Rolling balls clock

Copyright Chris Morgan. Clik on image to enlarge.

In November 1997, I though about building a clock with a pendulum activated by the fall of bearing balls. The first thing to design was the engine which was to sustain the pendulum’s  oscillation: how to get enough energy from a ball to make this possible? The balls had to be heavy enough to run a stable course but their size obviously determines the clock’s size.  As a trade-off between weight and size I arbitrarily picked a 19.8 millimetre diameter for the balls. I could picture the running clock in my mind, and I believed I could complete it in three months…
The engine’s first prototypes were just unbelievable, at that time I thought that I needed as much gain as possible to achieve the largest possible amplitude: after a month of effort I had achieved a 60 centimetre swing with a 1 meter pendulum. That’s when I stumbled across an old physics book from 1894 which explained that the isochronism of oscillations is only true for small amplitudes, not large ones. A month of work for nothing. I was now left with the reverse problem to solve: achieving the smallest possible amplitude with 20-millimetre balls. It took me another month to design the final version of the engine.

The engine

It runs on a simple principle: as it falls every ball pushes on the pendulum in the direction of oscillation. To make sure that the kinetic energy of the balls in the feeding ramp has no influence on the engine’s accuracy five right-angle speed-limiters slow them down before being delivered to the engine.
The Pendulum
The time reference is given by the pendulum’s length: about one meter for a ball that falls every second. The pendulum’s rods are made of invar a temperature-stable alloy. The bob is tooled in a 17 kilo brass ingot to provide enough inertia. Coarse tuning is done by screwing the bob up or down, fine tuning is done by adjusting the small auxiliary bob at the bottom.
To start the clock the pendulum must be swung with an amplitude large enough to admit one ball through the hole. This is done through a lever to achieve a first beat exactly in the oscillation plane.
As soon as the pendulum is released the movement is sustained by the seconds balls. The amplitude increases slowly up to a stable two centimetres after 10 minutes  

Building the chassis, the bob, the display was a matter of five months, I started in January 1998. The chassis size is mainly determined by the balls’ diameter. Sixty 19.8 millimetre balls come a little short of 1.2 meter (4 ft). Add 30 centimetres for the escape mechanism, 20 for the balls buffer, 10 for the moon phases and 50 for all the background mechanism. It all adds up to 2.2 metres high (7′ 2”)….
My initial idea was to use electricity, electronics and optical sensors for all functions other than the pendulum’s engine. But as months went by the idea of mixing mechanics and electronics was less and less appealing and at the end of may I finally decided to build an entirely mechanical clock.
That’s when many problems popped up. Consider this one for example: how is it possible to have the sixtieth second ball increase the minutes count and at the same time reset the seconds count to zero? This is the trick I devised. (Fasten your seat belts).

The quest for the lost minute…

The sixtieth second ball rolls over the fifty-nine preceding ones. Its fall is long enough (1 meter) to acquire the energy to open a gate at the bottom of the seconds tube. The same falling ball activates in passing a lever that temporarily blocks the seconds input to avoid that incoming balls are flushed with the outgoing ones. The weight of the 59 outgoing seconds is used to release a ball stored upstream of the escape mechanism. This one resets the lever that was blocking the incoming seconds and then it falls into the minutes tube. Easy as pie!
The same principle is used for the hours
All these problems kept me busy until… October! Now the next one was how to drive all the balls back up, and above all how to store them to avoid jamming since the balls come in randomly but must leave in an orderly fashion. I finally opted for a system that meets the constraints of simplicity, reliability and minimal footprint: a cylinder carved with 3 longitudinal grooves, inside a spiral spring.
Moon calendar and weekdays
Advancing the moon calendar and weekdays is done by using the energy of the falling 23 hours at midnight. They push on a lever that takes care of all these tasks.
As the moon does a full revolution in 29.5 days I’m using a 59-tooth wheel, advancing it by 2 teeth at midnight. The weekdays dial, the moon sphere and the ball elevator are the only rotating parts of the clock.
Regulating the buffer stock
Since balls go up randomly their number must be regulated to ensure enough supply upstream of the escape. There must be a little over 60 balls going up per minute with an automatic overflow path. Here is how it works: as long as the buffer is not full the ball goes through a curve with enough speed to avoid the overflow hole inside the curve, when the buffer is full the ball bounces back and falls through the hole down to the bottom tank.
Adding up the number of balls displayed in the tubes at midnight, 144, plus what is necessary to fill all the feeding paths, plus some safety margin the clock uses 450 balls.

Time setting

Every clock has a crown to set the time. On this one I have designed eight levers to implement the following functions:

  1. Release seconds.
  2. Flush seconds
  3. Add minute
  4. Release minutes
  5. Flush minutes
  6. Add hour
  7. Release hours
  8. Advance weekdays and moon

Using these 8 levers it is possible to activate the various counters and flush mechanisms to set the exact time (moon phase, weekday, hour, minute, second) while the clock is in operation.

Error overflow

What happens if something goes wrong ? For example the seconds input is blocked: the as balls continue to fall, one every second, they will overflow the rail and fall into the bottom tank and from there back up again, endlessly. A real disaster.
To make sure this does not happen I have designed an inertial failsafe system. If the balls follow the normal track they end up in the bottom tank. If they fall from the rail they are captured in a safety tank, above the first one and the clock stops after having exhausted the balls supply.

Some data…
1600 parts. 300 kilos (660 pounds). 95 ball bearings. Completed in 18 months.
Materials: brass, steel, stainless steel, silver, invar. Pyrex glass with sand engraving for the hours, minutes and seconds display. The cabinet is made of solid oak.