Compared to its robust downstream sibling, cable broadband's upstream has always looked like a 90-pound weakling.

But market shifts and new competitive threats from telco fiber services may soon change that. While most MSOs are still focused on offerings that emphasize the brawn of the downstream, they are considering options to bulk up the upstream as well. The question now is: How much upstream muscle do they need?

It's a question that can't be answered based on past experience. Up until recently, the argument for bigger cable data upstream capacity has been weak at best, mostly because the market simply hadn't demanded more. Meanwhile, early predictions that the peer-to-peer file-sharing craze would clog cable networks' return paths fizzled with the advent of subscription download services.

"I think what surprised us over the last 18 months is that for years we worried about the upstream, and what happened was that the downstream has become pretty fully utilized and we had to watch ourselves to stay on top of that," says Jay Rolls, Cox Communications' vice president of telephone and data engineering.

But the old saw holds that the only consistent element in life is change. Upstream bandwidth demand may be once again on the rise with the boom in digital cameras and new communications services such as video telephony and e-mail. To that end, Insight Communications will be debuting the Shutterfly photo sharing service this fall, and that could prompt an increase in upload speeds for Insight's high-speed data offerings.

"There are a lot more digital photos out there that people want to share with their friends and whatever," says Bryan Smith, product manager for high-speed data services at Insight Broadband. "And having those fast upload speeds makes a really big difference."

Al Johnson
Other cable operators are talking about new services such as video e-mail "and that's going to have a demand on the upstream," says Al Johnson, Xtend Networks' vice president of technology and advanced architectures. "So I think if you look on the horizon you will see that demand."

On top of all of that, cable operators are facing increased competition from telcos that are setting aside slower digital subscriber line products in favor of direct fiber service rollouts. That could give them a coveted edge in the bandwidth battle, and "for the first time, MSOs are going to see real competition on the triple play," notes Iain Drummond, interim president for Harmonic Inc.'s Broadband Access Network division. "So we do see the (telco) advertisements in a competitive sense that will be promoting the upstream availability as well."

Technology tools

Fortunately for cable operators, there are several technology options available that can beef up their upstream capacity.

One classic method to boost individual users' bandwidth allotment is to subdivide nodes—the smaller number of customer per node, the more bandwidth for each user. Drummond says that is already starting to happen in cable systems.

"We've seen over the past few years operators move from what had been 2,000 homes passed to 1,000, and clearly, people building in the end of 2004 and into 2005, 500-home nodes," Drummond says, adding that in areas where cable operators are facing greater telco fiber-based service competition, that node sizes are even smaller.

Harmonic has several segmentable node products, including its 3844 node that is capable with added electronics—an additional return path transmitter—of dividing into four logical nodes without the need for more installed fiber. Wave division multiplexing can be used to subchannel the existing fiber to the node and supply the added return path channels to the headend, where an added receiver can process the added pathway.

"So I've added a receiver at the headend and I've added a transmitter inside the node. I've done no construction, and now instead of having 42 MHz worth of return path bandwidth, I now have 84 MHz serving that same community," Drummond says. "And using our products you can go to a factor of four. It's a cost-effective and scalable way of adding bandwidth."

To further boost this capacity ramp, Harmonic recently unveiled a new digital coarse-wave division multiplexing product for the return path. CWDM subdivides fiber channels into multiple channels, further boosting the capacity.

In contrast to older analog versions, the digital CWDM gear is for the most part plug-and-play, cutting the amount of tuning needed to install it, Drummond notes.

"The operational cost of going out and setting up a new return path and balancing that return path is just much, much lower with digital than analog," he says.


Another technology that has promised greater upstream capacity for some time is DOCSIS 2.0, although it has had a bumpy start. DOCSIS 2.0 offers a return path boost because its options for S-CDMA or A-TDMA transmission schemes help shield data in the notoriously noisy return path frequencies. But while gear has been available since late 2002 and operators have been buying compatible cable modem termination systems (CMTSs) and modems, most have yet to fully turn on DOCSIS 2.0's capabilities.

BigBand Networks now claims the majority of the U.S. DOCSIS 2.0 deployments, with major customers including Adelphia Communications Inc., says Ashwin Moranganti, BigBand's director of product marketing. Adelphia is deploying DOCSIS 2.0 using 64 QAM, producing 15 Mbps in the upstream.

Why the sluggish deployment? The culprit lies in the older DOCSIS 1.0 and 1.1 cable modems mixed with newer DOCSIS 2.0-capable devices. With that blend, cable operators aren't able to turn on the full DOCSIS 2.0 capabilities.

"In order to take advantage of the higher width channels, you have to change the CPEs, too," Moranganti says. "We've seen that approximately 60 percent are 2.0, and the rest—40 percent—are still 1.0 models."

But that doesn't mean the DOCSIS 2.0 plant is lying inactive. With DOCSIS 2.0's advanced ingress noise cancellation, Cox has been able to extend 16 QAM modulation in the upstream to a greater portion of its plant, Rolls says.

"We are 60 percent 16 QAM. I don't think we'd be nearly that much if it wasn't for the ingress noise cancellation," Rolls says. "So that [DOCSIS] 2.0 gear has let us get to such a high number."

Three's the charm?

Greater upstream bandwidth also is in the offing with the proposed DOCSIS 3.0 specification that should be ready for release early in 2006. DOCSIS 3.0 includes provisions for bonding channels in the upstream as well as downstream, thereby multiplying the available bandwidth.

Ralph Brown
Channel bonding can boost the upstream not just because of its bigger combined pipe, "but also there is a lot of benefit in statistical multiplexing and load balancing that automatically takes place across those channels," says Ralph Brown, chief technology officer of CableLabs Inc. "So now the efficiencies in which you use that upstream capacity are more efficient. You don't need to worry about bouncing modems around if there is contention for the upstream and your ability to make use of that capacity is better."

It also could lead to a higher split for return path traffic, effectively widening the 5–42 MHz band usually allocated for the upstream.

"We today work with 42 MHz in the U.S. and 65 (MHz) is dominant in Europe. Now there has been some talk about a split somewhere in the ballpark of 80 to 90 MHz," Drummond says. "One day when we see some bandwidth reclamation I think that could be a means of dramatic expansion."

Going out-of-band

Still another option lies in out-of-band technology offered by outfits such as Xtend Networks. Xtend's scheme creates overlay capacity between 2.3 and 2.7 GHz that includes the ability to expand the upstream to as many as 10 additional blocks. These 10 blocks will provide 100 channels, and at 64 QAM or even 256 QAM "is a lot more than they (MSOs) can do today," Johnson says.

That contrasts with the existing cable data architecture, where return path channels are situated in the relatively small 5 MHz to 42 MHz range.

"What we alleviate is being able to help segment some of those return bands and bring them up, and we bring them back at a frequency where we don't see a lot of that outside stuff—the ingress," Johnson says. "We can't control what comes out of the premise—if you throw bad stuff in, you are gong to get bad stuff out—but the bands that we are operating in, the 2.3 to 2.7 GHz region, are very clean."

Upstream strength strategies

Despite all of these tools, however, the greater issue for cable operators is in deciding how much upstream bandwidth they really need. That's the case for Insight, which is now evaluating options for retooling its consumer Internet services— which now consist of a 4 Mbps downstream, 384 kbps upstream, and a premium tier at 6 Mbps/512 kbps.

Xtend Networks’ Xhub can support
as many as 10 upstream channels.
Aside from the avid gamer, a 4 Mbps down/384 kbps upstream product offers enough speed for the average user. But in markets where telcos have debuted FTTH service, cable operators have had to ramp up data speeds, Smith says.

A 1 Mbps service "may become the entry point at some level. That would not surprise me," Smith says. Telcos do have deep pockets and have started their fiber upgrade projects, and "if they can get all of their elements in place, they are going to be a formidable competitive force. And so I think that at some level, there really is a question of how much speed do you really need for what you want to do every day?"

Those changes may well come sooner than later. Insight still has work to do rolling out the latest DOCSIS iterations across its plant, and the recent launch of its digital phone service is still commanding much of the MSO's attention. But it will soon turn its gaze to the tweaking of data products, including speed. "We're going to definitely be out there within the next six months or so with some of these changes," Smith says.

Cox, meanwhile, has ramped its upstream channel width from 1.6 MHz to 3.2 MHz. That, combined with the upgrade to 16 QAM, generates four times the bandwidth capacity in the upstream, Rolls says.

The MSO also is splitting nodes, and it is decombining nodes, putting fewer on the same backhaul channel.

"All of that has allowed us to stay on top of the bandwidth growth," Rolls says. "The stuff that we have in our back pocket is 64 QAM and DOCSIS 2.0, and we have the ability to run dual upstreams so we can have more than one carrier in the upstream. There will be a continual migration to smaller node sizes—I think we will always be doing that, and then longer term it's DOCSIS 3.0."

That said, there really isn't any move toward a more symmetrical residential service with equal speeds downstream and upstream.

"I don't really see the demand," Rolls says.

In the long run, new user trends such as photo sharing and multimedia Web posting may increase demand for more data offerings with more muscular upstream speeds. But that won't happen overnight— hopefully.

"Probably the worry would be more of some kind of overnight sensation of some kind. As a network planner I worry more about that," Rolls says. "It's a traffic engineering challenge to stay on top of it. But we've got the systems in place that give us early warning signals."