While reaching the DOCSIS 3.0 summit remains a goal that the cable industry is bent on achieving, competitive forces from around the world are causing operators to consider an interim step called DOCSIS 2.0b that will allow them to offer downstream speeds in excess of 100 Mbps much sooner, and do so using some of their existing equipment.

At this point, CableLabs is being somewhat tightlipped about the details of 2.0b and what it will or won't support, since much of the feature set is yet to be decided. However, North American cable's R&D house did at least disclose more than the mere fact of its existence. It also let loose some details the week of May 8 during a closed door meeting with DOCSIS vendors and manufacturers.

"At the current time there is no formal specification for this. It's an outgrowth of some of the early downstream bonding work that's gone on in the industry," confirms CableLabs Chief Technology Officer Ralph Brown.

CableLabs, he adds, has been asked to provide a level of interoperability testing that gives vendors and operators the assurance that some of these early channel-bonding iterations will play nice with DOCSIS networks without the full label of DOCSIS certification.

Ralph Brown
When asked if a full 2.0b spec might be forthcoming, Brown says, "I think that's unlikely, mainly because our members, at the same time, are full speed ahead on DOCSIS 3.0. This [2.0b] is really being identified as an interim solution." Breaking stride on 3.0 by introducing a new set of specs for 2.0b "doesn't seem to make a lot of sense," Brown adds.

Because it's not a full-out specification but provides for interoperability, 2.0b is somewhat of a 'tweener. "It's neither fish nor fowl," Brown jokes.

So, what's in it?

CableLabs confirmed that one proposal on the table for 2.0b calls for a minimum requirement of two bonded channels, down from an original pitch that it combine a minimum of three.

Although the melding of three channels would give operators the opportunity to eclipse the magical 100 Mbps barrier (again, just in the downstream), engineers must also justify why they need those channels. Besides, chances are low that several channels are just languishing on the plant for channel bonding, particularly in systems that just gobbled up a handful of them for digital simulcast deployments. And, it is much easier to ask for two extra channels for bonding, and then settle for just one.

"Two may be all they can wrestle out of the hands of the HFC spectrum gods," says one engineer.

While 2.0b will be big on the downstream, it won't goose the upstream beyond the shared 30 Mbps they're theoretically supposed to be able to squeeze out of it. 2.0b might provide some help with commercial services, but it won't provide much aid in developing symmetrical, T-1-type services.

Brown also confirms that 2.0b will allow for the bonding of fewer than four downstream channels (the minimum requirement for the coming DOCSIS 3.0 specification) and provide the hooks for IPv6, a next-gen addressing system that will greatly expand the number of available IP addresses. That matters because many cable operators are quickly running out of addresses for the devices entering people's homes. A home that has, say, a cable modem, a DOCSIS Set-top Gateway (DSG)-equipped set-top, and an embedded multimedia terminal adapter (e-MTA) could, for example, require four IP addresses.

He says the value of CableLabs' involvement is providing the testing and interoperability needed so that vendors and operators don't embark down a proprietary path or head down a blind alley that doesn't link back and work with existing and forthcoming versions of DOCSIS.

The other value is that operators can begin to taste the benefits of downstream channel bonding well before DOCSIS 3.0 equipment finally becomes available.

Industry observers say North American cable operators can expect to see 3.0 products and deployments by the 2008 and 2009 timeframe. By then, 100 Mbps may not be the "magic number" anymore for a service operator. It might be 250 Mbps or higher.

2.0b is also being viewed as a nice stepping stone to 3.0, because it's less complicated and can leverage existing DOCSIS 2.0 cable modem termination systems (CMTSs) that are outfitted with a software upgrade.

"The fundamental capability on how to bond downstream channels [is] there now," says Doug Jones, the chief cable architect of BigBand Networks, the maker of the Cuda 12000 CMTS. But the modular CMTS and support for the forthcoming DOCSIS 3.0 spec is clearly on BigBand's product roadmap, he adds.

The delta between DOCSIS 2.0 and 3.0 is considered enormous. There's a lot of important stuff crammed in there, so much so that some in the industry refer to it as a "kitchen sink" specification.

"A lot of people have come to realize that 3.0 is probably the most complicated spec written yet in terms of levels of complications," says one industry engineer. "It drowns the complexity of DOCSIS 1.0 and 1.1."

"Some operators have come to realize they need channel bonding sooner and can't wait for full DOCSIS 3.0 CMTSs and modems," the engineer says.

The good news for DOCSIS 3.0 is that there is no maximum number of bonded channels it can logically fuse together in the downstream or the upstream. An 8×4 service, for example, would be fully compliant without any changes to the specification. The minimum number of four bonded channels (at 37 Mbps per channel) supported in DOCSIS 3.0 would give operators a potential throughput approaching 150 Mbps.

From a technical standpoint, that achieves what the operators thought they wanted in early 2004 and 2005. We're already half way through 2006, and it will be 2008 or 2009, according to most estimates, before DOCSIS 3.0 technology will reach the field and deliver big bandwidth services.

"The bottom line is that's too long," says one industry observer.

2.0b drivers

And why is that? It depends where one looks. The competitive threats driving the immediate need for downstream channel bonding and 2.0b vary in degree region by region.

In the United States, several cable operators are wrestling with telcos that are deploying fiber-to-the-premises technology to deliver speeds well above 20 Mbps and pushing the capabilities of single-channel DOCSIS technology. Verizon, FTTP's poster child, recently hit the gas with its FiOS Internet service in parts of New York, New Jersey and Connecticut and upgraded its high end tier from 15 Mbps down/2 Mbps up to 20 Mbps/5 Mbps. Its flagship tier was elevated from 5 Mbps/2 Mbps to 10 Mbps/2 Mbps.

While 2.0b will help operators meet or exceed those speeds, some are already trying out some non-DOCSIS technology to help out on the commercial side of the house. Cablevision Systems Corp., as one example, has already announced its use of out-of-band overlay technology from Narad Networks to deliver dedicated 50 Mbps services to some business customers in Oyster Bay, N.Y.

Matters are more pronounced in parts of Asia, including Singapore, South Korea and Japan, which have geographies that are well suited for super-fast VDSL services, as well as FTTP.

The picture is different in Europe. There, operators are seeking ways to offer video-over-IP via set-tops that support channel bonding.

The Genesis of 2.0b

The evolution of DOCSIS 2.0b can partly be stemmed to a decision made last year, when CableLabs selected "packet bonding" for DOCSIS 3.0 over a rival technique called "MPEG bonding." At the time, many in the industry pointed out that packet bonding won out because it offered operators a way to bond multiple channels via a software upgrade to existing cable modem termination systems.

Last April, about the time the news about the selection of packet bonding began to spread, there was also discussion about the other piece of the puzzle—the cable modem—and how there would be a need for new, condensed silicon that would reduce costs of existing pre-3.0 implementations, which tie together multiple cable modem chips, are expensive, and used in only very targeted deployments thus far.

The need for 2.0b has also arisen due to other pre-3.0 work going on in labs and even in the field. ARRIS has been solving the channel bonding problem since May 2004, when it first demonstrated the technique at the 2004 National Show in New Orleans. At this month's SCTE Cable-Tec Expo in Denver, ARRIS plans to demonstrate wideband modems intermixed with 2.0 modems working together on a PacketCable Multimedia network.

In the field, ARRIS is deploying or testing its wideband "FlexPath" platform with a number of MSOs in Korea and in parts of Europe, including a trial with NTL in the United Kingdom. The ARRIS Touchstone WBM650 modem uses silicon from Texas Instruments to fuse together four downstream channels.

On the CMTS side, ARRIS is presently looking to see if it will be able to build a software upgrade to support 2.0b, according to company CTO Tom Cloonan.

"We think we're ahead of a lot of folks. We've done an incredible amount of testing," he says.

Although TI and ARRIS are working together on some pre-3.0 work, the development of specific requirements now being linked to 2.0b can be traced to a digital set-top chip developed by Broadcom Corp., and unveiled in January at the Consumer Electronics Show.

That chip, the BCM3255, is capable of taking in three channels from anywhere in the cable spectrum. Near-term, it's expected to use one channel for the DOCSIS Set-top Gateway (DSG), and to use the other two for the delivery of more traditional digital video channels over MPEG. That seems to fit well with Comcast's plans for its RNG (Residential Network Gateway), which will use DSG for signaling, support advanced codecs, and have enough horsepower to run the OpenCable Application Platform (OCAP).

What's interesting about Broadcom's new set-top chip is that it can also operate in a mode that bonds up to three DOCSIS 2.0 downstream channels and handle, among other things, IP-based video applications and services.

cable modem reference board
Broadcom’s BCM93381 DOCSIS 2.0 cable modem reference
board with channel bonding could find a prominent
place in DOCSIS 2.0b development.
Broadcom has since taken a step further with the BCM3381, a cable modem chipset that is intended to bond up to three DOCSIS 2.0 downstream channels together with a single upstream. That's enough pop to give operators a blistering downstream pipe in excess of 100 Mbps—enough to allow MSOs to respond to emerging competitive threats.

Broadcom is presently sampling the chip, but expects to reach commercial production sometime this year, according to Jay Kirchoff, Broadcom's senior director of marketing, cable modem products. The company declined to disclose unit costs.

A more integrated modem chipset for channel bonding should, in turn, reduce modem unit costs compared to pre-DOCSIS 3.0 Wideband modems that are presently in use.

Today, a standalone DOCSIS 2.0 modem costs about $30. Operators, many in the industry say, would be happy with, or at least tolerate, 3.0 modem unit costs of $75. But, considering everything that is going into the 3.0 specifications, it's expected that the $75 target will be tough, if not impossible to hit, for some time. The conversion from 1.1 to 2.0 modems was "clean" because the price delta on those products was almost a mirror image.

Estimates for existing pre-3.0 wideband modems are in the range of $150 to $200 and typically tie together three or four modems with a common power supply and processor. They aren't offered in high volumes, but are being wielded to fend off competitive forces that are here now, not two years from now.

The impact on 3.0

Industry insiders say 2.0b interops and some lab trials will take place in 2006 among North American MSOs. The hope is that deployments start in 2007, but some see them slipping into 2008.

With those possible timeframes, there are varied levels of concern about whether the 2.0b project will somehow undermine or at least slow down the DOCSIS 3.0 effort.

"We think this [2.0b] gets the market almost accelerated, because [operators] can experience these capabilities earlier rather than having to wait for the full 3.0 silicon implementation," says Jeffrey Walker, senior director of marketing for Motorola's Connected Home division, which is working on an early channel bonding trial with StarHub of Singapore. "The other key thing is 2.0b is a stepping stone to 3.0. You're not looking at throwing things away."

Arris' WBM650
ARRIS’ pre-3.0 WBM650 wideband modem is
powered by silicon from Texas Instruments.
"Our initial glance shows that this [2.0b project] shouldn't slow us down a lot," adds an engineer with a DOCSIS vendor familiar with both projects. "It's not a monumental increase in terms of time. It may shift a month or two, but not a year."

Others aren't so sure. 2.0b, even as an interim step, will "definitely have a slowing effect" on DOCSIS 3.0, believes Peter Percosan, executive director of broadband strategy for the residential gateways and residential systems unit of Texas Instruments Inc.

"From our perspective it's a real distraction from getting next-generation DOCSIS into the field," he adds. TI, another major DOCSIS and PacketCable CPE chipmaker, bowed a "Wideband" reference design last August.

The "long pole in the tent" is the CMTS vendors, he says, adding that there is some concern that those companies will focus on obtaining near-term revenue from 2.0b rather than focusing on 3.0.

At the same time, Percosan believes operators could cap the development timeframe for 2.0b to ensure that 3.0 stays relatively on course. He says operators have asked TI not to focus on 2.0b, because it is not considered a long-term solution.

Although 2.0b will help in the short-term, TI hopes its emergence "won't punish us in the long-term," Percosan says.

He also suggests operators should try to solve some of their near-term speed issues with DOCSIS 2.0, which closes in on, but doesn't quite provide, a 40 Mbps symmetric pipe.

"With node management, [operators] could continue to have a pretty impressive product today. I don't know if they need bonded channels to support this," Percosan says. "We see that there are some possible short-term benefits to the operators [with 2.0b], but we're staying focused on the big prize."

Another concern about 2.0b is whether it will play well with earlier 2.0, 1.1 and 1.0 modems or be able to share the same spectrum.

Some have also questioned whether the same bandwidth set aside for 2.0b CMTSs and cable modems can also be used for the gear that's outfitted for the full-fledged 3.0 specification. Most are betting that everything is being done to ensure that the two are compatible at this level, as well.

Without it, 2.0b would be a "non-starter," Walker says.