[time-nuts] Mechanical 1PPS Oscillator Disciplining
phk at phk.freebsd.dk
Fri Jan 9 16:46:28 EST 2015
In message <DF7C7705-B1CC-4B8D-8BAC-D471E2AB5049 at bardagjy.com>, Andy Bardagjy w
>Does anyone know any other history about that particular piece of equipment?
I seriously doubt those claims of precision.
At datamuseum.dk we did a small booklet about the history of paper tape
as a storage medium some years back, and the Great Nordic Telegraph
Company was roughly half of the material.
The first thing to notice is that most cables were simplex, you could
only transmit in one direction at a time.
The turnaround was suprisingly slow. Because of the dielectric
absorption in many miles of cable you had to disconnect your (relatively
high voltage) transmitter, short the cable for some time, before you
could attach your (incredibly) sensitive receiver to the cable.
"Some time" depended on cable type and length of cable, but up to
five minutes were not unheard of. (The exact duration were often
determined by the clerk putting his moistened finger across the
Needless to say this made it a paramount matter of efficiency to
minimize turn-arounds, and therefore the general scheme of operation
was that one side would transmit until they had cleared their backlog
or until a certain maximum amount of time since last turnaround had
Some high-traffic cables ran on "clockwork" (minutes 0-15,30-45 A
to B, minutes 15-30,45-00 B to A) -- this made it possible to predict
how much papertape would be required.
During each "turning", the transmitter would be driven by papertape,
each roll as large as physically practical, but there would still
be a gap between individual messages on the tape and a longer gap
It was not uncommen for a high speed relay station to go through
five miles of papertape a day at rates of several inches per second.
This is even more astonishing when you realize that many of these
relay stations were remotely situatated, like for instance the
"Eastern & South African Telegraph Companys" station on the island
Bawe outside Zanzibar.
For particular long cables repeater stations were necessary and
since they only had two cables, there were never any doubt where
the messages would go. Most, but not all of these skipped the
paper-tape step, and had the receiving "undulator" drive the
transmit relay directly.
This is likely the kind of "syncronized" table described in the text.
The majority of stations had more than two cables and therefore
needed to make routing decisions, but messages would be batched
as early as possible to minimize the number of paper tape splices
What all this boils down to is that the syntonization requirements
were nowhere as dramatic as that text claims: +/- 5% were a very
Driving the 30Hz reed with a pendulum clock would trivially do this.
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
More information about the time-nuts