The Chronolith (technical)

Could the force of light alone move a 4 kilo pendulum? And if so, is it possible to make a clock out of it? It is to try to answer this question that I created this clock, which is so different from all the others. The frequency of oscillation of the pendulum is 1 second, two lamps placed on either side light up alternately, thus « pushing » the pendulum each time. The principle of its operation was discovered by Sir William Crookes in 1873. It acts not only as a motor for the balance wheel, but also as a brake, regulating its amplitude within a very precise range. The inside of the tube is under vacuum, with a pressure of about 0.01 bar. The balance is started by a magnet placed near the glass. The fine adjustment of the time base is done from the outside by turning the 4 planetary weights around the sphere of the balance using the same magnet. This avoids having to fill the clock with air, dismantle it and empty the air again for each adjustment. The clock thus consists of a glass tube, a pendulum, two cells (simple mica sheets), two relays and a quartz clock with the quartz removed. That is all. The whole clock is mounted without screws or bolts and is held together by the force of atmospheric pressure alone. To dismantle it, all you have to do is let air into the tube.

The idea for this clock came to me in April 2001, the material was ordered in May and the very first test started on Wednesday 7 November. The moment was crucial: if this test proved that the system could not work, I would be left with a glass tube and a big vacuum pump on my hands. The first proto? a suspended petanque ball, and an incredible 30 sheets of mica (simple transistor insulation), one side of which was blackened by candle smoke. A sensor was attached to the outside of the glass tube, I emptied it and started the clock at about 4pm. As it takes a good hour for the pendulum to stabilise, I went for a drink while waiting for the results. When I returned, the pendulum was still moving. As I was not sure if it was doing so from its own inertia, I went back for another drink. At 7pm the pendulum was still moving and the amplitude was constant, so I went back to celebrate. At 10pm, the clock was still working, and I was not so good…

This clock consists of a pyrex glass tube (diameter 190 mm, thickness 8 mm) inserted on a quartzite block impregnated with epoxy resin in order to ensure the impermeability to the passage of air. The large sphere visible on the photo is the dial, with its hour and second hands. It was then exchanged in January for a smaller one for aesthetic reasons.

The first tests showed an incredible reliability. Around 2 seconds off per month.

« This is the seventh and final version of the engine. It took a long time to develop, because it takes at least three hours each time before the balance reaches its cruising amplitude after the initial impulse. It was necessary to work first on the degree of depression of air, then on the shape of the engine so as to obtain the most gain possible. (It is important to realise that the clock cannot work if half a thimbleful of air enters it…) When this was done, I started to be able to lower the power of the bulbs from 35 watts to only 5 watts. This should give them a life of 32,000 hours, just under 4 years.

This clock is « organic », because it will adapt to unfavourable physical circumstances in order to keep its precision: it will compensate by itself for anything that could cause the amplitude of the balance to fluctuate. Let’s say the power of the lamps increases. The amplitude will automatically follow. The pendulum will therefore arrive with more speed in front of the cell, which will switch on the halogen bulb earlier, which will… brake the pendulum for a few millionths of a second, and then push it on as usual.

(19 December 2001) The last few days have been mostly spent refining the isochronism of the pendulum with respect to the pressure inside the tube. But as it is fixed to my workbench, each of my movements induces disturbances of up to 150 millionths of a second, which results in spikes on the screen of the measurement computer. It has become impossible to work any longer like this. So tomorrow I’m going to fix it to the wall next to the window, so that the clock is as stable as possible. And the accuracy tests can continue on a better basis…

(30 December 2001) The accuracy tests have started. First observation: it is very easy to adjust the length of the pendulum with magnets. Second observation: the clock is extremely accurate. No parasitic ripples, no incomprehensible fluctuations. The current variations in the vacuum in the tube lead to fluctuations of no more than 5 millionths of a second per beat. (These variations will disappear when the tube is finally assembled and currently has some losses. I have to switch on the vacuum pump every 9 hours, otherwise the clock stops…) My current little worries come from the way of measuring through the tube which is still 16 mm of pyrex glass. As the reflections of the measuring laser beam cause random inaccuracies, I took my measurements directly on one of the lamps. Each measurement taken separately can be unstable up to 8 seconds per day, but the average will be very reliable. On the sample below, taken over 4.9 hours, we can admire the stability of all the measurements: 1.9 seconds of shift per month. This measurement has since been confirmed by much longer samples. This clock is therefore the most accurate of all those I have built! So what started as a simple challenge (making a pendulum powered by the force of light) has become a reward. The adventure continues…

(sample of 4.9 hours. 0.0000007 seconds delay per beat of the pendulum. That means 1.9 seconds per month)

(20 January 2002) I received a new vacuum pump and a more efficient device to measure and regulate the air vacuum. This means that the clock will be running in an optimal configuration by the end of this week… Stay tuned!

(28 January 2002) All the tests this week were about time fluctuation tests with different air pressures. The results are mind-boggling. Here are some facts.

– The halogen light moves the pendulum from a vacuum of 6 X 10 power -1 bar.

– It still moves it at 1.4 X 10 power – 2 bar (maximum power of my pump)

– No noticeable difference in accuracy or fluctuation in the movement of the pendulum could be found within these two values!

– A half degree adjustment of the planetary gears results in a shift of about 0.00007 seconds per beat, i.e. about 1 second per month.

– The temperature differences in the workshop (a rise of about 4° during this week) do not seem to influence the running of the balance. This is very surprising to me: they should. This is probably due to the internal self-regulation of the isochronism, but this has yet to be confirmed.

(29 January 2001) I was stunned this morning: while the precision graphs were always flat, I discovered…

regular and incomprehensible peaks, which started 12 hours ago, and which appear every 2 minutes. The pendulum starts to accelerate to 1.99500 seconds, then slowly returns to its normal value. The total time of each disturbance is 22 seconds. This problem was quickly solved: it was simply that the base of the clock had moved during the night, and the pendulum was touching the glass wall every 2 minutes. But the resulting graph reveals something much more important: the constancy of the « pressure » on the pendulum. If the peaks had been arranged in a more anarchic way, it would have meant that this « pressure » was not constant and therefore a source of inaccuracy. This is not the case.

(3 February 2002 ) The next two weeks’ tests will be about accuracy in relation to temperature, light and depression level. This is what the poor Chronolith looks like in its torture room. The stone has been dismantled, the dial removed and lots of sensors and lasers are recording all the parameters…

(8 March 2002 ) The work is progressing and the Chronolith has finally found its final shape

(24 March 2002) Let’s talk about the defects of this clock, because they are far more interesting and instructive than its qualities. What happens, for example, if a light bulb goes out? And what is the effect of lamp wear on the amplitude of the pendulum? Since the radiometric effect is strongly dependent on the strength of the light, the pendulum should immediately slow down. But the clock corrects itself strongly. The picture below shows exactly that. Until the middle of the picture below, one of the bulbs is masked, which means 25% less power. After that, I remove the mask. You can see that the curve rises by about 5 millionths of a second per beat, or less than 0.5 seconds per day. It will therefore be very easy to compensate for any temporal drift by changing the voltage across the lamps. This will have the same effect as turning the weights, but without having to stop the pendulum.

This sample was taken with a lamp voltage of 7.3 volts, with a vacuum of 2.7 X 10 power -1 millibars. It is a laser that measures the passages of the pendulum through the glass tube. We can see random fluctuations between each sample of up to 20 millionths of a second. It is very important to know where they come from. They are not due to differences in the strength of the radiometric effect, as you might think at first. No: they are due to the way the Chronolith was originally designed. For the cells that register the passage of the balance wheel are infrared cells whose beam is reflected by a reflector placed on the balance wheel. And they are not accurate at all. If I put my measuring device directly on the output of the cells, this is what I get when I set it to the same scale…

We see here like a random number generator between 1.999300 and 2.000700 moving a very accurate clock. We also see very clearly what corrections will be made on the next one. It is enough to replace the infrared detectors by other lasers. The aesthetics of the Chronolith would suffer if I gave it one. This was the most difficult part of its construction. It would have been so much easier to take a bigger tube, to eliminate this stone, to avoid all these sealing problems… but no: the Chronolith is now as I wanted it, coherent, finished and as if it had sprung out of a line.

The pendulum

(April 21, 2002) The Chronolith is still being tested. These days I let it die slowly of asphyxiation by letting air in, while the control computer records all the parameters. It is still able to push the pendulum to a pressure of 8.5 X 10 power-1 millibars. Other tests underway include exposure to direct sunlight to assess its self-compensating ability. These tests should last another month, the time to acquire as much knowledge as possible, and then the clock will be exhibited for some time at the International Clock Museum in La Chaux de Fonds.

(1 July 2002 ) The Chronolithe won the first prize in the international competition of the Kinetic Art Organisation (Miami) in the category « Engineering Ingenuity ».

(23 September 2002 ) The Chronolithe is installed at the entrance of the Musée International d’Horlogerie in La-Chaux-de-Fonds.

And if you are curious to know about an unsolved physics problem that the Chronolith has raised, now is the time. These lines will only be of interest to scientists.

There is an still unsolved question about the Chronolith. Have a look to the diagram below…

A laser beam is cut every second by the pendulum’s course. My computer is set to capture the total time for 10 swings and record the average time for a swing in a data file. In the image above the middle of the picture is set at one second and the grey scale intervals are set to 0.000050 seconds.

Horologists are always looking for the flatest line. A higher line tells us that the clock is too fast, and a lower one shows that the pendulum’s beat is too slow. The graph above looks reasonably good but the software does not plot all of the points. A more detailed look at the data using Microsoft Excel is considerably more interesting. An heavy (5 Mo) Excel file with much more details of the datas recorded can be downloaded if asked.

This view of Chronolith’s data clearly shows that the pendulum is sometime 0.000020 seconds too fast and just after approximately 0.000020 too slow. Why? The histogram below shows the distribution of time intervals measured.

Histogram calculated by Bob Holmstrom

This cannot be the chance. The only thing I’m shure is that this fact happens more and more often when the air pressure insidethe pipe increases. If I let slowly some air penetrate inside the pipe until the radiometric effect vanishes, we can see that theperturbations
appears more and more often in a wider range until the pendulum stops.

Histogram calculated by Bob Holmstrom

The question why we can see this kind of phenomena has never been answered until yet.