A blog this week from Belal Hamzeh, CableLabs VP, R&D Development, Wireless Technologies, revealed that the specification has been completed for a Full Duplex DOCSIS 3.1. This sounds technical, but Telcos have for quite a while now been focusing on their one advantage, that G.fast was a time division technology, so up and down speeds could be programmable. Prior DOCSIS specs have all been frequency division, meaning that the upstream broadband over DOCSIS 3.1 was limited to a fraction of the downlink speeds.
A year ago CableLabs proposed to take a leaf out of the Telco G.fast book, and put out the idea of a DOCSIS where both ends were transmitting simultaneously across the entire spectrum, and each end cancelled the transmissions interference from the other end, with little or no performance penalty. So not time division or frequency division, but both at once.
The idea was like the magic trick G.fast has pulled in calculating crosstalk in a bundle of lines, and cancelling them. For G.fast it has been a 5 or 6 year fight to achieve this, but for CableLabs, partly because it has taken on the task in an era when Wireless CDMA technologies had already achieved a similar parlor trick, it can come far more swiftly. Whatever the truth of the mathematics behind it, the specification is out, and at IBC we saw companies offering pre-spec devices.
“In the United States, more than 90 percent of households are connected to an hybrid fiber-coaxial network, and consumers typically have higher download speeds than upload speeds. By enabling Full Duplex DOCSIS, the upstream and downstream traffic can flow at up to 10 Gigabits concurrently, doubling the efficiency of spectrum use,” said Phil McKinney, president and chief executive officer of CableLabs.
He’s neither precisely right nor even remotely close of course, cable always over-eggs its claims. First off the likely real world data throughput is about 73% of that total 10 Gbps with the rest of the PHY taking up housekeeping roles, so closer to 7.3 Gbps. Secondly some of that capacity will be lost, perhaps another 5%, in treating the opposing signal as interference. We reckon that each way might be closer to 6.8 Gbps usable in a fully configured CPE.
But that’s the other issue, CPE is expensive and the broader spectrum that you allow them to work across the more tuning logic you need, even if you use Full Band Capture technologies. Today in DOCSIS 3.0 this amounts to 300 Mbps in a typical real configuration, and in DOCSIS 3.1 Full Duplex it is likely to be closer to 1.5 Gbps in early devices. Cable can use all the way up to 1.79 GHz of spectrum for DOCSIS 3.1 services and tends to do this in clusters of 768 MHz (3840 OFDM sub-carriers of 50K each) with the potential to use multiple chunks of spectrum.
So while it is glib to talk about 10 Gbps both ways, in principle, citing PHY rate only, that’s what the spec allows for. What it will do is continue to answer the questions that Telcos are asking, with their easy shift between upstream and downstream speeds in emerging G.fast. Certainly Telcos expected to have an edge over cable in broadband speeds for some time – and this window of opportunity seems now to have become very short.
The blog talks about enabling symmetric multi-gigabit services over existing hybrid fiber-coaxial (HFC) technology and that’s about right, so expect Full Duplex to come to market offering 1.5 Gbps to 2.0 Gbps in both directions a first. And even that will be pretty revolutionary.
Also this week CableLabs also announced its Remote PHY Device Qualification Program which can now certify the emerging R-PHY Nodes or R-PHY Shelves, for formal testing (see separate article on Nokia this week).