fee, PHY, fo, Fumble
What with the accumulated hyperbole that's built up about the cable industry, high-speed data and cable modems going retail, it really can't be any surprise that the industry itself got caught up in the smaller-better-faster race that's engulfed the high-tech world. No sooner does one high-tech startup company bolt out of the starting blocks, than another entrepreneurial brainstormer shoots out of the starting gate looking to overtake early sprinters with some new technological innovation or marketing twist.
It's pretty much the same with the cable industry and standardized cable modems.
Shortly after the industry began its DOCSIS 1.0 certification last year, people started speculating about when DOCSIS 1.1 waves would start rolling. Almost in the same breath, people wondered aloud what a DOCSIS 1.2 specification might entail. Then, the race was really on. Operators, vendors and manufacturers jockeyed for position to see who would influence, if not lead, the 1.2 pack. All this despite the fact that the 1.0 effort only recently got up on its own steady, operational feet. Fortunately, the dust has cleared from the first flurry of activity on 1.2. The frantic pace has slowed down considerably before all concerned found themselves jumping the gun in one collective lunge. The potential, albeit more deliberate, spec has a good chance of keeping everyone in the cable industry on the fast track against other broadband competitors.Ready, set...go?
In November 1998, CableLabs selected Terayon Communications Systems (www.terayon.com) and Broadcom Corporation (www.broadcom.com) to author a detailed technical specification on a next-generation physical layer technology (PHY) for integration into the Data-Over-Cable Service Interface Specification (DOCSIS). The spec was to be based on Terayon's S-CDMA (Synchronous Code Division Multiple Access) and Broadcom's Advanced Frequency Agile TDMA (Time Division Multiple Access). However, the DOCSIS committee didn't want the two companies to act in a complete vacuum. They asked both contributors to work closely with the IEEE's 802.14 committee as well. "The DOCSIS committee asked us to go through the technical body of the IEEE 802.14, which is focused on the advanced physical layer, just to leverage the knowledge of the people who attend those meetings," says Zaki Rakib, CEO at Terayon. "And, so we did. We are in parallel while working with DOCSIS; we've been going through all the processes in 802.14. In fact, we took it on ourselves to reconcile any differences that existed between 802.14 and DOCSIS. It was good to see that DOCSIS 1.2 could end up being the first time we actually see convergence between the two bodies, i.e., IEEE and DOCSIS."
In the official CableLabs announcement, it was explained that the next-generation physical layer was "an important extension for DOCSIS because the robustness of these types of technologies will enable more cable systems to deliver higher speed two-way cable services and allow cable companies to increase the data capacity of their networks." That simple statement belied a more vigorous debate that went on behind the scenes for months before the announcement was made.Taking sides
According to a participant in the IEEE 802.14 effort who requested anonymity, the lively debate on the two technologies and the quality of the return channel had operators divided into essentially three camps. One camp, he says, took the position that said, "Hey, the return channel (CNR) is relatively high when we break the node sizes down to 250 or 150 homes per node. Therefore, we don't really have a capacity problem at all, and we don't need high PHY at all.' He notes this was in stark contrast to comments a year earlier when many of the same operators were saying, "Oh my God, we have a significant bottleneck in our architecture, i.e., the return channel that doesn't have a lot of bandwidth and it's noisy. So, we need an advanced PHY."
Another set of cable operators, he explains, were saying, "Yes, the return channel is cleaner, we have a higher carrier-to-noise ratio by virtue of these smaller node sizes, but we still have a bandwidth constraint. So, it's good to have a higher bandwidth PHY. We don't need S-CDMA because the noise is lower in smaller node sizes. But we should increase the symbol rate and the constellation size on the return channel so we have increased throughput." That group of cable operators, he says, was endorsing the high PHY, but only the TDMA portion of it.
A third group of operators took still another position. They were saying, "Now wait a second. Maybe some of our infrastructure has the carrier-to-noise ratio (CNR) that will allow you to run these higher order constellations, but the vast majority cannot support the higher constellations because the CNR is too low. Therefore, we need some robust modulation scheme to allow us to survive that noisier environment." The result of this intense debate, says one observer, was a "theoretical bakeoff" that decided both the 64-QAM/TDMA and the S-CDMA proposals made sense. What would such an advanced PHY bring to the table? According to Henry Samueli, CTO and co-chairman of the board for Broadcom, several important things.
"First of all," says Samueli, "it increases the data rate from 10 Mbps to 30 Mbps. And that principally benefits the cleaner cable plants. Because you won't be able to do the higher data rates and the higher modulations, whether or not it's S-CDMA or the 64 QAM/TDMA approach. That's irrelevant. You can't get to 30 Mbps unless you have a cleaner cable plant.
"Then, in the other direction, there is additional robustness for poorer plants. There are more powerful forward error correction techniques and interleaving techniques added to it to correct for burst errors. There are more powerful, adaptive ingress cancellers for nulling out RF interference. Things like that have been added to the spec to increase robustness."Backward into the future?
While the debate on the two advanced PHY technologies was lively, it was nothing compared to the wrangling that broke out this past summer. This debate went beyond the jousting for position on a new spec. It got down to the very essence of what a technical standard should be, now and in the future. According to one DOCSIS participant who also requested anonymity, a number of operators were unhappy with the fact that the proposed advanced PHY spec wasn't taking into account the specs that had preceded it. "The problem," he says, "is that some operators said that S-CDMA is not backward compatible. What that means is that should an operator have chosen an S-CDMA headend in order to deal with a plant that had not yet been converted to HFC, or had used very large nodes, or frankly had poor design implementation or maintenance, then DOCSIS 1.0 and 1.1 modems would not work in those headends.
"There were operators who had a real concern about that. It wasn't that they were concerned about the viability of the technology. They were concerned about the market confusion and the fragmentation of a business that was arguably started under the guise of true interoperability. And there were enough of those operators that basically said, 'We cannot support 1.2 as it is characterized today because of that lack of backward compatibility.'
"There was a misconception that there were certain cable operators that were out to kill the whole thing. That's absolutely not true. We're not trying to prevent anyone else from using any other technology to cut through crap. We're just not going to allow it to be called DOCSIS if it's not backward compatible, period."
The result was that there was a counter proposal put on the table. It called for S-CDMA to, in effect, be interleaved in the upstream with already accepted and approved modulation mechanisms. In other words, what if you could pick and choose between S-CDMA and DOCSIS?
"If you can truly time slice and there is really no overhead penalty in that time slicing, that seems to get around the problem," says one source. "If an operator doesn't need to take advantage of the additional robustness of S-CDMA, he would simply never allocate any time slices to S-CDMA. And yet, anyone who really thought they needed it, could allocate all the time slices they wanted to S-CDMA and still exchange the customers' modems if they wanted to, or alternate between the two of them." Then the issue became defining the true overhead from an occupied bandwidth standpoint, and to a lesser extent, the penalty of requiring this time slice capability in all modems.
As a result of this debate, the "official" DOCSIS 1.2 effort is on hold, although the advanced PHY effort continues. According to Rouzbeh Yassini, CableLabs' executive advisor on DOCSIS certification, once that effort meets certain criteria, it may earn a certification label. "At Cablelabs, we continue working, solving and understanding what it takes to have better efficiency, higher bits per hertz and more advanced features on DOCSIS. But any new enhancements must meet certain criteria. First, it can't impact the DOCSIS branding. That means it must be backward/forward compatible. Secondly, we must make sure it's made with a transparent technology introduction that does not impact the market introduction, i.e., people are waiting for something that's better or newer. And third, it must be based on proven technology.
"We do have a formal program called DOCSIS 1.0 and DOCSIS 1.1. We do not have, at this point, DOCSIS 1.2, and the reason is that those three criteria have not been met yet. But, what we're encouraging very strongly, i.e. to the vendor community and the industry, is to keep working and provide CableLabs a prototype of such a product. "To be specific, we will entertain looking at prototypes that have an advanced PHY, which is more bits per hertz, better efficiency, a better CNR environment. We've asked the guys to bring us a prototype that shows us such technology can coexist on the same RF channel that the DOCSIS products work (on).
"We'll take such a technology through four to six weeks of evaluation within CableLabs. We'll evaluate the technology, the good points as well as the bad points. We'll get together with the vendor community and the vendor authors and we'll discuss it. We'll take it to our CTOs, and if they deem it necessary and that it meets the criteria, then we will label it just like we did with DOCSIS 1.0 and 1.1."Moving ahead deliberately
Terayon's Rakib says the company is up to the DOCSIS challenge. "Of course, the DOCSIS (committee) and cable industry wanted to see that the advanced PHY really brings in all the expected advantages. One is resilience/robustness and the other is additional bandwidth or throughput. As you know, we are taking the upstream throughput from 2 Mbps to 27 Mbps. It's a very significant leap. And it does that without compromising the robustness or resilience to noise. "We have convinced them that it's not only going to be backward compliant, but it's also going to have a very, very small, if any, price deferential. They didn't want to see the complexity of DOCSIS 1.2 being translated into (extra) cost. We have demonstrated that this will not happen."
While there is considerable talk in the cable pipeline that Terayon is on its way to solving compatibility issues, apparently no formal demonstrations have yet taken place. The company seems to be working on a two-track development process that will simultaneously produce a backward-compatible "prototype" for CableLabs and the industry to evaluate, and an actual product for the company to manufacture. The estimated time frame for one or both? Mid-year 2000.
"From a Terayon perspective," says Rakib, "we're not necessarily going to wait for the certification date in order for us to come out with product based on 1.2. What we believe is that the (advanced PHY) specs will be finalized sometime this year. Therefore, what we bring out in the mid-2000 timeframe is a product we will call pre-certified as 1.2. We definitely will certify it as 1.1, and we will call it pre-certified for 1.2 given that we anticipate a delay in the certification itself."
It's not a cure-all Despite all the hype about an advanced PHY and its ability to "cut through crap," a number of cable professionals, including CableLabs' Yassini, warn that any standard developed along these lines is not some sort of killer application for those too lazy or too cheap to clean up their plants.
"At the end of the day," says Yassini, "I would say a cable operator who decides not to operate his cable plant for the long term is basically shooting himself or herself in the foot, because an upgrade is a fundamental (necessity) for packet-based, digital services that need to be deployed. This (advanced PHY) is no panacea (for upgrades) and it was never intended to be. "But what it does help is in the transition. This says that you can work in these types of (dirtier) environments while you're upgrading your system, and you won't have to take a second seat to the competition. If the customer demand is high, then you can deal with that.
"The bottom line is optimization. You upgrade your cable plant up to a point to solve some problems. You make the technology a little better to solve other problems. You get better network management as well. So, you don't spend 100 percent of your money on just one thing. These are all parts of the system's end-to-end optimization."It's not about technology
Where does this leave DOCSIS? Alive and kicking. The 1.0 certification is rolling along on its own momentum. Yassini predicts that 1.1 interoperability testing should begin in the March/April timeframe next year. Shortly after that, if the Terayon/ Broadcom hybrid spec takes physical shape and survives testing and evaluation, a DOCSIS 1.2 label may be in the offing. But many of those close to the situation say this preoccupation on technical minutiae is clouding more important issues. Modernizing the old cliche about not being able to see the forest through the trees, they say broadband professionals run the risk of not being able to see the content through the ICs.
As one DOCSIS participant put it, it's not about the technology. "There have been and there probably will be in the future proposals for enhancements to DOCSIS," he says. "And we would be stupid if we didn't consider every one of them given what we're really trying to do, which is really simple. We're trying to deliver content and services. We're not in the technology business."