Open Access
Add to BibSonomyAbstract
Extensive numerical investigations are undertaken to analyze and compare, for the first time, the performance, techno-economy, and power consumption of three-level electrical Duobinary, optical Duobinary, and PAM-4 modulation formats as candidates for high-speed next-generation PONs supporting downstream 40 Gb/s per wavelength signal transmission over standard SMFs in C-band. Optimization of transceiver bandwidths are undertaken to show the feasibility of utilizing low-cost and band-limited components to support next-generation PON transmissions. The effect of electro-absorption modulator chirp is examined for electrical Duobinary and PAM-4. Electrical Duobinary and optical Duobinary are power-efficient schemes for smaller transmission distances of 10 km SMFs and optical Duobinary offers the best receiver sensitivity albeit with a relatively high transceiver cost. PAM-4 shows the best power budget and cost-efficiency for larger distances of around 20 km, although it consumes more power. Electrical Duobinary shows the best trade-off between performance, cost and power dissipation.
© 2015 Optical Society of America
1. Introduction
Optical access networks based on passive optical networks (PONs) enable high-speed delivery of broadband services to individual and enterprise users, as well as various backhauling and fronthauling functionalities. In order to address the exponential increase of bandwidth requirement driven by applications such as HDTV, video-on-demand, online commerce and cloud computing, the standardized PONs have undergone a number of upgrades in data rates: from initial 125-Mb/s ATM PON (APON) to today’s fastest 40 Gb/s TWDM PONs (time- and wavelength-division multiplexed PON with 4 wavelengths of 10 Gb/s each) for Next-Generation PON stage 2 (NG-PON2) [1]. In addition, an optional extra capability of 8 point-to-point WDM channels in each direction is also considered for NG-PON2 [1]. Meanwhile, the co-existence of various PONs at different data rates in the same fiber infrastructure is preserved by using the WDM approach. As a result, the co-existence wavelength plan brings about challenges of future spectrum allocation. Considering the need for guard-band gaps, no spectrum is left anymore. Therefore, increasing the per-wavelength bit rate is a must for future upgrade beyond 40/100 Gb/s (as a matter of fact, 100 Gb/s is the maximum supported bit rate as discussed in NG-PON2), unless the requirement on full co-existence is relaxed. For such high single-wavelength data rates, the utilization of the conventional non-return-to-zero (NRZ) modulation causes significant technical difficulties in terms of both transceiver bandwidth capability and fiber dispersion.
To address the above-mentioned challenges, as alternatives to the NRZ scheme, advanced multilevel modulation formats featuring high spectral efficiency have been demonstrated. There have been demonstrations of 40-Gb/s TDM-PONs using optical Duobinary modulation which involves both, optical amplitude and phase modulation and only requires an NRZ receiver in downstream transmission over 42 km SMFs using a 20-GHz avalanche photodiode (APD) receiver and optical fiber dispersion compensation [2]. On the other hand, 26-Gb/s [3] and 40-Gb/s [4] three-level electrical Duobinary based PONs have also been demonstrated using 10-GHz receivers in the O-band and 25-GHz receivers in the C-band with optical dispersion compensation, respectively. To distinguish these two Duobinary schemes, in this paper thereinafter, three-level electrical Duobinary is referred to as electrical Duobinary, and amplitude and phase modulated Duobinary is referred to as optical Duobinary. In addition, PAM-4 at 10 Gb/s [5] and 40 Gb/s [6] has also been explored for access network applications, respectively. More recently, digital signal processing (DSP)-based high speed PON systems using complex quadrature amplitude modulation (QAM) have been proposed [7]. There have also been demonstrations of more complex orthogonal frequency division multiplexing (OFDM)-PONs supporting 108 Gb/s downstream transmission using polarization multiplexing [8]. Real-time downstream 20-Gb/s OFDM-PONs over 25 km SMFs have also been demonstrated [9].
For an NG-PON supporting a single-wavelength 40 Gb/s data rate, technical solutions with low cost and low complexity are extremely preferred. Therefore, intensity modulation and direct detection (IMDD) are desirable. Compared to DSP intensive schemes such as OFDM, (electrical/optical) Duobinary and PAM-4 are relatively simple, and thus are potentially cost-effective solutions. Although Duobinary and PAM-4 have been explored intensively as mentioned above, there has been lack of detailed investigations and fair comparisons of their achievable performances with an optimized transceiver bandwidth. To our best knowledge, previous demonstrations of high speed Duobinary used a Mach Zehnder modulator (MZM) [2–4]. Externally modulated electro-absorption modulators (EAMs) and their nonlinear effects have not been considered. They will be examined in this paper.
Moreover, as access networks are cost-sensitive, the adoption of a particular scheme is also determined by its cost and power consumption, which, however, to our best knowledge, has not been reported in published literatures. Our previous work [10] only compared electrical Duobinary and PAM-4 on equalized performance and cost without transceiver bandwidth optimization. Optical Duobinary was not considered and the fundamental limitations of un-equalized performance and power dissipation analysis are missing. In this paper, for the first time, we investigate and compare the performances of electrical Duobinary, optical Duobinary, and PAM-4 at 40 Gb/s data rates over SMF links in the C-band using simple un-equalized receivers. Optimization of transceiver bandwidths is performed to identify the feasibility of utilizing low-cost and band-limited components to support PON transmissions. In addition, detailed comparisons of transceiver costs and power dissipations are also performed to examine the cost- and power-efficiency of the modulation schemes targeted in the paper.
2. System architecture and simulation parameters
Figure 1 illustrates the WDM-PON architecture with downstream transceiver considered here. The transceivers use electrical Duobinary, optical Duobinary or PAM-4 modulation formats, respectively. For the electrical Duobinary case, the OLT transmitter is simply a conventional NRZ transmitter except that a pre-coder is incorporated in order to overcome error propagation. Being up-converted to an optical carrier by an external EAM, the resulting optical signal passes through a WDM multiplexer and then propagates through standard SMFs. In the remote node, a de-multiplexer, e.g., an arrayed waveguide grating (AWG), is used to route the signal of a specific wavelength to a targeted ONU. The ONU receiver consists of a photo-detector, a Duobinary shaping filter, and a Duobinary-to-NRZ convertor which consists of two-threshold slicers followed by an exclusive-or (XOR) gate [4]. Note that the Duobinary shaping filter is allocated in the receiver side rather than in the transmitter side in order to achieve better performance by efficiently filtering noise power [4,5].
Fig. 1 Architectures of next-generation PON using (a) electrical Duobinary, (b) optical Duobinary, and (c) PAM-4 formats for downstream transmissions.
For the optical Duobinary case, the corresponding transmitter involves a pre-coder [11], a shaping filter and a Mach Zehnder modulator (MZM) [11]. The shaping filter is designed in a way that the resulting three-level electrical Duobinary waveform, S(t), is generated. The driving signal is optimized with its peak-to-peak amplitude equal to 2 to produce the maximum signal extinction ratio, being the MZM switching voltage that changes the optical power from its maximum to minimum. The MZM is biased at leading to an output optical signal with two intensity levels as a result of both optical amplitude and phase modulation [10]. The output signal amplitude of the MZM operating in a pull-push manner is given by
where is the optical input continuous wave (CW) amplitude, S(t) has no DC. At the receiver side, a standard NRZ receiver is used. It is clear that optical Duobinary has a simpler receiver but more complex transmitter compared to the electrical Duobinary scheme shown in Fig. 1 (a).For the PAM-4 case, the transceiver architecture is similar to the electrical Duobinary scheme. The PAM-4 transmitter consists of an NRZ-to-PAM-4 symbol encoder and an EAM modulator. The receiver includes a photo-detector, a threshold device with three-threshold slicers to determine the amplitude levels, as well as a PAM-4-to-NRZ decoder.
For electrical Duobinary and PAM-4, the impact of the EAM nonlinear response and frequency chirp are considered. In modeling the EAM modulator, a measured response regarding the dependence of optical power and dynamic modulation factor on driving voltage is considered [12], as shown in Fig. 2(a) and (b), respectively. The output phase,, is governed by
where P(t) is the optical power and S(t) is the driving voltage, is the modulation factor that is named ‘Alpha’ in Fig. 2 (b) and depends on the applied driving voltage.Fig. 2 (a) EAM output optical power versus driving voltage and (b) modulation factor versus driving voltage.
In our numerical simulations, an EAM modulator bandwidth limitation is modeled as a Gaussian response and a MZM modulator has a typical driving voltage dependent cosine optical field modulation response as expressed in Eq. (1). Moreover, different from electrical Duobinary and PAM-4, an optical Duobinary transmitter usually needs a lowpass shaping filter in order to generate a 3-level signal. We model the frequency response of the shaping filter as a 5th order Bessel filter with a proper bandwidth that leads to an optimum receiver sensitivity and fiber dispersion tolerance [5]. All three schemes use an avalanche photo-diode (APD) and trans-impedance amplifier (TIA)-based optical receiver. Feedforward error correction (FEC) with a threshold bit error rate (BER) of 1.0 × 10−3 is considered. The transceiver bandwidths in simulations are optimized. For all three schemes, the optical launch power is set to 10 dBm. The APD has a responsivity of 3.2 A/W. The optical signal wavelength is 1550 nm in all cases, corresponding to a fiber loss of 0.2 dB/km, chromatic dispersion of 17 ps/km/nm, and a nonlinear coefficient γ = 1.3 W−1/km. The wavelength multiplexer/de-multiplexer is modeled as an optical band-pass Gaussian filter (order 2) with a 3-dB bandwidth of 40 GHz.
3. Simulation results
This section first identifies the optimum transceiver bandwidth for each scheme and then examines the properties of each individual scheme on aspects such as optical link power budget, transceiver cost, and transceiver power dissipation.
3.1 Transceiver bandwidth optimization
In this section, we look at the impact of the transceiver bandwidth on the performance of the optical Duobinary link and verify the validity of the system and component models by comparing with experimental measurements [2, 4]. Figure 3 shows the optical power sensitivity for a BER of 1.0 × 10−3 to the fiber dispersion subject to different transceiver bandwidths. Only chromatic dispersion is considered in the fiber model for Fig. 3, since fiber nonlinearities are not significant for short distances. The optical (De-)MUX filters are not included in Fig. 3 as they are not considered by [2]. This configuration serves for a comparison purpose and only applies to Fig. 3. As shown in Fig. 3(a), for a fixed shaping filter bandwidth larger than a quarter of the bit rate, the optical power sensitivity improves with increasing fiber dispersion until an optimum power sensitivity is achieved, beyond which the optical power sensitivity degrades significantly with increasing fiber dispersion. This agrees with experimental measurements [2], and the reason is that the transmitter low-pass filtering with a wider bandwidth leads to incomplete conversion from NRZ to three-level Duobinary. As a result, the ISI caused by a certain amount of fiber dispersion together with the bandwidth limited receiver help shape the waveform approaching optimum signal eye opening [2]. It also shows that the dispersion corresponding to the optimum optical power sensitivity decreases slightly with increasing the shaping filter bandwidth. Figure 3(a) also indicates an optimum transmitter shaping filter bandwidth of 13 GHz that leads to best dispersion tolerance. Based on the identified transmitter bandwidth, Fig. 3(b) examines the influence of the APD-TIA receiver bandwidth on the received signal tolerance to dispersion. Similarly, an optimum receiver bandwidth of 17.5 GHz is obtained. Without considering the optical (De-)MUX filters, Fig. 3 verifies the validity of the models used here by means of excellent agreement with experimental measurements with similar configurations [2].
Fig. 3 Receiver sensitivity versus fiber dispersion for optical Duobinary subject to various (a) transmitter lowpass shaping filter bandwidths (b) APD-TIA receiver bandwidths. (De-)MUX filters shown in Fig. 1 are not included.
When the optical (De-)MUX filters are considered, as shown in Fig. 4(a), the required optical power to achieve a BER of 1.0 × 10−3 becomes much less sensitive to the variation of the low-pass shaping filter bandwidth compared to the case without optical filters. This is attributed to the fact that the optical filters also contribute to shaping the signal, which shifts the optimum power sensitivity to the zero dispersion region. A small penalty of 1 dB is obtained when the shaping filter bandwidth varies from 13 GHz to 21 GHz, and here we use an optimum 17 GHz bandwidth due to its slightly better sensitivity. Note that the shaping filter can be removed if a MZM with proper low-pass filtering profile and required bandwidth is used instead [2], especially for the case of narrow-band optical (De-)MUX filtering in this paper. A similar behavior can be observed by taking into account the effect of the APD-TIA bandwidth to the receiver optical power sensitivity. The optimum receiver bandwidth is 17.5 GHz, beyond which the power sensitivity shows insignificant variation.
Fig. 4 Receiver sensitivity versus fiber dispersion for optical Duobinary subject to various (a) transmitter lowpass shaping filter bandwidths (b) APD-TIA receiver bandwidths. (De-)MUX filters as shown in Fig. 1 are included.
3.2 Influence of the EAM chirp
In this section, we identify the optimum transceiver bandwidths for 40 Gb/s electrical Duobinary and PAM-4, by considering both the fiber effects and EAM nonlinearities. Similar to the simulation process of optical Duobinary, the simulation identified optimum modulator bandwidth is 25 GHz (17.5 GHz) for electrical Duobinary (PAM-4), and the optimum APD-TIA bandwidth is 12.5 GHz (12.5 GHz).
Secondly, the impact of the EAM nonlinearity is examined and results are presented in Fig. 5, where the cases with and without considering EAM nonlinearities are compared for both, electrical Duobinary and PAM-4. Optimum transceiver bandwidths are used. For the case without EAM nonlinearities, an infinite extinction ratio is assumed for both schemes. For PAM-4, an unequal amplitude space is used in order to mitigate the nonlinear power versus drive voltage response as shown in Fig. 2(a). As shown in Fig. 5, for both schemes, the EAM nonlinearities bring about impact on two aspects: firstly, the dispersion tolerance is improved compared with the case without EAM nonlinearities. This is mainly attributed to the negative chirp of the EAM as shown in Fig. 2(b), which can partly compensate positive chromatic dispersion as witnessed by the optimum fiber length for achieving best receiver sensitivity in Fig. 5. However, the dispersion tolerance improvement for PAM-4 is not significant, because the EAM chirp together with chromatic dispersion also causes a distortion which behaves as eye diagram skew and PAM-4 is more sensitive to such distortion. The skew becomes stronger with increasing chromatic dispersion, which can be witnessed by the inset noise-free eye diagrams for PAM-4 signal at different SMF lengths with EAM chirp considered, leading to very limited improvement in dispersion tolerance compared with no-chirp case. The eye diagrams are obtained at the output of the APD-TIA. Secondly, to achieve negative chirp, an EAM usually needs bias at a large reverse voltage thus the system shows an extinction ratio penalty for short reach [12].
Fig. 5 Optical sensitivity at a BER of 10−3 versus fiber length for electrical Duobinary and PAM-4 subject to cases with and without EAM nonlinearities. The inset noise-free eye diagrams are for PAM-4 signal with EAM nonlinearities.
3.3 Optical link power budget
Based on the optimum transceiver bandwidth for each scheme identified in Section 3.1, this section investigates the optical power sensitivity of the targeted 40 Gb/s system for different fiber lengths by taking into account both linear and nonlinear effects of each component. The results are presented in Fig. 6. The achievable fiber length for each scheme is mainly limited by chromatic dispersion. For a short fiber length of 10 km, electrical Duobinary, optical Duobinary and PAM-4 have about 31 dB, 34 dB, and 29 dB power budgets, respectively, by considering a launch power of 10 dBm. Obviously, optical Duobinary shows the best power budget for short distances due to the lower number of amplitude levels of its detected signal and high extinction ratio it can achieve. Beyond 10 km SMF, the penalties for electrical/optical Duobinary grow dramatically and both can achieve about up to 18 km SMF transmission. Note that if the EAM were operated in the positive chirp regime, electrical Duobinary would support much less SMF transmission distances. As a result, only PAM-4 can support transmission over 20 km SMF, with a power budget of about 28 dB achievable at 20 km fiber length. This indicates that PAM-4 is preferable for longer transmission distances. To support even longer transmission distances, dispersion compensation in either electrical domain [4] or optical domain [2] or transmission at O-band [3] can be considered.
3.4 Transceiver cost
In this section, the cost of electrical Duobinary, optical Duobinary or PAM-4 transceivers is analyzed and benchmarked to a 10-Gb/s reference system. The 10-Gb/s reference transceiver cost and its constituent costs are calculated based on industry contributions that have been widely discussed in fora such as FSAN [13] and IEEE 802.3 [14]. The 10 Gb/s reference transceiver cost is referred to as a typical 10 Gb/s module for XG-PON1, which consists of the costs of major transceiver components including a laser and its driver, an APD and TIA, a linear amplifier (LA), clock and data recovery (CDR) as well as associated assembly and packaging. A good cost analysis of the 10-Gb/s reference system has been carried out in [13] by considering the cost scale up from G-PON to XG-PON1 as well as from XG-PON1 to 5-Gb/s upstream Duobinary and PAM-4 transceivers.
For all three schemes, the main assumption adopted in the analysis is that the cost of critical sub-components increase approximately linearly with bandwidth. Costs are also based on a forward-looking projection that assumes large volume production [13]. For example, the laser and APD costs scale up 1:1 with respect to bandwidth for all schemes. The optical Duobinary shaping filter and PAM-4 encoder costs are taken into account by the transmitter laser driver cost, which scales up 1:2 with respect to bandwidth, by referring [13]. In contrast, for electrical Duobinary, the laser driver has a cost scale up factor of 1:1 versus bandwidth since its transmitter is a simple NRZ transmitter. The CDR circuits contain mapping required by electrical Duobinary (which is a Duobinary to NRZ decoder chip) and PAM-4, whose costs scale up 2:1 with respect to bandwidth [13]. An optical Duobinary receiver is simply an NRZ receiver whose CDR scale up 1:1 with bandwidth. As far as the packaging costs are concerned, a moderate scale up factor of 0.6:1 is used with respect to bandwidth [13].
The modulator represents the major part of the transceiver cost. The 10-Gb/s reference transmitter is based on a directly modulated DFB laser (DML). The transmitter for electrical Duobinary or PAM-4 consists of a DFB laser and an EAM for good trade-off between performance and cost. For optical Duobinary, a MZM is needed together with a DFB laser due to the optical amplitude and phase modulation required. Not surprisingly, the OLT transmitter cost is about twice the cost of an ONU receiver for each scheme by assuming that OLT transmitter and ONU receiver have similar cost associated with the packaging and PCB, as indicated in Fig. 7. A smaller ONU receiver cost is beneficial for subscribers. The relative cost between DML, laser + EAM, and laser + MZM at the same bandwidth is assumed to be 1: 1.25: 2. This assumption is also based on the requirement that the lasers are narrow-band tunable for all cases considered in the paper.
Based on the above assumptions, 40 Gb/s electrical Duobinary, optical Duobinary, and PAM-4 using optimized transceivers have costs of approximately 3.0, 3.3, and 2.6, respectively, relative to the reference 10 Gb/s transceiver. Optical Duobinary has highest and PAM-4 least transceiver cost. This is due to the fact that PAM-4 requires only a moderate bandwidth EAM as identified in Section 3.2. In contrast, optical Duobinary needs the highest bandwidth receiver and an expensive MZM transmitter although with lowest modulator bandwidth (17 GHz). The above-mentioned results are summarized in Fig. 7. For a required link power budget and fiber length, it is worthwhile to choose a scheme by taking into account the best trade-off between performance and cost. By considering both Fig. 6 and Fig. 7, if the cost or transmission distance is the primary consideration, PAM-4 is the preferable choice. If power budget is the most important figure of merit, optical Duobinary can be considered for <15 km of SMF.
3.5 Transceiver power dissipation
This section further discusses the power dissipation of each 40-Gb/s transceiver. The results are presented in Fig. 8, where the relative power consumption of each component is considered. The power consumption of each constituent component of the transceiver is normalized to a 10-Gb/s XG-PON1 transceiver, whose power dissipation is calculated using parameters accepted by FSAN [15, 16], or published in literature [17–19]. The 10-Gb/s reference transceiver has a power dissipation of approximately 2.5 W, which is comparable to the power of a typical 10 GBASE-LR/ER XFP module.
In obtaining Fig. 8, the power consumption is scaled up with respect to data rate because of the fact that, for example, an EAM or MZM operating at higher bit rates is designed to operate at higher switching voltage, which corresponds to a higher power consumption [18]. For an EAM or MZM, the dominant power dissipation is from the driver and is mainly determined by the driving voltage root mean square (RMS) and the termination resistance [18]. A multilevel driving signal (associated with optical Duobinary and PAM-4 schemes) usually has a higher RMS compared to a binary NRZ driving signal (as with the electrical Duobinary considered) [18]. The estimated driver power consumption for the three schemes reflects this trend. The thermo-electric cooler (TEC) also accounts for significant power consumption [18]. The receiver power consumption is dominated by the TIA amplifier and CDR. The power consumption of the CDR also includes the symbol-to-bits conversion which involves threshold slicers and, for electrical Duobinary only, an exclusive-or (XOR) gate. As a result, optical Duobinary exhibits relatively less receiver power consumption compared to the other two schemes due to its simplicity as described in Section 2. Overall, electrical Duobinary and optical Duobinary transceivers have comparable power consumptions and PAM-4 transceiver consumes considerably more power.
4. Conclusions
We have investigated performance, cost, and power consumption of electrical Duobinary, optical Duobinary, and PAM-4 systems as candidates for high-speed NG-PONs supporting single-wavelength 40-Gb/s data transmission. Transceiver bandwidth optimization was performed and results show the feasibility of using 25-G transceivers to support 40-Gb/s transmission. The effect of EAM nonlinearities on the electrical Duobinary and PAM-4 performance has also been examined and it shows that an EAM with negative chirp can considerably improve dispersion tolerance, especially for electrical Duobinary. For short distances of 10 km, optical Duobinary is preferred if the power budget is the primary requirement, although it shows the highest cost. For transmission distances larger than 20 km, PAM-4 is a preferred cost-effective solution although it has a considerably higher transceiver power consumption compared to the other two schemes. Electrical Duobinary shows the best trade-off between performance, cost, and power consumption.
Acknowledgment
The research was supported by funding from the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme (FP7/2007-2013) under REA grant agreement n° 623515 via CEEOALAN project.
References and links
1. ITU-T, “G. 989.x, 40 Gigabit-capable passive optical networks 2 (NG-PON2),” to be available publicly.
2. D. T. van Veen, V. E. Houtsma, A. H. Gnauck, and P. Iannone, “Demonstration of 40-Gb/s TDM-PON over 42-km with 31 dB optical power budget using an APD-based receiver,” J. Lightwave Technol. 33(8), 1675–1680 (2015). [CrossRef]
3. D. V. Veen, V. Houtsma, P. Winzer, and P. Vetter, “26-Gbps PON transmission over 40-km using Duo-binary detection with a wow cost 7-GHz APD-based receiver,” in ECOC (IEEE, 2012), paper Tu.3.B.1 (2012).
4. V. Houtsma, D. V. Veen, A. Gnauck, and P. Iannone, “APD-Based Duobinary direct detection Receivers for 40 Gbps TDM-PON,” in OFC (OSA, 2015), paper Th4H.1.
5. S. Walklin and J. Conradi, “Multilevel signaling for increasing the reach of 10 Gb/s lightwave systems,” J. Lightwave Technol. 17(11), 2235–2248 (1999). [CrossRef]
6. J. L. Wei, N. Eiselt, H. Griesser, K. Grobe, M. Eiselt, J. J. Vegas-Olmos, I. T. Monroy, and J.-P. Elbers, “First demonstration of real-time end-to-end 40 Gb/s PAM-4 system using 10-G transmitter for next generation access applications,” in Proceedings of ECOC (IEEE, 2015), paper PDP4.4.
7. N. Iiyama, J. Kani, J. Terada, and N. Yoshimoto, “Feasibility study on a scheme for coexistence of DSP-based PON and 10-Gbps/λ PON using hierarchical star QAM format,” J. Lightwave Technol. 31(18), 3085–3092 (2013). [CrossRef]
8. D. Qian, N. Cvijetic, J. Hu, and T. Wang, “108 Gb/s OFDMA-PON with polarization multiplexing and direct detection,” in OFC (OSA, 2009), paper PDPD5.
9. R. P. Giddings, E. Hugues-Salas, and J. M. Tang, “Experimental demonstration of record high 19.125 Gb/s real-time end-to-end dual-band optical OFDM transmission over 25 km SMF in a simple EML-based IMDD system,” Opt. Express 20(18), 20666–20679 (2012). [CrossRef] [PubMed]
10. J. L. Wei, K. Grobe, and H. Griesser, “Cost-efficient high-speed modulation for next-generation PONs,” in Proceedings of Photonic Networks (2015), 16.
11. W. Kaiser, T. Wuth, M. Wichers, and W. Rosenkranz, “Reduced complexity optical duobinary 10 Gb/s transmitter setup resulting in an increased transmission distance,” IEEE Photonics Technol. Lett. 13(8), 884–886 (2001). [CrossRef]
12. W.-J. Choi, A. E. Bond, J. Kim, J. Zhang, R. Jambunathan, H. Foulk, S. O’Brien, J. V. Norman, D. Vandegrift, C. Wanamaker, J. Shakespeare, and H. Cao, “Lower insertion loss and low dispersion penalty InGaAsP quantum-well high-speed electroabsorption modulator for 40-Gb/s very-short-reach, long-reach, and long-haul applications,” J. Lightwave Technol. 20(12), 2052–2056 (2002). [CrossRef]
13. K. Ou, Y. Luo, and F. Effenberger, “XG-PON up-stream enhancement,” presentation at FSAN meeting, Las Vegas, USA. Unpublished (2010).
14. D. Lee and B. Y. Yoon, “Optic Cost Estimation for 10G EPON Downstream,” presented at IEEE P802.3av Task Force interim meeting. San Francisco, CA, USA (2007).
15. S. Dahlfort, B. Skubic, and D. Hood, “Power consumption of NG-PON2 components,” FSAN meeting, San Diego, US, unpublished (2011).
16. K. Grobe, “Component-level power consumption of WDM-PON w/ tunable laser diodes,” FSAN NG-PON2 Workshop, San Diego, unpublished (2011).
17. J. L. Wei, J. D. Ingham, D. G. Cunningham, R. V. Penty, and I. H. White, “Performance and power dissipation comparisons between 28 Gb/s NRZ, PAM, CAP and optical OFDM systems for datacommunication applications,” J. Lightwave Technol. 30(20), 3273–3280 (2012). [CrossRef]
18. R. S. Tucker, “Green optical communications—part I: energy limitations in transport,” IEEE J. Sel. Top. Quantum Electron. 17(2), 245–260 (2011). [CrossRef]
19. J. L. Wei, Q. Cheng, R. V. Penty, I. H. White, and D. G. Cunningham, “400 Gigabit Ethernet using advanced modulation formats: performance, complexity, and power dissipation,” IEEE Commun. Mag. 53(2), 182–189 (2015). [CrossRef]
