Industry sources say just about everything is on the table, including an option that would let MSOs insert a pristine, much wider upstream pipe above 1 GHz. While that might sound radical, it may end up being economically practical from a view that's looking out at least five years.
Panic hasn't set in, because HFC's current spectrum configurations can probably support the initial wave of Docsis 3.0 upstream channel bonding for years. But they wouldn't be enough to handle the type of capacity and speeds that might be needed much further down the road.
For now, it appears that cable's available spectrum could have enough juice to hit 100-Mbit/s bursts in the upstream, or somewhere close to it. Comcast Corp. (Nasdaq: CMCSA, CMCSK), for example, disclosed earlier this year that its upstream channel-bonding tests hit sustained speeds of around 75 Mbit/s. (See Comcast: Upstream Bonding Tests Yield 'Sustained' 75 Mbit/s .)
Today, there's a huge discrepancy between the amount of capacity available for cable's upstream and downstream. For North American Docsis, the upstream pipe is typically in the 5MHz to 54MHz range, with EuroDocsis configurations offering 5MHz to 65MHz. Noise renders the lower portion of those ranges (usually anything below 20MHz) generally unusable, although S-CDMA is starting to prove it can help make the spectrum viable for bonding or for faster single-channel Docsis data tiers. (See Moto: S-CDMA Starting to Spread and Moto Preaches Cable's Upstream Savior .)
But even then, there's not enough spectrum to give cable much of a shot at producing upstream bursts of much greater than 100 Mbit/s.
Cable has several options for getting into gigabit territory. Some techniques would expand the amount of usable upstream spectrum, an option called a Mid-Split. Others would create a pristine and much wider upstream path above 1GHz -- sometimes called a Top-Split.
A quantum leap in the making?
It's believed that the CableLabs group is working on how to get more upstream capacity from hybrid fiber/cable (HFC) plants as it becomes less economically attractive to continue splitting nodes and shrinking the groups that must share bandwidth.
Splitting nodes and moving to higher modulation schemes are generally considered incremental steps; this upstream capacity group, by comparison, wants a quantum leap, an approach that could work on a grand scale during the next five to 10 years, should the industry follow suit.
"This started as a science project," says an industry source who's familiar with the work. "It's now in the next phase, which will look at the financials and apply some business and use cases behind it."
Whether new Docsis specs or less formal recommendations will come out of this work is unknown, but the available options have been well documented. Of recent note, John Chapman, a Cisco Systems Inc. (Nasdaq: CSCO) fellow and chief architect for its access and transport technology group, weighed the pros and cons in a paper presented in May at The Cable Show in Los Angeles.
Table 1: Upstream Spectrum Options
|Name||Upstream Frequency Range||RF Bandwidth||Data Capacity|
|Low-Split||5�42MHz||37 MHz||120 Mbit/s|
|Mid-Split||5�65MHz||60 MHz||210 Mbit/s|
|High-Split||5�200MHz||195 MHz||770 Mbit/s / 1 Gbit/s|
|Top-Split||>1GHz||1GHz||2.2 Gbit/s / 3.6 Gbit/s|
The Low-Split pretty much spells out how things are done today and probably isn't a long-term contender. The Mid-Split would remap some upstream capacity where downstream services reside today, but unlike the High-Split and Top-Split options, would not get cable within spitting distance of 1-Gbit/s upstream speeds.
Mid- and High-Splits require plenty of planning, and, according to Chapman, make it imperative for an MSO to upgrade to 1GHz. (Most "upgraded" cable plant is built to 750MHz or 860MHz). "Thus, it would be a waste of capital to not include a downstream upgrade at the same time as the upstream upgrade," Chapman noted.
Chapman holds that the High-Split would offer fiber-like performance at one tenth the price of a fiber upgrade. "The disadvantage," he added, "is that it will take three to five years to bring to market, and the upgrade plan is the most challenging of all the options."
Although Chapman's paper offers plenty of guidance on the benefits and costs of these options, CableLabs declined to comment on its purported upstream project. "At this point in time, we're looking at a lot of different technologies to help our member companies," the organization said, in an emailed statement sent recently to Light Reading Cable.
Indeed, it was reported earlier this year that cable's R&D house had undertaken a project to see how the industry could push capacity to hit fiber-like speeds while preserving the HFC infrastructure. (See Cable Ponders Life After Docsis.)
— Jeff Baumgartner, Site Editor, Light Reading Cable