MARCEL BETRISEY CREATIONS  Sion Switzerland Phone: ++41 273238808 Mobile ++41 793801679 Email: see "Contact"


 
FLORENCE  (Ball clock)


 I decided to build this clock only to reuse and simplify the Anachrone, for such a good engine deserved to be optimised.

After several months of prototype testing I managed to simplify it to the point where one falling ball was communicating enough
energy to the pendulum for one hour. That's an improvement by a factor of 3599 over the Anachrone. The rest was mere
implementation...
 


 
 


To be torsion free the chassis had to be triangular (a rectangle experiences a 1/100 mm torsion per oscillation, which is enough to dissipate the pendulum's energy). The height was a consequence of the pendulum's length and the dial had to fit in a salad bowl. The materials I picked were steel with some brass as a warming touch. The pendulum rods are made of invar.

The ball upward movement is achieved this way: a rod pushes the ascending ball through a hole at the bottom of a cup sized to hold five balls. The balls in the cup spread apart then close again behind the ascending ball, preventing it to fall back. The ascending ball pushes on a column of seven balls in a glass tube and the top one falls into a labyrinth of 4 pins and 4 holes, arranged in such a way that in its movement downward the ball pushes the pendulum 4 times in the direction of oscillation. It then comes to rest at the bottom of the mechanism, waiting to be pushed up again when the minutes hand reaches the hour and is detected by an optical sensor. Another optical sensor counts the pendulum's beats and advances the seconds hand.

The design for this clock started in May 99 and ran for the first time in November, six months later. The accuracy is about 4
second per month.

"This one is my preferred clock because building on the Anachrone it simplifies it to an extreme. This engine concept gives me a
total freedom of implementation. I can now make a clock based on any set of parts: recycled helicopter, bike or printer parts,
there's no limit. This clock is really a step into new territories.
However I have used few recycled parts this time, two stainless steel salad bowls, a mahogany board, nothing more. The rest was tooled from scratch. So why recycle  salad bowls? To save time: with my limited equipment, tooling the dial casing out of stainless steel would have been a two day affair."
 

About the clock's accuracy

Initially this clock was not so much designed for accuracy as for taking advantage of a unique engine. My first surprise was that
its operation was perfect once the fine tuning was done. The second one, after a year and a half of testing was that it could
reach an accuracy of one second per month or better. I could make it even more accurate by housing it in a cabinet but this
would spoil its looks.
 

The chart below shows what happens over 1.8 hour with one measurement every 10 seconds.

Initially the pendulum is fast with a .9999 second beat. It goes faster and faster over 10 minutes, reaching .999827 s until the
hour ball falls, creating the first spike. The pendulum suddenly slows down  by 300 microseconds then speeds up again with
underlying fluctuations at one minute intervals. These fluctuations go on decreasing until the fall of the next ball. The only time
when the beat is precisely one second is on the exact half hour. The chart shows no perturbations created by movement around
the clock because I was not in the workshop at the time. Such perturbations can generate up to 15 microseconds beat
excursions.
 

Below is a recording over 2.2 days, sampled every 10 seconds.

Every spike on the graph is an hour ball falling. Every hour the pendulum's beat varies around one second. The overall impact is
a 7 second variability per day. This would not be acceptable for a normal clock  but Florence is designed to accommodate this:
on this sample the average beat is exactly one second and the yearly offset 0.0 second. This of course is an exceptional sample
because just moving around the clock will impact accuracy. With this source of perturbation taken into account accuracy can
reach 4 seconds per month. If you have the MicroSet software you can dowload and view the original datahere.