3 ms·

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 tha

• ajb 5y ago
  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.
  • varjag 5y ago
    There is no sampling/discretization on analog phone line. And since they mention crosstalk issues this has to be analog.
    • jcrawfordor 5y ago
      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.
• jcrawfordor 5y ago
  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.
  • varjag 5y ago
    Thank you so much for this detailed explanation.