Four or five years ago, when competitive pressures were leading many cable operators to consider offering telephony services over their networks, the process of upgrading cable networks to carry power for telephone services topped engineering discussions. Those conversations covered a range of issues, from the shift to 90 volts from 60 volts, to technician safety and the corrosive effects of passing power down drop cables.

But then the Internet happened, and along came high-speed Internet backbones like Excite@Home and Road Runner. The vendor community came alive and started to build cable modems. DOCSIS cable modems arrived. Digital video also burst on to the scene, with digital and advanced digital set-tops surging into the market at dramatically quick rates.

At about the same time-in the mid-1990s-competitive forces shifted. The telcos, as a group, backed away from distributing entertainment video over their networks. Freed from that threat, cable MSOs (with the exception of MediaOne and Cox) also retrenched, turning their attention to capturing data and digital video customers.

As a direct consequence, most MSOs halted the work they were doing on network powering. Why? Because when the telephony hype died down, so did the need to revamp the power architecture. After all, digital video and high-speed Internet access didn't require lifeline power. But the slowdown in outfitting broadband plant to carry power to homes is about to show the extent of that neglect.

Ma Cable is coming

It's impossible to discount, let alone ignore, cable telephony these days. It's what drove AT&T to buy TCI and MediaOne. Plus, the fact that six MSOs are now regionally organized and control 85 percent of U.S. cable subs is startlingly similar to the regional Bell operating company model, and what was known as "the seven sisters."

Telephony, in fact, is steering all of AT&T's network decisions, with digital video and cable modems serving almost as background items that will drive more customers to-you guessed it-telephony and bundled communication services.

AT&T is planning eight additional telephony pilots next year, some over cable, some via fixed wireless. On the cable side, AT&T next year will add circuit-switched phone service in Denver, Chicago, Pittsburgh, Dallas and Seattle to the 2,000 subs it already has in Fremont, Calif. (IP-based phone starts heavily in '01, AT&T executives say.) MediaOne rang into Massachusetts Oct. 13, launching digital phone service along the south shore area.

Cox is on the move, too. On Oct. 21, it started going after businesses in Orange County, Calif., from Newport Beach to San Clemente. It's a show of faith by Cox, because business telephony brings with it much more stringent technical and business considerations on the part of the MSO.

Power isn't present

That brings everything back to power. It's critical for cablephone, whether circuit-switched or IP, because phone service has to be up and running even when the power isn't, in order to compete effectively with incumbent telcos. The good news is that cable networks are in much better shape, technically, than they were a few short years ago. That's because much of the U.S. cable plant mileage has been expanded in bandwidth capacity and two-way activated to provide data service.

According to research conducted this summer by Paul Kagan Associates, MSOs as a whole will surpass the 65 percent mark for 750 MHz bandwidth by the end of this year; rising to 72 percent by the end of 2000. As for two-way plant, PKA estimates that 68 percent of U.S. cable subs will have access to it by the end of this year-a 20 percent increase from '98 two-way levels.

But there's bad news, too. The retreat by cable from telephony simply delayed powering decisions that should have been made previously. "Let's face it. When the Internet came, it was data, data, data—and powering issues got dropped like a hot potato," says Andy Paff, CTO of Lakewood, Colo.-based Worldbridge Broadband Services.

That means it'll be more complex, time-consuming and expensive to grow the number of current cablephone subscribers-about 300,000, nationwide, according to some estimates-into the millions of users.

"I work in the area where the hype keeps colliding with the reality of the existing network . . . and I think there's a long way to go," to resolve plant powering issues, says Tom Osterman, president of Seattle-based Comm/Net Systems, which provides on-site consultative services to several major MSOs on the issue of network powering.

Others well-versed in the nitty-gritty of plant upgrades agree. Paff says the powering issues can be resolved, "but the idea that there's 80 percent telco-ready plant out there? Forget it."

Local network assessments: HFC telephony vs. LEC.
How much power?

To provide phone service that competes with incumbent LECs on service, MSOs need to provide roughly eight hours of service following an outage. How does that translate into power design? Osterman submits that an admittedly extreme model-assuming 100 percent telephony penetration on a 1,000-home node-could require as much as 5 kilowatts of total power, assuming each house-mounted network interface unit (NIU) eats up just 5 watts.

That low power draw is available in contemporary circuit-switched cablephone gear, but not in cable modems or integrated set-top/modems. In-home devices like modems and set-tops have historically not had to operate when the power was out (because the PC and TV would be out, too), so there's been little pressure to reduce their power demands. As a result, most of those devices still operate at 25 watts or higher. And it could get worse as IP telephony becomes a reality and gets integrated into those devices.

And there's another kicker: Peak loading. To get an accurate representation of power draw, calculations must also include the possibility of every home within the 1,000-home node going off-hook at the same time. Under that assumption, the total power draw for the node could go as high as 7 to 8 kilowatts, says Osterman. "That may not seem like much, but it's an order of magnitude over what's used now," he says. The issue gets even more acute if an MSO needs to provide 5–7 kilowatts for eight hours. "That's a lot of batteries, and (requires) a big cabinet to do it," Osterman says.

Power of three

There are three basic methods to go about powering a network: Decentralized, centralized or home-powered. Most cable operators use decentralized power, meaning they install three or four power supplies (standby and non-standby) per neighborhood to provide juice to amplifiers and fiber nodes. But AT&T is about to buck that trend. Tony Werner, CTO of the MSO, says he'll switch to centralized powering in its major phone markets-a giant departure from its historical strategy.

"We think we can get to about a four-to-one reduction in the amount of power supplies we need by going to centralized power," Werner says, "which reduces maintenance issues dramatically.

Centralized power means finding the easement room to build large cabinets to house a generator, which keeps the phone live when the power is out. Including fuel (propane or natural gas), equipment cost and land acquisitions, centralized power for a single 1,000-home node could cost as much as $50,000, says Osterman.

Werner concedes that "the downside of centralized powering is the capital cost," and says AT&T is working to identify a balance between decentralized and centralized power, on an area-by-area basis. AT&T is already using centralized power in its Hartford, Conn. telephony market-active since '96-and says newer central power nodes are smaller, quieter and cooler. "The size of the new units is about 2 feet deep, 4 feet high, and 4.5 feet long," Werner says, adding that, "we're also developing another device that's very small-about 18 inches deep and high, and about 2 feet long," for installation in space-constrained areas.

Batteries are another option, and provide a quick solution to the power conundrum. A typical home-powered solution, in its simplest form, includes batteries and a power converter to provide power and backup time to the NIU. MSOs like the fact that the units are available, but find ongoing maintenance issues troublesome. Plus, depending on climate, batteries could need change-outs as frequently as every year.

"Clearly, it gets you into the market very quickly, and eight hours of backup is substantial-but there are install issues," Werner says.

No kidding, says Worldbridge's Paff: "Customer-provided powering has just got to go away," he says, quipping that while Worldbridge does battery install work for MSOs, he'd love to give it up. "You have to spend a day getting into the garage to get at the AC power-it's just a lot more expensive than it could be, with network powering. Frankly, it's god-awful."

Osterman agrees that install issues, particularly for home powering, aren't inconsequential. "To install these things is a real hassle," he says, explaining that oftentimes that entry point for the coaxial drop is located somewhere other than where the AC power unit resides.

Resolution could be costly

Whatever the solution-decentralized, centralized or home-powered-sending power down the cable network won't be cheap. One benchmark came from the AT&T/Time Warner telephony joint venture (now delayed until AT&T's absorption of MediaOne is completed), which stipulated that AT&T will pay Time Warner $10 to $15 per home passed for power-related gear, plus backup power for TWC headends.

That may not be nearly enough, industry experts say. Paff says tech staffers should somehow muster the courage to "tell their bosses and their boss' bosses" that it could take as much as $50 to $300 extra, per home-passed, to get plant telco-ready."

Then there are training issues. Most MSOs are cultivating core groups of technicians to handle telephony installations. But a lack of industrywide craftsmanship and install guidelines will start showing quickly, especially with AT&T's expansive plans. "Sure, there are ways to do this right-if you don't run a staple through the center conductor of the drop, if you pay attention," says Paff. "The reliability can be there, but it'll taking some doing."

Payoff is there

As far as AT&T's Werner is concerned, the technical issues pale next to the potential opportunity that telephony provides, and early technical results show remarkable reliability. In its Fremont trial, AT&T is already beating incumbent telco Pacific Bell on a long list of parameters. Werner set up the experiment using an ITU (International Telecommunications Union) telephony specification as a reference, and found that 100 percent of AT&T's lines met a "high-quality, toll grade" barometer, to 50 percent of PacBell's. Similarly, 100 percent of AT&T's lines met a "high-quality local call" grade, to 84 percent of PacBell's.

AT&T BIS engineers conducted the tests as new cablephone homes were installed. They ran the tests on PacBell's lines just before they disconnected them to hook up cablephones; AT&T's cablephone connections were tested directly afterwards.

AT&T also tested for dial-up modem connect speeds over its cablephone connections, and found that when connected to a 56 kbps phone modem, it actually linked at 49 kbps 92 percent of the time. Only 43 percent of PacBell's lines connected at that speed.

Tests for dial tone delay also came in consistently higher on AT&T BIS' hybrid fiber/coax plant than on the incumbent telco's local copper loops, with AT&T trumping every PacBell benchmark.

Long road ahead

The upshot of this will be a much sharper focus on network power issues, probably starting by the end of this year. MSOs serious about telephony will spark power design work, which could include node segmentation beyond today's average of 500 to 600 homes passed. But for the foreseeable future, we're unlikely to see a great amount of consensus on the subject.


Chip suppliers on the scent

Cable-delivered telephony will also emerge as an IP (Internet Protocol) application on cable modems and "combo boxes," like General Instrument's DCT-5000-none of which were designed to be frugal with power.The reasoning is sound. No power means no PC or TV, so it's unlikely that accompanying cable modems or set-tops would need to be functional.

But that all changes, if and when operators settle on IP telephony as a lifeline communications service. If ops need to provide juice to cable modems, set-tops or other IP "residential gateway" types of units, they'll need as much as 25 watts for each unit-a far cry from the 2 to 5 watts in today's circuit-switched NIUs.

Not to worry, says Broadcom CTO Tom Quigley. Broadcom, with 95-plus percent market share on chips used in both cable modems and set-tops, is already at work on ways to reduce the power draw in both devices.

The key, perhaps not surprisingly, is chip consolidation, which helps reduce the power draw in current chip-to-chip signal paths. Plus, Quigley says, CMOS chips are trending toward lower power draws, with smaller and smaller geometries. Chips sized at 0.18 and 0.22 microns, for example, run on 2.2 volts, internally, which helps.

Also on tap: A way to make some circuits within a set-top or modem go into "sleep mode" when they're not being used, much like today's PC monitors.