> I was wondering if the modems in that application were supplied by IBM
> (who appears to have developed the technology) or by AT&T. My
> understanding that AT&T's "Dataset" modem-telephones didn't come out
> until the 1960s.
I found some additional information on the above:
It appears the modems for the 1950s units were developed and
implemented by IBM, not AT&T. They used four signals to take
advantage of the 4 Khz range of a voice grade telephone line giving an
effective transmission rate of 1200 baud.
The information was sent from punched card to punched card. This was
an advantage over the prior method of converting it to paper tape and
back again for transmission.
The data was converted from Hollerith code to a special 8 bit code in
which there was always four bits to represent a character.
Considerable error checking and control protocols were included --
these were not present in the paper tape method -- and this was
considered a major feature of the system.
The passage said that AT&T strictly controlled attachments to their
lines; the IBM system was used mostly on private lines or leased
lines. As we recall, many large organizations, especially railroads,
maintained their own privately built and maintained telephone networks
and such users could of course attach anything they wanted. Railroads
could use this IBM system to send in freight car movements punched at
remote locations to a central site.
But I wonder if AT&T allowed private attachments to leased private
lines it supplied. I wonder if the rules were different for such
lines as opposed to the switched network. I also wonder if the
independent telephone companies were as strict as AT&T regarding
It was hard to tell from the passage just how many units were out
there actually in regular revenue service as opposed to specialty and
demonstration units. My guess is that there weren't very many in the
1950s; it was probably cheaper and adequate in those days to mail
source documents to central HQ to be keypunched there rather than
keypunch them remotely and transmit them. Reproducing cards is a slow
The early card-to-card systems used modified keypunch machines.
Now, around 1960 IBM began to offer a number of "tele-processing"
products and I suspect at that point volume did indeed grow. I don't
know what AT&T offered as modems in 1960 before their "Dataset" was
introduced. Around 1964 IBM introduced commnication systems that used
its new Selectric typewriter as a term. My own bank began to use them
Relatively early on IBM introduced an audio response unit from Touch
Tone queries. I remember another bank having a side Touch Tone keypad
next to a rotary phone for such inquiries around 1967. Supposedly one
customer for this system was AT&T itself to provide automated
route-rate information for operators. At the time I thought such
systems were a smart idea; of course today seeing how maddening it is
to use them I feel a little differently.