Translation Please: What’s Up In The Upstream?

As is habit (OK, obsession) with this column, it’s time again to peer into what’s up in the upstream path — that sliver of bandwidth used to send stuff upward, from the home to the headend, in cable systems.

Two things seem worth reviewing. One is the upstream path in light of the broadcasters’ DTV transition, and whether their all-digital maneuvers pave the way for a wider upstream path. Two is the notion of channel-bonding in the upstream, as part of the DOCSIS 3.0 cable-modem specifications.

Baseline review: The upstream path in cable systems, as defined by the Federal Communications Commission in the 1950s, stretches from 5 Megahertz to 42 MHz. In stretching terms, that’s like reaching your fingertips to your kneecaps. Not much of a stretch, in other words, considering the upper limit of available bandwidth is between 750 MHz and 1 GHz (1,000 MHz).

The upstream stretches no farther for one reason: Off-air broadcast channel two begins at 54 MHz. When the vast majority of us saw TV channels in analog, this mattered. Stuff couldn’t ride on top of each other.

Why: Even though broadcasters transmit over the open air, and cable’s carriage of broadcast channels run through wires, any spectral overlap caused visible problems. It manifested as a ghosted image of Walter Cronkite (or whomever), just over his shoulder. But last week, everything started to change. Broadcasters started to go all-digital, meaning no analog, in anticipation of the extended-but-still-real June 12, 2009, digital transition.

After June 12, broadcasters at large will no longer require an exclusive seat at 54 MHz. Other stuff could go there without causing interference. Like upstream traffic.

Is it time? Can the upstream path be widened? Definitely a step in the right direction, engineers say, but take a deep breath. Theory is one thing; practice is another.

For one thing, specific gear is required to separate upstream signals from downstream signals. The separation itself, “back in the days of analog,” required “notching” frequencies in taps (the things that allow fatter cable to adjoin to skinnier cable) and passives (things that pass signals along without requiring electricity).

To re-notch the nation’s taps and passives from 52 MHz (including guard band) to 150 MHz is, as engineers would say, not trivial.

But let’s just say you could. Awesome, right? 150 MHz divided by 6-MHz channel spacing equals 25 channels — in cable modem and DOCSIS 3.0 terms, plenty enough for a four-channel bond. That means plenty enough for a symmetrical service offering of 160 Mbps downstream, 160 Mbps upstream. Right?

Not exactly. Recall that the spectral location of the upstream path is a dark alley, with foot-deep potholes. And you’re driving a Mini Cooper.

When the road is full of holes, and you’re driving a Mini, you have three choices: Slow down, ruin your car, or buy a frequent cracker plan at a chiropractor.

The upstream path is that kind of noisy. For that reason, channel widths are slimmer — 1.6 MHz, 3.2 MHz, or 6.2 MHz, depending. (The latter width is as recent as DOCSIS 2.0.)

Modulation — the techniques used to imprint data onto a carrier — comes next. A noisy, bumpy path requires a sturdier (read: slower) modulation. Upstream, with hyper-vigilant plant maintenance, the very best you can do is 64-QAM — as opposed to 256-QAM on all-digital downstream channels.

The 64-QAM method maxes at 27 Mbps, compared with 38.8 Mbps for 256-QAM. So, a four-channel upstream bond would yield a raw burst rate of 108 Mbps. Not bad, assuming you have the channels to bond.

Can you channel bond so competitively without reaching up beyond traditional upstream boundaries? Here’s the math: 52-5 MHz=47 MHz of total available upstream bandwidth. Divide by a channel spacing of 6.2 MHz, to be able to do 64 QAM. That’s seven channels and change. Enough for a four-channel bond, assuming you have the channels to bond — meaning they’re empty.

Can cable operators achieve throughput symmetry in their downstream and upstream broadband offerings? Yes, with upstream channel bonding.

The trick, as with the downstream signal path, is how to best move stuff around to make room for the bond.

Stumped by gibberish? Visit Leslie Ellis