Coaxial Goes to War | Hacker News Reader

Coaxial Goes to War

computer.rip

112 pointsSniffnoy17 comments

Because phone conversations required around 15kHz of bandwidth (assuming no companding, which was not yet done at the time) this imposed a big limit...

Is that really true? In my home country the speech band was passed at 300-3000Hz, meaning 2x2.7KHz full duplex OTW. 15KHz sounds unnecessarily luxurious for speech.

Maybe this is counting both directions separately, and doubling again because of sampling vs nyquist? (Not sure the latter actually applies if you're doing everything in analogue, which they would have been at the start).

They probably also didn't have very sharp filters at the time, so couldn't optimally pack the channels.

There is no sampling/discretization on analog phone line. And since they mention crosstalk issues this has to be analog.

Yes, at the time the only real theoretical bandwidth requirement was the highest audio frequency they wanted to be able to pass, since this was all a trivial analogue setup. In practice there were a surprising number of extra constraints, some things you'd expect like the rolloff of the filters and amplifiers, but others are the kind of thing that only come up when you have very long cable runs with hundreds of amplifiers... they had to think very carefully about how specific harmonics would accumulate through all the amplifiers and place carriers accordingly, so that the harmonics that accumulated to a significant degree would fall in unused frequency ranges. All of this tended to increase the practical bandwidth requirements, or in other words get you fewer channels than you would expect on a given cable.

A sort of interesting issue is that the worst crosstalk problems were always at low frequency, so most of these carriers operated entirely above audio frequency. The audio frequency was thus an available circuit, but the quality was poor, so it was usually used for order wire or other internal purposes. This had the very convenient property on many open wire and cable systems that a lineman could connect their butt set directly to the pair and use it to call an exchange troubleshooting desk, the butt set didn't need to have any awareness or support for the carrier technology since the AF band was otherwise unused. This is very similar to how DSL works today.

Honestly that number is not a super reliable one, because it varied to a surprising extent by specific equipment. On early carriers the filtering technology was poor which meant that they had to provide a lot more room than the actual passband. On later carriers they started using companding and other techniques to get it way down to a couple kHz.

One of the issues which complicates this and that I didn't really mention at all is that AT&T identified different grades of circuits - toll grade circuits had a wide passband and the best possible noise properties and so could be used for any purpose including demanding ones such as radio network audio feeds (where the narrow passband of normal telephone lines would be completely unacceptable). On the other hand, it was common to have "pole pair" low-grade analog circuits that were mostly used for backconnection of voice calls from a tandem switch to a smaller local exchange office elsewhere on the line (basically add-drop on these toll leads was very hardware intensive so they would often centralize it in larger cities and then "double back" to towns further back on the lead, sort of like getting off the freeway in a rural area and having to take a frontage road back the other way for a while). These had far more relaxed quality requirements. Open-wire circuits tended to be almost entirely toll grade, but later carriers including cables typically had multiple grades with fewer channels of the high grades.

The scheme L-3 introduced of having either two voice mastergroups or a television channel presaged the later approach the telephone network would take, with hierarchical TDMA structures that allowed traffic planners to "pick and choose" how large of a slice they wanted to pull off. This continues basically to the modern day, but is clearest in the example of ISDN's DS0, DS1, DS2, etc that used to be commonly used for IP. Each successive DS is just a combination of more banks of DS0 (64kbps). The point is that they were sort of doing the same thing with the analog carriers but by choosing wire gauge, amplifier quality, etc. to fit the need.

There was a surprising range of different grades because of the surprising range of major uses of Long Lines at the time - radio and television relay were the most demanding, carrier telegraph and digital data had strict parameters to work reliably, voice and non-carrier telegraph could accept a pretty narrow bandwidth with relatively high noise.

Thank you so much for this detailed explanation.

That was a very interesting read on AT&T's coaxial long distance network and what it was used for.

There seems to be some homage being paid to the old L-3 system by the company that has been named variously L-3 communications, L3 Technologies and L3Harris "for the last initials of its founders". Probably an inside joke for those in the know.

https://en.wikipedia.org/wiki/L3_Technologies

Minimum radial bend is a big deal in wave guides. If you look at TV transmitter masts you may well see pretty fat cables and ducting. The ducting can be a wave guide or just weather cover for cables.

Thick Ethernet was coax, fat coax, marked where to put the taps for maximum signal strength.

... where to put the taps for maximum signal strength

It's not that signal strength varies along the cable. It's about minimizing the effect of reflections. Some reflection is unavoidable when you tap into the cable.

"Transceivers should be installed only at precise 2.5-meter intervals. This distance was chosen to not correspond to the signal's wavelength; this ensures that the reflections from multiple taps are not in phase.":

https://en.m.wikipedia.org/wiki/10BASE5

Once upon a time in a former life I had an industrial customer that used 10BASE5 and it ran all around their factory. One of the many jobs I did for these folks was installing vampire taps and AUI drop cables.

As you mention the optimal placing of a tap should be on these marks, or as close to as possible. Unfortunately the reality was that where these cables ran and where the taps needed to be fitted and where the dots actually existed never really aligned well, especially in high up confined crawl spaces. We soon discovered that 10BASE5 is pretty forgiving even if you install a tap in the least optimal positions, e.g. dead centre between the dots.

Previous installers even managed to install two to three taps more or less next to each other and there were no complaints about performance degradation.

But, one day one new tap takes the whole network down.

We used to joke that fixing campus CATV cable meant removing the end-tap over night to let the surplus electricity drip out. The thing is, during this time 2mbit telco cabling sometimes was a gas filled conduit, and your comms data frame included pipes with brass taps, and small buckets: they got leaks, and you had to drain water sometimes.

What a great read. Who would have thunk the history of coaxial cables could be so interesting?

There a lot of DCO Exchanges that still refer to facilities for all major fiber routes as part of the "Long Lines" infrastructure out of respect for these original coax runs.

That page reminds me so much of TOTSE (oh how I miss you) with a dash of Sick Sad World thrown in. Ahhh the nostalgic feels.

TOTSE was definitely very influential on my interest in the telecom system back in the early '00s. To this day, researching things like phone switch replacement projects I'll often end up finding text files that I remember being on TOTSE back when.

Reply on news.ycombinator.com