|
Jim Barthold
Unless you've been working on a federal grant studying the nocturnal habits of fishermen angling for striped bass in Papua, New Guinea, you've probably heard about cable's broadband pipe.
It's the fiber-fortified, eight-lane superhighway that's been built to bring you every imaginable future service, from hundreds of TV channels to lightning speed Internet access to telephone service.
What you probably haven't heard too much about is the copper-paved, two-lane, potholed, cowpath that brings those signals back. Called the upstream, this crucial part of tomorrow's cable plants is starting to get some attention - quite possibly just in time.
"It's definitely a hot area," said David Grubb, marketing VP for General Instrument Corp.'s transmission networks group.
It's hot because, while the downstream is a fire hose carrying signals in a bandwidth range from 50 to 750 or even 860 MHz and above, the upstream is generally a garden hose, restricted to the 5 to 40 MHz range. That's OK when the only thing coming back is the occasional burst from a pay-per-view order or a cable modem. It's not OK for telephone and large data files.
"I guess I wouldn't call it a problem," Grubb hedged. "People may need to go and add some equipment, like frequency stacking or WDM (wave division multiplex) products to add to that return capacity, but it's not like a complete reset for them."
It was enough, though, to have vendors salivating and showing off ways to make the most out of that bandwidth at last week's NCTA Show. Among the solutions, as Grubb suggests, are ways to consolidate the multiple signals coming from increasingly smaller nodes to hubs and then moving those signals from the hubs to headends.
"People are really getting around to being focused on it," agreed Paul Connolly, VP-marketing and network architectures for Scientific-Atlanta Inc.
S-A recently introduced a way to digitize the signals in the node before returning them to the hub. Digitization is aided, in that case, if the signals also undergo further consolidation through dense wave division multiplexing (DWDM).
The reverse problem is directly related to what's coming down the forward path. The more interactive services an operator wants to provide, or narrowcast, to subscribers, the smaller the number of homes needed in the node.
Node sizes, therefore, are dropping to fiber-fed areas of under 250 homes, and, in some experimental cases, as few as 50. These then use increasingly smaller coaxial cables to feed the subscriber homes.
"If too many people get on that one copper cable, it's going to slow down," warned Anthony Ley, chairman/president/CEO of Harmonic Inc. "That real limit is set by that piece of wire that connects to your house."
Ley pointed out that smaller nodes increase the number of signals coming back in the 5 to 40 MHz range.
"I have to bring all these 5 to 40s all the way back to my headend," he continued. "What's the most efficient, economical way?"
The best way is to make use of the fiber that the system has run deep into the network. Fiber's bandwidth, Ley emphasized, is virtually unlimited. But fiber is not cheap.
"You obviously want to start concentrating (the signals)," he explained. "One way is to try to digitize it. That helps do two (signals) a line. With DWDM we can do either four or eight (on top of that)."
In a traditional return architecture, the copper runs back to the node; fiber runs to the hub and from the hub to the headend.
"The most economical system is if you run it all the way back to the headend, because that's where you want the telephone switch," he said. "The fundamental limit is set by the coax cable. After that, it's how much fiber you have will determine what you do."
Frequency stacking is not the optimal way to go, said Bill Moore, VP/GM of fiber optics for Ortel Inc., which builds the lasers used to drive the signals through the fiber.
"It's incredibly expensive," he said. "It's actually cheaper to do it WDM."
In either case, lasers come into play and, with the exception of compact disc players, lasers also are not cheap.
"If it was point-to-point, then every node would have a turnabout laser and everything would be a home run," he said. "But the economics don't work out that way."
Mathematically, four lower powered 4 milliwatt lasers cost a lot more than a single high-powered 16 milliwatt laser. Moore recommends splitting the 16 milliwatt unit four ways.
"It's not that big of a deal, but some of it determines your reverse path architecture," he explained. "In the meantime you say, 'I have these hanging out there on one laser. How do I get the stuff back?'"
Via WDM, he suggests.
Dense WDM was first adopted by TCI in a move to optimize the use of its networks. Other MSOs are now climbing aboard the bandwagon.
"One system we announced was the Saratoga system of Comcast," reported Harmonic's Ley. "There they ran out of fiber for the return path. We put in DWDM; no problem."
For the most part, vendors are hanging their hats on DWDM, either from the node to the hub or all the way back to the headend. But, they concede, figuring ways to improve the return path is like adding shoulders to a two-lane road. It makes it more comfortable, but it's hardly a highway.
"Certainly, from an optical point of view, it's easy to deal with returns," said Tom Tucker, product manager of headend fiber optics at Philips Broadband Networks Inc. "We're kind of working toward providing a migrational solution."
There is time for that migration to take place. The heaviest traffic now coming back on cable systems is typically from cable modems. That, in turn, is generally short bursts seeking out Web pages and carrying e-mails. As more subscribers join the cable modem ranks and start to use the products' speed, their upstream demands will increase.
Cable modems use quadrature phase shift keying (QPSK) as the upstream modulation scheme. As demand increases, operators may employ more space saving schemes such as 16 quadrature amplitude modulation (QAM).
"You would use a higher modulation scheme to conserve the bandwidth and increase the data rates," explained Venktash Mutalik, Philip's advanced fiber optics staff engineer. "The only way you can do that is by improving the performance of the links."
Naturally, the most bandwidth-consumptive migration scheme is the one that covers telephony. Voice has no bursts; when you're on the phone you're on the network.
"The hybrid fiber/coax (HFC) architecture has been around for about five years. It was conceived with the idea of sending television signals to everybody," Ley recalled. "Four years ago, we thought about Internet penetration and it was like 1% is going to be a good number, 2% is going to make us a lot of money. Now you'd fall down laughing if you said 1% or 2%."
That also was before telephony.
"Now, if you talk about 30% penetration for telephony and modems, you need more return path," Ley said. "Basically, it's not a problem, it's just now you have to buy more equipment and think about how to deal with the fact that you have other constraints, like a limited number of fibers."
Even the vendors, as anxious as they are to sell products, concede that the bandwidth won't run out overnight.
"I think there's time," agreed Paul Whittlesey, senior director of product management for Antec Corp.'s network technologies group. "There's so many different ways to provision the network."
"It's (the cable industry) getting itself prepared for heavy use," agreed Ortel's Moore. "They're not behind the power curve."
And, he said, the industry has good reason to pursue the solution post haste.
"Your cable company is a monopoly," he pointed out. "It has 90% of the homes passed; 65% of the subscriptions. How do they get either their subscription rates up or another revenue stream? Cable modems and telephony are a great way."
It's just that they demand more of the return path.
"It may be bursty today, but it's going to change," he warned. "If you're going to install a two-way system, you better be prepared."
Back to this issue
|
