Comcast's Fellows Sees A Quicker Digital Future

With each new employer — Continental Cablevision Inc., AT&T Broadband, Comcast Corp. — David Fellows gets to play in a bigger sandbox. The executive vice president and chief technology officer of Comcast now oversees engineering operations for the world's largest MSO. In this interview with Broadband Week editors Matt Stump and Karen Brown, Fellows discusses the evolution of cable's plant, including the possibility of a quicker digital conversion than many anticipated. Looking ahead to this week's Cable-Tec Expo in Comcast's hometown (Philadelphia), the conversation took place on May 1.

BW: You spent all those years at Continental, engineering what were, at the time, pretty large cable systems. Then you moved to AT&T Broadband, where you had even more systems but less money to upgrade. And now, with Comcast, you have more plant than anyone in history and money to spend. What's that like?

David Fellows:
In terms of a company, Comcast feels a lot like Continental Cablevision, in that it feels like a cable company. We also happen to be in the high-speed data and telephone business. But you get the entrepreneurial feeling that bureaucracy is not going to get in the way of doing the right thing, which is quite enjoyable.

My entire career, I've asked myself one question before I accept a job: 'Do I think I can make an impact?' And given all of those aspects of Comcast, the corporation and the role of this office, if I can't make an impact in this industry, no one can. From that aspect it is very fulfilling.

I spent about a third of my career in telephony and I speak that language and all the acronyms. I spent a third of my time in the data business, and I speak all those acronyms. And I spent a third of my time in the video business.

So given I speak all three languages, the key is to go hire some smart people in each of those languages and I act as the anti-Tower of Babel. I tear down walls. I refuse to have silos here. The converged network, all the things we talked about in the dot-com era, they are still true. They still apply. The networks we are putting in place were not part of the dot-com boom and bust. Technology continues to march on. Moore's law still applies. Once you've digitized things, once you've put it on fiber, viewed with the correct perspective, it's all the same stuff.

BW: One could argue you have three silos, now, with MPEG [Moving Picture Experts Group] video, Internet-protocol data and circuit-switched telephony. How soon do those silos, or parts of those silos, migrate to, say, an all IP-network?

For a while, I've considered my middle name to be Baby Steps — David "Baby Steps" Fellows. We're going to take baby steps along that path. If I go back to AT&T Broadband, I had a converged network, in that the same piece of fiber and coax that went out to your neighborhood or home carried a converged bundle of services. But as soon as it got to the headend, it went on three separate networks, with three separate provisioning systems and a couple of billing systems. Here, we're concentrating on IP as a converged platform for the metro network, not just [hybrid fiber coax] out to the home.

With VOD, to start with, we put servers right at the edge, right where the piece of fiber went out to your neighborhood, and that meant we needed a separate server with all the video content at every place where I interfaced with pieces of fiber, at every fiber hub. But back at the headend, you didn't have all that much. What happened was with IP and gigabit Ethernet carrying IP, I can take those servers from the edge and centralize them. I can have one complex now with the full range of video on it. I go gigabit Ethernet out to that fiber hub, then HFC to your home.

If affordable gigabit IP had never come along, we would have deployed VOD at the edge. We would have looked at a higher revenue per view, but now I can use IP to centralize that. That means now I can offer thousands of hours of video, as opposed to the edge, where I might have been only able to afford hundreds of hours of video.

So I'm taking the video platform and putting it on the Internet platform. Separately, of course, a la Coatesville, Pa. [site of Comcast's IP telephony trial], and a different view in our hybrid deployment in Detroit, I'm taking voice and I'm putting it onto an Internet-protocol platform.

Now this is not the public best-efforts Internet. But I'm utilizing Internet technology and the price points for Internet technology, so I'm no longer running SONET [synchronous optical network] rings with telephony company pricing, I'm running resilient packet rings with gigabit Ethernet pricing.

On the back end and the provisioning side, I'm serving up IP addresses for voice MTAs [media terminal adapters] for telephone service, to cable modems, for data service and for the DOCSIS [Data Over Cable Service Interface Specification] modems inside my high-end [set-tops] for robust signaling purposes, the so-called DSG [digital signaling gateway] or the DOCSIS signaling gateway.

So everywhere here, I'm trying to break down silos. Now, silos weren't that bad. We made lots of money with silos. But you get to make, a, even more money and b, you get even more scale and simplicity. If I can make things simple and standardized, then I can increase speed and profits.

BW: In Coatesville, you're doing IP from scratch, while Detroit has circuit-switch with now an IP overlay. You have other preexisting circuit-switched markets with AT&T Broadband. Is Detroit the model for the evolution of the legacy circuit-switch markets and Coatesville the model for greenfield telephony markets?

Fellows: Coatesville is one vision of what voice-over-IP should be, and it's the one that's in the field that we're testing. If it turned out to be the perfect answer, I would deploy it in conjunction with what are known as reverse gateways, so I'm not throwing out all the NIU's (network-interface units) on the sides of homes and all the HDTs (host display terminals) that talk to those homes. I don't want to strand a lot of capital, and with something called a reverse gateway, I can do circuit switch from the home to the headend, go through a reverse gateway, come out IP and then go through a soft switch.

In Detroit, we're testing the opposite of that — IP in the access network — because I'm deploying new customers there, so I don't want to deploy circuit-switch product, so that's IP in the access portion. I then go through a gateway, come out GR303 into a class-5 switch that I've already invested in or already have sitting there.

Detroit was a model that said, 'If soft switches for some reason never prove scalable, I can still reduce a lot of the capital expense by using Internet protocol and DOCSIS modem technology on the customer-premise side, where there's a lot of numbers.' So I can get a lot of the capital savings I want out of a Detroit approach.

BW: Is that a likely scenario, that soft switches will never reach the class-5 switch capabilities?

At this point, I do not think that Detroit is a likely scenario. I may eat my words, but I don't think it's going to be the answer for two reasons. One, I do think the soft switches will come around. And secondly, if you think about the back office and provisioning — the things other than the plumbing — with the circuit-switch switch, you don't get a chance to converge those. With soft switches, they look a lot like servers, gateways look a lot like routers, so my provisioning systems see savings if I can go to more of a Coatesville approach.

In addition to Coatesville, we have a laboratory in Moorestown, N.J. I can make 100,000 phone calls in five minutes with the test equipment I have there. That's something I can't do out in the field.

BW: There seem to be increased discussions about moving to the all-digital network with a $35 price point thrown out for digital set-tops in, say, the next couple of years, and reclaiming lots of analog bandwidth. Where's your thinking on this?

If I look at my own evolution of thinking in this area, I originally had thought: 'Hey, there are 250 million analog TVs out there. We've got lots of bandwidth on our plant. We're going to have analog signals after everyone else in the world, including broadcasters, have gone digital. We'll be, what you might call, the last iceman out there, peddling analog channels to old analog TV sets.'

Then you start looking at high-definition TV. You start looking at video-on-demand. And you say bandwidth is kind of precious and I don't know if that's the best use of my bandwidth to preserve legacy television sets. So I then said, you're right, when we go all digital over the air, when we swallow that digital transition, in the midst of all of that turmoil, we'll go to all digital also and maybe even the timeframe is right for that.

I've now come to the point [where I] think, 'Hey, bandwidth is really precious. People like this video-on-demand in Philadelphia. In addition to just movies on demand, there's subscription video-on-demand, a whole tier of free VOD, and people really like it. I can use my bandwidth like that to make customers really happy.'

In addition, the government is sitting there in Washington trying to decide, 'How do we foster this over-the-air transition?' We've got the Powell plan. Comcast has been launching digital signals and high-definition signals. Maybe it fits into government's desire to have a quicker rather than later digital transition. It plays into our realization that this bandwidth can be used for some pretty cool things. The last remaining question is the price tag around that.

Normally, vendors come to me to talk about a $50 set-top, and I respond: Hey that's great, where's the $30 set-top?' But in that range, $30 to $50. If we could get a set-top box that was in that price range, it would probably be worth our while to go 100 percent digital sooner rather than later.

Set-top boxes look a lot more like computers than they do old clunky analog things, so that means they are on Moore's law. You've seen cable modem prices go below $50. So it's not out of the question we could get a sub $50 set-top here, sometime. There are more MPEG-ish type ways of attacking this set-top box. There are more IP or DOCSIS-based ways of going about coming up with this set-top box. But it is part of our thinking now and I think we'd rather go 100% digital sooner rather than later.

BW: Is a sub-$50 set-top achievable by 2005?

Fellows: Yes, I believe it is.

BW: On the digital conversion, Comcast would have 15 million homes that don't have digital boxes. You'd have to physically put a $35 box in all of those homes at some point, won't you?

Fellows: Let's say it's 30 million devices at $33 a piece. That's $1 billion. We actually have a billion dollars. If it's attractive enough, we have the money to do that. This is not an inconceivable thing.

BW: And the truck roll issue can be overcome?

More and more, we're moving to self-installs on cable modems, and that's working, and there is availability at retail. Again, you might not have to have, for the industry, 250 million truck rolls, or even 100 million truck rolls to service 250 million TVs.

BW: What's your thinking on current set-top box design? What do you want to see in those boxes?

I think about low-end boxes and high-end boxes. The low end I'm trying to drive lower. The DCT-2000s are becoming DCT-1700s and 1800s. If we can, we talked earlier about a $50 or $35 set-top. So on the low end, you want to drive it down and drive widespread access to our services.

At the other end, over time you see the high end getting even higher. You start with high definition and a DOCSIS modem inside that isn't doing very much. You'll see a next generation of that with [digital video recorder or personal video recorder] capability. We've got [Scientfic-Atlanta Inc.] in the field trialing their box in Arlington-Alexandria, Va. Later this year we expect a PVR box from Motorola, and as soon as we get our hands on it we'll deploy that solution.

Once you've deployed one PVR, you begin to think about multiple PVRs in the room and how can I get access to them. The high end continues to become higher as Moore's Law allows you to put in more capability as hard drives get less expensive.

But then there is the other dimension, which is the network dimension.

At the headend, you start up feeding in high-definition, feeding up lots of digital video, movies, free content. Eventually a question you're going to ask is, Network PVR or in-the-home PVR? The answer is, it's going to be both.

We've got PVR in the home as soon as we can get there. The fair-use [copyright] rules around that are a little bit clearer than putting that in the network. But we're also putting servers in the network with a lot of variety of content available on demand, in both short clips and two-hour movie clips.

Over time, some of that content will become high-definition on-demand. Over time, you want the storage in the networks, where it makes the most sense; in the home for something you've selected, and you want on-demand, and you'll look at an awful lot; and [at] the edge of the network for something a lot of people want to look at. Further back in the network for something that is not viewed that often — all the way back to Ted Turner's basement, where the one copy of the one movie one person a year is going to look at to get fed up on demand.

The competition we're going to face has got to put it in the home because their clear broadband pipe is mostly a one-way clear broadband pipe, not two-way. We will take advantage of our two-way pipe to give more variety of content on-demand and, when the digital rights management allows it, we'll store content where it makes the most sense.

BW: One last digital conversion question: You've talked about digital conversion as a way to open up bandwidth for VOD and HDTV. One could argue that the revenue models for both VOD and HDTV are unproven, compared to cable modems. Why make the switch to digital? Is it such a low price point that it doesn't matter?

Again, I don't need to make that decision today. If it turns out that VOD with the free dimension, four times as many people start using the purchase applications as opposed to the free applications, then I might say, 'Whoa, maybe I can afford a $100 set-top.' And I think you can get a lot of people to say you can get to that price point.

If, on the other hand, I can get to a $30 price point, then I don't need to make nearly as much money off of video-on-demand or other video services — HDTV for pay — in order to do the 100% digital.

Now if I happen to hit both — people love two-way interactive services and I can get the set-top price down below $50 — that's what we call a home run. It's like almost the opposite of Catch-22. We're going to get there.

This is one way we might get there, with lots of income. Another way we might get there is with very low priced set-tops.

It's much easier for me to predict 30 years from now than it is to predict three years from now. But we are going to get there way ahead of the over-the-air conversion [to digital]. There is going to have to be a driver in there, low price or higher revenue in conjunction with the transition. We'll ride the price points.

One other dimension here: you may have noticed the consumer-electronics deals that we are cutting both in Washington and with the PHILA [Point-of-Deployment Host Interface Licensing Agreement] license from [Cable Television Laboratories Inc.]. Every two-way TV set I can cause to be in the market through the retail channel directly from the consumer electronics manufacturer — that's a zero dollar set-top from my viewpoint.

Or if I have to put in a POD module, it's a $30 or whatever something dollar POD investment on my part. That's another one of these drivers, driving set-top prices down. Five million of these high-def screens will ship this year. If I can have that be 5 million, two-way digital capable, can-access-all-my-services consumer-electronics devices, then that's 5 million less set-tops I've got to go buy.