Gfast: First Products Certified, Ready for Deployment

　　While many homes and businesses rely on cable internet providers, many others (including me) don't — for numerous reasons. Although my 7 Mbps DSL is generally fast enough, many subscribers want higher download speeds with nearly as fast upload speeds. But the so-called "last mile" of copper is often the limiting factor.

　　Gfast (formerly G.fast) can deliver speeds approaching 1 Gbps under the right conditions. Now, the University of New Hampshire Interoperability Lab can certify Gfast products for compliance and interoperability. The UNH-IOL and the Broadband Forum have announced that the first eight Gfast products have now been certified, with more to follow. To find out what's behind Gfast, I spoke with UNH-IOL senior engineer Lincoln Lavoie on June 19.

　　Why Gfast and not fiber to the premises? "It's the cost," said Lavoie. "At a recent conference, I saw presentations that claim bringing fiber to the premises can be 50% to 60% of the overall delivery cost. If we can use existing copper, that's the business case for Gfast development."

　　"Although Gfast uses multiple carriers to carry data, it's fundamentally different than any form of DSL," added Lavoie. For one thing, Gfast is a half-duplex technology, meaning that bits can travel in only one direction at a time as opposed to DSL, which has different carriers for upstream and downstream. Half-duplex transmission makes more efficient use of the available 106-MHz bandwidth, and that's fine for sending and receiving data.

　　Service providers can designate the ratio of downstream to upstream service. A typical ratio is 80/20. "That's set at deployment time, not baked into the standards," said Lavoie. As with DSL, Gfast can adapt to the noise and loss conditions of the line. Gfast is also capable of retransmission at the physical layer; video can't afford retransmission at higher layers of the protocol stack. After all, you don't want to miss that football goal or baseball home run. DSL was all about download speeds, but that model has changed. With people now uploading content to social media and using cloud computing, upstream has gained importance.

　　The Broadband Forum's Gfast certification program consists of tests performed at UNH-IOL in Durham, New Hampshire. The first six companies to have certified products (eight in all) are Arris, Calix, Huawei, Metanoia, Nokia, and Technicolor. Tests verify interoperability and compliance to standards. The test bed (Fig. 1) consists of spools of wire, ranging from 20 m to 400 m, that can be switched into the loop. In addition, test equipment can inject noise on the line. Gfast devices must be able to adapt speeds to the line conditions.

　　The test bed is mostly automated, using equipment from Telebyte. The copper loops shown in Fig. 1 were built at UNH-IOL. A test takes between eight and 12 hours to run. Test plans call for specific cable models, and the test system adjusts the length of cable until its characteristics match those of the model. Tests start with power-spectral density (PSD) measurements on the transmitters. PSD measurements are important because the telcos need to assure that Gfast transmitters won't damage or interfere with other equipment or future deployments. For example, while current Gfast equipment operates at frequencies up to 106 MHz, future systems will go to 212 MHz. Thus, if there's noise between 106 MHz and 212 MHz, tests must account for those future deployments.

　　There are also regulatory requirements that vary by region. For example, FM radio, TV, HAM radio, and other services operate at or below 106 MHz and they vary by region. So Gfast equipment must be able to adapt by not using occupied frequencies to avoid interference from those signals. Tests include adding interference signals to see how the DUT adapts. When outside interference occurs, Gfast systems can drop their data rates as needed.

　　Will Gfast eclipse the various forms of DSL? That may depend on the progress of 5G. As I see it, the first deployments for 5G will deliver fixed (as opposed to mobile) internet access to premises. Therefore, the last 200 m of copper could go wireless with 1-GHz bandwidths. The race is on. Gfast is definitely ahead of 5G at this point, but I wonder if telcos will hold off deploying Gfast and wait for wireless.