_____                   _                  _____            _____       _ 
  |     |___ _____ ___ _ _| |_ ___ ___ ___   |  _  |___ ___   | __  |___ _| |
  |   --| . |     | . | | |  _| -_|  _|_ -|  |     |  _| -_|  | __ -| .'| . |
  |_____|___|_|_|_|  _|___|_| |___|_| |___|  |__|__|_| |___|  |_____|__,|___|
  a newsletter by |_| j. b. crawford               home archive subscribe rss
 ____                   _
|    |___ _____ ___ _ _| |_ ___ ___ ___ 
|  --| . |     | . | | |  _| -_|  _|_ -|
|____|___|_|_|_|  _|___|_| |___|_| |___|
     _____     |_|    _____       _
    |  _  |___ ___   | __  |___ _| |
    |     |  _| -_|  | __ -| .'| . |
    |__|__|_| |___|  |_____|__,|___|
    (a newsletter by j. b. crawford)
    home | archive | subscribe | rss


>>> 2024-09-08 private lines (PDF)

I have been meaning, for some time, to write about common carrier switching arrangements (CCSAs). These could be considered an early form of products like "virtual private ethernet:" a private telephone network that was served by the same switching machines that handled the public telephone system. A CCSA is, in effect, a "virtual telephone network." AT&T operated a number of these for both government agencies and large private organizations, and they might be viewed in a way as precursors to the large CENTREX-and-WATS arrangements that became a common fixture of state governments and school districts.

The problem is that I fear I am putting the cart before the horse. CCSAs, and even the fully private telephone systems they were intended to replace, are basically the extreme extension of the private line. Besides, private lines are an important part of the history of computing, as well: they were the pattern for the digital "leased line" services that formed the bulk of computer network connections through the early days of the internet.

Let's ease into it by starting with an important source in telephone history: the Bell System Practices.

Large organizations tend to function like religions. This is more or less overt, depending on the organization and the fashions of the time. For example, in the period we will mostly discuss in this article, a number of large companies published song books. IBM is the best known for this type of corporate spirit, ranging from their de facto hymnal to the "Think" signs customarily installed in IBM worker's offices. The television series "Severance" frequently referred to this aspect of post-war corporate culture, depicting a corporation that unified its employees through songs, quotations, and an internal museum. This type of pseudoreligious corporate culture is, of course, taken to an extreme in "Severance," but nonetheless resembles real practices that still echo through corporate America today.

Severance depicts its ominous corporation with exterior shots of the Bell Laboratories offices at Holmdel, a prestige design by Eero Saarinen that dates to an era in which the architecture of corporate offices was often an idealistic representation of their intended culture [1]. This is just a little bit ironic, as Severance is clearly patterned more after the computer industry than AT&T, which was comparatively subtle in its corporate religion. Still, there were aspects of the sublime: Saul Bass's pitch film for AT&T's 1960s rebrand devotes much of its length to a hagiography of the Telephone Men. Universal Service, AT&T's stated goal, was both as ambitious and ever-changing as "redemption," personified as the enormous, golden "Spirit of Communication" depicted in sculpture and mural at former AT&T offices. And AT&T had signs! Of course they had signs; the best known being the "Bell System Safety Creed" that hung in most work areas.

This is all to preface my use of the word "doctrine." The kind of shared culture that companies attempted to establish through song books is also an important part of practical operations. We associate doctrine mostly with religion, but more broadly, a doctrine is a set of codified, shared beliefs and practices that define the work of an organization. The military has doctrine in this sense, and AT&T has doctrine as well. Much of that doctrine was compiled into a lengthy, almost sacred text: the Bell System Practices, or BSPs.

BSPs are an invaluable source of information, and fortunately organizations like TCI have put a lot of effort into finding and preserving the remaining copies. Unlike publications like the Bell System Technical Journal (BSTJ), BSPs were for internal use only, and so the collection available today is scattered and incomplete. That's not the only challenge when using BSPs for research. Fitting the analogy to holy texts, the organization of the BSPs is arcane and has undergone a number of changes. For the latter half of the 20th century, BSPs were identified by nine digit numbers, separated into three groups of three digits. The first group identified the Division, the second the Section, and the third the Document.

BSP 000-000-000 provides an index to the Divisions. Each division begins with a section index, numbered NNN-000-000. Sections are typically grouped by their first digit, producing topics referred to by NNN-N. NNN-0 is devoted to indexes, NNN-1 is usually a "general information" topic. There are loose conventions for document numbering. For the weightier topics, -100 is often the overview or general procedures. There are as many exceptions as examples of these rules. Further complicating document lookup, BSPs were published roughly from 1930 to 1990, numbering conventions were changed and sections reorganized, and documents were replaced by updates identified by an issue number. The aspect of time can make it a frustrating and time-consuming process to match up a set of BSPs that not only cover a topic but represent a single point in time, rather than being confused by a decade of changes from one to the next.

We can hope that, one day, a monastic order of telephone enthusiasts will take up the compilation and, while they're at it, illumination of the BSPs. They will produce a Xanadu-esque compendium of AT&T doctrine navigable along dimensions of both time and space; a Grand Enfilade of communications technology. For my own part, I am perhaps alarmingly close but not quite ready to take a vow of silence, move yet further into the desert, and devote my life to the task. For the time being we'll have to settle for search engines, the indexing efforts of a few dedicated people, and no small amount of sweat.

Enough of that, let's take a look at divisions 310 through 312, which cover private lines for voice and data, as well as special switched services. A discussion of private lines starts rather naturally with 310-300-100, Two Point Private Line Systems: Two Point Private Line Telephone Circuits, Voice Only---Description.

Of course, we will not start there, because of the joy of BSPs: Naturally, I have searched several archives and not been able to find a copy of 310-300-100. We can fill in some of this gap from 310-300-300 and 310-300-500, though, Test Objectives and Test Procedures. 310-300-300 I1 (1975) 1.04:

The circuits discussed in this section are served over nonswitched facilities with no access to the message network. A two-point private line involves a channel between two terminal locations. The channel may or may not be entirely a metallic path.

A private line is, at its essence, a pair of wires that runs from one location to another. The locations are customer-specified, and the private line is not connected to anything else, only to the two service points. Some version of this service has existed for pretty much the entire history of the telephone, although it has often, as you would imagine, been very expensive.

The private lines covered in this section are for voice use only. This is important, because voice use implies a specific set of requirements. The topic of private lines, leased lines, and related services is actually a very difficult one to discuss succinctly. First, because there is a 100-year history of these services, and the technology, capabilities, and use-cases have all changed over time. Second, because there is a very diverse set of uses for private lines, each with a different set of requirements.

Voice use, for example, implies a customary telephone voice passband of about 300-3400Hz, and more precisely limited by the 8kHz sampling rate of digital telephony. This might strike you as surprisingly narrow, but it's enough for reliable speech intelligibility. It really is quite narrow though, part of the reason that "HD Voice" codecs now offered by VoIP technology sound startlingly different from traditional calls.

The fact that the circuit may not be entirely metallic is important as well. It might be a little unclear what a "nonmetallic" path would even mean, but remember that we are looking at a 1975 revision. By that time digital telephony was in full swing; a customer's "pair from one place to another" may very well be digitized, carried by the TDM switching network, and converted back to analog (analogized?) near the other end. The possibility of carrying private line service over the digital switching system radically reduced the cost of private lines, and enabled the proliferation of "leased lines" for use with computers.

But, it did have limitations. A private line of the 1930s would have been able to carry polarity reversal signaling, a common scheme at the time for alarm monitoring and remote control. A nonmetallic path was not electrically continuous, so it could not be used for any type of electrical signaling. It would only work for the signaling scheme it was provisioned for, namely, voice.

The BSPs describe a set of tests intended to ensure suitability for that purpose. There are standards for loss, frequency response, and noise. We also learn that private lines for voice use can be configured for ringing, so that one end can signal the other to pick up the phone. Elsewhere in the BSPs the supported types of ringing are enumerated, although they no doubt changed over time: ringing by a ring key (today we'd call that a ring button), by a hand-crank generator, by central office equipment, or no ringing at all.

I would like to cover this topic without getting too stuck in terminology, but it's hard to avoid because there is a lot of terminological confusion. For example, what is the difference between a "private line" and a "leased line" anyway? Well, it's mostly a matter of who's talking. AT&T seems to have always used the term "Private Line" except when discussing data services in the context of switched data networks, in which case they use "Leased Line" to refer to a fixed-capacity data connectivity arrangement. We'll get to that in a bit.

Another problem is how to describe the configuration of private lines. A private line provisioned for voice use, as described in 310-300, is not connected to any switching equipment but would be connected to battery power (AT&T called this "common battery," meaning that the phones did not require their own batteries to function). Private lines with common battery were sometimes referred to as "wet," while private lines without common battery were "dry." The latter were mostly used for non-telephone signaling applications.

Private lines might also be connected to other types of equipment to provide useful features. For example, you might wonder about how ringing applied by the central office would work. A popular type of private line for voice, especially later on, was the "Private Line Automatic Ringdown" (PLAR), also just called a "Ringdown" circuit. These private lines are configured so that picking up the phone at either end automatically applies ringing voltage to the other end.


Shameless cross-promotion: I uploaded a YouTube video yesterday where, among other things, I briefly mention that "hotline" is a heavily overloaded term that means different things in different contexts. One of the things that "hotline" often refers to is a PLAR. Mostly, though, I try out a very cheap PABX I bought on the internet and find out that, while not exactly great, it is certainly pretty good for the price.


Another option on private lines was whether they were two-wire or four-wire. A typical telephone uses only two wires, one pair. If you think about it, that's a little bit magical. The trick is a device called a hybrid transformer, installed in each telephone, that basically "subtracts" the signal it is transmitting from the signal on the line in order to isolate the "receive" signal. At a modern telephone exchange, another hybrid transformer splits them out once again for handling by the exchange equipment. This is a two-wire circuit, and has the advantage of lower-cost wiring. On private lines, you can also get four-wire service, with dedicated pairs for transmit and receive (also called, for obvious reasons, talk and listen). The major advantage of four-wire circuits is that they make "conference calling" much easier. When you put a lot of telephones onto one two-wire circuit, it becomes very difficult to manage the cumulative echo produced by the imperfection of the hybrid transformers. A full four-wire system avoids this problem.

What were these private lines for? Well, you can probably imagine a few applications for voice private lines, but keep in mind that for much of their history they were very expensive... prior to the use of TDM digital networks, you were basically paying AT&T for the installation and upkeep of however many miles of telephone wire were required to get between the two locations... plus carrier equipment, line conditioning, etc.

One of the cool things about the BSPs, and one of the reasons they were kept internal, is that it's not at all unusual for them to go into detail on specific customers. We can get a feel for the use-cases for these service offerings from the specific customers that used them heavily enough to merit BSPs. For two-point private line telephone service, that customer is the FAA. The FAA has long had a tight relationship with AT&T, rivaled only by the military for reliance on dedicated telephone infrastructure. Air-traffic controllers used voice private lines both to talk to each other across facilities, and to remotely operate the radios they used to communicate with aircraft.

Voice private lines were also popular in other situations where people across multiple locations needed to communicate in real-time. For example, television and radio networks often had private lines between studios and network control centers to aid the staff in coordinating the start and stop of programs from different locations. The telephone system was an obvious choice, because for the early history of radio and television most networks used "broadcast-grade" private lines (with larger bandwidths than voice lines, especially for television!) to distribute their programming to the member stations. The radio and television networks were a huge business for AT&T prior to the use of satellite transponders for the same purpose.

I know of some more exotic applications as well. Some critical infrastructure or large industrial facilities, and in particular nuclear power plants with their extensive emergency preparation requirements, had four-wire private line systems that linked their control rooms to contingency sites, disaster response agencies, and even the homes of their senior staff. These were configured as conference lines for use in coordinating an emergency response. Indeed, emergency management was another major application of private lines, and I will eventually write about FEMA's nationwide four-wire private line system, still used for dissemination of emergency warnings to state and regional emergency management agencies.

These conference lines are, as you have likely suspected, not really two-station private line systems. They are multistation voice systems, described in 310-405-100. TCI has a much older issue of this BSP, from 1957, and it describes the electrical resistor networks used to achieve the "conference line" configuration in which the talk pairs are mixed and sent on the listen pairs of all stations. 310-405-100 I1 (1957) 1.03:

One of the most important requirements of a multistation private line circuit is that the volume delivered to each receiver be substantially the same no matter which station is talking. The above holds true even when two or more circuits are switched together.

Indeed, and we are still struggling with this today!

1.04 is interesting:

Various signaling systems and combinations of signaling systems can be used to "call" the various stations on a private line circuit. Some types of signaling systems are: loudspeaker signaling, manual code ringing, ringdown signaling, 600-1500-cycle (2-tone) selective signaling, and code selective signaling.

Explaining all of those would be an article of its own, but it is an interesting note that multistation systems challenge the typical sense of "ringing" on a telephone line, and conventional ringing doesn't seem to have been very common on four-wire systems. A common replacement is what is described as "loudspeaker signaling" here, where the stations have an always-on speaker and you get someone's attention by... yelling at them, basically.

AT&T has never limited themselves to voice. There have long been various types of control or signaling equipment available, and 310-435 describes the SC2 Selective Control System available for use with multistation private lines. This telegraph-like system involves a control station sending series of coded pulses, which are detected by satellite stations that open and close relays in response. There are still a lot of products like this today, but the SC2 as a private line offering is a good time to make a point about the historic phone system: it used to be that all telephone equipment was the property of the telephone company. That included the SC2 control and satellite stations, and they had a terminal strip on the outside of the cabinet that exposed the relay contacts and served as the demarcation point between telco and customer property.

That has an interesting implication for these private line systems: the operating company needed a way to test and diagnose them. As a result, larger or higher-criticality private line systems usually included some kind of terminal equipment at the customer location that provided test functionality. On four-wire systems, a loopback test was the norm: some sort of signal sent on the line would cause a relay to shunt the talk pair to the listen pair at the customer premises, allowing the test board at the exchange to send signals all the way "around" the private line. The SC2 had a pretty complicated piece of terminal equipment, since it had to decode the signal pulses, and as a result there is an extensive test procedure with parts that can be performed remotely and parts that must be done on-site. The BSP cautions that arrangements must be made with the customer to disconnect their equipment before a telephone technician starts opening and closing relays to test.

A final common application of voice private lines was for "tie lines," or any of the other names they went by. These were lines that ran between two switching systems, providing something like a long-distance trunk between them. Imagine a corporation with two offices, each with a private branch exchange. The corporation could contract for a private line between the two PBXs that served as a tie line, and the operators at each location could then use it to directly connect calls between the two offices. Besides avoiding the operators having to dial those calls, tie lines could also save money over long-distance calling, if utilization was high enough. A number of services offered by AT&T and operating companies basically amounted to different ways of using tie lines, so Division 311 "Switched Special Services Systems" covers tie line configurations including WATS and operating company-managed PBXs.

Of course, if you have read this far, you are probably wondering about data. In the contemporary computer industry we associate this kind of private service entirely with the leased lines of the '70s and '80s. Division 312 covers Private Line Data Systems and Services, beginning with Electronic Telegraph Loops and teletypewriters. In the 1950s, "data" pretty much meant teletypewriter, and there is a dedicated customer section on Western Union who used AT&T circuits to extend their own network. The first half-dozen topics are devoted to various types of telegraph systems, including DC telegraphy (over continuous metallic circuits) and carrier telegraphy (combining multiple telegraph channels onto one private line using frequency division muxing).

312-8 is where we get to the good stuff, Data Sets. Data set was the term used by AT&T for what we now know as a modem, for example, the Bell 103 modem is more properly the 103 Data Set. The term "data set" predates acoustic modems and is somewhat more general, though, having been used by AT&T to refer to simple relay closure systems as well. Still, by the 1978 312-000-000 index there are a range of different data sets available, covering different speeds and applications. The systems covered in this section are intended for use with a complete private line data system, that is, the lines themselves were part of the system. They were typically private lines that terminated at dedicated equipment in the telephone exchange, where a data bus was used to carry signals between different lines.

It is in Division 313 that we find what we'll recognize today: Voice and Voiceband Data Circuits. These are data systems that operate over voice-type lines, using acoustic modem methods to encode data within the frequency response of those lines. This division is actually surprisingly short, for a reason. 313-100-100 I2 (1982) 1.01:

Telephone company testing of circuits that terminate in customer premises outlined in these sections includes only that portion of the circuit up to the network interface. The circuit testing procedure does not include customer premises terminal equipment.

In practice, a lot of voiceband equipment would be provided by AT&T, but they are establishing a clear separation of concerns between a voice-type private line and the modem used with it. It's important to understand that "network interface" as used here has a specific meaning within the telephone industry, it is the demarcation point between the telephone network and the customer's equipment.

Higher-speed telephone modems like the Bell 201, capable of 2400bps, could be used on four-wire circuits for full-duplex operation. Indeed, four-wire circuits were quite common for data use as well because they enabled full duplex operation. AT&T was also not the only option for voiceband modems. Division 314, Digital and Analogue Data Transmission Systems, is dedicated to a variety of modems and I49 (1983) includes familiar names like IBM and Data General.

This practice of using voiceband equipment to put data over private lines, without requiring extensive special equipment at the central office, started a shift in data communications practices that greatly blurred the lines between different types of service. Some of the confusion of terms we encounter today comes from these gray lines. The Bell 103 is a data set and can be used on private lines, but it can also be used on a conventional dial line. Similarly, the lines between voice and data were blurred. SAGE incorporated a nationwide computer network that operated over voiceband modems, often considered the first precursor to the modern internet. The digitization of the telephone network would further complicate definitions.


I have a long list of telephone-related topics to cover and I will probably never get to all of them, although I sure will do my best. It takes quite a bit of my free time to write these articles, and I'm also starting work on a more ambitious project around telephone history in more of a reference format. I'd appreciate your support in pursuing these projects---consider supporting me on ko-fi, which will also get you my subscribers-only newsletter EYES ONLY.


During the 1960s, AT&T introduced TDM digital trunks to the telephone network. Using digital technology, a large number of telephone calls could be digitized into samples and those samples multiplexed onto a single high-speed data connection between two telephone exchanges. This method of multiplexing was more reliable and less prone to noise than analog FDM methods, and it could be adapted to a wide variety of carriers. Over the following decades the telephone network underwent a wholesale conversion to digital, and it is now typical that the only "analog" parts of an analog telephone call are the last mile connections between the exchange and the customer premises. Modern telephone exchanges digitize calls at the line cards and it remains digital until reaching a line card on the other end. We were doing data over voiceband, and then we were doing voiceband over data.

The implied result, data over voiceband over data, was in fact very common and the apex of dial-up internet standards (v.90/v.92) assume that the underlying telephone connection is digital. The oddity of stacking multiple layers of digitization was far more apparent on private lines, though, where there was no requirement to retain compatibility with a standard telephone loop.

And thus the leased line was introduced. The closest thing I know of to a technical difference between a "private line" and a "leased line" is that a private line is assumed to be private over the entire span, while a "leased line," in practice, refers only to the last-mile. The actual telephone network is digital and using TDM or even packet switching methods, there is no need for a dedicated physical connection between two central offices to carry data. The leased line just allows a customer a way to insert data into the telephone network.

While v.92 assumes a bidirectional digital connection, it is limited to 56kbps because of properties of standard phone lines including the use of companding. Leased lines didn't have to put up with this limitation, they could omit the functions of a telephone line card and instead deliver a digital signal all the way to the customer premises. While the telephone network itself went through several major iterations of digital media (including the charmingly named "plesiochronous" network), the last-mile digital connection to customers has been well-standardized for a surprisingly long time: the T-carrier.

Let's talk quickly about the Digital Hierarchy, the scheme around which the TDM telephone network was designed. TDM involves packing samples into time slots in a round-robin fashion, so the digital hierarchy is similarly organized around cycles. As a hierarchy, those cycles get larger and larger. A telephone call consists of 8-bit samples at 8kHz, which multiples to 64kbps of data. That 64kbps channel, in the digital hierarchy, is referred to as DS0 or Digital Signal 0. By the somewhat arbitrary but pragmatic design of the digital hierarchy, 24 DS0s are multiplexed to form a DS1. These DS designations refer to the actual payload, not to the carrier technology used to transmit it. DS1, in the United States, was most commonly carried by Transmission System 1, commonly called T1 or T-carrier. 24 64kbps channels adds up to 1.544mbps, and that's exactly what a T1 delivered.

This connection was digital the entire way through, with no need for conversion to analog and the resulting quantization noise and bandwidth limitations. T1 was originally designed for trunk connections between telephone equipment, but it became quite natural to extend T1 connections to customer sites as a form of high-speed (for the time) data. T1 was the typical format of the leased line even into the '00s, and most people my age probably remember having dial-up service and coveting the remarkable speed of a T1 connection.

This is not to say that leased lines were limited to T1. There were higher-speed systems used within the telephone network that carried higher steps on the digital hierarchy, DS2, DS3, and so on, and these could be extended to customer premises as well. At the high end, larger businesses would be placed directly on a SONET fiber-optic ring via add-drop multiplexers, an arrangement capable of multi-gigabit speeds.

It's a little odd, actually, that the "leased line" we think of has very little to do with actual telephones. It's a digital network connection much like we use today, except that it functions on top of the provisioned-bandwidth, synchronous, TDM network originally built in order to carry telephone calls. There are probably still organizations today running ethernet over SDH for their internet connection, and it won't feel much different from anything else we use.

[1] The use of architecture as a symbol of corporate power, once a fundamental part of the computer and telecommunications industry, is largely lost today. There are more than a few reasons, but one of them is the development history of Silicon Valley. It is remarkable how underwhelming the offices of today's most powerful corporations are, consisting of scattered low-rise office parks with no identity besides the earth tones and angles of the 1970s. Of course, when modern tech companies do build prestige headquarters, they tend to be unspeakably ugly. I'm speaking mostly of Meta, I will give Apple's effort a mediocre grade.