[GreenKeys] Why and When 7.00 vs. 7.42 unit code?

[Jim Haynes]

I can't answer your late-term questions, and I thought I had told the
story of 7.00 vs. 7.42 many times already, but here it goes again.

For start-stop synchronization to work the receiver has to stop between
characters.  Morkrum accomplished this by running the receiving shaft
a bit faster than the transmitting shaft.  The transmitter sent 7.00
code and the reciving shaft completed its rotation by about the middle
of the stop pulse so it could stop to wait for the start pulse.

Prior to purchasing Teletype, Western Electric had their own line of
teleprinter equipment.  They used rotary faceplate distributors and
the transmitting and receiving faceplates were on opposite ends of the
same shaft.  Maybe this was a leftover from some kind of synchronous
operation - I don't know.  Anyway to allow the receiver to stop between
characters they had to delay the transmitting shaft.  This was done with
a latch and a relay.  When Teletype entered the picture they needed
interoperability with the W.E. machinery.  They measured the time delay
introduced by the relay and found it was equal to 0.42 of a pulse time.
So they required Teletype to stretch out the transmitted character stop
pulse to 1.42 unit.

W.U. didn't have the interoperability requirement, and presumably
appreciated the increased speed they could get by operating 7.00 code.
Also they were buying machines from both Morkrum and Kleinschmidt
when they were separate companies.  Perhaps Kleinschmidt never had
the requirement for interoperability with W.E. machines and thus was
7.00 code all along, but I don't know.

Teletype receiving machines have no trouble receiving 7.00 code since they
kept the principle of running the receiving shaft faster.  Hence there is
no difference in receiving machines whether receiving 7.42 or 7.00 code.
Only the old W.E. machines can't handle 7.00 code.  I don't know why the
7.42 was continued after the W.E. machines became extinct.  But it was.

Now when we move into modern times we have Teletype at 7.42 at 100 wpm,
or 10 chars/sec, which works out to 74.2 baud.  And the European standard
seems to have been 7.50 code at 50 baud for Telex.  I have no idea why
they went to 7.50.  And then ASCII with 11.0 unit code because the
Teletype equipment at 100 wpm couldn't get by with a single stop pulse.

Then the military decided to standardize on baud speeds.  They pushed 74.2
baud up to 75 to make a round number.  That's only a little ofer a one
percent change, so I assume equipment geared for 74.2 baud can tolerate
75 without hardly any trouble.  Maybe slightly reduced range.  So that's
how we get the speeds that are a power-of-two multiple of 75: 150, 300,
600, 1200, 2400,...  With 100 wpm ASCII being non-compliant, along with
IBM 2741 that runs at something like 134.5 baud.

I don't know how the Navy got a mixture of 7.00 and 7.42 machines - maybe
some equipment supplied under Western Union contracts.  And if you put
100 wpm gears in a 7.00 machine I guess you get 106 wpm.

When it comes to crypto, there were what seem like some terribly bad ideas
to me, but then I know nothing about crypto machinery.  You'll see in the
Teletype parts books keyboard parts for "synchronous pulsed operaton"
where there is an electromagnet to release the keyboard clutch of tape
reader clutch.  I have some pieces of stuff that are called "TD Steppers"
and these were used in encrypted systems to operate the tape reader
clutches and keyboard pulse magnets to assure that characters aren't
sent into the crypto machine too fast.

Teletype made some time-division multiplex equipment for the military
(AN/FGC-5, AN/UGC-1).  The multiplex had to run slightly faster than
the start-stop signals coming in, so now and then it got far enough ahead
that it had to insert a multiplex blank into the output stream.  In
ancient multiplex work the all spaces blank character was used for this
purpose, but in modern military systems the blank was considered a 
legitimate character.  So the multiplex actually transmitted a six-unit
code, where the sixth pulse simply told whether the character came from
the input or was a multiplex blank; multiplex blanks were no sent to the
output but simply caused a pause in the character stream.  And the real
characters were shorter than 7.42 code because the mux was running fast.

Apparently this was a problem for crypto equipment, which wanted a steady
7.42 character stream.  So they had developed a "printing telegraph
signal normalizer" to slow down the multiplex output slightly and
paste over the pauses.  See Dingley, "The Printing Telegraph Signal
Normalizer"  I.R.E. Transactions on Communication systems, Vol. 4
No. 3, 1956, p. 18.  And Steeneck, "Teleprinter Signal Normalizer"
Western Union Technical Review Vol 11 No. 1, January 1957, p. 10.

jhhaynes at earthlink dot net