The telephone industry wasn’t kidding when it named the newest version of its digital subscriber line (DSL) technology G.fast. Indeed, it's fast in two ways: The speed at which it delivers data over copper and the timeline the telephone industry is employing to bring it to market.
Of course, it is important for cable operators to stay abreast of what is happening in the DSL game. Though DSL is the technology used by cable’s competitors, the relationship to coaxial cable is unmistakable: In both cases, engineers are tasked with finding innovative ways to take a medium that is deeply entrenched and difficult to rip and replace somehow pushing throughput to keep up with exploding customer needs.
Both industries have done remarkable jobs.
Cable operators’ DOCSIS program is heading for a milestone with version 3.1 of the standard. G.fast represents a similar milestone for telcos. Keith Russell, the product marketing manager for access products for Alcatel-Lucent (Alcatel-Lucent (NYSE:ALU) laid out the history: ADSL was introduced in the 1999-2000 timeframe. It was followed by Very-high-bit-rate digital subscriber line (VDSL) technology in 2011 and, now, G.fast. The emerging standard is expected to be capable of transporting data at 1 Gbps.
G.fast, which is being written by the International Telecommunication Union, is not yet a standard. Michael Weissman, the vice president of marketing for chip maker Sckipio, said G.fast concluded the consent phase last December. At that point it became a recommended standard, which means that vendors can build chips with a high degree of certainty that they will pass muster for inclusion in products. Two vendors – Sckipio and Broadcom – did so, Weissman said.
Comments are solicited during the recommendation phase. The ITU received about 2,500 comments. Weissman said it generally would take about two years to wade through them all. But the need for G.fast was so great that the job is just about done. G.fast is slated to become a fully ratified standard one month from today.
G.fast puts a lot of tools to work. The biggest gain simply is in the amount of spectrum being used. Lincoln Lavoie, a senior engineer at The University of New Hampshire’s Interoperability Lab, said VDSL2 used a maximum of 17 MHz of spectrum to send data. The first iteration of the G.fast standard extends the range to 106 MHz and, in the future, doubles that to 212 MHz.
Vectoring is an existing interference cancelling technique that was not invented for G.fast - it is used in VDSL today - but clearly plays a role. A big problem in copper transmission is crosstalk between copper loops in the same bundle. Vectoring was designed to cancel this interference, Lavoie said. The analogy is to noise canceling headphones. Both sense the problem and send out signals of the same frequency to neutralize the unwanted signal.
A second enabling technique for G.fast is retransmission, which Lavoie said is a complement to forward error correction (FEC). The goal of both is to resend symbols that were corrupted due to noise. Lavoie said retransmission works at the physical layer and therefore is exceedingly fast. “It is really important in apps and real-time services such as VOD and broadcast television,” Lavoie said.
Finally, the transmission is changed. In previous iterations of DSL, frequency division multiplexing was used. G.fast utilizes time division multiplexing. Such a scenario – the entire spectrum given over to data moving in one direction or the other at a given moment in time – provides the telco with far more flexibility than FDM approaches.
From a statistical downside, the optimum recommended transmission distance of G.fast is significantly shorter than previous versions of DSL. Russell said VDSL2 with vectoring has an optimal range of 400 meters and operates moderately well to 1 kilometer and beyond. G.fast, however, is limited to 250 meters or less.
Reducing the optimal transmission length by about 60% is significant. But, proponents say, the fiber is moving ever deeper into networks. The real world ramifications of the loss of distance may be minimal.
Another advantage of G.fast is that it is a plug-and-play technology that generally doesn’t require a truck roll. “The self-install is a byproduct of the technical structure of G.fast and how it sets itself up, provisions itself and trains itself,” Weissman said. “The robustness of things that have been done in home networking means that anything related to the self-install is done from the [distribution point unit] to the home, not from the central office to the DPU.”
The industry is not wasting time. Broadband Forum CEO Robin Mersh said the goal is to have commercial deployments using G.fast in 2016. “That’s very fast if you plan backwards [from now],” he said. “At this point you don’t have chipsets in place. If you project it through the next year, it is a pretty aggressive timeline.”