I love antique clocks. They have a certain old-world charm that,
well, never gets old. The gentle sound of the quarterly chimes
exudes a real homey feeling and hearkens back to a simpler time. One
of the things I remember about visiting my great-grandfather's house
back when I was a small boy was the sounds of his clocks. Of course,
to him they weren't antiques - when he purchased them they were
brand new. Old railway clocks are especially cool; it is because of
the railroads' need to coordinate train movements across large
distances that we have standardized time today. The railroads also
pioneered some of the earliest methods for synchronizing clocks.
Railroad clocks simply exude history. A few years back
I stumbled across this
clock while surfing the web. It's an old railway station clock
from the Chicago & North Western Railroad.
It's a classic style "Drop Octagon" wall clock, often called a
"Regulator" because many of them had that word emblazoned across the
front. They are also known as "Schoolhouse" clocks because they were
very common in classrooms across the country. They were also
commonly found in railway stations; often the railroad would order
special versions with the railroad's name printed on the face and/or
special mechanisms to keep the clocks synchronized. A similar clock
appears in this photograph, taken in the office at the C&NW's
Proviso rail yard.
While I liked the idea of having such a clock in my home, I
didn't particularly relish the idea of having to wind it every
week, or reset it whenever it drifted or when daylight savings
time begins or ends. I wondered if there were a way to combine
the old-world charm of an antique clock with the modern
convenience of never having to wind or set it, all while keeping
perfect time. Perhaps I could upgrade an old clock with a new
mechanism that would constantly synchronize itself with an
atomic reference clock? Perhaps a stepper motor and some gears
could replace the clock's original pendulum escapement and drive
the clock's gear train electronically? Or maybe regulate the
period of the pendulum using a computer-controlled
electromagnet? As I continued to ponder, I realized that
the chances of finding an original C&NW station clock seemed
pretty low. Aside from the original web page that had prompted
my interest, I have never seen another one of these clocks
offered for sale anywhere. And even if I did find one, my
proposed modifications would probably destroy the resale value
of the clock, not to mention its historical significance. So I
decided on a different approach.
Reproducing an Old Station Clock
I decided to duplicate as closely as possible the look and feel
of the antique C&NW clock using modern components. I felt
this would exude the charm of the old world clock without
sacrificing any of the modern conveniences. As a starting point,
I found a modern reproduction of a similar drop octagon wall
clock on Amazon. Made by Seiko, it's a very nice clock in own
right, with a modern quartz movement, digitized quarterly
chimes, and a fake pendulum (it swings back and forth, but is
not involved in any actual timekeeping functions). More
importantly, it comes apart fairly easily, allowing me to
customize it and install a new (old) clock face. My first
thought was to use a photograph of the original clock's face,
but I didn't have any pictures without the hands installed, and
my Photoshop skills aren't sufficient to effectively remove them
from the photo. So I found a picture of a similar clock face
online, scaled it to the correct size, and added the C&NW
logo and text. I then printed my new clock face out onto
heavy-stock paper with a color laser printer and carefully cut
the face out. I then lay the new face right on top of the old
one, holding it in place with some Scotch tape. The clock's
bezel hides the tape and my slip-ups with the scissors. The
result looks like this:
With the exterior complete, the next phase of the project is to
update the interior.
Controlling the Clock
If you open up pretty much any modern clock with a quartz
movement you'll find something like this:
This movement is based on a Lavet
stepper motor. A controller circuit, typically powered
by a single AA battery, delivers a current pulse to an
electromagnet coil once per second, causing a small
permanent magnet to rotate exactly 180 degrees with each
pulse. The magnet is attached to a gear train which moves
the hands on the clock's face. With a minor modification, it
is possible for pretty much any embedded microcontroller
with General Purpose I/O capability to provide the pulses
and control the movement. There are lots of videos on
YouTube which explain how these movements work and how to
modify them; here's but one example:
Once the micro can advance the clock hands, I needed some
way of accurately tracking real-world time. One oft-used
solution is to receive and decode the WWVB
time signal broadcast by NIST from Ft. Collins, CO.
This approach has the advantage of being standalone,
requiring no network connection, but this advantage is
offset by the high cost and limited availability of WWVB
receiver modules. If the chosen micro has networking
capability, a second option would be to maintain the clock
using Network Time Protocol
software. The Network Time Protocol is one of the oldest and
most robust protocols in use on the Internet today, and is
more than sufficient for my needs. For this project, I
decided on the Raspberry Pi because it is cheap (as low as
$25), has plenty of GPIO pins, and it runs Linux which
allowed me to leverage existing software like gcc and the
NTP daemon. Because of its immense popularity, the Pi is
also supported by an immense ecosystem of add-on hardware,
which also proved useful in wiring up the interface circuit.
To avoid making any costly mistakes on my actual project
clock, I purchased a (broken) clock from a thrift store.
It's a miniature Grandfather clock which was sold by The
Bombay Company for significantly more than the $0.99 that I
paid for it. Although the battery in the back of the
movement was badly corroded, I took the chance that the
movement itself was still usable. The gamble paid off.
After removing all the corroded parts, I started by
soldering two wires onto the ends of the coil. I then built
an interface circuit called an H-bridge
to go between the Pi and the coil using an H-bridge module
from Adafruit. Because there is no way for the Raspberry Pi
to "read" the position of the hands, I also added a
ferro-electric RAM module
to provide non-volatile storage for the hand positions so I
don't have to re-set the hands after every power failure.
I wired it all together on a prototyping
hat and plugged it into a Raspberry Pi.
To protect the circuitry and facilitate mounting I purchased
an acrylic case
made by Sunfounder. Because the prototyping board added some
additional height when plugged in, the combination would not
fit comfortably inside the case. Fortunately this was easy
to remedy by adding longer standoffs. The case fits nicely
on the back of my "practice" clock.
Since the project clock is wall-mounted, my plan is to run a
cat-5e Ethernet cable to the location and install a
wall-mounting RF-45 jack similar to the
ones used to mount wall POTS telephones. By running
802.3af power-over-Ethernet to this jack I can provide the
clock with both power and network connectivity. A small PoE
splitter module provides regulated 5V power to the Pi.
There is plenty of space inside the project clock's case to
hide both the Raspberry Pi and the PoE splitter, leaving
none of this infrastructure visible from the outside.
Software
The NTP daemon does an excellent job of maintaining the
system clock to within a few milliseconds of the reference
atomic clock. Because the NTP daemon does all the heavy
lifting, the software which controls the positions of the
clock hands is extremely simple: the main loop reads the
current time from the system clock, and updates the
positions of the hands on the clock face to match. If the
hands are "behind" the actual time, the clock movement is
stepped repeatedly (and quickly) until the hand position
catches up. If the hand position is "ahead" of the actual
time, the Lavet mechanism is not stepped and the hands wait
for actual time to catch up with them. This is necessary
because the Lavet stepper mechanism is not designed to move
the hands backwards.Oddly enough, through experimentation I
discovered that it is possible to make a Lavet movement step
backwards by applying a relatively high voltage and current
to the coil; apparently the much-larger-than-normal magnetic
field that this generates causes the magnet gear to spin in
the wrong direction. Not wanting to impact the reliability
of my control circuit, I chose not to make use of this
"feature."
Remaining Steps
The "practice" clock has been running flawlessly for several
weeks now. The final phase of this project will be to modify
the Lavet movement of the real project clock and transfer
the Pi and PoE splitter into it. I then need to run cat-5e
cable to my selected mounting location and install the jack.
Fishing cables through finished interior walls is a PITA, so
it might be a while before I get around to fully completing
this project...
