Florence (technical)

Ball Clock

This clock was built to take the best of the Anachrone and improve on it, because it was necessary to optimise this motor which worked so well. After several months of testing on a prototype, I managed to simplify it to draw enough energy from the fall of a ball to move the pendulum for one hour. That’s 3599 times better than the Anachronous. Since it was possible, it was time to start..

The frame had to be triangular to avoid any torsion due to the passage of the balance wheel. (A rectangular frame has 0.01 millimetres of torsion per oscillation, which quickly cancels out all the inertia of the balance wheel) The size was determined by the length of the balance wheel, and the dial had to fit into a salad mixing bowl. The materials chosen were stainless steel and some brass to warm the eye. The balance rods are made of invar. The winding of a ball is done in the following way: a rod pushes the ball from the bottom and forces it through five others arranged in a circle. It then pushes a small column of 7 balls into a glass tube. As it does so, the 5 marbles close behind it and prevent it from falling back down. The seventh marble (the one at the top of the glass tube) goes into a maze made of 4 grids and 4 nails. In doing so, it pushes the balance wheel in the direction of oscillation each time, and lands at the bottom of the mechanism where it waits for an hour. The winding order will then be given by an optical sensor each time the minute hand is in the vertical position. Another optical sensor counts the passages of the balance and sends its impulses to the seconds hand.
Started in May 1999, it began to run in November, after six months of work.
The precision obtained is of the order of one second per month.

“This is obviously my favourite, because it takes the best of the Anachronous and simplifies it to a fault. This motor system gives me incredible freedom: I can now make a clock with any support. Helicopter parts, motorbike parts, printer parts, everything is now possible. This clock marks a step forward and opens many horizons. On the other hand, there are very few recycled parts. Two stainless steel salad plates, a mahogany board, nothing more. Everything else was machined from scratch. But why recover dishes? Simply because I lack the tools and because machining the back of the dial would have taken me at least two days if I had to cut it out of a stainless steel ingot.”

Some notes on precision

This clock was not originally designed to be accurate, but to be able to run on a unique motor. The first surprise was that it worked perfectly once the levels were set. The second surprise, which came after a year and a half of testing, was that it could have a drift of about 4 seconds per month or less. I could make it more accurate by carossing the clock, but the aesthetics would suffer.

On the graph below we can see everything that happens in 1.8 hours, with a measurement every ten seconds.

At the beginning of the graph, the balance is too fast: it beats at 0.999900 seconds. It accelerates more and more during ten minutes to 0.999827 until the hour ball falls, which produces the first peak. The balance suddenly slows down by 300 millionths of a second, then starts to accelerate by making a wave every two minutes. These waves will undulate for a whole hour, gradually weakening until the next hour ball arrives. The only time the pendulum beats the exact second is exactly halfway between two full hours, i.e. the half-hour. What is not visible on this sample are the disturbances produced by my movements around the clock (I was not in the workshop at the time). These can produce peaks or troughs of up to 15 millionths of a second.

Below we can see the same sample spread over 2.2 days, again with one measurement every ten seconds.

Each zig-zag point represents a falling hour ball. The pendulum thus slaloms around the second every hour. All these points calculated give us an instability of 7 seconds per day. This would be intolerable for a normal clock, but not for Florence, which is designed to accommodate it. This sample shows us that the average time is strictly 1 second, and calculates an annual average error of 0.0 seconds! This sample is obviously exceptional, as simply walking around the clock slightly distorts the measurements. If disturbances are included, the accuracy can be as high as four seconds per month.

Phaeno Science Museum, Wolfsburg (Germany) Exhibition from December 2017 to February 2019