Congreve

 2015-2021

A Congreve clock is one of the most eye catching clocks around. The rolling ball has a hypnotic attraction as it traverses its zig zag table.
It is beloved by model engineers as a standard exhibition piece and to complete one of these in metal shows true skill.

It was invented in 1808 by Sir William Congreve, not as a showpiece but as a genuine attempt to increase timekeeping accuracy by 'detaching' the escapement from the going train, the latter only being in motion for a very short moment every 15 seconds. Sadly it proved woefully inaccurate, largely because the motion of the ball is easily disturbed by dust collecting on the nearly horizontal table. Deviations of 15mins per day or more are common.

Sir William Congreve was a polymath who dabbled in many things, not the least perpetual motion and the development of rockets for warfare. Known as the Congreve Rocket they were used in several engagements in the early 1800's with mixed results, sometimes being more of a danger to the firer than the enemy.

Congreve clocks are very expensive, modern reproductions command a price up to £2000. My version costs perhaps a hundreth of that and has the advantage of being  more accurate, maybe in error just a few seconds a day. (Long term tests suggest better than that !)

 This is a traditional Congreve clock as sold by Devon Clocks  They make a wonderful range of reproduction historical clocks and instruments.

I originally made this clock in 2015 but in 2021 rebuilt with a better table and improved table tilting mechanism. 

The original and reproduction clocks have a spring driven fusee movement which both drives the table and the movement. The spring is very powerful and I feel unsuited to a wooden clock.

This version uses a small stepper motor to tip the table. The ball, when it hits the trigger wires at each end of its travel operates mechanical switches to command the servo to tip the table. The trigger switches and servo rely on  a small microcontroller mounted on the back of the clock for their operation.

It would be possible for the going train that moves the hands to be driven by the table motion but again for mechanical simplicity I found it easier to use a further servo to advance the seconds hand by 15 seconds when the table tips. This servo is also sequenced by the microcontroller.

To be absolutely clear it is the rolling ball which is the timekeeper, not the microcontroller. However using a microcontroller has another advantage in that depending on the time taken for the ball to traverse the table, the control system can adjust the tilt of the table to maintain an average 15 second traverse so that if conditions change, such as temperature, humidity or dust collection, an adjustment is made so that accurate timekeeping is achieved. This clock will never be good as a stopwatch, there is too much variation from one traverse to the next (maybe 0.2 seconds) but over longer periods it is capable of very good timekeeping.

Below is a video of it in operation.

Some constructional details

These are not meant to be comprehensive details, but just to fire the imagination.

The table:

For the ball to run regularly the table surface needs to be as smooth as possible. A laser cut 3mm skeleton  plywood track was found to be satisfactory.

There is a a 4mm diam. steel shaft supported by blocks glued to the underside of the table and this pivots in 8mm bearings in plummer blocks which are fixed to the base of the clock.

A bracket for the control arm is fixed somewhere along the rear edge of the table, positioned mainly by trial and error.

Two holes for the trigger wires to pass through the table were made at each end of the run.

The wheels and pinions:

The wheels and pinions were cut from laminate flooring as I find this to be cheap (I had leftovers ) durable and can take fine detail without disintegrating like most plywood.

 The wheels and pinions were drawn out on the computer to a Module of 1.5, printed, pasted onto the wood and cut with a bandsaw. Very fine detail cuts and 'nibbles' can be made with a bandsaw and the work progresses quickly. The interior detail was created using a scroll saw. Final fine finishing of the teeth profile was done by hand filing.

3.9mm holes were drilled for the centres and this gives a tight fit onto a 4mm shaft.

It is essential to get the holes accurately central on the gears, particularly so for the pinions. I find it easier drill a rough blank, support it in a lathe through the centre hole and turn the outside concentric before cutting the teeth.

The upper four gears are to give the 60:1 ratio to drive the minute hand. One of the 8 tooth gears is driven round in 15 sec segments and its shaft carries the seconds hand and hence goes round once a minute.

The remaining four gears give a 12:1 ratio to drive the hour hand from the minute shaft, Other ratios could be used but these give equal spacing to the three dials.


The power mechanisms:

The rear of the clock showing the microcontroller and servos

A small stepper motor is used to tilt the table. A major advantage of this is that such servos are very cheap and their movement can be controlled very precisely, and they are very easy to control. A model aircraft type servo motor  is used to drive the seconds hand shaft.

The microcontroller used is an Arduino Nano. This powerful little device is available for £4 or less from a number of sources.

In the picture above the microcontroller is the smaller device with the black diamond on it. The rest is board that makes the connections to the micro easier but is not strictly necessary.
The controller has its own power handling circuits so can be supplied by 7.5vvolts to 12volts dc. Thus it could be run from a 9v battery, but how long this would last I don't know. I use a 9v dc mains adapter.

The two servos plug connect to it with 3 wires each and the two trigger switches connect to it with 2 wires each.

The program to run the clock, of which more later, is written in a C++ type language and passed into the micro from a PC via the silver USB connector to the right of the black diamond.

The trigger wires are 2mm diameter steel rod bent through a right angle at the top and pivoted at the right angle. Thus when the ball strikes and moves the bottom from front to back the far end moves up and down. At the far end of each rod is a very simple switch consisting of a flat brass plate on the rod which can contact a 3mm screw. This screw can be moved up or down to set the contact distance and thus limit the movement at the ball end. Electrical connections are made from the rod and screw back to the controller.

The trigger wire and contacts seen from below.

The requirement for the motion servo is that it should rotate the seconds shaft 90 degree every time the table tilts, since the time taken for the ball to  traverse the table is 15 seconds.

This is achieved by having what is effectively a 4 pin lantern pinion on the seconds shaft and a pusher arm attached to the motion servo. On the return of the servo arm the pinion is prevented from reverse motion by the spring rubbing on the periphery of the wheel.

The seconds hand is connected to the seconds shaft and an 8 tooth gear on this shaft drives the rest of the gear train to give minutes and hours.



The control system:

The ball is the master controller. It is this hitting the trigger wire which sequences the rest of the actions and shows the time on the clock faces. Thus an accurate 15 second transit of the table is required.

There are a number of reasons why a consistent 15secs cannot be obtained in a static system where the angle the table can tilt through is fixed or manually adjusted. These can include temperature and humidity changes, dirt on the track and the surface the clock is standing on not being level.

But this is a dynamic system and this is where the accuracy of this clock comes from. The process might be something like this.

When the clock is started it is unlikely that the transit is exactly 15 seconds and will probably be different in each direction. The microcontroller records the transits in each direction for 6 transits (90seconds) and calculates an average period for each direction to the nearest millisecond. It compares this value with the target value which is 15,000 milliseconds and on the basis of the difference between these two values (the error difference) adjusts the tilt of the table to bring the transit time to 15 seconds. It may not be able to achieve this in one step but eventually will reduce the error difference to zero. It continually monitors the error value and keeps adjusting for external changes.

I have not had the opportunity to do a long term trial of the clock yet but over a 3 day period no time error was obvious by looking at the dials and comparing the time to a digital watch. I see no reason why this should not be a very accurate clock.

This type of control is used in industrial systems and is known as PID (Proportional, Integral, Derivative) control but only the proportional part was employed here. It was thought that the integral part might be necessary to improve long term accuracy but on the evidence so far it may not be required. Read up about PID control, it really is quite interesting.(But find a simple explanation)

This is a computer printout of the transit times both left and right over a 16 hour 57 minute period. Both left and right traces are shown here one in blue, the other in green.

Although the traces wander up and down and occasionally show  up to .25 second deviation on an individual run the average period appears to be  close to 15 seconds and shows no drift.

The time for the last transit measured, shown in the boxes below the trace, is given as 14.924 seconds for left and 15.025 for right but the average deviation for 2036 transits is 2 milliseconds to the left and zero to the right.

This calculates to be a .807 second total time error after 17 hours. Not bad eh!

The length of the track is about a metre and it is a sobering thought that the ball will travel about 2000 km in a year.

Time will tell how well it performs in the long term and I will post further information as I get it.