Saturated Collision Amplifier (SCA) is a novel amplification scheme that uses SOA saturation in order to maximize the output power and minimize the ASE noise and the polarization sensitivity. We demonstrate the SCA reach extension in a commercial single-wavelength XGPON1 prototype system where bidirectional optical budget of up to 50 dB is obtained. The traffic performances are compared between the SCA and the conventional SOA extender. The novel extension scheme is demonstrated also for two- and four-wavelength 10 Gbit/s unidirectional downstream configurations with 45 km and 100 km transmission distances with 58-dB maximum total optical budget for each wavelength.
© 2011 OSA
Passive Optical Network (PON) is a commonly used fixed access network architecture that consists of four basic elements: 1) the Central Office (CO) that hosts the Optical Line Termination (OLT), 2) the feeder line, 3) the passive splitting and the access line, and 4) the subscriber Optical Network Unit (ONU) (see Fig. 1a ). The PON is a point-to-multipoint arrangement, where the OLT shares the bandwidth by Time Division Multiplexing (TDM) between the ONUs.
In order to reduce the deployment and the operation costs of the access network, the PONs have evolved through various development stages with progressively increasing optical budgets . That is, increasing transmission distances, data rates, and splitting ratios. Recently ITU published a new PON standard ITU-T G987.2  for Time Division Multiplexed (TDM) continuous 10 Gbit/s downstream traffic at 1577 nm and burst mode 2.5 Gbit/s upstream traffic at 1270 nm (XGPON1).
The network operators and carriers have commercial interest to continue the increase of the PON power budget. However, the highest achievable budget is limited by sensitivity of receivers having acceptable cost. Therefore, the transmission distance, the splitting ratio, and the data rate cannot be increased indefinitely in a fully passive system. Various methods for the optical power budget increase, often referred to as “the reach extension,” have been proposed [3–5]. Among the proposed ones are the midway amplifiers (Fig. 1b), such as EDFA  and semiconductor optical amplifiers (SOA) [7–9], Raman amplification [10,11], and non-amplifier based coherent detection  techniques. For example, the use of a single SOA for optical budget increase of coexisting gigabit-capable PON (G-PON) and 10-gigabit-capable PON (XG-PON) is shown in .
Saturated Collision Amplifier (SCA) is an SOA based arrangement, consisting of a delay interferometer (DLI), a pair of circulators, and the SOA (see Fig. 2 ) [13,14]. The DLI transforms each bit of the differentially phase coded input signal to two amplitude modulated outputs, ‘mark’ and ‘space’, i.e., signals with opposite polarity. The DLI outputs are coupled to the SOA via the circulators. The complementary signals travel through the gain medium from opposite directions and are coupled out again via the circulators.
The SCA is intended to operate in saturation region, i.e., with relatively high optical input powers. The conventional SOA performs poorly in saturation, because the low-level ‘space’ experiences higher gain than the high-level ‘mark’, resulting in closed signal eye. In SCA the low- and high-level signals travel through the gain medium simultaneously from opposite directions and thus experience the same gain. The gain saturation provides the highest available output power, minimized polarization dependence, and reduced ASE noise. In addition, the differentially coded input signal has optical power on each symbol slot and the saturated amplifier has thus relatively stable input power. Therefore, the SCA exhibits minimal bit-pattern or transient effects, which often plague the conventional SOAs.
In this paper, a network configuration for downstream time- and dense-wavelength-division multiplexed (TDM/DWDM) PON using the SCA is proposed. First we investigate bidirectional single-wavelength reach extension with a commercial XGPON1 prototype using conventional SOAs. The results are compared with a case where the downstream extender is replaced with the SCA. Second, we demonstrate multiwavelength SCA operation in 2- and 4-wavelength DWDM setup. The transmission performance and the optical budget extension are assessed with bit error rate (BER) measurements.
2. Bidirectional XGPON1 reach extension based on conventional SOAs
The experimental setup for XGPON1 reach extension consists of an OLT and two ONUs produced by a major system vendor. The reach extender is composed of two SOAs between the Coarse Wavelength Division Multiplexing (CWDM) multiplexers that amplify separately the up- and the downstream signals (see Fig. 3 ). The small signal (−30 dBm) gain, noise factor, and gain peak wavelengths of the SOAs are, respectively, 20 dB, 6.5 dB, and 1290 nm; and 35 dB, 10 dB, and 1550 nm for up- and downstream directions.
We define the “feeder budget” as the optical budget between the OLT and the reach extender, and the “access budget” as the budget between the reach extender and the ONUs. In order to evaluate the optical budget increase, the architecture is “in-line” type but results could be extrapolated for amplifiers at the Central Office (CO). Variable Optical Attenuators (VOAs) are inserted in the architecture in order to vary for the feeder and access budgets. The setup includes optical fibers with total length of 20 km, which brings the prototype system to its current highest tolerable round-trip time limit. Triple play services were set and running throughout the experiments.
We measured the transmission performance with bidirectional Ethernet traffic (1518 bytes) between the OLT and the two ONUs. The bit-error-rates were not studied, because the system rejects the whole frame when an error occurs. Therefore, we chose to measure the number of lost packets. Without amplifiers, the optical budget of the XGPON1 is limited between 13 dB to 30 dB (maximum 30 dB for the upstream and 32 dB for the downstream). This fits with the class N1 of the standard ITU-T G987.2, which specifies an optical budget between 14 dB and 29 dB.
The packet loss map is depicted in Fig. 4 that was obtained by varying the feeder and access budgets in 1 dB increments. The transmission is considered error free when no packet loss occurs. Downstream transmission performance is shown in Fig. 4a, upstream in Fig. 4b, and bidirectional traffic in Fig. 4c (a combination of the 2 previous cases). The working areas are limited by the SOA gain, noise and the XGPON1 receiver properties (such as sensitivity and saturation behavior) .
The maps show a feeder budget increase of 19 dB and 17 dB for the access budget classes N1 and N2, respectively; and indicate a possibility to reach a total optical budget of 48 dB. For the upstream direction the feeder budget increase remain limited between 19 and 22 dB for the class N1, and 18–20 dB for the class N2.
The transmission on a PON system is obviously bidirectional and the combined results (Fig. 4c) indicate incompatibility with any standard class when using the conventional SOAs for the up- and downstream extension. A total optical budget of up to 46 dB is obtained, but the resulting feeder budget ranges equal to zero due to a high overload of the OTL receiver of the XGPON1 prototype. It could be expected that a higher optical budget could be obtained, using the same downstream SOA but in the SCA configuration.
3. XGPON1 reach extender based on SCA and DPSK to ASK conversion
An experimental setup with the downstream SCA is illustrated in Fig. 5 below. The SOA of the SCA is of the same type that was used in previous experiments. While the SCA requires the DPSK input, the ASK modulated OLT transmitter is equipped with an ASK-to-DPSK transponder (Fig. 6a ). Otherwise the experimental setup is the same than in the previous case. Especially, the ONUs are intact and do not need demodulators, since the SCA converts the DPSK input into ASK output.
The ASK-to-DPSK transponder receives the signal from the XGPON1 OLT and modulates the DPSK signal on another CW carrier at 1577 nm. The output power from the DPSK transmitter is amplified onto + 10 dBm optical power (which is still below the eye safety limit of the OLT) using an L-band EDFA.
The downstream packet loss map is shown in Fig. 6b. While the conventional SOA upstream amplifier is unaltered, the upstream performance remains intact, the same as in previous case. The map of Fig. 6c illustrates the working area of the bidirectional transmission. As evident, the use of the downstream SCA has brought the possibility to reach optical budget of up to 62 dB (31 dB range for the feeder and the access), thus providing compatibility with class N1, N2 and E (18–33 dB) of the XGPON1 standard.
In bidirectional operation (Fig. 6c) the performance is limited by the conventional upstream SOA. Unlike in previous case (conventional SOA up- and downstream extension), the current results show the compatibility with the class N1 and N2 for total optical budget of up to 50 dB. The SCA outperforms the conventional SOA extender in terms of the access budget, which is because higher output power from the extender can be used. A more direct quantitative comparison of the two cannot simply be made here due to the difference in the used transmitters. A comparison of the two amplification schemes has been presented in greater depths in .
4. Multiwavelength SCA reach extension
Saturated SOAs are usually ill-suited for multiwavelength operation since the multiple wavelength signals are expected to be strongly coupled inside the highly nonlinear gain medium. Hence, the noise on channel 1 will affect the signal on channel 2 and vice versa. This would probably be an obstacle in long-haul optical links that host a cascade of amplifiers and many wavelengths. However, in passive optical networks the link rarely contains many other amplifiers than the extender itself and the signal OSNR is expected to remain high. The multiple-wavelength operation would allow additional flexibility and upgrade scenarios. For instance, if the output power of the extender outperforms the locally needed access budget for a closed customer group (e.g. 128 customers to be served by a given TDMA protocol), extra wavelengths can later be added without traffic interruption to serve further parallel PON trees while sharing the same feeder line infrastructure.
In order to test the suitability of the SCA scheme in WDM environment we experimented with TDM/DWDM PON downstream system (see Fig. 7 ). The OLT consists of DPSK transmitters which offer the wavelengths from 1 to N for downstream each to its own TDM PON tree. In the remote node, all DPSK signals are demodulated and amplified simultaneously by the PON extender. The DWDM signals are then demultiplexed by an AWG demux and the wavelengths are sent to respective individual TDM PON trees. The optional input of the PON extender can be used for protection.
The measurement setup is shown in Fig. 8 . The transmitter (Tx) generates a DPSK with RZ carving (33% duty cycle) at 10 Gbit/s (PRBS of 231−1) by using two Mach-Zehnder modulators, one for phase coding using push-pull setup and other for the pulse carving. As only one DPSK modulator is available, all considered wavelengths are modulated by the same transmitter and the signals are de-synchronized with a demux-mux pair that are connected with varying fiber lengths. The signals are boosted by an EDFA so that each wavelength has about 0 dBm power at the input of the feeder line. The used AWG mux/demux have a 100-GHz channel spacing and a 3-dB bandwidth of 55 GHz.
The ONU receiver consists of an avalanche photodiode and an electrical clock recovery. In back-to-back operation, the receiver has a sensitivity of −30 dBm for BER of 10−9 and −35 dBm for BER of 10−3. The feeder budget is defined as the optical budget between the OLT and the remote node, while the access budget corresponds to the optical budget between the PON extender and the ONU.
We first considered a suburban scenario with 20 km feeder and 25 km access fibers with two-wavelength operation (1552.5 nm and 1557.4 nm). The SOA current is optimized at 350 mA with an output power of 10 dBm. A complete BER map was measured for 1557.4-nm signal (see Fig. 9a ). The other wavelength had about the same performance. We can see that the feeder budgets of the 10G-PON E1 and E2 classes (access budget of 33 dB and 35 dB) for an error free transmission (BER<10−9) are extended by 15 dB and 11 dB, respectively. This corresponds to maximum total budget of 48 dB. If we consider the use of forward error correction code, the BER can be degraded to 10−3. In that case, with feeder budget of 15 dB, the access budget can be extended to 40 dB. This access budget allows a transmission over an AWG demux (4-dB loss), 25 km of fiber (5-dB loss), 1:256 splitting ratio (24-dB loss), two filters/circulators (one at extender output, and other at ONU input) to separate up- and downstream signals (2x2 dB) and an extra budget of 3 dB. Consequently, with two wavelengths and two outputs of the SCA, the number of served clients can be up to 1024.
Four-wavelength operation was tested at grid of 200 GHz: 1552.5 nm, 1554.1 nm, 1555.8 nm, and 1557.4 nm. The SOA current is maintained at 350 mA, the output power of each wavelength is thus decreased by 3 dB compared to the two-wavelength case. The BER performances of the two- and four-wavelength operations are compared to each other. Figure 9b shows BER curves at feeder budget of 10 dB, the DLI input powers are about −7 dBm and −4 dBm for the two- and four-wavelength cases, respectively. As can be seen, the two curves overlap, implying minimal power penalty due to additional wavelengths. Figure 9c depicts the BER contours of 10−3 for two- and four-wavelength configurations. The access budget decreases by 3 dB when the number of wavelengths is doubled. Consequently, the number of clients is not increased with the number of wavelengths, but with the same number of clients, the splitting ratio for each wavelength can be reduced and the data rate per client therefore increases.
A long reach rural scenario is also considered, in which the SCA extender is placed at 75 km distance from the OLT and 25 km distance from the ONUs. A preamplifier and a dispersion compensating fiber (DCF) are put prior to the delay interferometer (Fig. 10a ). As the feeder loss is considerably increased, the preamplifier is needed in order to keep the SOA of the SCA in the saturation regime. The preamplifier SOA is identical to the one used in the SCA. The preamplifier SOA injection current is 92 mA. The DCF, between preamplifier and DLI, compensates 65-km SMF transmission. Otherwise the setup was similar to the previous one. Four-wavelength operation is measured.
Figure 10b shows simultaneously the BER contours of 10−3 for four wavelengths in 100-km configuration. We obtained approximately the same performance for all wavelengths. In comparison to the previous scenario, the feeder budget is tremendously increased thanks to the use of the preamplifier. The access budget is however decreased due to amplified spontaneous emission noise (from the preamplifier) and the residual chromatic dispersion. The BER curves at a feeder budget of 28 dB are depicted in Fig. 10c. At BER of 10−3, the worst case access budget is 30 dB. The total optical budget is 58 dB in this long reach scenario.
We have demonstrated the operation of the SCA extender in downstream direction for single wavelength XGPON1 and for two- and four-wavelength TDM/DWDM PON. The single-wavelength setup utilizing commercial XGPON1 prototype with bidirectional traffic and SOA upstream and SCA downstream extenders achieved total optical budget of 50 dB that is compatible with class N1 and N2 of the XGPON1 standard. The two- and four-wavelength configurations were tested and compared in 45 km suburban scenario, using a setup to measure BER as a function of feeder and access power budgets. The SCA extender was found to perform equally well in both cases. In the two-wavelength case, an optical access budget of 40 dB is obtained, which is sufficient to support 25-km transmission and 1:256 splitting ratio, for both wavelengths. Should both outputs of the SCA be utilized, one amplifier could thus support 1024 clients. When more wavelengths are added, minimal power penalty was observed, allowing flexible network upgrade without traffic interruption and infrastructure changes. A long reach rural scenario was also considered with 100-km transmission and four-wavelength operation. The total optical power budget is 58 dB. Similar to the bidirectional XGPON1 test, full-duplex transmission can be performed by using an SOA for power budget extension of the upstream also in these multiwavelength cases.
The work done by Orange Labs was financially supported by French National ANR-TRILOB project.
References and links
1. R. Davey, J. Kani, F. Bourgart, and K. McCammon, “Options for future optical access networks,” IEEE Commun. Mag. 44(10), 50–56 (2006). [CrossRef]
2. ITU-T G987.2, “10-Gigabit-capable passive optical networks (XG-PON): Physical media dependent (PMD) layer specification” (2010).
3. P. Chanclou, J.-P. Lanquetin, S. Durel, F. Saliou, B. Landousies, N. Genay, and Z. Belfqih, “Investigation into optical technologies for access evolution,” in Optical Fiber Communication Conference, OSA Technical Digest (CD) (Optical Society of America, 2009), paper OWH1.
4. Z. Belfqih, P. Chanclou, F. Saliou, N. Genay, and B. Landousies, “Enhanced optical budget system performance of an burst extended PON at 10.7Gbit/s over 60km of fibre,” in 34th European Conference on Optical Communication, 2008. ECOC'08, paper Th.2.F.4 (2008).
5. ITU-T G984.6, “Gigabit-capable passive optical networks (GPON): Reach extension” (2008).
6. N. Genay, T. Soret, P. Chanclou, B. Landousies, L. Guillo, and F. Saliou, “Evaluation of the Budget Extension of a GPON by EDFA Amplification,” in 9th International Conference on Transparent Optical Networks, 2007. ICTON '07, paper Mo.P.20 (2007).
7. L. Spiekman, D. Piehler, P. Iannone, K. Reichmann, and L. Han-Hyub, “Semiconductor optical amplifier for FTTx,” in International Conference on Transparent Optical Networks, 2007. ICTON '07. 9th, Rome, Italy, Mo.D2.4 (2007).
8. D. Nesset, S. Appathurai, R. Davey, and T. Kelly, “Extended reach GPON using high gain semiconductor optical amplifiers,” in National Fiber Optic Engineers Conference, OSA Technical Digest (CD) (Optical Society of America, 2008), paper JWA107.
9. F. Saliou, P. Chanclou, N. Genay, J. A. Lazaro, F. Bonada, A. Othmani, and Y. Zhou, “Single SOA to extend simultaneously the optical budget of coexisting GPON and 10G-PON,” in 36th European Conference on Optical Communication, 2010. ECOC '10, paper Tu.5.B.5 (2010).
10. D. Nesset and P. Wright, “Raman extended GPON using 1240 nm semiconductor quantum-dot lasers,” in Optical Fiber Communication Conference, OSA Technical Digest (CD) (Optical Society of America, 2010), paper OThW6.
11. B. Zhu, “Entirely passive reach extended GPON using Raman amplification,” Opt. Express 18(22), 23428–23434 (2010), http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-18-22-23428. [CrossRef] [PubMed]
12. H. Rohde, S. Smolorz, E. Gottwald, and K. Kloppe, “Next generation optical access: 1 Gbit/s for everyone,” in 35th European Conference on Optical Communication, 2009. ECOC '09, paper 10.5.5. (2009).
13. A. Tervonen, M. Mattila, W. Weiershausen, T. von Lerber, E. Parsons, H. Chaouch, A. Marculescu, J. Leuthold, and F. Kueppers, “Dual output SOA based amplifier for PON extenders,” in 36th European Conference on Optical Communication, 2010. ECOC '10, paper P6.18 (2010).
14. H. Chaouch, E. Parsons, A. Tervonen, M. Mattila, W. Weiershausen, T. von Lerber, S. Honkanen, J.-Y. Yang, A. E. Willner, and F. Küppers, “All-optical processing of RZ-DPSK signals using counter-propagating pulses in a saturated SOA,” Opt. Commun. 284(10-11), 2576–2580 (2011). [CrossRef]
15. R. Maher, L. P. Barry, and P. M. Anandarajah, “Cost efficient directly modulated DPSK downstream transmitter and colourless upstream remodulation for full-duplex WDM-PONs,” in National Fiber Optic Engineers Conference, OSA Technical Digest (CD) (Optical Society of America, 2010), paper JThA29.
16. R. Vodhanel, A. Elrefaie, M. Iqbal, R. Wagner, J. Gimlett, and S. Tsuji, “Performance of directly modulated DFB lasers in 10-Gb/s ASK, FSK, and DPSK lightwave systems,” J. Lightwave Technol. 8(9), 1379–1386 (1990). [CrossRef]
17. W. Jia, J. Xu, Z. Liu, K.-H. Tse, and C.-K. Chan, “Generation and transmission of 10-Gb/s RZ-DPSK signals using a directly modulated chirp-managed laser,” IEEE Photon. Technol. Lett. 23(3), 173–175 (2011). [CrossRef]