard considers a new wavelength locking method to relax the requirements of ONU Tx wavelength accuracy and stability.
The main feasibility concern of an ONU Tx is how to achieve the combination of tunability, high power budget, and minimum power consumption, while using the most cost-efficient technologies. ONU Tx based on EML and inexpensive DML can both be effective options if a dispersion mitigation technique is applied to DML.
The wavelength multiplexer at the OLT may be based on either AWG or TFF technology, while the specific choice is vendor dependent. For TWDM, cyclic AWGs may further enable the use of a low-cost thermally tuned ONU laser. Examples of OLT configurations based on AWGs or TFFs were described.
Because PtP WDM is specified to transport a client service with minimum or no payload modification, an AMCC channel is necessary to provide essential management functions. Two implementations are possible, transparent and transcoding, depending on how the AMCC content is transported. For TWDM, AMCC is used for an ONU to communicate US during the activation phase due to the multi-wavelength feature of NG-PON2 systems.
Over the course of 2014/15, several NG-PON2 prototypes and field trials were announced and conducted, with commercial products expected by the end of 2015. Technology innovations are ongoing with the intent to drive costs down dramatically. It is envisioned that future enhancements of NG-PON2 system design will benefit from technology advances in other application areas, and will likely take shape in increased wavelength numbers, higher line rates, or the use of advanced line codes. With its unique capabilities of providing flexible 40 Gb/s bi-directional, multi-wavelength connectivity for residential and business services on one single PON infrastructure, the NG-PON2 system is well positioned to support new opportunities in future optical access networks.
Acknowledgment
The authors thank members in the FSAN and ITU-T SG15/Q2 groups for the invaluable discussions and collaboration during the NG-PON2 standards development process.
References
1. “40-gigabit-capable passive optical networks (NG-PON2): Physical media dependent (PMD) layer specification,” ITU-T G.989.2 Recommendation, 2014.
2. N. Fujiwara, H. Ishii, H. Okamoto, Y. Kawaguchi, Y. Kondo, and H. Oohashi, “Suppression of thermal wavelength drift in super-structure grating distributed Bragg reflector (SSG-DBR) laser with thermal drift compensator,” IEEE J. Sel. Top. Quantum Electron., vol. 13, no. 5, pp. 1164–1169, 2007. [CrossRef]
3. W. Poehlmann, D. van Veen, R. Farah, Th. Pfeiffer, and P. Vetter, “Wavelength drift of burst-mode DML for TWDM-PON,” J. Opt. Commun. Netw., vol. 7, no. 1, pp. A44–A51, 2015. [CrossRef]
4. C. Antony, G. Talli, P. D. Townsend, J. Bauwelinck, D. W. Smith, and I. Lealman, “High extinction switching of SOAs for in-band cross-talk reduction in PON,” Electron. Lett., vol. 44, no. 14, pp. 872–873, 2008. [CrossRef]
5. R. A. Kumar, R. Murano, M. J. L. Cahill, J. M. Aufiero, and G. D. Bartolini, “Cross-talk performance analysis of low cost 8 × 10 Gb/s tunable receiver for TWDM,” in Optical Fiber Communication Conf. and the Nat. Fiber Optic Engineers Conf. (OFC/NFOEC), 2013, paper NTh3F.5.
6. “High sensitivity 10Gb/s surface mount coplanar APD preamp receiver AT10XGC,” Bookham Data Sheet, BH12852, Rev 2.0, Jan. 2007.
7. K. Grobe, M. H. Eiselt, S. Pachnicke, and J.-P. Elbers, “Access networks based on tunable lasers,” J. Lightwave Technol., vol. 32, no. 16, pp. 2815–2823, Aug. 2014. [CrossRef]
8. “Specification for SFP+ 10 Gb/s and low speed electrical interface,”SFF Committee SFF-8431, Rev. 4.1 and Addendum, Sept. 15, 2013.
9. “10Gb/s limiting APD receiver AT10XGC,” Data Sheet D00010-PB, Issue 06, Oclaro, Jan. 2013 [Online]. Available: http://www.oclaro.com/wp-content/uploads/2014/08/AT10XGC-Datasheet-D00010-PB-06.pdf.
10. M. N. Sysak, J. W. Raring, G. P. Morrison, D. J. Blumenthal, and L. A. Coldren, “Monolithically integrated, sampled grating DBR laser transmitter with an asymmetric quantum well electroabsorption modulator,” in Int. Semiconductor Laser Conf., Sept. 2006, paper TuB6.
11. M. Gokhale, J. Johnson, and J. Zhang, “Performance of optically amplified 10 G sources,” in IEEE P802.3av 10GEPON Task Force Meeting, Monterey, California, Jan. 2007.
12. C. Caillaud, P. Chanclou, F. Blache, P. Angelini, B. Duval, P. Charbonnier, D. Lanteri, G. Glastre, and M. Achouche, “Integrated SOA-PIN detector for high-speed short reach applications,” J. Lightwave Technol., vol. 33, no. 8, pp. 1596–1601, Apr. 2015. [CrossRef]
13. Y. Matsui, D. Mahgerefteh, X. Zheng, C. Liao, Z. F. Fan, K. McCallion, and P. Tayebati, “Chirp-managed directly modulated, laser,” Photon. Technol. Lett., vol. 18, no. 2, pp. 385–387, 2006. [CrossRef]
14. W. Poehlmann, B. Deppisch, Th. Pfeiffer, C. Ferrari, M. Earnshaw, A. Duque, R. L. Farah, J. Galaro, J. Kotch, M. F. Lau, D. T. van Veen, and P. Vetter, “Low cost TWDM by wavelength-set division multiplexing,” Bell Labs Tech. J., vol. 18, no. 3, pp. 173–193, Dec. 2013. [CrossRef]
15. L. Bertignono, V. Ferrero, M. Valvo, and R. Gaudino, “Photon ranging for upstream ONU activation signaling in TWDM-PON,” J. Lightwave Technol., to be published.
Yuanqiu Luo (S’02–M’06–SM’11) received a Ph.D. in electrical engineering from the New Jersey Institute of Technology, Newark, NJ, USA, in 2006. She is currently a senior staff engineer in the Optical Access Network Department of Huawei Technologies, Bridgewater, NJ, USA. Before joining Huawei, she was with NEC Laboratories America, Princeton, NJ, USA. She has been heavily involved in the R&D effort of optical access networks, such as the first XG-PON1 field trial, PON as wireless backhaul, XG-PON1 and NG-PON2 protocol design, and the first NG-PON2 prototype system. Dr. Luo is a co-editor of ITU-T Recommendations G.987.3, G.989.2, and G.9802. She is a clause editor of IEEE Standard 802.1AS. Dr. Luo is an editor of IEEE Communications Surveys and Tutorials and an editor of Springer Photonic Network Communications. She has authored more than 60 publications in journals and conferences. She serves on OFC, HPSR, AccessNets, and other conference subcommittees. She is a co-author of the book Passive Optical Networks: Flattening the Last Mile Access (IEEE Comsoc Pocket Guides to Communications Technologies). She holds 19 US patents. Dr. Luo was honored with an IEEE Standards Award in 2011.
Hal Roberts (M’12) received a B.A. with a major in physics from Colgate University, Hamilton, NY, USA, in 1976 and an M.S. from Thayer Engineering School of Dartmouth College, Hanover, NH, USA, in 1979. He joined ADC Telecommunications in 1985, first developing optical components and then a PON-based access system for last-mile telephony transport in the late 1980s/early 1990s called “Homeworx.” Subsequently, Hal was responsible for leading a team that developed a new RF access technology for the Homeworx system to allow operation on a hybrid fiber coax system. This resulted in the first OFDMA-based access solution. Mr. Roberts is listed as a key contributor to the DOCSIS 2.0 and 3.0 DRFI and went on to the development of digital DAS systems for mobile front-haul. He left ADC and joined Calix in 2009 and since then has worked on PON and wireless technologies and participated in FSAN and ITU-T [T15/Q2]. Mr. Roberts has worked on progressing NG-PON2 with over 30 contributions related to the PMD and was editor of G. supplement 51 on PON protection. He has over 40 patents in wireless, optical, and coax access technologies.
Klaus Grobe (M’94–SM’13) received Dipl.-Ing. and Dr.-Ing. degrees in electrical engineering from Leibniz University, Hannover, Germany, in 1990 and 1998, respectively. He has worked over 20 years in the fields of lightwave guides and WDM. This includes positions as a technical staff member at German and pan-European network operators. In 2000, he joined ADVA Optical Networking SE, where he now works as a senior principal engineer in the CTO office in Munich, Germany. He is one of the main authors of Wavelength Division Multiplexing—A Practical Engineering Guide (Hoboken, NJ, Wiley, 2014) and has authored and co-authored more than 100 technical publications as well as three further book chapters on WDM and PON technologies. His research interests include next-generation broadband access networks as well as high-speed WDM transport. He holds 25 (pending) patents. Dr. Grobe is a member of the German VDE/ITG and ITG Study Group 5.3.3 on photonic networks. He served on the OFC subcommittee 10 in 2009–2012 and works in FSAN and ITU-T–SG15-Q.2.
Maurizio Valvo received his M.Sc. degree in electronics engineering cum laude from the University of Naples, Italy, in 1991. In the same year, he joined CSELT, the Center for Research in Telecommunications, now Telecom Italia Lab–Turin, where he is currently active. He has focused since the beginning of his career on broadband access networks and in particular on PON systems, which he also contributed to specify, develop, and test during the 1990s in the framework of several European-funded projects. He has led research projects with the objective to specify and test, both in the laboratory and in the field, innovative access network technologies based on PON, xDSL, HFC, WiMAX, and free-space optics. Currently, Mr. Valvo leads the laboratory for fixed access network innovation and a research project for the innovation of the Telecom Italia optical access network. He holds four patents and is co-author of three books and several papers.
Derek Nesset (M’01–SM’13) received a B.Sc. degree in physics from Birmingham University, UK, in 1989 and an M.Sc. degree in telecommunications engineering from the University of London in 1995. He joined BT in 1989 and spent several years developing photonic components for fiber-optic communication systems. Following this, he worked on advanced fiber-optic system technologies up to 100 Gbit/s, including the first field demonstration of 40 Gbit/s transmission over BT’s fiber infrastructure. In 2000, Derek joined Marconi, where he was responsible for the ROADM subsystem development for ultra-long-haul DWDM. He returned to BT in 2003 to pursue research interests on enhanced PON systems for fiber access. Most recently, he has focused on next-generation PON technologies and standards and actively contributes to progressing NG-PON2 in both FSAN and ITU-T. He leads research into future optical access networks within BT’s Research & Technology organization near Ipswich, UK. He also chairs the NG-PON task group in FSAN. Mr. Nesset was awarded an honorary professorship at Bangor University in 2014. He has contributed over 90 journal and conference publications and has eight patents. He has participated in OFC, OECC, and ACP conference subcommittees and chaired the optical access subcommittee for OFC 2014.
Kota Asaka (M’99) received B.S. and M.S. degrees in electrical engineering from Waseda University in 1996 and 1999, respectively, and a Ph.D. degree in physics from Kitasato University in 2008. In 1999, he joined NTT Photonics Laboratories, Atsugi, where he engaged in research on photonics integrated circuits and long-wavelength VCSELs. From 2005 to 2008, he engaged in research on OCT by using SSG-DBR-LDs with Kitasato University. From 2009 to 2012, he worked on developing low-cost and small optical subassemblies for optical access networks. He served as the working group secretary of TC86/SC86C/WG4 of IEC, and he engaged in the international standardization of fiber optic active components from 2009 to 2012. He is now a senior research engineer with NTT Access Service System Laboratories, Yokosuka, NTT Corporation. He has engaged in the international standardization and related activities for next-generation optical access systems in ITU-T Q2/SG15 and FSAN since 2013. Dr. Asaka is a member of IEEE ComSoc, the Institute of Electronics, Information and Communication Engineers (IEICE) of Japan, and the Japan Society of Applied Physics.
Harald Rohde (M’03–SM’08) received his Ph.D. in physics at the university in Innsbruck/Austria in 2001. Since then, he has worked in different departments of the Siemens AG, Nokia Siemens Networks, and Coriant. He worked on research for optical transmission systems, covering a broad range of topics. Currently, his main interest lies in next-generation optical access networks. Dr. Rohde is author or co-author of more than 70 publications and holds more than 50 patents. He is a co-author of the book Core and Metro Networks (published by Wiley). He has been actively involved in the standardization forum FSAN since 2008.
Joe Smith has worked in the fiber optic industry for nearly four decades. He joined ITT in 1976 and worked on initial fiber optic concepts, manufacturing techniques, and validation of new fiber optic technologies. Subsequently, he worked on the development of the world’s first fiber optic system at 1.5 Mb/s, followed by systems for 45 Mb/s through 2.5 Gb/s in transport for ITT, Alcatel, and Alcatel-Lucent. Mr. Smith has participated in standards in various forums for 25 years in transport arenas for ANSI, Bellcore, FSAN, and ITU, where he currently holds co-editorship responsibility for the NG-PON2 PMD Layer, G.989.2.
Jun Shan Wey (M’95–SM’11) received a Ph.D. in electrical engineering from University of Maryland, College Park, in 1995. She began her career in the research faculty at the Lab for Physical Sciences, University of Maryland, focusing on high-speed optical communication and signal processing. She joined Siemens in 2004 to work on IPTV product development. Later, with Nokia Siemens Networks and, subsequently, Coriant, she became active in the standardization for IPTV and next-generation optical access networks. She participated in both FSAN and ITU-T on developing the NG-PON2 standard and was a co-editor of the G.989.2 recommendation. In 2014, she founded Lightnotes Consulting to advise clients on competitive and go-to-market strategies for emerging technologies. Dr. Wey co-chaired the IPTV Interoperability Task Force of the ATIS IIF and MSF groups. She was a co-editor in the Verizon VICAS Forum. She currently serves on the WDM-PON forum board and the OFC technical program N4 subcommittee.
Frank J. Effenberger (M’94–SM’01–F’15) received a Ph.D. from the University of Central Florida in 1995. After completing his doctoral work in 1995, he worked at Bellcore, where he analyzed all types of access network technologies, especially passive optical networks. He witnessed the early development of the FSAN initiative and the APON standard. In 2000, he moved to Quantum Bridge Communications, where he managed the system engineering group. This work supported the development and standardization of advanced optical access systems based on B-PON and G-PON technologies. In 2006, he became director of FTTx in the advanced technology department of Futurewei Technologies USA. He remains heavily involved in standards work, and in 2008, he became the chairman of ITU-T Q2/15. He and his team work on forward-looking fiber and copper access technologies, including the 10G EPON, XG-PON, and 40G-PON. Notably, his team supported the world’s first trials of XG-PON and 40G-PON. In 2011, Dr. Effenberger was named as a Huawei Fellow, and in 2012 he was promoted to VP of access research. In 2015, he was named as a Fellow of both the OSA and the IEEE. He holds 60 US patents.
