Who's the Fastest in Broadband?

10/21/2005 8:00 PM Eastern

One of the more palpable trends in the broadband Internet community is the leapfrog war between cable and telephone companies over connection speeds.

It's most visible in the downstream direction, or in terms of how swiftly packets can be pumped into homes. In multi-Megabit chunks, the rate of speed available to a cable or DSL (digital subscriber line) modem tends to ascend regularly. It's a current measure of competitive agility.

An active debate among broadband technologists, for instance, is whether a Moore's Law parallel is brewing in bandwidth — where downstream speeds double every 18 months, while prices remain fairly static.

Apply that hypothesis to Cablevision Systems Corp., which offered a maximum downstream speed of 10 Megabits per second (Mbps) last October. That would mean rocketing to 20 Mbps next spring.

The speed war implications are vast. Is there a ceiling? Who's best positioned to be fastest: cable or the telcos? If you're a cable system manager, how do you budget to be the fastest guy in town next year – without squeezing a speed trigger that isn't really necessary until '07?

Tough call. Over the past nine years of broadband's ascendancy into the consumer mainstream, one fact persists: No matter how much there is, people always find ways to consume more. Broadband is like fresh water. Or gasoline.


Figuring out theoretical maxes for varying broadband providers isn't the easiest math in the world, but it isn't astrophysics, either. This week's translation will focus on how to do the “max speed” calculations for the cable side.

The amount of downstream speed available to customers of cable modems depends on two variables: Bandwidth allocation, and system architecture.

Say an operator allocates one 6-Megahertz digital channel to broadband Internet services. (Most do.) Say that operator is slinging data toward homes using a form of modulation known as 256-QAM (quadrature amplitude modulation). That's equivalent to 38 Mbps of data traveling downstream, on that channel, toward homes.

This is where system architecture comes in. That high-speed channel travels from the headend to a “serving area group.” In general, a serving area is comprised of four, 500-home nodes, all of which are “sharing” the 38 Mbps available in that high-speed channel.

Let's assume that 60% of the homes passed by that 2,000-home serving area group are cable customers. That's 1,200 homes. Let's further say that 40% of those homes use a cable modem for Internet access. We're down to 480 homes, sharing the 38 Mbps of raw capacity on that broadband channel.

It's tempting, but incorrect, to next divide 38 Mbps by 480 customers, to establish how much bandwidth people really get. In practice, people hardly ever partake in the Internet in the same exact way, at the same exact time.

You might be reading a Web page, while your neighbor is downloading a fat operating system patch. Two streets over, somebody's teenager is clattering messages into a dozen different instant message windows. Each of you is using bandwidth differently.

Geography matters, too. College towns devour bandwidth. Retirement communities snack on it. That adds to the trickiness of the calculations.


Cable operators have three immediate options when it comes to retaining their position as “the fastest guy in town.” One is to revisit those serving area groups, to decouple them from the 38 Mbps spigot. Most operators do this in two steps: Unhook two of the four nodes sharing the channel, then the other two, as bandwidth needs dictate.

As resources go, disaggregating serving groups is a matter of buying “blades” for the headend portion of the broadband system, known as the cable-modem termination system (CMTS). Pricing varies predictably, but a safe estimate is $25,000 per blade, or around $12,000 per downstream port.

After that, a further option is to “split the node,” which really means adding lasers and receivers in such a way that 250 passings share the 38 Mbps, instead of 500. Split it again and that's 125 homes.

The third option is to dedicate a second (or third, or fourth) channel to broadband Internet. That's part of what DOCSIS (Data Over Cable Service Interface Specification) 3.0 is all about: Gluing four or more channels together, summing the throughput (38 Mbps x 4), blasting it down to individual modems.

In combination, those options (when carefully applied and in a rational market) will suffice to position the cable side as the “fastest guy in town.”

Next time: How to do the math on telco-delivered speed maximums.

Stumped by gibberish? Visit Leslie Ellis at