Want to know what's on the minds of the engineers who work for mid-sized cable operators these days? We did, mostly to determine if high-speed data and digital video are real and making their way into second-tier cities. To find out, CED Editor in Chief Roger Brown and Contributing Editor Leslie Ellis hosted a telephone-based roundtable discussion with Dan Liberatore, senior VP of engineering at Adelphia Communications; Ken Wright, chief technology officer at Intermedia Partners; Pete Smith, senior VP of engineering at Rifkin & Associates; and Tom Jokerst, senior VP of advanced technology at Charter Communications. An edited transcript of the conversation follows.
CED: What are your top three current priorities?
Liberatore: My number-one priority is upgrades and rebuilds and getting everything up to 550/750 MHz, two-way capability. Number two and three (are) probably digital video and cable modem deployments.
Wright: I'm pretty much done with rebuilds and digital video. In terms of priorities early in the year, four come to mind: Cable modem launches with the same aggressiveness that we've done with digital video; internal cross-pollination (of training and information from) where we've successfully deployed cable modems and operated return plant; taking digital video and control platforms to the next level (by) evaluating some of the interactive applications that can ride on that platform, like video-on-demand, Web surfing and e-mail; and (focusing) on some of the key aspects of customer satisfaction from a technical perspective-things like reliability, service call percentage, repeat service calls, service and installation response time.
Smith: I have 50-some upgrades to do next year. We've launched digital (video) now in three markets. We're going to launch another dozen next year. We're pretty well along the way of having that model work for us, in terms of training and how you go about doing it. Seven systems are launched with cable modems. Another dozen (are coming) next year. We've just written a launch bible on how you go about doing all that stuff.
Wright: By year end (of 1998), we'll have rebuilt about 75 percent of our homes passed. The vast majority of that will be two-way, as well. Our two-way activations are following behind our upgrades, with about a one-month lag.
Smith: Most of our larger systems are done, or very close to being done. The urban ones are 750 MHz. When you get out to the smaller markets, most of those are 550 MHz rebuilds or upgrades. I'm guessing we have 50 percent or better rebuilt. Now we're beginning to attack the smaller systems.
Liberatore: All our core markets are pretty much done, except for where we've made acquisitions around the fringes. We're in the 50 percent range, in terms of completion. Our goal is to be done in another two to three years. We also have some go-back, where we didn't turn on the two-way.
CED: This whole issue of running out of bandwidth - what about in the reverse? Looking at the 5–40 MHz spectrum, between the parts you can't use and the extra services there - is it wide enough?
Wright: That's where the segmenting of the network really benefits us, because of frequency reuse. The more we break our networks up, and subdivide the nodes, the more we get those frequencies back to reuse for other homes.
Smith: We're probably headed more to a user demand approach, rather than a broadcast approach. In the past, we just added channels and everybody got them. Now you may be looking at more video-on-demand types of service, and that is perfect for frequency reuse types of applications. I calculated it one time. If I needed enough bandwidth to do a VOD application in a market of 50,000 or 60,000, I'd need 1.5 GHz, or something like that. But with frequency reuse, I only need about 18 MHz.
Liberatore: For the last five years, we've been building an architecture that has a fiber node, it has a post amplifier, and rarely some line extenders after that. Maybe a two-amp cascade. We've been averaging under 100, to, in dense places, a couple hundred homes per node. So we're not too concerned about it. Because of the frequency re-use things, the headends are gong to get a little more complicated.
CED: Are you starting to do any traffic modeling, with respect to streaming video and how much bandwidth it will need, or constant bit rate services? All that seems to be shouldering for space in the upstream path-is that an issue for you?
Wright: What we're all doing has to do with some of the early traffic engineering that was done. Everybody had their own model of how many homes per node you had to limit yourself to. But the reality is, on the return traffic today, we're taking eight nodes and combining that into one return laser, coming back. But if we're fortunate enough to use up all of that return bandwidth, that's a great thing. I think we've all built extra fibers out to those nodes. We can subdivide the nodes beyond that.
Smith: Remember, most of the return traffic these days is bursty. It's not occupying the bandwidth 100 percent of the time. Even on cable modem service, when you look at the bandwidth usage, most of it is still going through the customer, and not coming back from the customer.
Wright: So as long as there are people wanting more bandwidth, that means revenue, and that means we can afford to subdivide the nodes or whatever we have to do. I think we've left ourselves a lot of room to migrate ourselves to more and more return demand.
Liberatore: The key is to figure out how to actually manage that and build by that traffic rate, rather than the flat rate thing we're doing today.
Smith: The weakness in all this stuff is probably called "billing systems." We can right now tell how much bandwidth-how many megabytes or gigabytes-people actually use. The ability to get that through some kind of table into the billing system that now bills people based on that is one of the limiting factors right now.
Liberatore: Most everybody today is doing flat-rate billing, particularly in their data service. That's all we have at this point. The next step is tiered billing. Then maybe per-packet billing later. I don't know that anybody can do that today. As for traffic modeling, we've tried to do that. We're pretty comfortable that we're OK.
Jokerst: The one case where it's pretty easy to break the upstream model is picture phone, or heavy videoconferencing. That is presumably symmetrical traffic. What may be a correct model assumption is that everything will vary.
Some nodes in a network will greatly exceed, and others will fall well below the model. It depends on demographics and a lot of externals that are tough to consider in a model.
CED: What is your current thinking on the debate known as clean plant vs. the use of filters, for the upstream signal path?
Jokerst: Philosophically, I'm opposed to the mass deployment of filters. I believe in using filters for rifle shots, rather than shotgun. In other words, where you have a specific problem with a specific drop, or even a specific MDU (multiple dwelling unit) complex, install a filter. We have some major urban areas with 600 homes per mile, 400 of which are MDUs. So filters there, on an interim basis, are a very useful tool. I think the biggest problem with using filters comes when people go out and buy the devices, take them home, and self-install them. But we'll still have to roll a truck to that customer, if at some point along the way we've installed a filter on the tap. That's just a fundamental problem I can't get past. I'm trying to not create a bunch of future truck rolls by installing high-band filters on the tap ports.
Wright: When it comes to filters and traps, our motto is "just say no." We're not doing it. We're making the return plant work without traps. I think there are a lot of reasons you want to do that. If you put traps out there, you're just hiding the problems and becoming your own worst enemy, because as penetration rates for new, two-way services grow, so do the problems, as you start to remove the traps. From day one, we took the approach that we're not trapping.
Smith: I'm the opposite. I'm using traps. That's because, when we upgrade systems, we do not necessarily replace drops along the way. We replace obviously defective drops, but my view of it is, I want to make sure that the expense goes along with the revenue here. I understand the DOCSIS argument, although my view is that that'll take a while before it becomes a big issue for us. The only place it's hurt us is in home demonstrations. You can't have a person walk in with a laptop and show them the Internet service, because there's a filter on the drop. But we figured out ways to get past that with more public demonstrations.
Liberatore: We've taken a strong position on filters, too. I agree with Ken. We do not use filters or traps unless there are specific places where people have to go and troubleshoot. What happens is, we say no, and whenever the field guys need filters, they go chase down one of Pete's trucks and borrow some (laughs).
CED: Let's move into new services, and there are a bunch of them. Maybe we could start with you ranking them: Digital video, high-speed data, telephony, VOD. What's your priority for '99?
Jokerst: You just named it. They're all "priority one."
Wright: For us, the first priority was digital video. That's why we'll be largely done with that by year-end. We'll have it up and running in our top 15 headends, which serve 88 percent of our subscribers. Our second priority is cable modems. We're not as far along with them. But we have three markets that are commercial now, and another seven where we've either installed or are installing the routers right now. In those seven launches, we'll be launching with DOCSIS modems. Behind that is the interactive services, whether it's video-on-demand or Web surfing and e-mail to the TV. Telephony, we're still interested and we're still watching, but that's a number-four priority for us.
Jokerst: Our first priority has been cable modems, followed by WorldGate and digital, equally weighted. With VOD, we've got one place where we're in the process of installing a server on a demonstration basis, for ourselves internally. We haven't settled out yet what we'll use for content. I do see the immense opportunities with VOD, as servers get better and less expensive. As we get digital platforms deployed, that opportunity is right there. Telephony comes in last.
Smith: Our priorities are similar. We're doing digital video. We're doing cable modems. The technology for both exists, and it works, and we figured out the plan. We have a VOD test going with Diva and a proprietary box, but we're hoping that will be integrated into existing boxes soon. The results so far are very encouraging. I can't discuss specifics because of non-disclosures, but in terms of what people will buy, and how much they'll pay, and their satisfaction-it's extremely high.
Liberatore: Last year, before the holiday season, we did 25 digital headends. For the last year, we've tried to increase the penetration. That was for the (General Instrument) DCT1000s. This year, we're installing some of the interactive boxes, and trying to get up this interactive curve. Whether it's simple Internet, or some of the things that others are doing, like Wink, we're just starting as we get this interactive box in place.
(As for) VOD, we're doing a small Diva (test) in the Philadelphia area. From a modem standpoint, we did the same thing last year: About 17 sites with a one-way product, and six sites with a two-way product. The idea was to start to develop the business, the marketing strategies, and all that. This year, it's been an issue of getting those sites, and new sites, over to DOCSIS and refining the business practices. We're doing a small thing with telephony in Florida.
CED: On the DOCSIS side, how much of a gating factor is that standard, given the delays? And once the certification occurs, do deployments really ramp up rapidly, or what?
Wright: As I said earlier, we have three commercially launched markets, and seven more waiting. Those seven are not small systems. So, yes, we have seven headends where we're waiting. The routers are in place. We are not going to launch with proprietary product. We'll launch with DOCSIS product. As soon as we have DOCSIS product, we go to market, in front of an awful lot of homes.
Liberatore: We're doing the same thing. We're not deploying any more proprietary systems. I guess what we're really trying to do here, is tiptoe through this, and gauge whether we need to wait for the final certification, or get out there now. We're trying to understand the risks involved with taking the latter approach. If I could (have) launched sites (last) November with DOCSIS, I (would have done) that. I (got) pressure to do that.
Smith: There's a model out there that should teach us all about waiting for a standard. It's called DSL (Digital Subscriber Line). They've been waiting for a standard for years and years and years, and look where they are. There are a few trials. No major deployments. It's a little dangerous to wait for a standard. The standard will be good for us, obviously, in the long haul. But I'd prefer to get the customer now, and get them satisfied that they've gotten a terrific level of service. So, we didn't wait for DOCSIS. We're certainly interested in becoming compliant, and we'll move in that direction. But I'm not going to wait at all, for any standard, at this point.
Jokerst: My personal opinion on DSL is that I am not a naysayer. I believe that it's going to work, albeit potentially at something less than the full data rates they're discussing. The telephone industry is quietly and busily working to get DSL deployed, to defend the small business and SOHO (small office/home office) customers, which today represent a significant revenue opportunity to them. I don't think the cable industry is as focused on those customers as we should be. In my opinion, those are the customers we should get first.
We're deploying non-DOCSIS modems. At this point, I'm very glad we're doing that. In my opinion, we're deploying a DOCSIS 1.1 class product, with which we're able to deliver multiple levels of service. Out of about seven deployments, I have one that is DOCSIS. We did that intentionally because of our belief in the retail model. At this point, we've been very disappointed with the retail success.
Wright: Let me just jump in on the whole retail thing. I would hate to see us come away from this sending the message that there are problems with retail. Circuit City and Radio Shack have their eyes on this, and they're very interested in being involved. As far as the price issue, Tom, what I'm seeing is that right out of the gate, the DOCSIS modems are significantly lower in price than the proprietary modems. I think it takes very little time at mass production for those costs to go lower. There are a lot of people lining up to make that happen.
Jokerst: From my standpoint, the greater amount of pain is maintaining two platforms from the headend.
Wright: I totally agree. That's why the next seven launches are going to be DOCSIS only. We're not going to put proprietary gear out there. We're not going to support two platforms in those launches.
Liberatore: Let me shift here. Tom's talking about running two platforms (and with cable modems) there's probably not a whole lot we can do about that. One of the big issues for me relates to the digital (video) platform. This idea of "harmony" and particularly the key-sharing provisions of that. It's not getting done. It's forcing us to duplicate bandwidth in a digital environment, until somebody gets that thing resolved. I don't know if you guys have any problem with that, but I'm looking for some team players to go knock on some doors.
Jokerst: Well, the interesting thing there is, there's plenty of upstream spectrum-but it shouldn't be wasted. Having two downstream conditional access and security and billing interfaces and all that stuff is an issue. Where it becomes really wasteful is if we have to double what we should have to allocate in the reverse path, to accomplish multiple vendor interoperability.
Liberatore: That's two 6-MHz channels at best. Two 3-MHz channels in the DOCSIS world. In the digital world, on the downstream, if you have seven or eight tiers, that now becomes 14 to 16 extra channels that you've got to put on. In my mind, it's because our two vendors (General Instrument and Scientific-Atlanta) haven't gotten together to solve that problem. In the analog days, we were able to operate multiple vendor environments by piggy-backing scramblers. We're worse off in a digital environment, until OpenCable happens. But in an immediate timeframe, we're getting back to where we don't want to be, with proprietary vendors.