Insiders Say Next-Gen DOCSIS Will Be ‘Dumb Pipe’ over Coax

Steve Craddock, EVP and CTO, PCT International

Steve Craddock, EVP and CTO, PCT International

February 19, 2010 – As Google ratchets up efforts to spur national investment in fiber access networks, the cable industry is gambling its emerging IP delivery architecture tied to ongoing use of coaxial cable will provide a much lower cost and faster means of meeting next-generation service and bandwidth requirements.

Details of the cable agenda are starting to emerge as vendors, operators and CableLabs ratchet up efforts to reach agreement on a new generation of DOCSIS – now loosely referred to as “DOCSIS 4” – that will enable distribution of high volumes of video content across all consumer devices at the lowest costs possible. As previously reported (January issue, p. 1), the goal is to accomplish this transition to ultra-broadband capacity in the return as well as downstream paths without resorting to massive spending on fiber access.

The strategy, already a dicey gamble in light of Verizon’s ongoing buildout of GPON (gigabit passive optical network) infrastructure, looks riskier than ever in light of Google’s campaign to influence the FCC’s emerging broadband plan in the direction of promoting wide-scale use of fiber links. Google, with plans to reach up to 500,000 households with its own networks, has issued an RFI in search of communities to serve as test beds for demonstrating all the benefits of a network that could deliver a 1 gigabit-per-second stream to every household.

Citing potential applications such as 3D imaging from health clinics, fast downloads of high-definition movies, distance learning in 3D and other futuristic concepts, Google says in a statement on its Web site, “We’ve urged the FCC to look at new, and creative ways to get there in its National Broadband Plan – and now we’re announcing an experiment of our own….Our goal is to experiment with new ways to help make Internet access better and faster for everyone.”

Along with encouraging development of bandwidth-intensive applications, Google says it will test new ways to build fiber networks and share results with the outside world. It will be an open-access network “giving users the choice of multiple service providers.”

Knowing it will be a long time before anyone, including Verizon, gets to anything approaching ubiquitous availability of 1 gbps per household over fiber, the cable industry is banking on its ability to move quickly to a head-turning level of broadband capacity and functionality well ahead of most competitors. Strategists believe a 1 gig IP service delivered across a relatively small coaxial service area is doable and will be more than adequate to meeting consumer demand for a long time to come.

Just what DOCSIS 4.0 (not yet an official moniker at CableLabs) will look like is far from settled, although it appears the case for basing it on commodity-priced gigabit Ethernet technology is gaining strength. “Why wouldn’t we want to take advantage of the commodity pricing of GigE?” asks a senior industry executive, speaking on background. “What you want is a big dumb pipe that allows you to get everything to and from the home as cheaply as possible.”

Indeed, just about all PCs and Macs of recent vintage come with 1000Base-T NICs (GigE network interface controllers), as do many other devices. Moreover, various models of DOCSIS 3.0 modems are equipped with 10/100/1000Base-T ports, allowing modems to communicate at whichever Ethernet tier is best suited to devices and network capacity in the home. For example, Motorola’s Surfboard 6120 cable modem with GigE port is now on sale through retail stores and online for around $90 or under.

In the minds of many cable strategists, DOCSIS 4 architecture eliminates the need for the traditional DOCSIS CMTS (cable modem termination system). Instead, the media access control (MAC) specification that emerges for the new generation of DOCSIS would be layered into edge routers and other devices positioned in conjunction with the primary, secondary and local storage points across the cable CDN (content delivery network).

According to sources who are willing to discuss the emerging architecture on background, the new MAC would serve as the interface between the data pipe and everything else so as to ensure the specific quality-of-service parameters attending each type of service are maintained across the distribution network. The new MAC would have to “talk” to the traditional DOCSIS MAC, MPEG MACs and other data protocols and translate them into commands at the Ethernet MAC layer, but it would not necessarily need to be inserted as the MAC header into the Ethernet stream itself, as is the case with today’s DOCSIS transport. This would do away with the need for cable-specific end devices.

All the intelligence associated with user authentication, billing, operations support, usage policies, functioning of applications, advertising placements, security, etc. could be handled in the IP domain, allowing operators to position intelligence in support of such operations wherever they wish. These controls would all be passed transparently through the system to communicate with network and premises devices.

Of course, this elegant, simple approach to next-generation cable IP architecture introduces a bevy of big challenging issues. One has to do with how the industry would go about adding capacity over existing coax to transport at gigabit speeds in both directions. Another, related issue has to do with the quality of existing connections in the coax distribution plant.

And there are issues associated with how the new gigE IP service and legacy services are managed at customer gateways for distribution to devices connected to home networks. However, there will be plenty of flexibility to design these units as MSOs see fit within the general parameters of the new IP infrastructure (see February issue, p. 15).

As for the capacity issue, there’s no doubt that, as noted by CableLabs CEO Paul Liao in previously reported remarks (January, p. 1), “That coaxial cable that runs outside your house can carry five gigabits per second easily to your home if we harness it correctly.” Indeed, the RF spectrum currently in use over coax is sufficient to carry that load in all-IP mode if one eliminates analog and traditional digital TV signals. Raw digital capacity, measured at about 38 megabits per second per 6 MHz channel using current RF modulation techniques, easily tops 5 gbps on 1 GHz cable plant, and comes close to that mark over the more commonly deployed 750 and 860 MHz plant.

But, in the interest of keeping costs down, rather than convert overnight to delivering 5 gbps worth of IP-based services, operators must find a way to introduce a very big IP stream, such as the aforementioned 1 GigE, without interfering with the legacy service streams in that first 1 GHz tier. This is doable, as evidenced by products of the past from Narad Networks and Vyyo that were meant to deliver GigE-levels of service to business customers over the RF spectrum above 1 GHz.

Narad is defunct, but at least some of Vyyo’s technology survives in the product mix of Javelin Innovations, the firm begun by Vyyo founder Davidi Gilo, who bought out Vyyo in 2008. Cox Communications is the one known customer for the taps used with the original UltraBand Vyyo platform, which utilizes 500 MHz bandwidth segments in the 1-2.75 GHz tier. Javelin has been quiet since first demonstrating its 2.75 GHz taps at last year’s Cable Show, and on its Web site does not refer to any offering of the active components that are essential to operating above 1 GHz.

The way RF attenuation works, the higher the frequencies, the greater the attenuation over any given distance. So there’s no way to get around the need for adding in-line amplifiers to support high-tier RF throughput unless operators radically reduce those link distances by pushing fiber deeper into the coax plant.

But steps taken to expand the capacity on coax will be a lot cheaper than what it would take to hook everybody up to fiber, especially in areas where underground plant is involved. Vendors are already making this case with cost-effective “fiber deep” solutions that allow operators to turn existing amplifier stations into miniature fiber nodes, thereby reducing amplifier counts between the node and any household to two, one or even none. As previously reported (March, p. 14), Videotron, in the largest announced fiber deployment to date, is upgrading its Montreal system with the fiber-deep solution from Aurora Networks to cut node serving areas to an average of 125 households.

While such efforts presently are aimed at reducing the number of households contending for broadband bandwidth and dedicated on-demand TV channels, they also open up the possibility of using RF transport above 1 GHz for an all-IP tier of services. Thus, the cost-benefit case for fiber deep has a near-term component tied to legacy services and a longer-term benefit tied to migration to all-IP in competition with all-fiber networks.

Eventually, with elimination of the legacy services in favor of nothing but IP delivery, cable operators would be able to exploit the full multi-gigabit capacity of the coax network across small node service areas to deliver a per-household dedicated service stream that easily rivals the 2.88 gbps throughput of GPON, which is passively split onto fibers serving 32, 64 or 128 households, depending on neighborhood densities. Thus a 5- to 6-gbps stream shared across 125 cable households in a fiber-deep deployment would provide about the same dedicated bandwidth per home that can be achieved over a 64-home split on a GPON network.

In fact, insofar as all those PON fibers are only extended to paying customers whereas the coax passes non-subscribers as well, the actual customer count contending for bandwidth anywhere south of 100 percent penetration on the coax would produce a higher rate of throughput per premises than GPON offers on a 64-home split and, depending on the cable penetration ratio, would come close to matching the throughput of a 32-home split. Moreover, given the fact that GPON operates at 1.44 gbps in the return, use of a high bandwidth tier for 1 GigE return over coax would leave the two strategies on par in that direction as well.

GPON is presently being deployed by Verizon in ongoing FiOS buildouts and will eventually replace the previously deployed BPON (broadband PON) networks, which deliver downstream throughput of 622 mbps across 32 households. But, in a cautionary note to cable strategists, the PON evolution doesn’t stop at GPON, as recently demonstrated by Verizon with field tests of the IEEE’s XG-PON or 10G GPON platform, which is close to being finalized as a standard.

“From the earliest stages of the FiOS design, we knew we could repeatedly and progressively leverage the immense capacity of fiber to carry more and more data in support of customer applications,” says Mark Wegleitner, senior vice president of technology for Verizon. “Now we’re already working on the best way to take the next leap forward in capacity.”

Much as GPON is designed to operate over existing BPON infrastructure, XG-PON, delivering 10 gbps in the downstream and 2.88 gbps in the return, is designed to coexist with GPON, he notes. In one of two tests of the technology, which used gear supplied by Huawei, the XG-PON stream was merged as a separate wavelength onto an already-operating live fiber running FiOS with GPON at 2.5 gbps to provide a total downstream flow of 12.5 gbps. At the customer’s house, the XG-PON and GPON signals were received by ONT devices, showing that the network can simultaneously deliver the GPON and XG-PON signals on the same fiber at the two speeds, without any degradation in service. The other test was run over a separate fiber.

Verizon says it will soon test other vendors’ XG-PON solutions. Sources at Motorola say they have trials with unnamed customers planned this year, but that only about ten carriers worldwide are showing even “tire-kicking” interest in the technology at this early stage. Nonetheless, they project there will be broad global interest and the start of commercial deployments by late 2012.

The argument from many cable strategists, including Liao, is there is not likely to be any real need for that much bandwidth for a long, long time to come. With a 6 gbps downstream IP feed split across two-thirds of the customer households within a 125-home serving area, cable could deliver 75 mbps of dedicated bandwidth per household, even if every customer was on the network at one time. That’s enough bandwidth to support four unicast MPEG-4 HDTV channels together with aggregate Internet usage across all devices at 50 mbps. Needless to say, in real-world applications when only a percentage of customers are online or watching TV at any given time, the bandwidth available to each household would be much greater.

But to get to these levels of eventual throughput over coax plant or even to be able to support an initial hybrid topology where GigE operates in the downstream and return paths at the high RF tier, operators will have to spend money on tightening connections on the coax. “Most cable operators have decided a frenetic push to fiber-to-the-home isn’t necessary,” says Steve Craddock, who recently retired as senior vice president of technology at Comcast Cable and now is serving as executive vice president and CTO of PCT International. “But the problem is a lot of that plant needs to be tightened up and tied down. This is really important if you’re going to use the higher frequencies in the RF spectrum on outside plant or in the home.”

Recognition that solutions designed to meet these requirements are going to be paramount to cable’s future success was a big reason Craddock chose to go with PCT over “more sexy” options that were open to him, he says. The company’s “hardening-the-drop” products include high-performance drop cables, drop amplifiers and unique connector tools and components which, together, address the leakage and corrosion that are key contributors to HFC maintenance costs.

“In terms of operations and capital costs, the last mile is most of it,” Craddock says. “[PCT has] what it takes to create a hard drop, and that’s what cable needs.”

If the cable industry is going to compete into the future on the strength of a coaxial last mile, he adds, operators can’t be burdened with traditional cost factors. “If and when Verizon finishes building out their fiber network, the cable guys will have to reduce op ex if they’re going to stay competitive,” he says.

But if it’s going to take better connections and new drop cables like the environmentally hardened Triton cable recently introduced by PCT, not to mention deeper fiber extensions, just how much will cable operators be saving versus taking the all-fiber plunge? “That’s a good question,” Craddock says. “But a better question is whether you want to be digging up streets and pushing fiber through Aunt Amy’s petunia patch. Cable guys don’t want to go there if they don’t have to.”