[GreenKeys] Further to Dataspeed Type 5

[Jim Haynes]

You'll note that the tape reader is a CX with adjustable guides for 
different widths of tape.  Since the tape reader contacts close only
momentarily, the sender used Western Electric dry reed relays to store
the bits for the modem.  The timing channel simply had to reverse state
for each character.  My boss Frank Biggam was a brilliant circuits man
and invented a circuit to do the reversing, which resulted in patent
3,351,819.  This required a SPDT reed relay, and Western Electric didn't
make one, so I think we used one made by General Electric.  G.E. has
been in and out of a lot of different fields of businesses.

One of the options devised by Warren Foxwell and me was called the
identifier-recognizer option.  Without this, the sender left overnight
ready to send would spill its data if it were called up by anything
that would handshake with the modem.  The modems were one-directional,
but had a reverse-channel option intended to insure that the telephone
connection was still connected.  AT&T never stated what data rate
could be used over the reverse channel, but told us a number that we
could design to.  We had a circular piece of PC board etched with
fourteen bit segments, which could be coded in the field for marking
or spacing.  The identifier in the receiver would send out its code
over the reverse channel, and the recognizer in the sender would compare
with what was encoded in its disk, and allow transmission to proceed
if they agreed.  The comparison was done using a relay made by IBM,
which had two equal operate windings plus a mechanical latch with its
release winding.  The reverse channel signal went to one relay winding
and the local recognizer bit went to the other.  If both are spacing
there is no current in either and the relay remains released.  If both
are marking there is current in both relay windings, which are wired
to oppose each other, so no magnetic field and the relay remains
released.  But if one signal is marking and the other is spacing then
there is current in only one relay winding and the relay operates and
latches up and prevents any data being sent.  The latch release coil is
energized just before a comparison is to take place to be sure the
relay is released at that point.

There was another option involving a time-delay relay, and I no longer
remember what it did, but it was put in there to meet the need of Univac,
who didn't want the computer to have to set a timeout timer to signal
some kind of unexpected event.