History AT&T early modem developments? [telecom]

As someone who has worked with and for modem vendors since 1965, I have also tried to find a good written "History of Modems". Most people today think modems were invented at the same time, and just for, the Internet.

By the mid 1960's, many types of modems were available not only from AT&T/Bell, but such as Rixon, GTE Lenkurt, Milgo, Codex, Paradyne, Anderson-Jacobsen, Vadic and others.

Most "history" articles on the Internet today tend to be the same basic, general info re-hashed. I have found the following sites to be of some value re the really early development stages.

--reed

Lack thereof? AT&T had their own huge TWX data network at the time.

Not relevant, as most of the leased private line voice systems out there were leasing from AT&T.

--scott

AT&T's TWX network dated from at least the 1930s. But it was direct pulse from the subscriber.

AT&T had a separate tarrif for "20 ma" and "60 ma" lines in which pulses were directly transmitted. These were discontinued some time ago, but were used to drive all the teletypewriters out there.

I don't understand. What I'm saying is that back in the 1950s the quality of the dial-up long distance network may have been inferior to that of private lines.

As mentioned. the bandwidth of voice long distance in that era was narrowed to 2-3 Khz instead of 4 Khz in order to squeeze out more capacity on limited toll circuits during the war.

***** Moderator's Note *****

I don't think your take on the "60 speed" TWX is correct: AFAIK, they used ordinary phone numbers connected to regular business exchanges, and that would require modems.

TELEX, OTOH, _did_ use DC signalling, with specially constructed step exchanges.

Who is our resident TWX expert? Comments?

Bill Horne Temporary Moderator

(Please put [Telecom] at the end of the subject line of your post, or I may never see it. Thanks!)

Right. This is what I mean by a "data network."

Note that the TWX network by the 1950s was using FSK modems to transmit data, so that many different data circuits could be put onto one voice line. The idea basically came out of the radioteletype world. The loop line to the customer was DC, but the cross-country trunk was FSK.

When we talk about "modems" we are talking about two different things. First of all, there are high speed modems that are intended for use on wideband leased lines. That's what IBM came up with... they had some repeating machines where you could put a deck of cards into a reader and have copies spit out of a punch many miles away. This was a big deal for far-flung business organizations and was the predecessor to RJE. This is basically a data network, running over analogue lines.

THEN, we have the dialup modem, which is really just a cheesy workaround to get low-rate data over voice-grade lines, so people could use existing infrastructure for data.

Most of the private line networks were built with trunks leased from AT&T. The railroads had their own voice networks running on their own cables, but if you called from one Ford Motor Company office to another Ford Motor Company office over Ford's internal voice network, you were doing it over lines leased by Ford from AT&T.

By the sixties and seventies, many folks were using dialup lines to connect their local TWX machine to the network. This is fine for outgoing messages, but the remote service has no way to inform you of urgent incoming messages.

--scott

That's what I meant. The customer loop was DC; after that was carrier. But to the customer that was all transparent.

Indeed, from the Bell System History it appears they used a crude form of modulation very early on to superimpose telegraph signals over toll lines separate from the voice signal. This got tricky with loading coils and repeaters since the telegraph signal had to be handled differently than the voice signal.

Exactly. But I don't think they were "wideband", just regular voice grade. From a telegraph/data point of view, a voice grade line had far more capacity.

What I'm trying to find out is the history of this. Apparently these came out in trial form in 1958 which a production product in 1960. It seems there would've been a demand for these sooner, but maybe not.

I can't believe a company would lease a private line just to send data at the end of the day. Perhaps such companies used the lines for intermittent message activity.

TWX goes back to the 1930s (see the Bell System history vol 1). As Scott said, they used DC lines for the local loop; no modems at the customer end. Later on they may have used modems.

Teletypewriters connected to computers used modems, though I believe some early data services did use TWX or Telex.

***** Moderator's Note *****

The "crude form of modulation" you're thinking of was called simplexing or duplexing, and didn't involve any modulation. Transformers, called "Repeat Coils" in phone lingo, were wired into toll circuits to provide "phantom", "Ghost", or "Wraith" circuits, deriving as many as seven circuits from four pairs of wire. Do a Google search for "phantom circuit"; you'll see plenty of links.

The information I've seen via Google leads me to believe that the 60-speed TWX network could be used either way, although the connections were switched by operators and may have been DC to the customer premise..

After ~1961, both 60 speed and 100 speed (110 baud, my mistake) machines used modems end-to-end, and could dial their calls directly.

Bill Horne Temporary Moderator

(Please put [Telecom] at the end of the subject line of your post, or I may never see it. Thanks!)

The computer industry developed extremely rapidly in this time frame. Just a little earlier it was limited to things like the SAGE air defense system and the SABRE airline reservation system. In the latter IBM multiplexed a number of ticket agent terminals over a single voice grade leased line. And there were the IBM card transceivers. But quite soon there was a big demand for the kind of on-line data collection and dissemination you have in mind; and the dialup network rapidly achieved the ability to handle that kind of traffic for most customer locations.

In the same time frame there was the rapid shift from person-to-person communication, which TWX was designed for, to person-computer communication. At first you would send data at the end of the day by TWX, say, and the message would be handed to a keypunch operator to prepare it for the computer. Pretty soon you could send the data, perhaps punched into paper tape at the originating point, direct to the computer room as paper tape, and then directly into the computer as a stream of bits.

There was also a brief period when special-purpose computers took over control of telegraph systems from wired-logic electromechanical switching offices. IBM had the 7740; G.E. the Datanet-30; Collins the C-8400; and there were a few others (and the DOD had AUTODIN). These could be used just to run a message switching communication network, or to connect such a network to a data processing computer, perhaps through a shared-access disk drive. What I mean by "brief" is the transition from what is mostly a people-to-people message passing network to a people-to-computer sort of operation which could handle message passing on the side. Special purpose communication front-end computers continued after this period, in addition to wired-logic equipment for attaching mainframe computers to communication lines.

This topic covers a lot of ground.

Both AT&T and WU had carrier systems, or what we now call frequency division multiplex, that carried multiple telegraph circuits on a single voice channel, from the 1930s. WU also made extensive use of time division multiplex until the 1950s.

AT&T introduced TWX in 1931, using manual switchboards and DC loops from the switchboard to the subscriber. No doubt carrier systems were used to get inter-city trunks. This made the telegraph companies, W.U. and Postal, really mad, because it cut into their telegram business for the large business customers; but they didn't offer an effective competing service. Telex began in Europe about the same time, and differently in different countries. It was especially well developed in Germany, using a dialup network completely separate from the voice network, and using DC loops between the subscribers and the switching offices.

You could lease telegraph grade or voice grade private line networks from AT&T, and telegraph grade from W.U. These might be as simple as a line between two points, or a complete telegraph message or voice switching system with hundreds of points. Over short distances you could send any sort of DC signaling over a telegraph line; but over long distances there were regenerative repeaters in the line. These limited operation to Teletype standards of 5-unit code, 60 or 75 wpm. You could connect about anything you wanted to a voice grade private line (assuming you didn't transmit at too high a level) and in later years they could be conditioned (equalized) for data transmission using third-party modems. The SAGE air defense system had a lot to do with getting the Bell System into this environment of transmitting high speed data over leased lines. Unlike private lines, TWX and the voice switched network came under the "no foreign attachments" doctrine, meaning you couldn't legally connect anything of your own to the line; everything you needed was supplied by the telephone company.

In 1958 W.U. introduced Telex to the U.S., using switching equipment made by Siemens in Germany.

Some TWX subscribers were served using modems, which were single channels of frequency division multiplex carrier system. This was just for the convenience of the telephone company which might not have a suitable DC circuit going where the customer needed it. Or which might want to server several TWX subscribers over a single voice-grade line.

Circa 1959 the Bell System offered a modem, at the time called a "Digital Subset" for lease to subscribers. These were used, for instance, in the FAA ADIS weather data collection system, which operated a nationwide backbone circuit at 600 wpm. About the same time the Bell System set up a teletypewriter switching system for Delta Airlines, operating over the voice switched network using modems. The terminal stations had to be served out of #5 crossbar offices, so in some cases there were some long foreign-exchange lines.

During the same time frame the voice switched network was rapidly approaching a state of completion, in which customers could dial all long distance calls themselves and have the billing done automatically. (But as late as the 1970s I lived in a village where you could dial your own calls but then had to give your number to an operator for billing purposes.)

The billing problem was one reason TWX stayed so long with manual switchboards, while Telex was dialed in most of the world. Nearly all TWX and Telex calls are long distance calls. In the U.S. customers expect an itemized bill showing the place called and the number of minutes charged for. This was easy to do with manual switchboards. In Germany instead of an itemized bill the customer gets charged for a number of pulses. When a Telex connection is made a pulse counter associated with the caller's line is connected to a source of pulses, the pulse rate depending on the distance called.

As the voice switch network neared completion the Bell System decided to convert the whole TWX system to use dialup service over the voice network. The volume of TWX traffic was tiny compared with the rapidly growing volume of voice traffic, so it could be handled by the voice network without any substantial additions. There was also a desire to introduce ASCII in place of Baudot (Murray) code, but the substantial number of Baudot subscribers had to be accomodated. The end of all this was to develop modems to place at the customer locations and operate the Teletype machines over the switched voice network. There were some special arrangements to allow Baudot machines to call only other Baudot machines directly, and to go through speed/code converters to connect with ASCII machines, and vice versa.

These early TWX modems appear to be terribly overdesigned; but one must remember that the voice network was not all that good in the early 1960s. The modem design tried to guarantee a satisfactory 110 baud connection for almost all calls between any two points on the network.

It was also circa 1960 that Bell System people began to realize there was a substantial market for dialup data transmission between arbitrary kinds of business machines, even at the rates charged for voice calls. This was the result of rapid growth in online computer sytems, beyond the special cases like SAGE and airline reservation systems. Since Bell couldn't supply all the business machines, they altered the no-foreign-attachments rule by offering to lease dialup modems of various kinds. There were the 103 series at 150 baud, full duplex; the 201 series at 2000 baud synchronous, one way at a time, the 202 series for 1200 baud asynchronous, one way at a time. And there were modems for fax and Telautograph and other odd things.

W.U. tried over and over to get the government to force the Bell System out of the telegraph and data business. Government policy seems to have been to keep W.U. with exactly one foot in the grave at all times.

Very informative and helpful response, as always. Thanks!

ANI (automatic number idenitification) was a pain for the Bell System to install because it was a cumbersome retrofit in most cases. All switching systems were designed to go forward, not backward. There were various schemes to 'hunt back' to find the originally subscriber line (which could require a directory crossmatch). The equipment was expensive for its value, although as time went on it got cheaper. The switching history goes into detail on this.

In the 1970s my office in downtown Philadelphia still had ONI. I was surprised it was used for suburban message-unit calls, our house in the city didn't have that.

I don't know the actual percentages, but by 1970 a good portion of the Bell System still did NOT have DDD. (Heck, a tiny percentage was still manual in 1970). In the 1970s they introduced new long distance rates where the initial period was one minute (instead of 3 minutes) and the rates were steeply discounted for overnight and weekends. At that time the tariff allowed that calls from non-DDD areas would still get the DDD rate, as would operator calls where the customer had trouble dialing.

(The Bell system published a statistical summary every year that broke down these sort of details, plus some details were in the Annual Report.)

(It's ironic that at that time some exchanges were stll manual while some had IDDD (international direct dial). I thought IDDD required ESS, but only #5.)

Would you recall what proportion of TWX subscribers wanted ASCII ("4 row") instead of Buadot ("3 row"). Did the Teletype model 33 cost more than a 32 or earlier model?

It seems most stuff the Bell System did was designed for very heavy duty service and long life. Some might say "overdesigned". The history writeup on ANI units, for example, suggests that as time went by they made simpler and cheaper units, esp for SxS and panel offices that didn't have much of a remaining life expectantancy. Likewise for Touch Tone units. The gear devised for PBX usage was cheaper since the "service requirements weren't as demanding" and the "risk of a wrong interpretation wasn't as big".

Oslin's history of Western Union mentions this repeatedly. It did seem rather unfair to allow the Bell System to compete in the telegraph transmission business against Western Union. I was surprised to read that the Bell System had as much private telegraph mileage as did Western Union. I would think this competition would prevent Western Union from developing necessary economies of scale.

It was also foolish for the business community to have two separate switched telegraph networks, although I think toward the end (before the WU aquisition)they provided an interface to each other.

Please provide more details on WU's TDM system(s). I didn't know that TDM was in use before the 50's, so I'm very curious about these.

[snip]

The modems used for the 4-row side of TWX had an important feature not found in equivalent units from other manufacturers: because they had to inter-operate with the 3-row machines running at 60-speed, they could respond to the presence of BOTH a mark and a space tone being received at the same time, which was a RESTRAIN signal that turned off the tape reader and locked the keyboard until the 60-speed machine at the other end of the connection could catch up. There was, of course, a speed and code converter in the middle of such calls, but it had a very limited buffer.

[snip]

Bill

The first time I saw a modem, as a teenager in 1968, I was fairly unimpressed. I had already been a ham operator since age 13 and thought to myself "Look at that. RTTY by telephone."

- Ron

In article , Ron Kritzman wrote:

And that could lead us into another long story, teletypewriting for the deaf. There is a good book "A phone of our own : the deaf insurrection against Ma Bell" by Harry G. Lang, Gallaudet University Press. The quick summary is that it was a deaf ham, Bob Weitbrecht, W6NRM, who was an avid RTTY operator and also transmitted TTY by phone to some of his nearby neighbors over the phone. When Bob became friends with Jim Marsters, a deaf man who was simply unable to copy Morse code and get a ham radio license, he attempted to use his modem over a distance of 300 miles or so. The modem was simply single-tone on the spacing signal only, which worked very well over local phone distances and avoided the need for a send-receive switch (or for a full duplex modem as Bell was using for TWX). The problem with greater distances was echoes. Bob was a genius at doing things simply; and discovered that adding a marking tone at a fairly arbitrary frequency would drown out the echoes and allow space-tone-only copy. He cut off the marking tone at the end of every character to preserve the break-in properties of space only signaling. Bob's modem used an acoustic coupler to the telephone handset to avoid the no-foreign-attachments rules of the phone company. At first the Bell System had no use for this at all, thinking that their TWX service was exactly what the deaf people needed. Later they came to realize that deaf people were paying quite well to make long-distance telephone calls. Still later the political climate changed in favor of subsidizing assistive technologies for the handicapped.

***** Moderator's Note *****

If I'm not mistaken, I think the TDD's use FSK, not single-tone signalling. The carrier is on only when the user is typing, so it allows for "semi break in", and can work with answering machines.

Bill Horne Temporary Moderator

(Please put [Telecom] at the end of the subject line of your post, or I may never see it. Thanks!)

As a reminder, we have the Western Union Technical Review on file here. I don't know if it's indexed or if there's an issue specifically on this subject, but it is interesting going through them. The first page is a contents of the issue. Please see:

I'm not familiar with that one. Is that vol 4? I think more books were planned but then divesture stopped the effort.

Initially, the Teletype 33 ASRs we used for computer time sharing terminals had the built-in modem and automatic dialing; and had an orange lamp "REST". We never knew what it was for. Also, for some reason our machine was Touch Tone, which was unusual in March 1968 (for me, marks forty years of this stuff.) It was also all number, which for us used to lettered exchanges was a little cumbersome.

Our later Teletype 33s didn't have the automatic modem. Some had the acoustical coupler. Some had the interface block and a modem which was activated by pulling up the 'hold' knob on the phone. (They used a two-line phone which had a pull-up knob.) Apparently this arrangement was cheaper than renting the TTY from the phone company. (But maint service was free, an important consideration given high school kids were rough users).

***** Moderator's Note *****

I'm surprised that you were able to get that model of 33 ASR to work with time sharing systems, because AFAIK all the "TWX" machines (if you had a RESTrain light, that's what it was) were wired for half-duplex and local echo. I know, because I used the one at Back Bay Toll in Boston to work with Ward Christianson's bulletin board in Chicago, and it ddoouubblleedd eevveerryy cchhaarraacctteerr I typed, because in addition to the local echo, the BBS echoed everything it received.

Although it was possible to set the bulletin board so as to turn off character echo, there was no way to do it on a call-by-call basis, so Ward couldn't cure the problem for me: either your time sharing guys had gotten that figured out or all your machines were doing local-echo.

The first time I logged on (from school) with a 300 baud modem, Ward cut in and said "Hey, Speedy!". ;-)

Bill Horne Temporary Moderator

(Please put [Telecom] at the end of the subject line of your post, or I may never see it. Thanks!)

IBM developed that in 1941 in response to the War Department's request. They needed to centralize record keeping and inventory control for the worldwide operations. Both telegraph lines and shortwave radio were utilized.

The book, "Building IBM" by Pugh has some background on that project.

I don't know much about shortwave radio, but I think by WW II the transmission principles were pretty well understood. I'm pretty sure individuals could buy shortwave receivers to listen to broadcasts from other countries. I don't know if SW was sent via FM or AM.

The Bell System used a combination of shortwave and longwave to send overseas telephone calls until they got the undersea cables in the

1950s. The Vol 1 book previously cited goes into considerable detail about radio wave propagation and reception. Apparently there were certain times of the day and certain days of the year when atmospheric conditions were more favorable toward longwave than shortwave. My impression was than on overseas calls a technician had to carefully monitor the signal and make compensating adjustments. An awful lot of experimentation was required on every component of the overseas system to optimize it.

They also had a crude encryption by some frequency shifting, but a skilled listener could decipher it. Then they were a little more sophisticated, but a skilled person could build a device to decipher it. (The book goes into the technical details which are beyond me.)

Transmitting telephone signals was much harder than transmitting someone at specially designed station. The signals had to mesh with the landline network and instruments in common use at the time.

The opening of the translatlantic cable must have been a tremendous boon, but I suspect calls were stilled handled on a delayed rather than a demand basis.

***** Moderator's Note *****

Long and Short-wave broadcasting was always AM, since FM requires too much bandwidth to be useable in the longwave and shortwave broadcast bands, and it was invented long after broadcasting had started anyway. In later years, there have been some attempts to save on bandwidth and power by using transmitters with single-sideband technology, with a carrier left in to allow old AM receivers to work. It never became a common standard, although you'll sometimes hear military broadcast stations using it.

I think the "encryption" you're thinking of was actually a single-sideband transmission: AT&T had a radio system that sent a number of SSB signals together, the same way that L carrier did. I don't recall the details offhand.

Bill Horne Temporary Moderator

(Please put [Telecom] at the end of the subject line of your post, or I may never see it. Thanks!)

I don't have any information on this handy. Probably the 33 didn't cost much more than a 32, because there were just a few added parts and the 33s were being made in much larger quantities than the 32. As for subscribers' wishes - well it was still in the era of "whatever the phone company supplies is what you want." I don't know that the customers were given a choice - perhaps all new TWX installations were ASCII, and then the old ones were gradually changed over to ASCII.

True, but I meant overdesigned in the sense of dealing with signals of poor quality - high transmission loss, poor phase linearity, echoes, etc.

There is a lot of information in the Western Union Technical Review, which is in the Telecom archives.

Recall that time division multiplex goes back even earlier than 1870, but was standard with the Baudot printing telegraph system of that period. The technology involved rotating faceplate distributors and high-tech (for the time) means of keeping them in synchronism at the sending and receiving ends. This was under continuous development until 1950 or so, with W.U. having a lot of mux circuits. Teletype Corp. had a multiplex system in its catalog through the 1930s. There was even radioteletype operation of Western Union TDM equipment during WW-II.

The signal sounds like FSK, but in fact it is single-tone signaling. The other (mark) tone is only there to drown out the echoes, and is of rather arbitrary frequency. Tho in fact the patent has frequency- shift in the title. (pat. 3,507,997)

One shortcoming of Pugh's book is that you would think that only IBM had any kind of radio-typewriter transmissions. In fact RTTY, radioteletype using ordinary Teletype equipment, was much more common before, during, and after WW-II.

To pick up on a couple of points...

When I said the Model 33 probably cost about the same as the Model 32, I was only talking about the bare machine. The TWX version of the Model 33 and Model 35 machines was quite costly because of the complex call control unit that connected to the 101 Data Set (modem). There was a 99 wire cable between the two, which around Teletype was considered the epitome of awfulness. Then there were so many variations of the TWX call control unit: rotary dial, TouchTone dial, card dialer, and loudspeaker versus hand receiver.

There was a plan, which I believe was never implemented, to have a set of call-progress lights on the machine that would show dial tone, ringing, busy tone, reorder, connected, etc. I don't believe they ever developed a modem that could reliably recognize all these conditions.

I was told that some of the Bell operating companies avoided the high price of the TWX Model 33/35 equipment by buying the simple private line models and attaching them to 103 Data Sets, which had all the telephone stuff internal to the modem package.

The 100-series Data Sets, or at least the 103s, had strapping options allowing either of the two tone pairs to be used for originate vs. answer direction of transmission, and for either tone of either pair to represent mark. These three options allowed the same modem to be used for eight different mutually-incompatible services. One of these was TWX, another was Data Phone which allowed voice and data transmission and was charged for at voice rates. Then there were WADS (wide area data service) and something called WADS-Prime that was yet another service. WADS was shot down by the FCC; and I never knew what WADS-Prime and the other services were intended to be.

Concerning time sharing, a lot of the early time sharing systems were half-duplex. I well remember the computer prompting for Login: and then it would prompt for Password: and type a bunch of overstrike characters to create a black mess so that your password would not appear readable on the paper.

Jim,

I am always amazed at how much I don't know about what I don't know.

Since the patent claims both partial (11 ms) and standard (22 ms, 31 ms stop) marking tones, let's call it a draw: it isn't "real" FSK, but I thought it was. ;-)

Bill