BIPON: A More Energy-efficient TDM PON
The bit interleaving passive optical network (BIPON) features a new protocol that cuts energy consumption in time division multiplexing (TDM) PON optical network units (ONUs) by a factor of 10 compared to today’s protocols.
Innovating to save energy
As optical networks accelerate from 2.5 Gb/s to 10 Gb/s, and even 40 Gb/s in the future, the energy requirements of the optical network units (ONU) at each customer’s premise will also quadruple. As part of the GreenTouch™ initiative focused on improving the energy efficiency of Information and Communications Technology (ICT) networks, Alcatel-Lucent Bell Labs researchers invented the bit interleaving protocol for time division multiplexing (TDM) passive optical networks (PONs), the most widely used fiber access technology.
GreenTouch is a consortium of leading ICT industry, academic and non-governmental research experts dedicated to fundamentally transforming communications and data networks, including the Internet, and significantly reducing the carbon footprint of ICT devices, platforms and networks. By 2015, its goal is to deliver the architecture, specifications and roadmap --and demonstrate key components -- needed to increase network energy efficiency by a factor of 1000 from current levels.
In today’s broadcast-and-select based TDM PON system, every ONU on the PON must perform intensive processing on all of the data that is broadcast from the optical line terminal (OLT) in the central office to select the data relevant for the respective ONU. When this energy-intensive processing is complete, 97% to 99% of the original bits are dropped because they are intended for other ONUs on the PON. With the bit interleaving protocol, every ONU still receives every bit that is broadcast on the PON. But it can immediately determine which bits are intended for other units and drop them before they undergo heavy processing. Bell Labs testing confirms the bit interleaving protocol reduces power consumption in the ONU by a factor of 10 — from about 2 W to less than 200 mW — compared to today’s conventional TDM protocols as specified in the standards. Figure 1 illustrates a bit interleaving passive optical network, or BIPON. A BIPON runs the bit interleaving protocol in the downstream direction from the OLT in the central office to the ONUs in the customer premises. New ASIC chipsets that support the protocol are built into the OLT and the ONUs.
Simplified circuitry, more efficient data flow
The bit interleaving protocol reduces energy consumption in the ONU by reducing:
- Clock speed requirements
- Data processing requirements
- Voltage requirements
- Memory requirements
Figure 2 shows the downstream data flow from the OLT in the central office through each ONU for XGPON, which delivers 10 Gb/s throughput.
- The orange boxes on the left illustrate each stage that all of the data broadcast from the OLT must pass through before the irrelevant bits are dropped by the ONU.
- The green boxes on the right illustrate the stages that the remaining 1% to 3% of the data undergoes before it is put onto the customer’s LAN.
It’s easy to see by the number of orange boxes in Figure 2 that most of the processing effort is expended on the total amount of data for all ONUs on the PON. Processing requirements are significantly reduced after the irrelevant bits are discarded — the green boxes. Because the data is coming into the ONU at the XGPON rate of 10 Gb/s, the clock and data recovery stage and the deserialization stage shown on the left side of Figure 2 operate at very high clock speeds. The higher the clock speed, the more energy the ONU consumes. The stages after deserialization operate at lower clock speeds. However, energy consumption is still high due to the massive parallel processing that must be applied to the data after it is deserialized. In particular, the forward error correction (FEC) decoder and payload frame receiver represent a large part of the electronics required to process today’s protocols. In contrast, Figure 3 shows the downstream data flow from the OLT through the ONU when the bit interleaving protocol is used. Now there are only 2 orange stages as opposed to 8.
With the bit interleaving protocol, only the data intended for the ONU is sampled (also known as decimation). That means the high clock speed required to handle the incoming PON line rate is needed only for a very short period of time during clock recovery and the decimation process. It also means there is no longer a need for massive parallel processing because further processing is applied only to target user data which is typically only a small fraction of the total data available on the line. With much less data to process, the processing stages — represented by the green boxes — can now operate at the clock speed matched to the customer’s rate, typically 10 Mb/s or 100 Mb/s, which is much lower than the line rate of 10 Gb/s. Lower clock speeds and the elimination of parallel processing requirements reduce energy consumption in the ONU. Lower clock speed also allows for more aggressive voltage scaling circuit design techniques. Hence, additional energy savings can be achieved. The end result is that energy consumption in the ONU is now proportional to the bit rate the customer is consuming, rather than the full incoming PON line rate. As data flowing in processing blocks match the outgoing rate, buffer memory requirement is significantly reduced or eliminated. This further lowers power consumption — because it means that data is processed “just in time.” It does not need to be buffered to compensate for the “burstiness.”
BIPON frame structure
The bit interleaving protocol maintains the 125 μs frame time used in the GPON and XGPON standards for ease of comparison. However, it interleaves the data in the frame structure to allow early selection of relevant bits at the customer’s data rate and clock speed. Figure 4 illustrates the bit interleaving protocol frame structure.
Like all packet frames, the bit interleaving protocol frame structure includes a header and a payload section. In this case, the header section contains a synchronization code word and a unique identifier for each ONU on the PON. This identifier allows the ONU to read only the header information in its own “lane.” The header bits further contain information about the payload in the frame, more specifically, upstream and downstream bandwidth allocation. This information effectively tells the ONU where its payload bits are — at what offset and bit rate they are being sent in the payload section. The header section also contains an optional operations, administration and maintenance (OAM) message field. After the ONU has complete information about the location of the bits it is supposed to receive, it can adjust its sampling hardware (also known as decimator) to take only its own bits off of the line for further processing. It does not touch any bits intended for other ONUs on the PON.
Key innovations enable the bit interleaving protocol
A number of innovations enable the efficiencies delivered by the bit interleaving protocol. These innovations incorporate contributions from IMEC Ghent University in Belgium, a GreenTouch partner. They include:
- Developing a new algorithm for synchronization and descrambling
- Eliminating the deserializer
- Introducing the decimator
- Maintaining statistical multiplexing efficiency
Developing a new algorithm for synchronization and descrambling The bit interleaving protocol includes a new synchronization and descrambling algorithm that can be executed on an individual bitstream basis. Adapting the synchronization pattern to an interleaved format is necessary to completely eliminate the deserializer, a major power consumer in today’s ONUs. In the bit interleaving protocol, the synchronization pattern is implemented in 2 parts — a constant part and a part that is specific to each ONU. The protocol then runs a 2-step algorithm which synchronizes the ONU to its own “lane” of bits. Scrambling ensures there are sufficient transitions between the zeros and ones in the data being transmitted to maintain clock synchronization. In today’s TDM PON protocols, scrambling and descrambling are applied at the high PON line rate on consecutive bits — bits that are not necessarily intended for the same customer. However, because the bit interleaving protocol decimates irrelevant data before the scrambling process, a new algorithm is required. In the bit interleaving protocol, descrambling is applied to a sequence of non-consecutive bits that are sampled by decimation — at different bit rates. This is because every frame received can be at a different rate and a different offset. Eliminating the deserializer Eliminating the need for a power-hungry deserializer in the ONU is central to the energy savings in the bit interleaving protocol. In the TDM PON protocols used today, the deserializer aligns the incoming serial bits into a parallel arrangement so they can undergo parallel processing at a lower clock speed. That means it must run a very high clock speed to match the bit rate on the incoming PON line. The deserializing process consumes a lot of energy. Even the interface between the deserializer and the media access control (MAC) section of the ONU consumes significant energy. Because the bit interleaving protocol applies decimation before all of the bits are captured, deserialization is no longer required. Introducing the decimator The decimator is a flexible data sampler, which is much simpler than a typical deserializer. It allows for selection of the relevant bits in a dynamic way on a frame-to-frame basis. It also incorporates a very energy efficient clock and data recovery mechanism that was developed by Bell Labs in partnership with IMEC Ghent University in Belgium. Maintaining statistical multiplexing efficiency It is well known that simple circuit-based TDM interleaving is more energy efficient than packet-based communications. But a static circuit-based TDM is less efficient in statistical multiplexing than the packet-based approach. The bit interleaving protocol includes a dynamic bandwidth allocation mechanism that allows for changes in user bit rate on a frame-by-frame basis. It uses a bandwidth map within the header section to tell the ONU the bit interval and offset for the coming payload section. Because the bandwidth map can be optionally carried in each frame, bandwidth can be flexibly allocated to maintain the statistical multiplexing efficiencies delivered by packet-based technologies. Dynamic bandwidth allocation is important because different customers on the PON will have different requirements and traffic load variations. For example, businesses may be subscribed to the maximum bit rate but consume that maximum for only very short periods of time. With dynamic bandwidth allocation, service providers can give high-end customers larger bandwidth when they need it, but reallocating the bandwidth to other low-end customers at another instance when demand is low.
Toward more energy efficient networks
Reducing energy consumption in ONUs is an important step toward reducing the overall energy consumption of communications networks. We’ve already made progress. Existing standards for 10 Gb/s XGPON define sleep modes that reduce ONU power consumption by a typical factor of 3. However, as Bell Labs’ research into the bit interleaving protocol confirms, there is more that can be done to reduce energy consumption in ONUs. The Alcatel-Lucent BIPON design can combine the bit interleaving protocol with sleep mode, which leads to further energy savings. In addition, because most of the circuits in a BIPON scale with the customer’s data rate, savings from rate-adaptive voltage scaling on a session-to-session basis may be possible in the future. As optical networks evolve to 40 Gb/s, energy-saving innovations such as the bit interleaving protocol will become even more crucial for sustainability. As part of our commitment to GreenTouch and increasing energy efficiency in networks, Alcatel-Lucent will continue to collaborate and innovate with service providers, other vendors, industry organizations, research institutions and universities to develop eco-sustainable network solutions. To contact the authors or request additional information, please send an e-mail to email@example.com.
- Energy consumption in the upstream direction from the ONU to the OLT is already proportional to the customer’s LAN rate so optimization is not needed.