Prior to the KSR33 Teletype were the model 15, which I used
extensively, the model 14 "transmitter-distributor" (tape reader), and
model 14 "reperf" or "typing reperf" (tape punches).  The typing reperf
typed the characters as well as punching the tape so it was human
readable.  Prior to this was the model 12 which was actually a bit more
electrical than the model 15.  The model 12 printer used a mechanical
distributor to do parallel to serial conversion and vice-versa.  The
model 15 did it all mechanically.  Quite amazing!
        The model 15's I used were 60 words per minute.  The start bit
and the five data bits (using Baudot code) were 22 mS long.  The stop
"bit" was 31 mS long.  This was CLOSE to the 1.5 stop bits many serial
ports can now do.
        The speed was set by a synchronous motor (though some used
governor regulated motors).  I believe the shaft would finish one
rotation in 154 mS (7 bit times).  The transmit end would hold the line
in the mark condition for another 9 mS to allow for slight speed
variations between transmit and receive.  During the stop bit, the
selector magnet would be pulled in and the armature would put a stop in
the way of the rotating shaft.  When a lump in the shaft came around, it
would get stuck on the magnet armature, stopping the shaft rotation and
slipping the clutch.  When the magnet was released during a stop bit, the
shaft was free to rotate again.
        The model 15 originally ran 60 mA through series wired "pulling"
magnets.  In the mark condition, the magnet would pull the armature to
the core.  In the space condition, the armature would be released.  Later
machines had "holding" magnets.  A cam on the shaft would move the
armature up to the core just prior to bit sampling time.  The cam would
then quickly drop away.  If the armature stuck, it was a mark.  If it
fell away, indicating no magnet current, it was a space.  the use of
holding magnets allowed the loop current to be reduced to 20 mA for
series wired magnets, or 60 mA for parallel magnets.
        These machines also used high voltage DC (about 150 VDC) to run
the magnets.  You could get enough current with 12 VDC, but it took too
long to build up the current after a keyboard switch contact closure due
to the magnet inductance.  Use of high voltage with a series resistor
resulted in a faster rise time when the keyboard switch closed (since
T=L/R).
        As someone else said...  "the good ol' days?"

Harold