1.

INTRODUCTION

As the popularity of online services grows, the demands for high-speed transmission rates in access networks increase year by year.¹ During the COVID pandemic, stable-connection and low-latency networks were used more often.² Passive optical network (PON) recommendations are published for two sectors by the Institute of Electrical and Electronics Engineers (IEEE) and the International Telecommunication Union (ITU). The current recommendations focus on 10 Gbit to 100 Gbit.³ 100 G is obtained by four different wavelengths with 25 Gbit.⁴ Research papers have addressed the physical layer of PON recommendations, such as the modulation format, line code, and reach of the system. We present a complex solution for real-time control data processing of a 10 Gigabit-capable symmetrical passive optical network (XG(S)-PON). The transmission convergence layer of the PON has been discussed according to implementation, but our purpose is to implement the frame structures and apply them to our self-developed field programmable gate array (FPGA).⁵

2.

MEASUREMENT SCHEME

The paper proposes an XG-PON analysis system consisting of a transmitting frame interceptor and a software analyzer. A schematic diagram is shown in Figure 1a. The interceptor is highlighted in green, where the XG-PON frames are eavesdropped using a 2-by-2 3 dB coupler to capture both upstream and downstream directions. Both captured signals are converted from the optical domain to binary using a newly designed FPGA-based network interface card. The rest of the system performs software analysis and is highlighted in yellow. The proposed newly designed FPGA card is shown in Figure 1b. There are 4 SFP+ cages on the motherboard with the possibility of connecting optical network units (ONUs) and optical line termination (OLT) units for the GPON, XG-PON and XG(S)-PON standards, including combinations of them. The downlink and uplink must be connected to the board separately. The individual directions can also be analyzed independently, and in particular, for the downstream direction, it is not necessary to have information about the distribution of time windows sent by the individual ONUs. A peripheral component interconnect express (PCIe) interface for data transfer between the Jetson module and the FPGA and a nonvolatile memory express (NVMe) connection to the Jetson module are used. The Jetson module is wired as a PCIe Gen4 root point for speeds up to 16 Gbps and provides various combinations of controller wiring widths of ×8, ×4, ×2, and ×1. The FPGA PCIe endpoint enables the maximum transfer rate according to the Gen3 specification.

Figure 1:

(a) Schematic diagram of the proposed analyzer deployed in a real XG-PON. (b) Newly designed FPGA-based network interface card used for XG-PON frame capturing.

3.

MEASUREMENT RESULTS

The results of data processing are shown in Table 1. The allocation identifier (Alloc-ID) represents the unique ONU identifier (ONU-ID). The first value of Alloc-ID equals ONU-ID, and ONU-ID is obtained for OLT during the activation process.⁶ The first column contains the unique identifiers of Alloc-ID (14336, 10, 2570, 3082, 14337, 9, 2569, and 3081). Based on these values, it is possible to determine the number of unique ONU-IDs (9 and 10). Values higher than 1024 are reserved if more than a single Alloc-ID is needed for an ONU, for example, if the ONU handles more transmission containers (T-CONT). An ONU can report buffer occupancy if a flag bit is present; if not, the dynamic bandwidth report upstream (DBRu) is not transmitted. The next column identifies the StartTime of upstream transmission (bit expression). The GrantSize field contains a 16-bit number for the total length indication of the XGTC payload with the DBRu field. GrantSize should be 0 during the activation process of the physical layer operation administration and maintenance (PLOAM) grant only. The forced wake-up indication (FWI) field indicates forced ONU wake-up if a bit is present. In our case, FWI is 0, which means that the ONUs are not in low-power mode. Burst Profile indicates the index of a burst profile on the OLT side. The last column represents the hybrid error correction (HEC) field of the frame.

Table 1:

BWmap allocation for separate Alloc-IDs.

4.

CONCLUSION

The proposed system is capable of dealing with real-time data analysis of passive optical networks. The downstream generates data at up to 10 Gbit/s with 125 μs frame durations. The ONUs receive all data due to the point-to-multipoint (P2MP) topology but process only data with matching ONU-IDs. The remaining data are dropped. The final solution is capable of extracting unique fields of the frame (BWmap allocation, control messages, etc.).