When I started Covad fifteen years ago, we envisioned a rapid ascent to a broadband future via DSL. I had also done analysis for Bell Labs which strongly suggested that a Fiber to the Neighborhood approach (FTTN) would be the best for the long term future of the copper phone companies; they would use DSL for the last mile, after the fiber, and would steadily push fiber all the way to the home in new build and rebuild areas. It nonetheless took almost 15 years for AT&T to finally put fiber nodes near my home.

I was finally getting a chance to take a ride on the Information Superhighway.

You would think that after 15 years of DSL, Ma Bell could get it right. But nooooo. They promoted a deal to switch from cable modems. My home is in a challenging spot: at the long end of three different runs of copper to three switching stations; 45,000 feet. Nine copper miles! (Copper snakes around neighborhoods; the central office is only three miles away, as a drone flies.) Most homes are within 12 kilofeet, and even in the ‘burbs, within 25 kilofeet. Our distance is a challenge for voice as well as data.

AT&T rolled fiber as FTTN (fiber to the neighborhood), like Malcolm Turnbull is proposing for Australia’s broadband network, the NBN. The prior government had proposed fiber to the premise, but the economics are challenging. Google is fibering up cities, but in a fashion not available to a country-wide program: they are cherry-picking the easier areas, and “fiberhooding” by first generating enough buying interest before building out. Free does something similar in Paris — ask for interest, then go into the neighborhood with the most interest. A government program, however, is pulled by constituents to serve as broadly as possible. Hence, the FTTN approach, later fibering to the premise.

My current experience shows how challenging even that purportedly prudent plan could be. My closest fiber node turns out to be 4,500 feet (as the copper flies) away. A copper mile, just about. A bit long, but theoretically enough for 25 megabits. Should be easy, right?

They came, they installed, it worked; but Internet was spotty. And we had hiss on the line for our phones. My family complained. Tough being the IT guy with a picky family.

Second truck rolled, said he wouldn’t do anything because we had a home PBX. I was out of town. He left. My family wasn’t happy.

Third repair guy came, very helpful, identified a fault in the long copper run of the old phone lines. He switched the broadband port on the fiber termination to improve broadband. It got better, but the hiss remained. It turned out he had inadvertently switched the wires for our two phone lines, so the numbers were messed up. My family hadn’t noticed, but still heard the hiss. They complained. Time and patience were running out.

Fourth repair guy came, he fixed the fault in the long copper run; he swapped the lines back; but couldn’t defeat the hiss. Tried multiple ways of rearranging copper pairs and isolating problems. We determined that our home PBX was amplifying the signal and turning a minor hiss into a major problem. He turned off broadband — and the hiss disappeared. So the hiss came from the broadband running on the same lines as phone service.

When I started Covad, we understood that the copper pair allowed about 1 MHz of spectrum to be used — like a radio over wire. The bottom 25 KHz was used for voice, and everything above could be used for data. (Covad stood for COmbined Voice And Data.) As the loop lengths get longer, the top end frequencies drop, lessening the throughput — but as chips get faster, the algorithms get better, and more can be packed in the wire. Chips have gone from ADSL through several generations to VDSL. We experimented with VDSL back then, but the chips weren’t ready. Today, with VDSL, two copper pairs can deliver 25 — 50 Mb. I should be getting around 25 Mb over a copper mile.

But filters are supposed to separate the lower voice band from the data band. Data sounds like a hiss. That is what we were hearing. The digital hiss of broadband Internet! Sounds like the long-lost cosmic echo of the Big Bang. The Cosmic Hum.

The fourth AT&T tech followed a new approach: instead of bringing in the data and voice over the same copper, he is hunting for unused copper pairs. He will then bring the data in on separate lines. It should work fine — if he can find unused pairs.

This used to be a serious challenge. Back when I first upgraded off dial-up, I tried ISDN. I used two pairs, and got around 128 Kb. I then installed Covad, and got their lowest tier of broadband — IDSL, around ISDN speeds. Then I switched to cable modems, and initially got 3 Mb. It has gotten better, up to 25 Mb or more, as the chips improve. I kept the ISDN lines for a while for additional (and better sounding) phone lines. At one time we had three phone lines, two ISDN and two data lines into our home. Those were the days of homes having kid lines, fax lines, data lines, etc. — and the copper conduit only has a finite number of lines. So a neighborly competition would ensue to grab the most. But now, with mobile phones, most homes drop the kid lines, drop the fax lines, use cable for broadband; or if they use AT&T, tend to use VoIP, not old fashioned POTS lines, for phone service.

As a consequence, I was confident the fourth tech would find two spare pair, and remove the hiss. But nooooo again. The conduit has 25 pair, and not all are used; but many have decayed badly over the years, and were unusable for broadband. In the end, we got one clean line and one noisy line working. Success of a sort. But my speed has inexplicably dropped from around 20 Mbps to 7 Mbps. If the family found out, they would really complain. Back to dial-up, it would feel to them!

More to do to make this really work. By rebooting everything I got it back up to the 15-20 Mbps range. But you get the idea: a painful experience so far. And as the IT guy, I was constantly at risk of disappointing the family. Anything, anything but that!

The broader question is how AT&T, and by extension the NBN, should build out their network. It might be that the cost of managing around the decaying copper is not worth the purported savings; the political fallout of disappointed families would be resounding.

Google may have the better idea. They are doing it the Internet Way, not the Telco Way: all IP network; fiber to the home; leveraging low-cost gigabit ethernet gear; basing voice, video and data services off IP. The Telco Way in contrast is to embed complexity into the network and manage around it, to the consternation and pain of customers — and disappointment of families in their IT guy.

The best analogy for the Google Way is how Tesla re-invented the electric car. The core insight was to use the vast low-cost scale of laptop battery cores, and build up from there. The Auto Way has been to design a custom lithium battery to fit the car, then add range extenders and other complexities to make it more car-like. The result is that Tesla has a cost and simplicity advantage over the auto makers — and a cult-like following by its customers. Do we have a Cult of Comcast, or Adoration of AT&T, in the US?