Energy-Efficient Optical Access Networks Supported by a Noise-Powered Extender Box

480

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 17, NO. 2, MARCH/APRIL 2011

Energy-Efficient Optical Access Networks Supported by a Noise-Powered Extender Box Bernhard Schrenk, Student Member, IEEE, Francesc Bonada Bo, Johan Bauwelinck, Member, IEEE, Josep Prat, Member, IEEE, and Jose A. Lazaro, Member, IEEE

Abstract—A method for energy-efficient amplification by reutilization of optical noise as pumping power in access networks with extended loss budget is presented. The amplified spontaneous emission (ASE) of an optical amplifier at the customer premises is thereby reshaped and reused as a natural pump source for Erbiumdoped fibers (EDF) inside the signal distribution elements of an access network. This increases the number of served customers and enables also transmission at no extra cost, by just recycling optical noise. Three scenarios, which differ in the design of the optical network unit (ONU), are evaluated and show that up to 4000 users can be served in passive optical networks with adequate signal quality for certain configurations of the ONU. Typical transmission constraints due to the use of cost-effective reflective ONUs, such as Rayleigh backscattering effects, are reduced due to the introduction of a distributed amplification scheme. Besides, full-duplex 10 Gb/s transmission on a single wavelength with simultaneous EDF pump generation is demonstrated with low-cost ONUs, based on reflective semiconductor optical amplifiers. A comparison with traditional amplification techniques is given, and the relationship between the different power consumption for the pump delivery is discussed. Index Terms—Optical access, optical fiber amplifiers, optical fiber communication, optical noise, optical pumping, passive optical network (PON).

I. INTRODUCTION IBER-TO-THE-HOME solutions for optical telecommunication are gaining increased attention in the scientific and commercial field of the access segment, mostly fuelled due to its capability to deliver high-bandwidth triple-play services over the last mile toward the customers [1]. At the same time, a convergence between metro and access networks takes place, aiming to cover a wide geographical area with a high-customer density with a single network [2]. Especially, the passive optical network (PON) has been found as the most cost-effective solution, since there are no electrically powered components located in the optical distribution network between the central

F

Manuscript received March 1, 2010; revised May 5, 2010; accepted June 16, 2010. Date of publication August 11, 2010; date of current version April 6, 2011. This work was supported in part by the European FP7 EURO-FOS and Scalable Advanced Ring-Based Passive Dense Access Network Architecture projects, and by the Spanish Ministerio de Ciencia e Innovaci´on under Project TEC2008-01887 and Formaci´on de Prefesorado Universitario program. B. Schrenk, F. Bonada Bo, J. Prat, and J. A. Lazaro are with the Department of Signal Theory and Communication, Technical University of Catalonia, 08034 Barcelona, Spain (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). J. Bauwelinck is with the Department of Information Technology (IMEC/INTEC), Ghent University, 9000 Ghent, Belgium (e-mail: johan. [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JSTQE.2010.2053839

office of the service provider and the customer premises. As a result, the capital and operating expenditures are inherently lower, and can be further lowered once the share of the deployed fiber-optic components is increased between a higher number of customers. By incorporating wavelength- and timedivision multiplexing techniques (WDM/TDM), the access network, which is then referred to as being hybrid, evolves at the cost of higher optical losses to a dense network with a certain topology. These topologies are typically ring + tree or trunk + tree type [2]. The trunk/ring segment is thereby in principle used as feeder for a comb of WDM signals, while the tree section incorporates a TDM splitter to locally distribute the data signals among a high number of customers that are equipped with a cost-efficient colorless optical network unit (ONU) [3]–[6]. With the extension of the network reach and the tree split, significant optical losses are introduced at the light path of such hybrid networks. As a solution to this, means of optical amplification are introduced inside the network, most commonly at the interconnection points between the WDM and the TDM segment [4], [6]–[8]. In addition, the customer premises equipment, referred to as the ONU, is designed in a way so that a strong optical output signal is provided. The ONU is, therefore, typically equipped with a cost-effective optical amplifier having a small form factor. A promising candidate for this kind of amplifiers has been found with the semiconductor optical amplifier (SOA) or its reflective version, which are both suitable for photonic integrations and capable to provide a moderate to high gain for the incident optical signal, while offering their capability for intensity modulation at low- and high-data rates at the same time [9]. In this way, networks with considerable customer density and a reach up to 100 km have been demonstrated in their technical feasibility [4], [10], [11]. On the other hand, the focus is thereby mostly put on the transmission of data signals rather than on environmentally related considerations, such as the energy consumption of the incorporated amplification techniques. Considering this additional aspect, a wide variety of approaches exists to face the optical transmission losses; while some network proposals take advantage of electronically powered amplifiers that are used as extender boxes [4], [8], others are remotely pumping rare-earth-doped fiber amplifiers inside the fully passive fiber plant from the optical line terminal (OLT) by transmitting a strong optical pump, taking also benefit of Raman amplification [11], [12]. With the expected mass deployment of the continuously migrating access networks, the energy efficiency of the applied

1077-260X/$26.00 © 2010 IEEE

SCHRENK et al.: ENERGY-EFFICIENT OPTICAL ACCESS NETWORKS

Fig. 1.

481

Scheme of the optically noise-powered extender box, placed between the two tree splitting stages with split 1:N and 1:M.

amplification scheme has to be considered as a main contributor to the overall environmental footprint of the network. As a solution for an energy-efficient PON, we propose to reuse a SOA-based amplifier inside the ONU, by feeding its spontaneously emitted noise as a natural byproduct in a controlled way toward an Erbium-doped fiber (EDF) that is placed inside the optical distribution network. With this technique, significant signal amplification can be obtained inside a PON without introducing a significant increase of neither the complexity of the ONU nor its power consumption. The rest of the paper is organized as follows. Section II presents the technique to obtain an optically noise-powered extender box and explains the required modifications for the ONU and the PON. Sections III–V show three scenarios for different ONU designs and evaluate their performance. Section VI gives a comparison with traditional techniques, and points to the individual advantages and energy efficiencies. Finally, Section VII concludes this paper.

II. OPTICALLY PUMPED EXTENDER BOX WITH NOISE REUSE FROM THE CUSTOMER PREMISES EQUIPMENT Following an approach, where a concentrated fiber amplifier is inserted in the lightpath, the required pump for the EDF can be either transmitted from the OLT or the ONU. While the first option suffers from the high-transmission losses for the pump on the long way toward the EDF, a scheme, where the ONUs provide a pump toward the amplification stage, is advantageous. One possibility to add pumping capabilities to the ONU relies on the incorporation of optical sources for the pump waveband of the EDF [13]. Although this might be seen to be in agreement with a colorless ONU design and retains the possibility for its mass deployment, this method is inefficient when

considering the extra cost and power consumption of the light sources. Alternatively, a SOA can fulfill the task of pump generation, while it is also used as amplifier for a weak data signal. This functionality as dual-waveband amplifier can be established by laying a seed loop over the SOA and parts of the fiber plant, as it is illustrated in Fig. 1 and was previously demonstrated [14]. As long as the seeding process that occurs at a useful EDF pump wavelength, does not prevent the concurrent amplification of the weak data signal, a beneficial effect of recycling the amplified spontaneous emission (ASE) can be obtained. This, in turn, allows a more efficient use of energy, without dismissing the previous functionality of the SOA. As a prerequisite, the SOA gain has to cover not only the wavelength range of the employed data signals, but also the pump waveband, and should be additionally tailored in its spectral response to optimize the efficiency for the pump wavelength. In this way, additional active or complex elements at the ONU are avoided, while the driving requirements, and therefore, the power consumption of the SOA stays the same as if it would be solely used for the amplification of the data signal. The widespread usage of SOAs enables this technique to be deployed in a variety of ONU designs, as will be demonstrated in Sections III–V for ONU transmitters that are based on a combination of SOA and reflective electroabsorption modulator (REAM) or on solely RSOAs, and for ONUs containing a SOA as preamplifier for the downstream reception. Despite the low-saturation power of SOAs, this technique can offer pumps that are strong enough to obtain a significant signal gain in EDFs. The benefit is taken from a seed-loop design, which allows to add multiplexers for the generated pumps to the loop. The ASE of the SOA can be sliced with filters in the fiber plant, and fed back to the ONU in a spectrally well-defined window. In this way, the SOA is reshaping its own ASE toward

482

Fig. 2.

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 17, NO. 2, MARCH/APRIL 2011

PON architecture with dual-stage power splitter including an intermediate EDF amplifier, pumped via the ASE output of the SOA + REAM-based ONU.

this spectral window, which coincides with a preferred pump wavelength of the EDF. The granularity of the seeded pumps is not only limited by the need for a sufficiently strong ASE feedback toward the SOA (i.e., a spectrally broad feedback), but depends also on the network design of the tree segment. Considering the case that the EDF is placed inside the power splitter to reduce its net-splitting loss, as shown in Fig. 1, the number of ONUs that are bundled in a single-seed loop is determined by the split M of the second splitting stage. In a simple topology, where half of the ONUs are used for forward pumping of the EDF and the other half for backward pumping, the number of required pump wavelengths is, hence, M/2. In detail, the ASE of several ONUs is split off the data signals in front of the 1:M power splitter by waveband WDM couplers, and redirected to a loop that is constructed by a circulator. A bigger part of the ASE is dropped to the EDF by couplers (CS 1 and CS 2 in Fig. 1) with a fixed ratio of, e.g., 80/20. Multiplexing devices (M1 and M2 ) allow to combine a series of seed loops that are spanned over different ONUs, whereby the multiplexer operates inside the pump waveband. These multiplexers are responsible for shaping the ASE of the SOA (or RSOA), since they provide a wavelength-specific feedback to each ONU. Inside the ONU, the seed loop is spanned over a SOA with the help of a circulator and waveband couplers, or simply fed together with the data signal into a RSOA, depending on the chosen ONU design, as will be explained in the following sections. Typical pump powers that have been achieved with this technique are up to 6.4 dBm for SOA-based and 5.1 dBm for RSOA-based ONUs with saturation power levels of 10.2 and 11 dBm for SOA and RSOA, respectively, as will be reported later in Sections III–V. Considering a split M = 4, this already leads to an overall EDF pump of 12.4 dBm. In conjunction with low-doped EDFs, such as the HE980, this leads to signal gain values >10 dB [6]. Although a slightly higher gain can be achieved with remotely pumped EDFs that are placed before the power splitter and delivered with a stronger pump from the OLT, the high-concentrated splitting loss is avoided when the pump is delivered from the ONU due to a more distributed amplification scheme. This, in turn, results not only in an improved optical SNR (OSNR) for the upstream signal, but also benefits from a lower influence of Rayleigh backscattering (RB) effects in the feeder fiber of the power splitter, where the downstream can cause severe degradation for the upstream transmission [14].

III. SCENARIO 1: REFLECTIVE ONU TRANSMITTER BASED ON SOA AND REAM To evaluate the transmission performance of such a selfpumped PON, first ONU implementation with a SOA and a REAM as transmitter was chosen. Focus is given only on the upstream transmission, although full-duplex transmission could be easily obtained by adding a downstream signal on a separate wavelength in another waveband. The experimental setup of the hybrid PON, shown in Fig. 2, corresponds to a tree-like architecture, in which a 1 × 40 arrayed waveguide grating (AWG) with a channel spacing of 100 GHz is preceding a dual-stage power splitter with an overall split of N × M = 32. This dense PON is, therefore, capable of serving more than 1200 customers. In addition, a dual single-mode fiber (SMF) trunk of 25 km length is placed between the OLT and the AWG, and the tree splitter is connected to the AWG via a 10-km-long feeder SMF. RB effects are therefore considered for the TDM segment after the AWG. An unmodulated optical carrier at 1559.6 nm is transmitted from the OLT toward the ONU, where the upstream data is continuously imprinted via the REAM, using a data rate of 10 Gb/s and a pseudorandom bit sequence (PRBS) with length 231 – 1. Upstream reception at the OLT occurs with a PIN diode that is placed after an EDF-based preamplifier with a noise figure of 4.7 dB and a 200 GHz bandpass filter. A dispersion compensating fiber (DCF) with a dispersion of −671 ps/nm compensates for the pulse broadening along several fiber spans, which is pronounced due to the chirp of the REAM. Since the REAM is subject to high-intrinsic losses of ∼16 dB, SOAs are placed at the ONU to act as preamplifier and booster. Since a single integrated SOA + REAM device with high-saturation output power was not available, this configuration with a dual SOA was chosen to avoid cross-gain modulation between the outgoing REAM signal and the incident unmodulated optical carrier. The preamplifier SOA is not only intended to level the received optical carrier, but also to seed the EDF pump in the 1480 nm waveband. Waveband splitters for the S- and C-band are, therefore, inserted around the SOA and at the outputs of the second splitting stage. Two 1 × 2 pump multiplexers for the 1480 and 1490 nm channel were inserted in the two seed loops for the forward- and backward-propagating EDF pump. These wavelength windows are compatible with the 15 m long HE980 EDF, which was inserted between the 1:8 and 1:4

SCHRENK et al.: ENERGY-EFFICIENT OPTICAL ACCESS NETWORKS

483

Fig. 3. (a) Spectra of the injected EDF pumps that can be obtained in one of the channels by a single ONU, and (b) delivered pump power for different power values of the received upstream carrier.

splitting stage. Due to a limited availability of components, the EDF pumps of the three other ONUs that contribute to the overall pump seed were emulated by laser diodes that were set to the obtained pump power values of the shaped ASE from the other three ONUs. The spectra of the seeded pumps for both channels are shown in Fig. 3(a), for the case that a single ONU is seeding in one of these channels. The spectra are, thereby, located at the maximum overlap between the ASE spectrum of the SOA, which has a maximum at 1492 nm, and the transmission of the pump channels [see Fig. 3(a)], and are therefore shifted from the nominal wavelengths of 1480 and 1490 nm. The shared SOA gain medium for the shaped ASE and the upstream carrier lead to a reduction of the EDF pump once the power of the upstream carrier increases. However, for reasonable input values below 12 dBm for the carrier, there is no significant reduction for the pump, as can be seen in Fig. 3(b). Power values of 6.4 and 5.9 dBm can be obtained in the 1480 and 1490 nm channel, respectively. For these pumps, the EDF provides a gain of 8.9 and 10 dB for the unmodulated carrier and the modulated upstream signal, and reduces the net splitting loss of the 1:32 splitter, typically ∼15.5 dB, to just 6.6 and 5.5 dB, respectively. The rest of the EDFs, which are connected to the other output ports of the first splitting stage, cause additional noise contribution for the upstream. For this reason, an ASE source was added to account for that noise accumulation. This might be seen as a disadvantage of the self-pumped PON, but is negligible once the splitting ratio of the first splitting stage is low or the noise degradation from the reflective ONU transmitter is already dominating the upstream OSNR, as will be addressed in more detail in Sections IV and V. The evolution of the power levels for down- and upstream is shown in Fig. 4, together with the OSNRs for the upstream. Three different power levels of PTX = 4, 7, and 10 dBm have been considered for the launched upstream carrier (point A in Fig. 2). For a launch of 7 dBm, the power levels and OSNRs are 8.2 dBm and 53.3 dB at the input of the power splitter (D), and −14.8 dBm and 33.2 dB at its output (G) when reaching the ONU. While a stronger launch of 10 dBm does not benefit from

Fig. 4.

Signal evolution for three different launches of the upstream carrier.

additional output power at the EDF, the OSNR of a weaker launch of 4 dBm is additionally degraded. The strong OSNR degradation to 23.1 dB for a launch of 7 dBm at the booster SOA (I-G) is caused by the high-intrinsic loss of the REAM (H-I), which was 16.8 dB. However, the net gain of the ONU (G-H-I-G) is 8.1 dB. On the return path, the signal reaches the power splitter (G) with –6.7 dBm, and the feeder fiber (D) with –12 dBm and an OSNR of 22.1 dB, already considering the noise accumulation from other EDFs. The signal arrives with −26.3 dBm at the OLT receiver and sets the available power for upstream reception. For a weak launch of 4 dBm, not only the received power is lower with –30 dBm, but also the OSNR is reduced with 19.2 dB. The influence of RB is given for the feeder fiber and the EDF. The optical signal-to-RB ratio (OSRR) at the feeder fiber (C) is 26.8 and 24 dB for a launch of 7 and 10 dBm, respectively. Since the RB coefficient, defined as the ratio between the Rayleigh backscattered light of a fiber and the launched light into the fiber, is −24.3 dB for the 15-m-long EDF, and therefore, higher than for the SMF, which had a RB coefficient of −34.2 dB for its length of 10 km, an extra OSRR contribution has to be taken into account for the EDF amplifier. However, due to the high loss of the first splitter (i.e., the low EDF input power), this OSRR is 28.7 dB, and therefore, proves that the RB contribution of the EDF is not dominating. The transmission performance was evaluated in terms of bit error ratio (BER) measurements, for which an attenuator (AU ) was placed in front of the optical receiver at the OLT. The high-loss budget of the PON restricts to reach low BER values of 10−10 for launched upstream carriers of 4 and 10 dBm (see Fig. 5). This is caused by error floors, which derive from the low OSNR and RB in the feeder fiber, respectively. However, for the optimum launch of 7 dBm, a BER of 10−10 can be reached. The power margin, defined as the difference between the available power and the reception sensitivity, is then 1 dB and could be increased to 7.7 dB for the incorporation of a Reed–Solomon (255, 239) forward error correction (FEC),

484

Fig. 5.

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 17, NO. 2, MARCH/APRIL 2011

BER measurements for different upstream carrier launches.

which allows reception at a BER of 10−4 , as proposed for the gigabit-PON (GPON) standard [15]. Note that if the EDF-based extender box inside the splitter is removed, the high-loss budget prevents data transmission due to a low received upstream signal power level below –40 dBm and/or a low OSNR and OSRR, even at the FEC level. IV. SCENARIO 2: ONU TRANSMITTER BASED ON LASER DIODE + EAM In the second case study, the reflective transmitter of the ONU is replaced by a laser diode that was integrated together with an EAM. Together with a booster SOA, this allows to launch a powerful upstream signal from the ONU, which does not suffer from OSNR degradation due to modulation of an already noisy optical carrier as in the previous scenario. The investigation aims at achieving high-splitting ratios, by investigating the best combination between the first split N and the second split M at the tree segment. The PON is, therefore, reduced to a TDM tree without WDM segment (see Fig. 6). The 10 Gb/s downstream is provided at 1531.9 nm via a Mach–Zehnder modulator, while the 10 Gb/s burst-mode upstream was transmitted at 1548.22 nm with a GPON compatible 125 μs frame and a duty cycle of 1:4, meaning a packet length and booster-SOA gate of 31.5 μs. A PRBS of length 231 – 1 was chosen for both data streams. Separation of down- and upstream is therefore possible with red/blue splitters inside the C-band. The same OLT receiver and the same EDF as in Section III were used. Since the booster-SOA at the ONU is intended to be gated for burst-mode transmission, it is not possible to reuse it for seeding a pump. Instead, the ONU receiver is implemented by a combination of SOA + PIN diode that substitutes the typically used APD. The SOA had a noise figure of 8.9 dB, and the optical reception bandwidth for the PIN diode was 20 nm. The seed loop is spanned in a similar way over the ONU, while at the second splitting stage just four ONUs are connected. The obtained pumps at 1480 and 1490 nm are similar to the ones obtained in Section III. With power levels for the received downstream below –15 dBm, there is also no significant degradation for the pump seed due to the shared SOA gain medium.

The obtained EDF gain for the upstream signal is shown in Fig. 7(a). Since the upstream arrives strong at the EDF for a low second split M, it sets the saturation point and clamps the gain, which is also low in this case. The gain of the weaker downstream is then quite independent of the first split N. Just for the case that the first split N is very low and the second split M is high at the same time, the downstream has significant influence on the upstream gain. Note that the upstream gain is slightly higher due to the burst-mode transmission and the resulting EDF gain transients, since the chosen packet length of 31.5 μs is small enough to still have some excess gain of ∼0.5 dB at the end of the packet (see Fig. 6), while no severe asymmetries have been observed in the eye. For the BER contour measurements, the two splitting ratios 1:M and 1:N were emulated with variable attenuators, whereby a excess loss of 0.5 dB for each splitting stage was considered. The splitting loss for each attenuator was independently increased in steps of 3 dB (i.e., doubling the splitting ratio). The downstream BER, presented in Fig. 7(b), shows a strong tilt in its iso-BER curves. This is caused by the OSNR degradation in the EDF between the splitting stages. For a certain split N (e.g., 32), the second split M does not cause additional penalties unless it reaches higher values (e.g., 128), meaning a further OSNR degradation due to the SOA in ONU receiver. For N × M = 16 × 256, an overall 1:4k split is possible for a downstream BER of 10−10 . Preferring higher second splits, a PON split of 1:2k is possible for a series of split combinations. With FEC, the total split can be extended to 1:8k (N × M = 16 × 512, 32 × 256, 64 × 128). For the burst-mode upstream with ASE accumulation, the OSNR worsens for higher first splits N, so that a split M > N is preferred. This means not only less employed EDFAs in the fiber plant, but leads to a more distributed amplification scheme at the same time. The iso-BER curves for the upstream, shown in Fig. 7(c), follow this OSNR degradation and are shifted toward a reduced total split. Nevertheless, reception at a BER of 10−10 can be obtained for an overall split of 1:1k for not too high first splits (N × M = 8 × 128, 16 × 64, 32 × 32, 64 × 16), while a total split of 1:4k can be achieved with the help of FEC, preferably for higher second splits (N × M = 8 × 512, 16 × 256, 32 × 128, 64 × 64). Considering full-duplex transmission, the maximum split is limited by the upstream reception, allowing a PON split of 1:1k at a BER of 10−10 (N × M = 8 × 128, 16 × 64, 32 × 32) and 1:4k with FEC (N × M = 8 × 512, 16 × 256, 32 × 128, 64 × 64). Although these high splits limit the guaranteed bandwidth per user, it shows that a high-customer density of PONs is not restricted by the split, making such PONs attractive for very short reaches with increased data rates. For the particular case of N × M = 16 × 64, corresponding to a loss budget of 39.5 dB far beyond the GPON class C++ or the next-generation PON (NGPON) class PR30, margins of 3.4 dB at a BER of 10−10 and 7.8 dB with FEC are given for the downstream. Considering the upstream transmission, slightly larger margins of 5.4 and 11.6 dB are available despite noise accumulation, for a BER of 10−10 and with FEC, respectively.

SCHRENK et al.: ENERGY-EFFICIENT OPTICAL ACCESS NETWORKS

485

Fig. 6.

Self-pumped PON with an ONU transmitter based on laser diode + EAM. The inset on the right shows the burst-mode upstream data packets.

Fig. 7.

(a) EDF gain for the upstream. (b) Continuous downstream BER. (c) Burst-mode upstream BER considering noise accumulation across the tree.

V. SCENARIO 3: RSOA-BASED ONU TRANSMITTER In the third case study, the ONU transmitter is reduced to a single RSOA, which is advantageously used for data signal remodulation and EDF pump generation. The PON architecture, shown in Fig. 8, corresponds to the one that is also used in Section III, but with a swapped splitting ratio N and M for the first and second splitter, respectively. This is motivated by the expected lower pump output from the ONU, requiring a larger number of pump wavelengths, and therefore, a higher splitting ratio in the second stage. Down- and upstream are continuously modulated at 10 Gb/s on a single wavelength in the L-band. The OLT receiver was slightly altered by using an APD instead of the PIN diode. At the ONU, the downstream, which is modulated with reduced extinction ratio (ER), is split by a 50/50 coupler (CO ) and remodulated by the RSOA, which had a gain spectrum centered at 1561 nm and a small signal gain bandwidth of 39 nm [9]. This provides not only the possibility to amplify an L-band data signal, but also to seed a pump at 1540 nm. This, in turn, is compatible with, e.g., a 50-m-long C-band HE980 EDF amplifier, which can offer also amplification over a wide wavelength range in the L-band when being pumped around 1540 nm [16]. The shaped ASE in the C-band is bypassed around the CO splitter with C/L waveband couplers not to suffer from additional loss in the seed loop. For a practical implementation with burst-mode traffic, a switchable loss element (SO ), which could be also implemented by a cheap attenuator, would have to be inserted

in the ONU at the data signal path between the CO coupler and the RSOA. Otherwise, all inactive ONUs would also amplify and reflect the incident downstream signal toward the tree, and thereby cause severe crosstalk for the upstream signal of the active ONU. The 1 × 4 multiplexers M1 and M2 in the optical distribution network had a bandwidth of 200 GHz for each of the four channels that were located at 1538.19, 1539.77, 1541.35, and 1542.94 nm. The obtained pump levels were 4.5, 4.7, 5.0, and 5.1 dBm for a RSOA bias of 200 mA (inactive ONU), and 2.2, 2.7, 3.1, and 3.3 dBm for a bias of 120 mA, which is used for data modulation, as in [17]. The gain for the data signal at 1585 nm was 22.5 dB for an input power of –20 dBm, while a pump was seeded in addition. Fig. 9 shows the obtained EDF gain at 1585 nm and the OSNR for the case of unidirectional amplification. Different signal and pump power levels were measured at different pump wavelengths. For a pump power in the range of the multiplexed pump seeds, a gain of ∼10 dB, quite independent of the pump wavelength, can be achieved. The obtained OSNR values correspond to a noise figure of ∼6 dB. With the reduction of the net 1:32 splitting loss from 15.8 dB to just 5 dB, the self-pumped PON allows fullduplex transmission even with reflective ONUs based on RSOAs. A strong downstream is provided to the ONU and guarantees also a high OSRR of 30.7 dB (including RB in the EDF), while the upstream reaches the OLT with a power of 22.1 dBm and an

486

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 17, NO. 2, MARCH/APRIL 2011

Fig. 8.

Self-pumped PON with RSOA-based ONUs and wavelength reuse. The L-band data signals are amplified by an EDF that is pumped in the C-band.

Fig. 9.

EDF gain at 1585 nm and OSNR for unidirectional amplification.

OSNR of 27.6 dB. Note that although the downstream OSNR is 33.4 dB, the reception suffers from the fact that no ASE filter was used after the EDF, to keep the lightpath as simple and transparent as usual TDM-PONs. Noise accumulation from the other three EDF stages across the tree has been also emulated by the ASE source in Fig. 8 and it has been verified that it is not a problem. As can be seen in Fig. 10, the RSOA noise overcasts the noise of a single EDF stage substantially, so that after the accumulation, the difference in the ASE background, marked as ΔASE , is > 6 dB. The BER measurements, acquired with the help of the attenuators AD and AU and taken for a PRBS of length 231 – 1, are shown in Fig. 11. The downstream is constraint by the unfiltered ASE background of the EDF, though still providing a sensitivity of –22.8 dBm at 10 Gb/s for an ER > 10 dB and a standard APD receiver. For an ER of 3 dB, an expected penalty of 5.4 dB is suffered over an ER > 10 dB at the FEC level, leaving a power margin of 1.4 dB. The upstream of the self-pumped PON suffers from error floors that are introduced from the present downstream pattern, which is only suppressed by the optical gain saturation effect of the RSOA. Better performance could be achieved by incorporating additional downstream cancellation techniques [18]. However, transmission is possible with FEC, having a penalty of 5.5 dB over the downstream-less case, and a margin of 2.2 dB. The replacement of the feeder fiber with equivalent

Fig. 10. Upstream spectra after amplification with a single EDF stage () and after the feeder fiber without () and with noise accumulation (䊉). (blank markers: inactive ONU, filled markers: active ONUs).

Fig. 11. BER measurements for the downstream (DS, dashed lines) and the upstream (US, solid lines), taken for different downstream ERs.

attenuation did not enhance the upstream performance, even at low BERs, proving that the downstream crosstalk is the dominant impairment and more critical than RB. The demonstrated PON allows to serve >1000 customers with full-duplex transmission at 10 Gb/s over loss budgets equivalent to the GPON class C or NGPON class PR30, considering

SCHRENK et al.: ENERGY-EFFICIENT OPTICAL ACCESS NETWORKS

the signal launch from OLT and ONU and the reception sensitivities for down- and upstream, the seeding condition of the RSOA and the nominal power splitter loss.

487

to additional 1.6 W, since no share among multiple network customers is given. VII. CONCLUSION

VI. COMPARISON WITH OTHER AMPLIFICATION SCHEMES The energy efficiency is assessed with a simple approach similar to [19] or [20]. Comparing the additional required power at the ONU to seed the EDF pump, the difference to a standard ONU design is mainly found in the continuous gating of a SOA that is originally intended to be switched in burst-mode (scenarios 1 and 3). However, considering the power hungry 10 Gb/s electronics at the ONU that are used for higher layer functionalities and also the modulator driver for the high-frequency data signals, this continuous SOA bias contributes just a small part of ∼0.3 W when compared to a typical overall ONU power of 5 W [19]. Note that this contribution depends strongly on the SOA design, since an overlap between the SOA gain spectrum and the pump wavelength of the EDF is required to achieve optimum energy efficiency. The semiconductor material should be, therefore, optimized by means of band structure engineering [21]. Considering alternative amplification schemes that deliver a pump from the OLT [6], [11], or schemes, where electrically powered extender boxes or protocol terminators are located inside the fiber plant [1], [4], [8], the high-power consumption that is needed for these added subsystems requires extra power supplies and environmental conditioning, and contribute therefore with some overhead [19]. Utilizing commercial equipment that is shared among 32 users per tree, these amplification schemes contribute with ∼1–4 W/user, depending on the exact realization, and eventually raise the capital and operating expenditures due to the need for cabinets in the field. Especially, the use of electronics at remote network nodes, applied to perform network protocol translation [1], shows up with a high-additional power requirement of 4 W/user. A strong reduction of this additional amount of energy consumption to just 1 W/user can be obtained by keeping the data in the optical domain, just performing optical amplification with local electrical supply [4]. However, the required pump laser diodes and their thermal conditioning prevent to reach lower power consumptions. On the contrary, remote pumping techniques allow to centralize electrically powered equipment [6]. Considering strong commercial pumps, the required extra power per user can be found about 0.4 W for the previously presented network architectures, and is therefore, comparable with the proposed technique. However, it shall be noted that a remote pumping scheme from the OLT has to overcome the transmission losses once the reach of the network is extended, and requires stronger pumps in such a case. Although the use of pump laser diodes at the customer premises [13] would provide a more effective pumping scheme, it is inefficient in terms of energy consumption, since an extra energy overhead is introduced due to the buildup of the inversion in the pump laser diode and also due to its extra thermal cooling for the case of a discrete ONU design. This has direct impact on the energy consumption per user, which raises

A novel pumping scheme that reuses the ASE from a SOA- or RSOA-based ONU has been evaluated. Together with a seeding loop and a multiplexing technique for combining the pumps of different ONUs, EDF gain values of ∼10 dB can be achieved and reduce the splitting loss in the TDM tree of a hybrid PON. For the case of nonreflective ONUs, the loss budget can be increased to 39.5 dB, while maintaining data transmission, providing service to > 4000 users in purely TDM-based PON. The concurrent operation of a RSOA as modulator and as pump source has been demonstrated, together with full-duplex 10 Gb/s transmission over a single wavelength, without adding pumping capabilities to the OLT nor placing electrically powered amplifiers in the fiber plant. Considering alternative amplification schemes of access networks, the proposed pumping scheme has been found to be a good candidate for energy- and cost-effective PONs. It extends the use of common infrastructure to a significantly higher number of users, by no extra energy cost, just the reuse of optical noise as pump for optical amplification. ACKNOWLEDGMENT The authors would like to thank G. de Valicourt and R. Brenot from Alcatel-Thales III–V Lab for the fruitful discussions and the RSOA supply. REFERENCES [1] J. H. Lee, S.-H. Cho, H.-H. Lee, E.-S. Jung, J.-H. Yu, B.-W. Kim, S.-H. Lee, J.-S. Koh, B.-H. Sung, S.-J. Kang, J.-H. Kim, K.-T. Jeong, and S. S. Lee, “First commercial deployment of a colorless gigabit WDM/TDM hybrid PON system using remote protocol terminator,” J. Lightw. Technol., vol. 28, no. 4, pp. 344–351, Feb. 2010. [2] L. G. Kazovsky, W. Shaw, D. Gutierrez, N. Cheng, and S. Wong, “Nextgeneration optical access networks,” J. Lightw. Technol., vol. 25, no. 11, pp. 3428–3442, Nov. 2007. [3] G. Talli and P. D. Townsend, “Hybrid DWDM-TDM long-reach PON for next-generation optical access,” J. Lightw. Technol., vol. 24, no. 7, pp. 2827–2834, Jul. 2006. [4] S. Appathurai, R. Davey, and D. Nesset, “Next generation fibre-to-thehome solutions,” in Proc. BROADNETS2008, London, U.K., Sep.2008, pp. 232–235. [5] F.-T. An, K. S. Kim, D. Gutierrez, S. Yam, E. Hu, K. Shrikhande, and L. G. Kazovsky, “SUCCESS: A next-generation hybrid WDM/TDM optical access network architecture,” J. Lightw. Technol., vol. 22, no. 11, pp. 2557–2569, Nov. 2004. [6] J. A. Lazaro, C. Bock, V. Polo, R. I. Martinez, and J. Prat, “Remotely amplified combined ring-tree dense access network architecture using reflective RSOA-based ONU,” OSA J. Opt. Netw., vol. 6, pp. 801–807, Jun. 2007. [7] J. M. Oh, S. G. Koo, D. Lee, and S. J. Park, “Enhancement of the performance of a reflective SOA-based hybrid WDM/TDM PON system with a remotely pumped erbium-doped fiber amplifier,” J. Lightw. Technol., vol. 26, no. 1, pp. 144–149, Jan. 2008. [8] F. Saliou, P. Chanclou, F. Laurent, N. Genay, J. A. Lazaro, F. Bonada, and J. Prat, “Reach extension strategies for passive optical networks,” J. Opt. Commun. Netw., vol. 1, no. 4, p. C51–C60, Sep. 2009. [9] G. de Valicourt, D. Mak´e, J. Landreau, M. Lamponi, G. H. Duan, P. Chanclou, and R. Brenot, “High gain (30 dB) and high saturation power (11 dBm) RSOA devices as colorless ONU sources in long-reach hybrid WDM/TDM-PON architecture,” IEEE Photon. Technol. Lett., vol. 22, no. 3, pp. 191–193, Feb. 2010.

488

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 17, NO. 2, MARCH/APRIL 2011

[10] B. Schrenk, F. Bonada, M. Omella, J. A. Lazaro, and J. Prat, “Enhanced transmission in long reach WDM/TDM passive optical networks by means of multiple downstream cancellation techniques,” presented at the ECOC2009, Vienna, Austria, Sep., We.8.5.4. [11] R. Kjaer, I. T. Monroy, L. K. Oxenløwe, P. Jeppesen, and B. Palsdottir, “Bi-directional 120 km long-reach PON link based on distributed raman amplification,” in Proc. LEOS2006, pp. 703–704. [12] J. A. L´azaro, J. Prat, P. Chanclou, G. M. Tosi Beleffi, A. Teixeira, I. Tomkos, R. Soila, and V. Koratzinos, “Scalable extended reach PON,” presented at the OFC2008, San Diego, CA, Feb., OThL2. [13] M. D. Feuer, R. D. Feldman, J. L. Zyskind, S. C. Shunk, J. Sulhoff, and T. H. Wood, “Remotely-pumped self-amplified star network for local access,” in Proc. OFC1996, San Jose, CA, Feb., pp. 146–147. [14] B. Schrenk, F. Bonada, J. A. Lazaro, and J. Prat, “Fortistis: Split extension in dense passive optical networks by inline amplification with remote ASE-shaped pump delivery via colorless optical network units,” in Proc. OFC2010, San Diego, CA, Mar., JThA33. [15] Gigabit-Capable Passive Optical Networks (G-PON): Transmission Convergence Layer Specification, ITU-T Recommendation G.984.3, 2003. [16] B. H. Choi, H. H. Park, and M. J. Chu, “New pump wavelength of 1540nm band for long-wavelength-band erbium-doped fiber amplifier (L-Band EDFA),” IEEE J. Quant. Electr., vol. 39, no. 10, pp. 1272–1280, Oct. 2003. [17] B. Schrenk, G. de Valicourt, M. Omella, J. A. Lazaro, R. Brenot, and J. Prat, “Direct 10 Gb/s modulation of a single-section RSOA in PONs with high optical budget,” IEEE Photon. Technol. Lett., vol. 22, no. 6, pp. 392–394, Mar. 2010. [18] J. H. Yu, N. Kim, and B. W. Kim, “Remodulation schemes with reflective SOA for colorless DWDM PON,” OSA J. Opt. Netw., vol. 6, pp. 1041– 1054, Aug. 2007. [19] J. Baliga, R. Ayre, W. V. Sorin, K. Hinton, and R. S. Tucker, “Energy consumption in access networks,” presented at the OFC2008, San Diego, CA, Feb., OThT6. [20] A. Lovric and S. Aleksic, “Power efficiency of extended reach 10G-EPON and TDM/WDM PON,” presented at the OFC2010, San Diego, CA, Mar., NMC4. [21] I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, “Band parameters for III–V compound semiconductors and their alloys,” AIP J. Appl. Phys., vol. 89, pp. 5815–5875, Jun. 2001.

Bernhard Schrenk (S’10) was born in Mistelbach, Austria, in 1982. He received the B.Sc. and M.Sc. degrees in microelectronics from the Technical University of Vienna, Vienna, Austria, in 2005 and 2007, respectively. He is currently working toward the Ph.D. degree from the Polytechnic University of Catalonia, Barcelona, Spain. His Diploma thesis was focused on photonics. He was at the Institute of Experimental Physics of Prof. A. Zeilinger, where he was involved in the realization of a first commercial prototype for a quantum cryptography system, based on an entangled photon source, within the European Secure Communication based on Quantum Cryptography (SECOQC) project. His current research interests include next-generation passive optical networks for long-reach and high-data rate fiber-to-the-home access networks. Mr. Schrenk was engaged in several European projects such as Scalable advanced ring-based passive dense access network architecture, Building the future optical network in Europe, and Euro-FOS projects.

Francesc Bonada Bo was born in Ripoll, Spain, in 1982. He received the M.S. degree in telecommunications engineering from the Universitat Polit`ecnica de Catalunya, Barcelona, Spain, in 2007, where he is currently working toward the Ph.D. degree and his research focuses on remote Erbium-doped fibers (EDF) amplification for long-reach and high-splitting ratio next-generation passive optical networks. Mr. Bonada Bo was engaged in several European projects (Scalable advanced ring-based passive dense access network architecture, Building the future optical network in Europe, and Euro-FOS) on EDF optical-burst-mode amplification and remote node design.

Johan Bauwelinck (M’02) was born in Sint-Niklaas, Belgium, in 1977. He received the M.Sc. degree in applied electronics and the Ph.D. degree in applied sciences and electronics from Ghent University, Ghent, Belgium, in 2000 and 2005, respectively. Since 2000, he has been a Research Assistant in the Information and Technology Design Laboratory, Ghent University, where he is currently a full-time Tenure Track Professor. His research interests include high-speed, high-frequency (opto-) electronic circuits and systems, and its applications on chip and board level, including optical access networks, automotive optical networks, and RF design for wireless communication and ranging.

Josep Prat (M’83) received the M.S. degree in telecommunications engineering and the Ph.D. degree from the Universitat Polit`ecnica de Catalunya (UPC), Barcelona, Spain, in 1987 and 1995, respectively. He is currently a Full Professor in the Optical Communications Group, Signal Theory and Communications Department, UPC. In 1998, he was a Guest Scientist in the University College of London. He is the editor of the book Fiber-to-the-Home Technologies and Next generation FTTH Networks. His current research interests include broadband optical communications with emphasis on high-bit-rate optical systems, access networks, and wavelength-division multiplexing transmission design, and impairment control. Dr. Prat was involved in several European projects (Euro-Fos, Building the future optical network in Europe, ePhoton/One, Layers Interworking Optical Networks, Management of Photonic Systems and Networks, Management of Optical Networks, Switchless Optical Network for Advanced Transport Architecture) on optical transport and access networks, and leads the FP7 Scalable advanced ring-based passive dense access network architecture project on nextgeneration fiber-to-the-home networks.

Jose A. Lazaro (M’01) received the M.Sc. degree in physics and the Ph.D. degree from the University of Zaragoza, Zaragoza, Spain, in 1993 and 1999, respectively. His research was focused on Erbiumdoped waveguide amplifiers. He is currently a Ramon-y-Cajal Researcher at the Optical Communications Group, Department of Signal Theory and Communications, Universitat Polit`ecnica de Catalunya (UPC), Barcelona, Spain, where he is teaching in the School of Telecommunications Engineering of Barcelona and the Master in Photonics. He was a Senior Researcher at the Optical Transmission and Broadband Technologies Laboratory, Aragon Institute for Engineering Research, Huesca, Spain. He was a R&D Engineer in ALCATEL, Stuttgart, Germany in the Department of Passive Optical Components from 2000 to 2002, and in the Department of Optical Transmission Systems from 2002 to 2004. Dr. Lazaro has been involved in several European projects (e-Photon/One, Building the future optical network in Europe, EURO-FOS and Scalable advanced ring-based passive dense access network architecture) on optical transmission, monitoring, subsystems, and access networks.