Because it marked the day that the first three specs for the all-important DOCSIS 3.0 platform were publicly issued, August 4, 2006 will likely go down as a key date in the evolution of cable technology. Speedier (much speedier) Internet services, a much beefier IP address book, a big stepping stone toward IPTV, and enhanced security are but a few of the features that will be supported by the system spawned from this massive stack of documentation.

'Inside DOCSIS 3.0'

Now comes the hard part–turning those tree-killers into something that is not only deployable but also interoperable.

Shaw Communications of Canada has already trumpeted passionate support for DOCSIS 3.0,issuing a statement in mid-August that it is committed to implementing it "as quickly as possible."

Here we'll try to gauge exactly what is possible and when,and delve into some preliminary steps operators will have available to them well before 3.0 enters the product and deployment phase.

Although deployments of the full 3.0 platform remain far on the horizon,analysts are already predicting that it will be a boon for equipment suppliers as soon as 2007.

ABI Research, for example, expects penetration of DOCSIS 3.0 CMTSs to reach about 60 percent by 2011.Penetration of DOCSIS 3.0 CPEs (modems, set-tops, eMTAs, etc.) will be a tad slower by that time–just under 40 percent.

But getting to that point will still take some doing.In addition to the three issues specs (security,PHY layer, and MAC and upper layer protocols), CableLabs is also working on a fourth spec that will cover OSS.

The R&D house is also starting work on the intricate DOCSIS 3.0 test plans. Once those are written, they must then be automated.

The process of writing and automating those tests plans is "probably more work than writing the specs themselves," says Michelle Kuska,VP of broadband access at CableLabs.Those next steps, she adds, "are pretty dull, but a lot of work."

"In the beginning, it will be a lot heavier task for sure, but it will dampen over time," agrees Peter Percosan, executive director of broadband strategy for Texas Instruments, a supplier of DOCSIS silicon and software. The high level of complexity,however,will not add weeks and weeks to a certification process,he adds.

But there will be plenty of features to consider for the full DOCSIS 3.0 lineup, which easily dwarfs the earlier versions of the specs (see "Kitchen sink," p. 22). "It's an incredibly large set of specifications,which makes it a daunting task to implement everything," notes Jeffrey Walker, senior director of marketing for Motorola Inc.'s Connected Home division.

John Chapman
Operators,meanwhile,are expected to start with,or at least prioritize, some subset of features of DOCSIS 3.0 before biting off on all of them. Topping the list:Downstream channel bonding and IPv6. "We think those are the two hot features within 3.0," says John Chapman,the chief architect and distinguished engineer for the cable business unit of Cisco Systems Corp.

On the way to 3.0

Although what was once termed "DOCSIS 2.0b" is now all but deadand buried as an "official" project, there is still much interest for products that support a subset of the DOCSIS 3.0 feature set–particularly channel bonding and IPv6. For the sake of interoperability, CableLabs will test pre-DOCSIS 3.0 products.

In the U.S., some operators will leverage pre-3.0 channel bonding to meet competition in pockets where telcos operate fiber-to-the-premises (FTTP) networks. But those deployments likely will be targeted and use gear produced in low volume.

By bonding in the downstream today, "that will be at least as fast as or faster than what the telephony companies are providing," Walker says. "They [cable operators] want to be sure to maintain a high-speed competitiveness–even it's for positioning."

In the U.S. "those are pretty much niche markets–but in some cases there is interest and need for potentially a 100 Mbps service in the nearer-term," says Jay Kirchoff, senior director of marketing for Broadcom Corp.'s broadband communications division.

Jay Kirchoff
For that segment, Broadcom has created modem silicon called the BCM3381, which bonds up to three channels in the downstream, and uses just one unbonded DOCSIS 2.0 upstream channel. By Q4 Broadcom also plans to have a pre-3.0 reference design that adds voice to the equation. The BCM3381 is designed to work with existing DOCSIS 3.0 CMTSs, but be compatible with forthcoming 3.0 gear so that those modems don't become stranded assets.

Broadcom expects to hit lab trials in Q3 and deployments in Q4. But the resulting modem won't be cheap. Initially, operators targeted a $75 device, with expectations that it will drop to $35. But there will be big surprises if it starts at $75, or ever hits $35.

The bandwidth discussion used to be about big MP3 files that could be downloaded. Now, it's about bandwidth-sucking streaming video phenoms such as YouTube.

"The move for more capacity hasn't slowed down," says Doug Jones, chief architect, cable, for BigBand Networks, maker of the Cuda CMTS.

BigBand is pushing the needle with an Early Access Program (EAP), which will identify two or three "key" accounts in North America and the EMEA and Asia Pacific regions. It hopes to gather participation from three MSOs in Europe, up to two in North America and two or three in Asia, and could have some of those announced by publication of this article.

"But we're not trying to slow down 3.0," Jones insists. Adding downstream channel bonding with the existing CMTS silicon "is a straightforward and easy move for us," he adds.

ARRIS is also doing more than a fair bit of pre-3.0 work. At the CMTS level, ARRIS can start channel bonding with a software upgrade to its flagship Cadant C4 chassis. It can do that now, the company says, largely because packet processing is handled in the CMTS hardware.

ARRIS believes its existing base of C4s will be able to support all of those high priority components, including discreet downstreams and upstreams. The company also sees a key role for its downstream-heavy D5 device. Although only in demo mode today, ARRIS has shown scenarios in which they can work in tandem, with D5 handling downstream and the C4 taking care of the upstream.

"Our goal is to make the transition to 3.0 as economic as possible for our existing customers," says Stan Brovont, ARRIS' VP of marketing and broadband business development.

In the interim, ARRIS is focusing efforts on its pre-3.0 FlexPath platform, which bonds channels in the downstream only, and uses silicon from TI. Looking ahead, ARRIS will support a range of 3.0 channel bonding configurations, including 2 downstreams/2 upstreams, 3x3, 4x4 and 4x1.

Other operators are looking to early channel bonding implementations to support commercial services, Brovont points out.

"The sense of urgency is probably a little different," between U.S. MSOs and others around the world in terms of setting any pre-3.0 wideband strategies, Brovont says.

"The Asia community was screaming for this stuff probably a couple of years ago," says Cisco's Chapman.

Still, there other ways operators are boosting speeds without implementing pre-3.0 channel bonding.

For one, they can add a new QAM for a group of 500 homes sharing a 6 MHz channel, and split that to 250 homes on that channel. "Now you can do 10 Mbps with relatively small costs," says Gil Katz, director of cable solutions and strategy for Harmonic Inc.

Another near-term option is to leverage (but not bond) an additional 6 MHz channel for high-speed data (HSD). In addition to increasing the capacity that can be shared by a given group of homes, another benefit is that operators will not have to introduce a new class of modem to support the strategy.

Instead of 500 homes listening to just one 6 MHz channel, by adding another 6 MHz channel for HSD, the available capacity of those two channels is then split by two different 250-home batches.

"Effectively you increase the capacity of the modem by two," Katz says. "I believe this will be very hot next year because it's so simple and so cheap, and solves the immediate problem."

It starts with silicon

But the long-term problem will be addressed by DOCSIS 3.0, and that effort starts with the companies that make the chipsets.

Although most of the specs are already issued, not all of the specific requirements for DOCSIS 3.0 are determined. And that generates moving targets that could possibly extend product timelines for silicon vendors and leave them to ponder what should or shouldn't be integrated right off the bat.

Even if operators think they'll switch on DOCSIS 3.0 in 2008 or 2009, internal analysis at Broadcom shows that operators will need more than the minimum four upstream/four downstream configuration.

Also, unlike earlier DOCSIS specs, 3.0 is not just about fixed modems and e-MTAs. Mobility and support for dual-mode handsets that tap cellular and WiFi networks will also need to be part of the mix.

"We're building components and deciding what's necessary for the 3.0 modem. But it's not clear to us what the real requirements will be for a device that has some longevity once it's deployed," Kirchoff says. Broadcom is noodling those decisions now, but hopes to deliver a product that has at least three to four years of life to it.

For TI, meanwhile, the company is now concentrating on the execution of three arcs–finishing the silicon, completing the "system" (silicon and software), and creating test environments.

TI hopes to have a reference design ready for the second half of 2007.

While TI and Broadcom have served as the primary competition in the sector of late, Conexant Systems is back on the DOCSIS radar.

Although the company did take a step back from the DOCSIS modem and embedded MTA market, it has since refocused itself on DOCSIS silicon for the set-top.

Eric Rayel
"What's largely driving our roadmap today is the set-top, first and foremost," says Eric Rayel, the director of marketing for Conexant's cable set-top product line.

Part of that decision is tied to success the company has already had with DOCSIS-powered set-tops with early movers in Europe and Korea. NTL, for example, is using Conexant DOCSIS silicon for a basic high-speed Internet connection. Some operators in Korea, thanks partly to government influence, moved ahead with an OpenCable infrastructure with the DOCSIS Set-top Gateway (DSG), rather than proprietary out-of-band signaling techniques. Samsung and Humax, two Conexant set-top partners, are also shipping set-tops with the company's DOCSIS 2.0 silicon for the Japanese market.

Conexant's next investment will be in set-top modem silicon that will incorporate IPv6 and bond up to three channels in the downstream, with no upstream bonding–very similar to Broadcom's strategy with the BCM3381. Conexant expects to announce a specific pre-DOCSIS 3.0 set-top product with wideband capabilities later this year.

For a full DOCSIS 3.0 product, Conexant is targeting mid-2008, though that timeframe will be driven by how quickly its set-top partners want to drive the technology in the U.S. market. Conexant has not announced a wideband set-top partner, but its closest affiliation in the U.S. is Motorola, which uses Conexant chips in the all-digital DCT700 set-top box.

Once silicon is geared up, it will next be up to the modem and CMTS vendors to take it to that next important step–creating the end products.

Today, Motorola's BSR 64000 can bond up to four channels with the existing CMTS cards that supply two downstreams and eight upstreams. It has demonstrated bonding with 145 Mbps.

Motorola hopes to have 3.0 CMTS silicon in-hand by Q3 2007. Then it will take another year before a product will be ready for full qualification testing, estimates Walker.

At the CMTS level, Cisco is zeroing its 3.0 effort on the uBR10012 chassis, rather than the more widely deployed uBR7246.

Because of channel density plant requirements, operators will need to move from the medium-class to the high-end CMTS once they start to use 4x4 channel bonding scenarios, Chapman explains.

Regarding wideband, "the first priority is the 10012," with the necessary external edge QAMs, he adds. As for the 7246, he notes that Cisco will continue to evaluate the market and decide whether there's a good business reason to do it on that particular platform.

Despite providing some features that represent just a piece of the bigger picture, most agree that it's time to put the nose to the grindstone on the full DOCSIS 3.0 specs.

"It's work at this point. It's staffing and it's funding," says Jones of BigBand.

Early adopters
There's not much yet to speak of in North America, but here's a glance at some announced early Wideband work happening elsewhere around the globe.
Operator Market Notes
StarHub Singapore Using Motorola modems and CMTSs to cap speeds at 100 Mbps-plus with a couple hundred customers.
KCTV JEJU Broadcasting JEJU Island, South Korea Deployment of 100 Mbps-plus services with ARRIS' FlexPath Wideband CMTS/modem combo.
Hanaro Telecom South Korea Working with ARRIS.
J:COM Japan Some small trials with Motorola.
UPC Broadband Europe Markets not disclosed, but the operator has set trials with ARRIS, as well as the Wideband trio of Cisco Systems (CMTS), Linksys (modem) and Scientific Atlanta (modem and edge QAM).
NTL United Kingdom Kicked off trial with "friendlies" in March. ARRIS is the announced modem and CMTS partner.


Where's the kitchen sink?

DOCSIS 3.0 is a massive set of specs with features galore. Here's a snapshot of some of the major ones.

Channel bonding: Specs call for a configuration of at least four downstream and four upstream channels, though a variety of configurations will be available and supported in the years ahead.

IPv6 : IPv6 is a next-gen addressing scheme that will live alongside the current generation of IP devices as the pool of existing IPv4 addresses exhausts itself (for more detail, see CED Sept. 2006, p. 32). Clearly, there aren't enough IPv4 addresses available to manage the number of IP devices that will need to be managed.

Presenting during a recent CED Web cast on the subject in conjunction with Juniper Networks (an archive is available at, Alain Durand, the director and IPv6 architect for Comcast, explained that his company would need 100 million IP addresses alone just to manage a fleet of IP-enabled set-tops for about 20 million customers–with the assumption that there are 2.5 set-tops per home, and 2 IP addresses per set-top. And that figure does not even address MTAs. With the growth of cable modem and VoIP services at Comcast, he added, the MSO exhausted its private IPv4 address pool last year. Comcast has its IPv6 work well underway, and will start only where IPv6 is necessary–the management and operation of edge devices such as DOCSIS modems, set-tops, and PacketCable eMTAs.

IPv6 has already taken hold in areas that were running low on IPv4 addresses earlier on, particularly in Asia. J:COM of Japan, for example, has already been using IPv6 for a few years. Some networks in India are already predominantly IPv6, noted fellow panelist Mallik Tatipamula, Juniper's senior director of service provider strategy and planning.

E-router: Related to this is the e-router, a developing spec that will turn the 3.0 modem into a router and do away with the NAT (network address translation) headache. NAT, in existing routers, makes it difficult if not impossible for operators to manage services that hang off the home network. With the e-router, a subnet can be applied in the home that is made more powerful with IPv6 address spacing. "In the old days, DOCSIS was a bridge. Today, the notion is that 3.0 becomes a router that's more in tune with the market and the data that we'll be pushing around," says TI's Peter Percosan.

Enhanced IP multicast: This will enable the ability to provision and manage IP multicast, and allow customers to subscribe to a session, and to allocate the proper amount of bandwidth for that particular session. This will allow the operator to provision and manage the IP connection, rather than just permit it, and pave the way for video over IP multicast sessions. Cisco's John Chapman calls this a "sleeper feature" of DOCSIS 3.0.

Enhanced security: Supports AES traffic encryption, which is stronger than DES, and could solve the modem cloning problem. With this, a CMTS can also check the validity of a modem on the network, and therefore, prevent theft of service.

Enhanced network management: Takes over where a legacy feature called "modem slapping" left off. If a network becomes too noisy, it can cause modems to reboot themselves multiple times. But once that "slap list" (or "cable modem diagnostic log," if you prefer the kinder, gentler term) was pinpointed, it followed that an operator could then determine where the plant is bad. This practice is now standard in 3.0. "What started as an operational problem has turned into a feature," Chapman says.

Physical layer enhancement: The upstream frequency (on plants that use 6 MHz channels) has been extended from 5-42 MHz to 5-88 MHz, with 88 MHz serving as the split between the upstream and downstream. This basically doubles the amount of spectrum available to use for upstream channel bonding, though diplexers will need to be changed. "It's a minor upgrade to the plant," Chapman explains.

Business services over DOCSIS: With more upstream, operators can now retarget the DOCSIS plant toward the lucrative commercial segment, and deliver symmetrical (or close to symmetrical) services to business customers, including Layer 2 VPN and Layer 3 T-1/E-1 offerings. –JB

The M-CMTS: Related but not required

Although there are some linkages, the modular CMTS is not a specific requirement for DOCSIS 3.0, and is in fact a separate spec.

As it's put together, the M-CMTS focuses on edge QAMs that can handle video and DOCSIS traffic and share the long-term migration to an all IP environment. But an important feature of this approach is the ability to disassociate and decouple the CMTS upstreams from the CMTS downstreams. If channels can only be bonded in the downstream, it makes no sense for operators to buy cards with upstream capacity that will go unused. Another derivative benefit: it should drive down the costs of the edge QAMs themselves.

But DOCSIS 3.0 can be supported through an integrated CMTS (I-CMTS), as well.

For its part, BigBand will support M-CMTS and I-CMTS implementations. On the first, the company will match its Cuda CMTS platform with a universal edge QAM that it is already using for switched broadcast. Eventually, that "universal" edge QAM will support VOD, video and data simultaneously.

"Operators will want a little bit of both," says Doug Jones of BigBand. "Sometimes the field wants one, and corporate wants another. You can do DOCSIS 3.0 either way," he adds, with the caveat that the modular method will provide operators with savings and flexibility they can't get with an integrated approach.

"We've put our stick in the ground, and we think the M-CMTS will be the most economical way to go," adds BigBand Director of Strategic Marketing David Ronald. "But we're not ruling out the I-CMTS approach." –JB