Standard cable coax systems have a total capacity of 4.7 gigabits when used just for data. Most of the bandwidth today is used for TV. As the TV side goes all digital, data can claim more of the pipe.

First generation cable modems shared a single 6 MHz or 8 MHz channel for a maximum speed of 35-50 megabits, shared.

DOCSIS 3 initially bonded 4 channels for 150-200 megabits downstream. The systems now coming out of the labs bond 25 channels for a shared gigabit. The typical cable system has 132 6 MHz channels, but carrying 100 HD TV channels and the remaining analog TV takes up many of those. Upstream is typically limited and offered to customers as no more than 5-6 megabits. Higher upstream speeds are practical, but would require more changes to the system and are very rarely implemented.

  Intel began sampling gigabit chips in 2012 and Arris has shown working gear at trade shows.  Kabel Deutschland CTO Glanz  is optimistic he’ll begin serving customers by the end of 2013.

  DOCSIS 3.1 is expected to be 1 gigabit (shared) upstream and 10 gigabits downstream, according to John Chapman, Cisco’s chief cable architect and a member of the committee. Manufacturers are targeting 2015 but were two years late with DOCSIS 3.0. It uses more efficient OFDM coding and I believe will use higher frequencies as well. Most systems will require (modest) upgrades to CMTS, amplifiers and other gear. The cost should be moderate (? < $200 most places) but real.

How we got to higher speeds.    

John Chapman, Cisco’s chief architect, stunned everyone in 2005 when he defined the path to the gigabit. That was 25 times faster than anything shipping at the time. Anton Wahlman, Wall Street’s best broadband analyst, spread the word. Brian Roberts, Comcast CEO, was scared of Verizon fiber and told the DOCSIS committee to go full speed ahead. Chapman and the other DOCSIS engineers had the standard in place by 2007 but it’s taken years for the chips to be ready.

  It required world class engineering to bring DOCSIS to a gigabit but there is no magic or rocket science. The fat, shielded coaxial cable has great capacity. The termination systems have been ready for several years, at least in lab trials. Two problems needed to be solved: a modem that could handle a gigabit and freeing up the channels. A chip that can run at a billion bits a second is pretty demanding. But Moore’s Law has made chips more powerful every year. The time has come.

  Freeing enough channels would have been nearly impossible a few years ago when half or more of the channels were used for analog TV. There was a real squeeze as competition forced cablecos to carry 100 HD channels. Cablecos responded by reducing the number of analog channels and most are well on the way to going all digital.

  The digital TV also became far more bandwidth efficent with switched digital. At any given moment, 300 or 400 of the “500 channels” are not being watched by any home on a given node. The less popular channels are taken off the feed that goes to every home and “switched” in when someone tunes in. There’s a slight delay, but that’s an ok tradeoff for the increased capacity.

   U.S. cablecos are still challenged to find room in some systems so will probably be late to the gigabit unless competition like Google forces them to it. The Europeans and Asians haven’t developed as many channels so have less of a problem. Korea faces fiber competition so is a likely pioneer. In Germany, the cablecos have taken millions of customers from the telcos with higher speeds. Kabel Deutschland brags “We offer twice the speed of DSL at the same price and always will.” That’s their main sales point and gives them incentive to pioneer the highest speeds.


Cable is shared, so how fast will gigabit cable be in the real world?

Sharing works remarkably well. 36 megabit shared early DOCSIS was remarkably consistent at 10 megabits even with hundreds of homes on a node. Only occasionally did a node get overloaded and need to be split. If everyone was streaming Netflix in HD at the same time that would change, but the empirical evidence is that’s much less common than expected. The actual results on sharing - now with 150 meg DOCSIS 3 - are remarkably good. I didn’t believe that for years myself.

   My opinion, based on a decade of experience with cable systems and discussions with network managers, is that users will be able to download faster than 500 megabits 95+%  and probably 99% of the time on “gigabit” cable. Bandwidth demand always grows so that will change over time. Predictions are tough, especially about the future, but I’m pretty sure it will be hundreds of megabits almost all the time for a decade in any well-maintained cable network.


Why Upstream will rarely be more than 5 megabits for years

DOCSIS has long had provisions for faster upstreams but cablecos have chosen not to deploy them,  Cheap partial upgrades have made 5 meg upstreams practical. DSL has been even slower on the upstream so there’s little competitive pressure. Some cablecos don’t want to give customers faster uplinks, which would allow, for example, two homes to use a slingbox to share a single cable TV subscription. (Yes.)

   Most important is that the faster upstreams will have a moderate cost to install. The NY TImes reported going to DOCSIS 3 downstream for 50 and 100 megabits cost as little as $20. But cable upstream is currently confined to limited frequencies. Channels would need to be rearranged in a major way and some systems might need significant upgrades.

  Verizon in the U.S. has moved to a standard 50 down, 25 up for their fiber in recent promotions and appears to be winning customers. That might inspire cablecos to upgrade sooner.