Swimming upstream

When the concept of delivering telephone service over cable TV operators' hybrid fiber/coax networks first came up several years ago, the cable industry seemed eager for the telecom reform gun to go off.

Since then, realizations about HFC plant limitations have caused much of the industry to stall its telephony plans before the race ever started.

Cable operators once announced that all they had to do was upgrade to two-way and install some return amplifiers to carry voice over their networks, but they now find themselves grappling with their Achilles' heel: the thorny return path.

HFC networks have always had a noisy return path, but until now that hasn't been a problem for cable operators, says Tom Hall, senior advisor of broadband networks at Northern Telecom.

The HFC network structure ensures that once noise gets into the system, it's impossible to get it out. Because the return path consolidates the signals from every home on the network and delivers it to the headend, it funnels any noise introduced into the system upstream into the headend (Figure 1).

And noise has ample opportunities to introduce itself into the cable plant. Most prevalent is narrowband ingress, which is noise interference caused by household appliances.

Consumers hear the ingress effect as static "pops" that interrupt the call. If the ingress is bad enough, it may even terminate the call. Complicating this is the fact that more remote homes deliver a weaker signal back to the headend and can be affected by ingress injected further up the line.

Cable operators can take steps to minimize ingress and its effect on the plant. Proper maintenance of the cable plant can minimize both narrowband ingress and impulse noise, which is caused by flaws in the outside plant.

"The plant is made of real-world materials, so it can corrode after a time," says Tim Frendberg, product manager at Scientific-Atlanta. "Cable operators need to be vigilant about maintaining connectors and other places where noise can get into the plant. That can improve the noise situation.

But cable operators have no control over their wires once they go past the customer's front door. Moreover, even homes that do not subscribe to telephony service can contribute to the noise level in the plant.

One way to address this is to install block filters between the drop cable and the house cable at homes without telephony service. Some vendors have taken that approach a step further. In Ericsson's broadband HFC access system, the network interface unit at the side of the home functions as a filter, effectively isolating the home from the access network.

But even with filtering capabilities, ingress won't go away with time.

Noise can be such a problem that it's impossible to find more than 1 MHz of bandwidth clean enough to carry a signal, says Ron Smith, vice president of operations at Motorola's Multimedia Group.

Moreover, increasing penetration and adding subscribers means that more noise can penetrate the network, says Nortel's Hall.

"Every time you add an additional house or subscriber, you have the opportunity to introduce a new noise footprint," Hall says. "You just can't keep all the noise out.

To QAM or not to QAM How to cope with the remaining noise after all possible precautions have been taken is today's hot topic among telephony-over-cable vendors. In developing an upstream solution, vendors need to decide on a digital modulation technique and a multiplexing protocol.

Choosing a modulation scheme entails striking a balance between the most efficient delivery of the signal and immunity to noise. One of the most common modulation schemes is quadrature phase shift keying (QPSK), a version of phase shift keying that uses four phase states.

Another modulation technique is quadrature amplitude modulation (QAM), which exploits the ability of different quadrature components of a single-frequency carrier wave to carry two input signals. QAM is available in several different versions, the most basic of which is 4-QAM.

QPSK and 4-QAM are comparable in terms of efficiency and noise tolerance. Both can operate well in systems with a fairly low carrier-to-noise ratio-which refers to the carrier's strength in relation to the network's noise-meaning each technique is well-suited for noisier environments.

QPSK and 4-QAM are also similar in terms of efficiency, measured in bits per hertz, or the amount of telephony traffic that a frequency can handle. Both schemes feature bits-per-hertz rates in the 1.5 to 2 range.

Raising the number of phase states in QAM modulation increases the bits-per-hertz rate, allowing a given frequency to carry more traffic. For example, ADC Telecommunications' Homeworx system with 32-QAM has an efficiency level of 5 b/Hz.

The tradeoff is that a higher bits-per-hertz rate results in a higher carrier-to-noise ratio, meaning that the system is less able to tolerate noise.

"To run a higher-order modulation scheme, you have to address the quality of your plant," says Hall. "The plant needs to be cleaned up a good bit more than it would need to be with QPSK [or lower-level QAM] because you're talking about a much higher sensitivity to noise.

While most of the industry agrees that the modulation scheme of the future will be a higher level of QAM, many have opted to begin offering QPSK or low-level QAM and migrate up as they proceed along the learning curve and as cable operators clean up their plant.

"Everyone started with a low-density scheme because no one knew what to expect," Hall says. "Once manufacturers get a better understanding of the upstream path, modulation schemes will go up higher.

One vendor is already putting that migration path into effect. Tellabs' Cablespan 2300 product currently uses QPSK as its modulation scheme, although the company is also developing a version that uses 16-QAM. It will be available in the first quarter of 1998.

"Cable operators that have a plant that is conducive to supporting higher modulation schemes can use [the 16-QAM solution]," says Wayne Partington, group manager of product marketing for network access systems at Tellabs. "If the plant is not optimized to support the carrier-to-noise performance required for higher modulation schemes, they can get into the business using QPSK and then can migrate to a higher-order modulation capability.

Cable operators can use other ways to increase the amount of traffic that the return path can carry. Dynamic bandwidth allocation, otherwise known as RF concentration, allows network operators to accept more subscribers than it has bandwidth to support. The key is to provide each subscriber with only the bandwidth that he or she needs at that minute, as opposed to assigning a certain amount of bandwidth to each subscriber whether or not it is being used.

A key enabler of dynamic bandwidth allocation is a new switch interface standard, GR303, which allows network operators to set up and tear down connections to the end user, says James Crowley, director of advanced planning in the access products division of DSC.

GR303 provides the ability to signal all the way to the network interface unit at the side of the house. This ensures that the subscriber gets a connection when he picks up the phone, and it allows network operators to put bandwidth in a "pool" to be used when needed.

The traditional TR-008 switch interface, on the other hand, is not well-suited for dynamic bandwidth allocation because it typically does not signal all the way down the network. This means that the switch could set up a call but fail to allocate the bandwidth all the way to the home.

While TR-008 is suitable for limited deployments, cable operators will run into problems with the interface when they begin to increase their penetration and capacity because bandwidth capacity must be dedicated to each customer, says Crowley. Cable telephony vendors are currently working with major switch manufacturers to incorporate the GR303 standard into their products, he says.

FDMA or TDMA? Another key decision that vendors must make is how to multiplex the signal in the upstream path. Cable TV systems traditionally use frequency division multiple access (FDMA) downstream, which means that a single channel is assigned to each carrier. Each carrier is modulated to varying frequencies so they can be transmitted simultaneously.

Another alternative is time division multiple access (TDMA), in which a 2 MHz channel band is shared among approximately 32 DS-0s via an assigned time allocation on the carrier.

FDMA proponents cite the simplicity of the technology. "There's less technical risk in deploying FDMA," says Tim Frendberg, product manager at Scientific-Atlanta. "We know it will work on a cable system because it's what works there today.

With TDMA, each carrier must be sent at a certain time to ensure that it does not collide with other carriers, Frendberg says. In addition, TDMA has ranging requirements, meaning that signals from homes farther down the network take more time to reach the headend than signals from closer homes. Because of this, cable operators must perfectly synchronize the system, which can be difficult in a cable plant, he says.

"Temperature, routine maintenance or a reconfiguration of the system due to a broken fiber could throw off the synchronization," Frendberg says. The result could be interrupted or dropped telephone calls. FDMA does not pose that risk because every channel has its own frequency, he says.

Another advantage of FDMA is that the technology is more scalable than its alternatives. Dealing with increased telephony penetration or adding data services to a system using FDMA is as simple as reallocating channels, says Wyley Robinson, technical manager at Lucent Technologies.

"Because FDMA doesn't care what's on another frequency, a variety of different services can coexist," Robinson says. "It's different from TDMA in that you don't have to have some amount of coordination from one frequency to another to avoid interference. Data or additional telephony channels can just coexist peacefully with other frequencies.

However, TDMA proponents argue that their technology is better than FDMA for "bursty" telephony traffic.

"TDMA lets you share resources between a lot of different users who don't send traffic, all the time," says Ingmar Dahlqvist, HFC business line manager at Ericsson. "When you have bursts of traffic it's better to share in the time domain rather than the frequency domain.

And FDMA can actually contribute to noise in the network, says Tom Mitchell, director of advanced product development at Arris Interactive, a joint venture between Antec and Nortel.

Remote transmitters are constantly active in FDMA systems, which can cause a buildup of emitted out-of-band noise, Mitchell says. With TDMA, however, each upstream transmitter is active only for the duration of its allotted time slot, which on average is only 4% of the time.

A matter of perspective One major advantage to FDMA is that it allows the spectrum to be more "finely sliced" than TDMA does.

Dividing the bandwidth into narrow carriers means that any given carrier is less likely to be hit by noise if there's interference in the return path. Scientific-Atlanta's FDMA-based CoAxium system, for example, features 50 kHz carriers, while most TDMA-based systems slice the return path spectrum into 1.5 to 2 MHz windows.

However, FDMA requires complex filters between channels to prevent interference, which results in less-than-efficient bandwidth use. A technology adopted by ADC Telecommunications, orthogonal frequency division multiplexing (OFDM), eliminates that problem.

By constructing the wave forms orthogonally, OFDM eliminates the need for filters between tones, allowing the tones to be stacked together so closely that they overlap, says Hal Roberts, senior principal engineer for ADC.

And because it doesn't require those guardbands between channels, OFDM allows the bandwidth to be divided into 9 kHz windows, which are even smaller than those offered by most FDMA systems, Roberts says.

"We decided that the best way to deal with dirty upstream spectrum was to slice it up as fine as we could," he says.

Another vendor has taken OFDM one step further. DSC is using OFDM technology combined with discrete wavelet multitone technology (DWMT) from Aware.

While regular OFDM uses digital Fourier transforms as the modulation engine, DWMT uses digital wavelet transforms, which allows for even smaller subdivision of the bandwidth, says Rick Gross, lead project manager for the HFC program at Aware.

The result is a 4 kHz carrier, which is far smaller than the 2 MHz carrier of most TDMA systems and even significantly smaller than the carrier size that traditional OFDM offers, says DSC's Crowley.

"[Aware's] technology allows us to subdivide the spectrum into very small segments that we can deal with individually," Crowley says. "We can easily notch out signals being hit by interference without affecting large portions of bandwidth.

While the obvious benefit of a smaller channel is that it's less likely to pick up interference, Crowley and other proponents of FDMA-based technology argue that smaller carriers are an important enabler of another technique vendors use to combat upstream noise: spread spectrum technology, otherwise known as frequency agility.

Both TDMA- and FDMA-based solutions use frequency agility, a technique in which the carrier that is hit by interference "jumps" to another channel. Without frequency agility, the carrier is forced to absorb the impact of the noise, which can result in static, interruptions and, if the interference is bad enough, termination.

"When situations like ingress arise, we have to be able to react to them, particularly when we're talking about high-reliability services like telephony," says Tellabs' Partington. "If the interference is creating problems with voice or data transmission, then we have to be able to move the payload accordingly and reallocate those carriers into another part of the spectrum.

But the smaller channels that FDMA enables make frequency hopping more efficient, says Scientific-Atlanta's Frendberg.

"If you're trying to get across the city during rush hour, would you rather be riding a motorcycle or a bus?" he asks. "When you're talking about 1.5 or 2 MHz channels, it's harder to move those relatively big blocks around than if you're talking about a 50 kHz channel.

An eye on the network Regardless of which combination of modulation and multiplexing they use, all vendors agree that network management is key to reliably delivering telephony service over an HFC network.

Scientific-Atlanta's CoAxium product, for example, allows the NIUs on the side of the house to communicate directly with the headend, letting the network operator identify specifically which calls are encountering interference. CoAxium also records whenever a call jumps from one frequency to another and which NIU was involved, allowing the operator to correlate the information to look for systematic problems.

Similarly, Nortel's Cornerstone product automatically samples the return path every two and a half seconds and can display that data in a three-dimensional form that shows frequency, time and amplitude.

Cornerstone can also help cable operators identify which nodes are contributing a lot of noise to the system, which could be resolved by simply reallocating resources within the network, says Arris' Mitchell.

"This tells us we had a frequency hop at 2 a.m., so we can look at the return path spectrum between 1:30 and 2:30 a.m. and see what happened to cause that hop," Mitchell says. "One of those charts is worth 10 truck rolls."

Even as cable telephony vendors are developing ways to conquer return path problems, the cable industry is having second thoughts about delivering telephone service over its hybrid fiber/coax networks.

Cable operators are finding that ramping up to offer telephony over their systems is like opening a box to find another one inside-every time they seem to resolve one issue, another one pops up to take its place.

For example, while cable operators have finally acknowledged that grooming the return path to offer telephony is a tricky business, they're finding out that they need to revamp their back office systems as well.

Customer management and billing becomes much more complicated when multiple systems operators branch out beyond one-way video services, says Mike Huseby, partner at Arthur Andersen's global communications and entertainment group.

"Cable operators have a lot of work to do on their operations support systems," Huseby says. "They need to be able to bill for one-stop shopping and gather customer intelligence. Their systems are not at a point where they can support these things.

"The back office stuff is definitely not trivial," agrees Chuck McElroy, vice president of residential broadband services at Cox Communications. "There's a lot of network management aspects associated with the back office of telephone operations that have to be tied to operations support systems. We're learning as we go how complex the support of telephony services is.

Regulatory issues are also turning out to be a stumbling block. With the Federal Communications Commission's interconnection order still tied up in court, MSOs are reluctant to enter a market governed by nebulous rules.

"There's still a great deal of unclarity in the regulatory structure for the competitive landscape," says a spokesman for Time Warner Cable, which has cut back its residential telephony rollouts. "Until that is resolved, we don't think that it's a wise thing to make multimillion dollar investments to provide residential telephone service when we don't know what kind of financial dynamics will drive that business.

And with its financial situation precarious these days, the cable industry is questioning whether directing its limited resources at telephony is the thing do. The telcos have scaled back on their plans to enter the video arena, lifting some of the pressure from cable operators to capture part of the telephony market.

"The cable industry doesn't see telephone companies moving into video at the same pace that it appeared they were at a couple of years ago," says John Aronsohn, senior analyst at The Yankee Group. "There's a clear cause-and-effect relationship there.

The immediate threat from direct broadcast satellite carriers, combined with the fact that upgrading for digital is much easier than upgrading for telephony, means that most MSOs are focusing on rolling out digital cable services.

"DBS has sharpened cable's focus on its core product," Huseby says. "Its channel capacity and quality of signal has created enough of a threat for cable that it needs to respond by deploying digital service on a more rapid and broader basis than it expected to.

DBS has also led MSOs to concentrate on delivering cable modem service, Aronsohn says.

As a result, most cable operators have scaled back their telephony plans considerably. Tele-Communications Inc., for example, is rolling out its PeopleLink service only in three regions-Arlington Heights, Ill.; Fremont/Sunnyvale, Calif.; and Hartford, Conn.

While TCI denies that it is backing off from its telephony plans, Time Warner Cable openly admits that it is limiting its telephony rollouts. The MSO has opted to concentrate on the business market, restricting its residential telephony plans to its system in Rochester, N.Y.

"For the foreseeable future, we're focusing on the competitive access switched business services area," says the spokesman, adding that the MSO will hold off on residential service until it sees the results from Rochester and irons out the regulatory issues.

Cox remains the leader in cable telephony, charging forward with its trials in California and Connecticut, but McElroy admits that it's a hard market to be in.

"I can't say that [delivering telephony] is difficult or unattainable, but it's not simple," he says. "Some of the cable providers have definitely backed off or slowed down their plans to roll out telephony, but we haven't reached the point at which we say this is not an attractive business to be in."-SL