Open Access
Add to BibSonomyAbstract
We demonstrate a real-time 25-Gb/s PON prototype with ethernet-PON MAC/PHY, O-band transmitter, and cost-effective APD receivers. With applying parasitic inductance and capacitance reduction, the frequency response of 25-Gb/s APD ROSA with TO46-pacakge is improved to support high receiver sensitivity around −25 dBm at the BER of 10−3. The 30 dB power budget of 25 Gb/s downstream is achieved at the BER of 10−3. With long-term ethernet packet transmission, 25 Gigabit and 10 Gigabit ethernet traffics are successfully transmitted through the 20-km SMF over 14 hour’s observation window. Furthermore, QoS and bandwidth re-assignment function of the 25-Gb/s PON prototype are successfully demonstrated with respect to residential, business and mobile backhaul services in ONUs.
© 2016 Optical Society of America
1. Introduction
Data traffic in optical access network is rapidly increasing due to various types of smart mobile devices and services. Increasing data traffics require a continuous expansion in access network capacity, and passive optical network (PON) plays a critical role for accommodating residential and business service as well as mobile fronthaul/backhaul. The demand is reaching 1 Gb/s for residential service and multi-Gb/s speeds for business service. The throughput for IMT2020 (5G) networks requires 20 Gb/s per cell [1]. Recently, there have been substantial efforts to find a practical and low-cost single wavelength solution upgrading 10 Gb/s to 25 Gb/s per wavelength. By combining wavelength division multiplexing (WDM) with 25-Gb/s time division multiplexing (TDM)-passive optical network (PON), the capacity could be easily expanded to 100 Gb/s which was targeted in 100G-EPON in IEEE 802.3ca [2,3]. However, previous works on high-speed PON were mostly focused on the investigation of technical feasibility in 25-Gb/s modulation format for downstream transmission in optical layer to overcome chromatic dispersion limitation in C-band. These include optical duo-binary, electrical duo-binary, and pulse-amplitude modulation (PAM)-4 [4–6]. However, there was few report on 25-Gb/s PON with MAC and cost-effective optical modules.
In this paper, we demonstrate a real-time 25-Gb/s PON prototype with ethernet-PON MAC/PHY, O-band transmitter, and cost-effective APD receivers. 25 Gigabit and 10 Gigabit ethernet traffics are successfully transmitted through the 20-km SMF over 14 hour’s observation window. QoS and bandwidth re-assignment function of the 25-Gb/s PON prototype are successfully demonstrated.
2. Real-time demonstration of QoS guaranteed 25-Gb/s PON prototype
Figure 1 illustrates the configuration of future optical access network based on PON. A typical PON uses a single fiber from the central office to the splitter, and employs a single wavelength to communicate with ONUs. Most PON technologies utilize TDM access to share the optical distributed network (ODN) among its users. A current PON such as 10G-EPON provides maximum 10-Gb/s bandwidth in down and upstream signals. Considering growing number of subscribers and increasing bandwidth demands, the future optical access network would be migrated from TDM to WDM/TDM, and there have been substantial efforts to study WDM/TDM hybrid access network [7–9]. To implement future optical access networks cost effectively, there would be several issues to be resolved. First of all, operating speed of each wavelength would be 25 Gb/s, and the future optical access network based on PON should be upgraded by utilizing already deployed ODN. Thus, power budget equivalent to the legacy TDM-PON and chromatic dispersion limitation due to high speed optical signals should be resolved in optical layer. In addition, the future optical access network would accommodate residential services and business service as well as mobile fronthaul/backhaul services through the same ODN. Thus, QoS guaranteed MAC and flexible bandwidth management should be resolved in the MAC layer to copy with the data traffic variation caused by the time of day or service types.
For demonstration of future optical access networks, we implemented a real-time and QoS guaranteed 25-Gb/s PON prototype, as shown in Fig. 2(a). The 25-Gb/s PON MAC/PHYprototype is implemented by extending IEEE 10G-EPON standard. The MAC/PHY board is composed of a single FPGA and ten 10GbE UNI ports, which could be expandable up to four 25-Gb/s PON MAC/PHY. The implemented MAC/PHY board shown in Fig. 2(b) is utilized to emulate OLT/ONU functions. To accommodate 25-Gb/s downstream and 10-Gb/s upstream in legacy ODN (PR30, > 29-dB power budget) cost effectively, we developed APD-ROSAs for the PON prototype. The same TO-46 package is used for the 25-Gb/s and 10-Gb/s burst mode APD-ROSA, as shown in Fig. 2(c). To meet the power budget of PR30 and to improve OE bandwidth, low-noise and low parasitic inductance designs are utilized in the APD-ROSAs. For a low-cost single wavelength solution operating at 25 Gb/s, we use an externally modulated laser (EML) based O-band downstream transmitter since it has an advantage of compact optical transceivers for OLT and ONU. The output of 25-Gb/s signal from OLT is connected to EML nominally operating at 1309.08 nm, and it is transmitted through 5-km of single mode fiber as a feeder. The output power and extinction ratio of downstream signal are measured to be + 5 dBm and 7.2 dB, respectively. To test real deployment environment of optical access networks, we intentionally set different transmission distances from OLT to ONU1 and ONU2 as 20 km and 6 km, respectively. In addition, we assume a service scenario that ONU1 supports high bandwidth service for business or mobile fronthaul/backhaul, while ONU2 supports low bandwidth residential service. At the ONUs, the downstream signals are detected with the 25-Gb/s APD based receivers. For the upstream transmission, 10-Gb/s burst mode directly modulated lasers (DMLs) operating at 1270 nm are utilized, which is compliant with IEEE 10G-EPON standard. At the OLT, 10-Gb/s BM APD is used to detect upstream burst packets from the ONU1 and the ONU2.
Fig. 2 Real-time demonstration of QoS guaranteed 25-Gb/s PON prototype (a) experimental setup (b) Ethernet PON MAC/PHY board, (c) 10-Gb/s burst mode APD-ROSA and 25-Gb/s APD-ROSA.
For a real-time 25-Gb/s PON demonstration, PON protocol for ONU registration and bandwidth allocation was implemented by extending IEEE 10G-EPON standard. Thus, when ONUs were connected ODN, each ONU was automatically registered to OLT through handshaking of multi-point control protocol (MPCP) messages, and then granted bandwidth was assigned to each ONU. A packet generator & analyzer was used to investigate the performance of ethernet packet transmission and multiple QoS function in the implemented 25-Gb/s PON prototype. We measure the packet loss rate (PLR) by using 128-bytes ethernet packet. In addition, bit error rate (BER) of 25-Gb/s downstream was measured by implementing PRBS generator and checker in the FPGA to evaluate the performances of optical layer and MAC/PHY layer simultaneously. The implemented PON prototype was operated at 25 Gb/s basis. By adding more optical modules and MAC lanes, the capacity of the implemented PON prototype could be upgraded with WDM technology.
3. Results and discussions
Figure 3(a) illustrates a configuration of 25-Gb/s APD-ROSA including TO-stem and flexible PCB (FPCB). For a cost effective 25-Gb/s APD-ROSA, we used a TO-46 package which could be available off the shelf component. To ensure high frequency operation, it is important to compensate parasitic inductance and capacitance on signal path which will cause electrical reflection due to impedance mismatch. For the bandwidth enhancement for 25-Gb/s operation, we simulated signal path optimization. Thus, the optimization of chip-to-package and package-to-FPCB connections were carried out by compensating parasitic inductance with intentional capacitance at each signal connection. Figure 3(b) shows the frequency response of the simulated electrical reflection (S11) and transmission (S21) at the interface between TO-stem and FPCB. The corresponding 25-Gb/s simulated electrical waveforms are shown in the insets of Fig. 3(b). These simulations are conducted at the protruded extra-lead pin length of 0.4 mm and the bond-wire length of 650 μm. The results show that the electrical return loss (S11) was reduced 12 dB at 17 GHz with parasitic LC reduction. With the help of reduced return loss (S11) characteristics, timing jitter in the 25-Gb/s eye-diagram was also decreased.
Fig. 3 25-Gb/s APD-ROSA with TO-46 package (a) configuration (b) frequency response and simulated 25-Gb/s eye-diagrams.
Figures 4(a) and 4(b) represent the measured PLR performances of 25-Gb/s downstream and 10-Gb/s upstream when downstream and upstream channels are operating through the system simultaneously. For all cases, the PLR curves are well behaved with no indication of power penalties due to high chromatic dispersion tolerance in O-band transmission. The open symbols in Fig. 4(a) show the measured BER curves for the downstream signal. Since we used packet length of 128-byte, there should be three orders difference between BER and PLR at the same received power. The measured BER and PLR curves are almost overlapped, which indicates that there is no performance degradation in MAC/PHY layer and the system performance is mainly limited by optical layer. The receiver sensitivity of 25-Gb/s downstream is measured to be −25 dBm at BER of 10−3. Thus, the power budget of downstream is 30 dB at BER 10−3 which is corresponding to BER of 10−12 with applying forward error correction (FEC). The PON prototype supports 20 km transmission (8 dB, 0.4 dB/km) and 64 split (21 dB, 3.5 dB/two split).
Fig. 4 Performance of downstream and upstream in the 25-Gb/s PON prototype (a) PLR and BER of 25-Gb/s downstream (b) PLR of 10 Gb/s- burst-mode upstream when LSR is 14 dB.
The effects of loud soft ratio (LSR) of burst packets from ONU1 (soft packet case) and ONU2 (loud packet case) on the PLR was evaluated, as shown in Fig. 5. Since the territorial locations of each ONU are different, an input power to OLT BM receiver was varied as the ONU’s location. For the evaluation of different ONU location, LSR of burst packets, we intentionally set the different power level between OLT and ONU1 by using an optical attenuator, as shown in Fig. 5(a). Figure 5(b) shows the waveforms of optical input signals and electrical output signals of the 10 Gb/s burst mode receiver. The optical input signals were successfully converted into the same output voltage level without distortion even at 22 dB of LSR. Figure 5(c) indicates a measured PLR of upstream signal from ONU1 as a function of LSR. The measured power penalty was maintained less than 0.2 dB for LSR up to 22 dB, which is higher than the requirement of IEEE 10G-EPON standard (i.e. 15 dB).
Fig. 5 Performance of upstream in the 25-Gb/s PON prototype (a) measurement setup for effect of LSR on PLR performance in burst-mode upstream signal (b) waveforms of optical input and electrical output: 10 dB of LSR case (i) and 22 dB of LSR case (ii) (c) the effect of LSR on PLR performance in burst-mode upstream signal.
To evaluate long-term system reliability, we set the received powers of APD-ROSAs to achieve downstream PLR of 10−9 and upstream PLR of 10−11, respectively. The measured PLR was maintained constantly over 14 hour’s observation window as shown in Fig. 6. Consequently, 25 Gigabit and 10 Gigabit ethernet traffic were successfully transmitted over the 20-km SMF.
We implemented QoS guaranteed multiple-class queues in the ONU to investigate the accommodation of residential, business, and mobile fronthaul/backhaul services in the PON prototype. The queues were composed of high-priority output (HPO) and low-priority output (LPO). To verify the QoS function of the ONUs, we assumed two scenarios as shown in Table 1. One was intra-ONU priority scenario that HPO guarantees throughput in the case of traffic congestion. The other was inter-ONU priority scenario that ONU1 for business or mobile services was assigned higher bandwidth than ONU2 for residential service. Even though two classes of priority were demonstrated in the prototype for the performance evaluation, the more QoS classes could be easily implemented by adding additional classees of priorities.
Table 1. QoS Test Scenarios
Figures 7(a) and 7(b) show the measured throughputs of ONU1 and ONU2 in scenario I. We increased the output traffic of packet generator which connected to ONU1 and ONU2 from 1 Gb/s to 10 Gb/s according the requested bandwidth, and then measured throughput and packet drop. When the requested bandwidth was less than the granted bandwidth, there was no packet drop. On other hand, when the requested bandwidth was exceeding the granted bandwidth, the throughput was saturated at the granted bandwidth for both ONU1 and ONU2. As a result, LPO packet was dropped first under traffic congestion, whereas HPO packet was transmitted until the capacity of HPO packet reached at the granted bandwidth. In addition, the throughput of ONU1 was higher than that of ONU2 since a higher bandwidth was assigned for ONU1 which assumed to serve business or mobile services.
Daily data traffic would be varied based on the time of day and types of supporting services at ONU. We assumed a bandwidth allocation scenario (scenario II) that business and mobile service will require more bandwidth in daytime than the bandwidth in nighttime. Figure 8(a) illustrates service scenario II where ONU1 and ONU2 were used for business/mobile services and residential services, respectively. The requested bandwidths for ONU1 and ONU2 were set to 5.5 Gb/s and 3.5 Gb/s in time1 (nighttime), respectively. In daytime, the requested bandwidth of ONU1 was increased from 5.5 Gb/s to 8.0 Gb/s, whereas the bandwidth of ONU2 was decreased to from 3.5 Gb/s to 1.0 Gb/s. Thus, by re-allocating more bandwidth to ONU1 at time2 (daytime), the business or mobile service in ONU1 was guaranteed while the residential service in ONU2 was restricted. Figure 8(b) shows measured real-time throughputs of ONU1 and ONU2 as a function of time. The measured throughputs were successfully varied according to the bandwidth re-assignment. All these results indicated that the implemented PON prototype provides QoS function with respect to various services in ONUs.
4. Summary
We successfully demonstrated the first real-time 25-Gb/s PON prototype with ethernet-PON MAC/PHY, O-band transmitter, and cost-effective APD receivers. With applying parasitic inductance and capacitance reduction, the frequency response of 25-Gb/s APD ROSA with TO46-pacakge was improved to support high receiver sensitivity around −25 dBm at the BER of 10−3. Without using an optical amplifier, the measured power budget of 25-Gb/s downstream was 30 dB at BER 10−3. 25 Gigabit and 10 Gigabit ethernet traffics were successfully transmitted through the 20-km SMF over 14 hour’s observation window. The results of performance evaluation with QoS guaranteed ONUs supporting residential, business and mobile backhaul services under various traffic conditions indicated that the 25-Gb/s PON prototype provided QoS and bandwidth re-allocation functions. We expect that the total capacity of PON prototype could be expanded from 25 Gb/s to 100 Gb/s by using WDM technology.
Acknowledgment
This work was supported by ICT R&D program of MSIP/IITP [B0132-16-1004, SDN based wired and wireless converged optical access networking].
References and links
1 . 5G Vision and requirements, IMT2020 (5G) Promotion Group, 2014.
2. IEEE P802.3ca 100G-EPON Task Force, “Physical layer specifications and management parameters for 25 Gb/s, 50 Gb/s, and 100 Gb/s passive optical networks,” (http://www.ieee802.org/3/ca/).
3. C. Knittle, “IEEE 100 Gb/s EPON,” in OFC 2016 (2016), paper Th1I.6.
4. Z. Ye, S. Li, N. Cheng, and X. Liu, “Demonstration of high-performance cost-effective 100-Gb/s TWDM-PON using 4x25-Gb/s optical duobinary channels with 16-GHz APD and receiver-side post-equalization,” in ECOC 2015 (2015), paper Mo.3.4.4.
5. V. Houtsma and D. van Veen, “Demonstration of 25 Gb/s TDM-PON with 31.5 dB optical power budget using only 10 Gb/s optical components,” in ECOC 2015 (2015), paper PD.4.3.
6. J. Gao, “Demonstration of the first 29dB power budget of 25-Gb/s 4-PAM system without optical amplifier for next generation access network,” in OFC 2016 (2016), paper Th1I.2.
7. S.-G. Mun, E.-G. Lee, J. H. Lee, H. Park, S.-K. Kang, H. H. Lee, K. Kim, K.-H. Doo, H. Lee, H. S. Chung, J. H. Lee, S. Lee, and J. C. Lee, “Demonstration of time- and wavelength-division multiplexed passive optical network based on VCSEL Array,” ETRI J. 38(1), 9–17 (2016). [CrossRef]
8. S. Kaneko, T. Yoshida, S. Furusawa, M. Sarashina, H. Tamai, A. Suzuki, T. Mukojima, S. Kimura, and N. Yoshimoto, “Demonstration of load-balancing operation based on hitless dynamic wavelength allocation on symmetric 40-Gbit/s λ-tunable WDM/TDM-PON,” J. Lightwave Technol. 33(3), 645–652 (2015). [CrossRef]
9. D. Nesset, “NG-PON2 technology and standards,” J. Lightwave Technol. 33(5), 1136–1143 (2015). [CrossRef]
