By Fang Liming, Wang Xiang & Liu Jianhua
The fourth-generation broadband (4GBB) project set out in 2009 to create an “ultimate DSL technology” for copper, culminating in 2012 with G.fast. And while G.fast is certainly a quantum leap over what came before, Huawei thinks that copper has more left to give, and envisages the development of a fifth generation (5GBB) standard for the medium.
5GBB: Next-gen broadband
As copper technologies have evolved, optical fiber has moved closer to the end user (Figure 1), with 4GBB standards specifying a distance of 20 to 200 meters away.
According to ITU definition, the maximum spectrum for G.fast is 106MHz, and can provide 1Gbps access rates. G.fast is now well recognized in the industry, with vendors now directly or indirectly announcing plans for its commercialization, and targets now being set for the end of 2015.
But then what? History demonstrates that generational leaps in copper technology are achieved every eight years (Figure 2), and work on G.fast began in earnest in 2009. Thus, work on a fifth generation will probably commence sometime in the next two or three years, but what are the challenges involved and are the goals worthy of the effort?
Technical and engineering challenges to 5GBB
Capacity limits
Increased capacity requires increased frequency. To realize a capacity leap on par with that for G.fast, the frequency must increase to 1GHz. However, high-frequency transmission over copper is very lossy, so the lengths will have to be shortened. And what’s more, the 1GHz band overlaps with many other radio frequencies, and that means a lot of interference. Some frequency bands will have to be lowered or even restricted in order to prevent this.
Testing and capacity simulations show that, in the absence of interference, a 30 meter-long loop of category-5 twisted pairs can reach a speed of 12Gbps, with a drop line rate of 10Gbps. At 50 meters, the capacity drops to 5Gbps. However, reality brings background noise, bridge taps, and radio interference into the equation, and these are among the problems that must be resolved before 5GBB over 30m loop line is commercialized.
Installation & maintenance
A shorter length of copper usually means an increase in the number of remote installations.
According to TNO research statistics (Geographic Deployment and Cost Studies for G.fast, 2014 TNO DSL Seminar), there are currently 2,100 fiber-to-the-curb (FTTC) nodes in Amsterdam, covering 360,000 home users, with 192 users supported per node. If G.fast is to be deployed, that the maximum copper length would decrease to 150 meters, as would the node capacity (48 users). Thus, 10,300 fiber-to-the-distribution-point (FTTdp) nodes would be needed, a fivefold increase.
In a 5GBB scenario, with copper lengths shortened to 50 meters or less, the number of nodes balloons to nearly 100,000, a fiftyfold increase and a nightmare in terms of installation and maintenance.
5GBB access architecture
As you see, without a major rethinking of the access architecture, 5GBB just isn’t viable. FTTH, even in the most difficult situations, would still be preferable to 100,000 FTTdp nodes. However, traditional FTTx architecture is dual-level, making remote DSL services complex. Both hardware resources and software resources for management, control, and data processing are needed, with centralization and simplification of the remote gear a must.
Centralized access networks
Centralized network architecture would move remote device functions to the central office, leaving certain interfaces connecting the remaining remote modules with the centralized modules so that network functionality remains consistent. In terms of precedent, hybrid fiber coaxial (HFC) network architecture enables all modem functions to be centralized in the hub equipment room, with remote devices basically functioning as amplifiers.
In theory, remote DSL devices can utilize multiple interface categories. In addition to the commonly used Layer-2 interfaces, interface allocation is also an option. Different interfaces have different characteristics. Analog interfaces ("Layer-0 interfaces") allow for the simplest and most reliable remote devices. Hybrid and analog circuits provide better forward compatibility.
5GBB network architecture
"Layer-0" interfaces can simplify remote devices to the greatest extent, transforming them into protocol-free analog circuits, while hardware simplification reduces the size, power consumption, and cost of devices, making large-scale deployment much easier. Hardware simplification also minimizes device faults, thus increasing device reliability and cutting maintenance costs.
If the maximum analog capability of a remote device is 1GHz, then regardless of the DSL standards used by the central office, as long as the analog bandwidth does not exceed this number, there is no need to upgrade remote devices. Carriers can simply upgrade the DSL line cards in the central office, saving a lot of the network upgrade costs.
If all DSL line cards are in the central office, signals can be processed using pooled resources. On the one hand, flexible DSL resource allocation can deliver statistical multiplexing, which saves DSL line card resources. DSL upgrade can be realized by replacing algorithm software, without the need for line card replacement.
What's more, concentration of line cards facilitates joint signal processing. So far, due to interference and geography, lines between the central office and remote sites cannot be accelerated through Vectoring. However, line card concentration would make Vectoring viable.
5GBB network topology
The access rate per port for 5GBB networks will be 1-to-10Gbps, but most users won’t need such high speeds all the time. If multiple users can share one port, a lot of network construction costs would be saved, with the power consumption of network devices reduced greatly. Therefore, 5GBB networks should use point-to-multipoint (P2MP) virtual topology, with crosstalk channels used to share information.
Underlying technology of 5GBB
5GBB will take advantage of more advanced digital signal processing technologies for coding and modulation to further improve spectrum utilization and the access rate. Power consumption is another consideration with 5GBB.
Modulation-demodulation technology
From the perspective of modulation-demodulation, with ADSL, discrete multi-tone (DMT) has been widely used with excellent results. DMT's frequency orthogonality mode effectively decouples crosstalk, reducing the difficulties in crosstalk cancelation, especially with Vectoring. Thus, we see no need for a change with 5GBB.
However, 5GBB shortens the copper loop. If the maximum distance is 50 meters, the available frequency range for copper can be 500MHz to 1GHz. Therefore, the specific DMT parameters must be adjusted and uplink/downlink duplex solutions should be considered. A proper solution should be developed that provides high performance at a modest cost.
Coding technology
Channel coding is another way to improve spectrum utilization, and 5GBB will certainly use more advanced and complex coding technologies to increase the net coding gain. One option is low-density parity-check (LDPC) coding. Practices have shown that LDPC coding provides higher performance than TCM+RS, and it has been adopted for the latest technical standards such as LTE-A, DOCSIS 3.1, G.hn, and DVBS2. LDPC Coded Modulation (LCM) is another option, but either would provide at least 1.5dB more gain than TCM+RS.
Polar code would be a more radical solution. Proposed by Erdal Arikan in 2007, based on the channel polarization theory, polar code is currently a focus of study, and reports already indicate superiority to LDPC.
Energy saving
To reduce the power consumed during idling, ADSL2 defines the L2 mode. G.fast adopts a discontinuous mode, so the power consumption is zero during idle times. However, the digital signal processor’s consumption is not reduced, so we recommend burst mode for 5GBB. No signal would be sent, with the entire signal link in an idle state, when there is no service data transmission. Power consumption in digital signal processing and the static power consumption of the digital signal processor should be reduced as well.