Need For Speed

Cable's newest broadband hot-rods — systems capable of blasting data down to subscribers' homes at more than 100 million bits per second — still need to get their tires kicked by major operators before they're ready to hit the street.

The next-generation of cable-modem technology, CableLabs' Data Over Cable Service Interface Specification 3.0, should be ready for early deployment in 2008. Some executives, though, said the “wideband” technology won't hit its stride until at least mid-2009 and cautioned that initial DOCSIS 3.0 modems will be pricier than expected.

By far the specification's most attractive feature is the promise of faster downloads.

And in some cases, 100-Megabit per second-plus cable modems can't arrive quickly enough. For the near term the key competitive threat is Verizon Communications' FiOS fiber-optic Internet service, which the phone company is offering at up to 50 Mbps in six states.

“I think DOCSIS 3.0 is the cable industry's answer to FiOS,” said Rouzbeh Yassini, an early cable-modem pioneer who is CEO of venture-capital firm YAS Broadband Ventures.

Added John Sweeney, Scientific Atlanta director of product strategy and management: “It's a clear shot across the bow that cable can do the kind of bandwidth FiOS is marketing.”

DOCSIS 3.0, in addition to ideally helping operators retain high-value triple-play customers with far faster Internet pipes, also will boost the bandwidth cable can provide to small and midsize businesses, particularly for upstream applications like Web-server hosting. Moreover, 3.0 could eventually provide a boulevard for video content delivered via Internet Protocol.

But until cable operators deploy DOCSIS 3.0 — called wideband by the marketing inclined — they won't be able to match the fastest FiOS offerings.

Comcast, for one, plans to have deployed next-generation modems in systems serving about 20% of its homes passed by the end of 2008, chief technology officer Tony Werner said last month at the CableNEXT conference. Comcast won't detail which markets will be targeted or the services it may offer, but the initial areas are almost certain to be neighborhoods where FiOS Internet is currently available.

Cable deserves credit for anticipating the need to crank up bandwidth when it began to develop the channel-bonding features about three years ago, said Mike Paxton, an analyst with research firm In-Stat, a sister company to Multichannel News.

“Cable saw the big picture. They said, 'By 2010 we need to have this widely available,'” he said. “When everybody's pushing in the right direction good things can happen.”

Ultimately, Paxton noted, a fiber-to-the-home network like FiOS is more efficient than a hybrid fiber-coax network. “But by developing this technology,” he said, “it gives cable a pretty potent weapon to respond to what they're seeing from their competitors.”

At this point, Paxton said, cable-modem suppliers are anticipating that volume shipments of DOCSIS 3.0 units will begin in September or October 2008. That's about a quarter later than many previously expected, but he said such slippage is “common enough when you're talking about a new technology like this.”

“I don't see anything with red flags in terms of rolling out the products,” Paxton said. “But maybe there was a little bit too much enthusiasm about how quickly this would happen.”

Sweeney said Scientific Atlanta is in “a ramp-up mode” to produce its first DOCSIS 3.0 modems, with the initial models expected to roll off the production line in the early first quarter of 2008.

“The big issue will be getting orders in place to justify production,” he said. The question will be how long it will take cable operators to evaluate the new products in their labs, a process he said could take up a good part of the first quarter.

Yassini believes operators won't begin wide commercial deployments of DOCSIS 3.0 until mid-2009 at the earliest.

Historically, he said, for a new cable technology, it has taken about 18 months to go from factory production to field implementation. “If you can do it more quickly, more power to you,” he said.

“Cable operators have to keep making money and they're going to continue deploying what works,” Yassini added, referring to current-generation DOCSIS 2.0 cable modems.

With the first production units close to shipping, it's also becoming clearer that DOCSIS 3.0 modems will be more expensive than perhaps previously anticipated.

The street price of a 3.0 modem will be around $100, Sweeney estimated, adding that quantity has a big effect on the per-unit price. Current DOCSIS 2.0 modems cost an average of $65 to $70, according to In-Stat research.

“It's not the $10-to-$20 delta [between 2.0 and 3.0] the cable operators were encouraging the industry to hit,” Sweeney said.

DOCSIS 3.0 delivers higher bandwidth by virtually combining several 6-Megahertz channels in a digital cable system, to act as if they were one. The specification requires modems to support four downstream and four upstream channels, theoretically capable of sending data at around 160 Mbps down and 120 Mbps up.

One of the reasons DOCSIS 3.0 modems cost more is that to be able to bond four channels up and down, they require much more memory than their 2.0 cousins. For example, while a 2.0 modem may have needed 8 megabytes of memory, a 3.0 unit will require around 32 MB.

As Sweeney explained: “You have to park the packets in memory and then reassemble them. In the data rates we're working with — 160 Mbps — that's a lot of bits you have to reassemble.”

Over time, vendors will improve modems so that future models should require less memory, said Peter Percosan, executive director of broadband strategy for Texas Instruments' residential gateway business. TI is supplying DOCSIS 3.0 chips to Arris, Scientific Atlanta and Motorola, among other vendors.

“The initial focus was functionality and stability,” he said. “There's a lot of opportunity to shrink the memory footprint.”

In general, the DOCSIS 3.0 specification is much bigger and more complex than the preceding generations, said Tom Cloonan, chief technology officer for Arris's broadband division.

The entire suite of specifications DOCSIS 3.0 encompasses — when printed out — yield a stack of paper more than 6 inches thick, whereas DOCSIS 2.0 was closer to 3 inches, according to Cloonan.

“I don't think there's ever been a delta this large,” he said. “There are some anchor features like channel-bonding that everyone knows about. But if you look at the list it's a mile long.”

Other major features of DOCSIS 3.0 include support for multicast IP video, Advanced Encryption Standard security and IPv6, the next-generation Internet addressing scheme that provides an astronomical number of unique addresses.

The large number of DOCSIS 3.0 features prompted CableLabs to establish three tiers for certifying cable-modem termination systems: bronze, which will provide downstream-channel bonding; silver, which adds upstream bonding and other features; and full. The idea was to allow 3.0 CMTS systems that support the most desirable feature — faster downloads — to get to market more quickly.

This month, CableLabs is completing certification testing of the first wave of DOCSIS 3.0 gear. Vendors that have submitted cable-modem and CMTS products for the tests include Arris, Cisco, Motorola and Scientific Atlanta.

“The first time through [on a new specification] is always the hardest,” CableLabs chief technology officer Ralph Brown said. “As always with products that are newly coming to market, there are software bugs and things that have to get ironed out.” In addition, Brown noted, the 3.0 certification tests require making sure the systems are fully compatible with earlier versions of DOCSIS.

Overall, the certification process has been “business as usual,” according to Brown. Leading up to this first DOCSIS 3.0 wave, he added, CableLabs sponsored several interoperability events for vendors, including a three-week event in September.

After testing is completed, a certification board comprised of cable operator executives reviews the results and issues recommendations on whether individual cable modems or CMTSs comply with the specification. The certification board may request that a product undergo additional retesting after a vendor addresses any issues.

“The results — passing or failing by the vendors — are a matter of how prepared [the vendors] are and how compliant they've developed the products to meet the specification,” Brown said.

That means CableLabs-certified DOCSIS 3.0 products may be ready as early as January, or it could be a few months longer than that.

“DOCSIS specifications are complicated, and sometimes it takes more than once to get through certification. Early specs can be read differently,” Cloonan said. “We're hopeful [about receiving certification], but we're realistic.”

TI's Percosan expects to see the first DOCSIS 3.0-compliant embedded multimedia terminal adapters (eMTAs), which combine voice and data services, show up in the next two CableLabs certification waves in January and February.

After the current wave is completed, he said, TI and its cable-modem vendor partners will be “heads down” in developing eMTAs that support the PacketCable quality of service specification to ensure phone calls aren't disrupted.

“Deploying a best-effort data modem is relatively simple,” he said. “If the quality-of-service features aren't there, there's a big problem for voice modems.”

Meanwhile, perhaps because of potential delays with full DOCSIS 3.0-certified products, some cable-modem vendors believe there's still an opportunity for prestandard “2.0-plus” gear that provides channel-bonding using existing CMTS infrastructure.

Paris-based Thomson, for example, in late November announced the availability of the DCM465 cable modem, which can bond up to three downstream channels to provide download speeds of 100 Mbps with a software upgrade to a CMTS platform.

Charles Roederer, Thomson's general manager of cable for the Americas, said the company is also developing a fully 3.0-compliant modem, set for delivery in the second half of 2008. But the 100-Mbps “2.0-plus” unit is for cable companies that want to provide higher bandwidth tiers in the short term.

“We developed this at the direction of some of the operators,” he said. “In the event that DOCSIS 3.0 has any road bumps, they want to have something in their back pocket in case FiOS is taking a bigger chunk of market share.”

The DCM465, based on a Broadcom chip set, is “forward-compatible” with DOCSIS 3.0 CMTSs. Thomson expects to initiate trials of the cable modem in the first quarter, but the vendor has not disclosed prospective customers.

Arris has had a “2.0-plus” cable-modem offering since 2005, Cloonan said, but the volumes of the product have been relatively small. “I think they'll continue to be small,” he said. Operators are “not going to deploy a lot of that because it's legacy equipment they'll have to support for years. Most people are waiting for the standards-compliant product — and it's on its way.”

Besides, Cloonan said, another major reason cable operators want to move to DOCSIS 3.0 is to be able to provide more bandwidth to existing customers.

Current-generation CMTSs provide a fixed ratio of upstream and downstream channels. Arris's C4, for example, provides DOCSIS 2.0 access modules with 2 downstream and 12 upstream ports on each line card. Some of the new CMTS designs for DOCSIS 3.0 will permit flexible downstream-to-upstream port ratios by decoupling the downstream and upstream channels.

That will allow an operator to mix and match: Arris plans to sell separate 12-upstream and 16-downstream line cards for the C4, with support for up to 16 cards in a single CMTS chassis.

Operators “will be buying more channels because they're going to be increasing the bandwidth per subscriber,” Cloonan said. “Even if their subscriber base stayed the same — which it won't — they'd still need more bandwidth to keep those subscribers happy.”

But packaging up juicy higher-bandwidth tiers remains one of the leading business drivers for DOCSIS 3.0. Operators are generally “looking at offering 50 to 100 Mbps,” said John Chapman, a research fellow in Cisco Systems' cable business unit.

Once cable systems reclaim more spectrum — for example, by retiring analog TV channels that each consume 6 MHz — even higher speeds will be in the offing with DOCSIS 3.0, because the specification allows for bonding together as many channels as a given cable modem can handle.

“You can double or triple bandwidth when you go to all-digital plant,” Chapman said.