The cHipp Toggle clock

A DIY, free pendulum, electrical master clock project.


Inspired by Matthias Hipp's idea of 1842.

You may have landed on this page because you "Googled" Hipp Toggle. Somehow this webpage is the No1 hit. So here's a definition of the Hipp Toggle.

If an unpowered clock pendulum, is set to swing, it will describe an arc that will slowly decline in amplitude. The decline is affected by three interferences.

1) the suspension on which it hangs. This is usually a spring
2) The resistance of the air through which it passes. This varies with barometric pressure.
3) Any interference to it motion that its motion that occurs during its arc of swing. This may be mechanical or magnetic.

Matthias Hipp's original Hipp toggle was a simple device ( a metal toggle) that was activated when the arc of swing fell to a certain minimum amplitude. It then momentarily closed a contact, which powered an electro magnet, which accelerated the pendulum slightly on that swing. The arc increased and the Hipp toggle effect would not be invoked again until the amplitude had fallen again to the set level.

Matthias Hipp proposed this idea in 1842. Below is Frank Hope-Jones description of the Hipp toggle. It's reproduced from his 1949 book "Electrical Timekeeping". Frank Hope-Jones was the designer and manufacturer of the 20th century Synchronome master clocks (which did not utilise the Hipp toggle)

....But one cannot speak of merit in electric clock inventions without giving pride of place to the Hipp "Butterfly" escapement which is illustrated in fig. 37. An apparently free /pendulum has an armature attached to its lower end above an electro-magnet fixed to the case. The pendulum carries a little freely swinging vane. This vane trails backwards and forwards over a notched block mounted on the end of the upper spring of a pair of contact springs until, as a result of the gradually decreasing arc, it engages the notch and forces the top spring vigorously into contact with the lower. The notched block, or the electro magnet, is placed a little off-centre, and the contact is so adjusted that the attraction takes place at the correct phase of the pendulum's vibration, the optimum position for increasing the arc and causing the vane to trail over and beyond the notch until the action is called upon to repeat itself.

Also: Observe that, though fluctuations of battery power vary the value of the impulse, nevertheless the frequency of the impulses is increased automatically and in exact proportion to their lack of strength; hence the average arc is reasonably constant.
A further outstanding merit of the invention is the comparative freedom of the pendulum, due to the wide intervals at which contact and 'impulses take place. Add to this the concentration at zero of such interference as exists, and
you get such a good timekeeper that it was actually tried in observatories for precision purposes.

You must read Frank Hope-Jones's interesting book for more.

The Hipp Toggle principal was not widely used but the British Post Office used it in their telephone exchange clocks master until about 1980. There is a good explanation of the Hipp Toggle Post Office No. 36 clock at . If necessary have a look ; It may help understand what follows.

My Clock

I had wanted to make a pendulum clock of my own for a long time. It was always going to be largely electronic because my mechanical skills are not great. My inspiration came from the master clocks with one second pendulums that were so popular in large buildings during the first three quarters of the 20th century. The Synchronome clock designed by Frank Hope-Jones being one of the best. The story of other clocks I have is HERE

The time keeping of the Synchronome clocks - claimed to be within 2 seconds per week - would be the standard I would aim to achieve.

I have never been an proper horologist so my "bible"has been the book referred to above and I have tried to abide by the important priciples outlined by HJ. I began making my clock in March 2015, using cmos ICs (chips) There are other electrically operated pendulum clocks described on the internet and I don't claim that mine is any better than them. I don't know how to program a micropressor so I didn't consider using one.

My "toggle" effect is produced with opto-interupters, rather than mechanical devices. When something interrupts the infra-red beam which passes between the two small pillars, an output pulse is generated for as long as the beam is broken. The type I used have a comparator chip on a small pcb and were sourced cheaply from eBay. (as low as 99p)

A short length of soft sleeving which I will call the "flag" is attached to the 4ba rating thread beneath the bob so that it passed between the the pillars of three opto-interupters as the pendulum swings back and forth, creating the operating pulses as it passes.

The optos are arranged from right to left in the order A, B, C. When the pendulum is stopped the flag hangs midway between optos A and B - ie. at Bottom Dead Centre (BDC). An important point is that the electro magnets are co-incident with opto A; slightly to the right of BDC.

When the clock is running, as the pendulum swings from its rightmost extremity, it first passes through opto A, then opto B generating a short pulse from each one as it goes. A and B are 0.5" apart, centralised on BDC.

Opto C is positioned 3/4" to the left of BDC and is the Hipp Toggle sensor. It works like this. Assume the pendulum has sufficient amplitude to swing right through opto C at its leftward extremity. A pulse is generated. It then stops and begins to swing back to the right. As it does it passes through C again, generating a second pulse.

So on every swing that has sufficient amplitude to pass through opto C twice, two pulses are generated in quick succession. These are fed to a cmos counter chip (CD4013, a) which is wired to divide by two. In other words the output of the 4013 counter is clocked HIGH then LOW.
This is the digital equivalent of the toggle in the mechanical version passing over the V block.

All the while the pendulum arc is slowly decreasing and eventual (now, after 52 seconds in my clock) there will come a time when the flag does not pass right through opto C but enters it, reverses, and proceeds to the right again generating only ONE pulse. This will clock the counter HIGH and and it will remain HIGH as returns to the right. This is the equivalent of the mechanical vane catching on the V block in a mechanical Hipp Toggle clock.

Then as the pendulum continues its swing back to the right it soon passes through opto B (1/4" from BDC) generating another pulse. A basic chip known as a NAND GATE (IC3a, CD4011) detects that it now has two HIGH signals on its input terminals - one from opto C which was left in the HIGH state and a HIGH from opto B as the flag arrives. With these two HIGHs on its input its ouput goes LOW, is inverted in IC3b to make a HIGH which triggers a flip flop (IC2b) to HIGH which remains HIGH until the flag passes through opto A which resets it again to LOW.

Thus the HIGH on THIS flip flop (ic2b), as the pendulum travels between opto B and A, is the drive period control for the electro magnet's drive pulse. This drive pulse is switched on with a transistor and the Electro-Magnets are powered during that section of the arc (B>A) giving the pendulum an impulse which maintains the pendulum in motion.

On the next swing the pendulum will again pass right through opto C because of the slightly increased swing amplitude and the process will repeat in as many seconds as the energy given by drive pulse will allow. (now 50 seconds in my clock)

A feature of this design is that the pendulum arc is closely controlled by the position of opto C relative to BDC. ie. .75" This works out to be about 2.55 degrees or 1.28 semi arc. This is quite a small arc which is good for time-keeping and the system has been so reliable that If I was starting again I might try an even smaller arc. (the semi arc of a Synchronome is about 1.76 degrees I think)

Here's a short video showing the pendulum as it was in an earlier implementation when the "flag" was in front of the bob which was a serious mistake. The LED on Opto C can be seen to blink twice until, after about 30 seconds, it only blinks once (circled). A drive pulse is generated as described above. Please note that sometimes the quickest of the double blinks appear a single blink due to the frame rate of the video.

The circuit of the electronic cHipp Toggle section.

The purpose of Ic1 (6x Schmidt trigger inverters) is to speed up the pulse edge transitions from the opto-isolators. Used in pairs,the original pulse polarity is maintained.

The completed electronic cHipp Toggle unit

We now have a drive pulse positioned 1/4" either side of BDC which, when amplified by the transistor switch, powers the EMs for 220 milliseconds; that period being set by the distance between opto A and B.

Frank Hope Jones in the Synchronome clock, powered his pendulum at fixed 30 sec intervals with the potential energy of a gravity lever hanging up on a catch so that the energy imparted as the bob passes through BDC was always the same.

The pendulum impulse in this clock is delivered magnetically to the armature beneath the bob. The force of attraction of a magnet varies as an inverse square law of the decreasing distance to the approaching armature.

In my clock, the energy that pulses the EMs does not come directly from the battery but from a capacitor charged to precisely 5 volts from a voltage regulator. This ensures that exactly the same energy is available on every impulse. The capacitor value was found by trial and error to empty its charge during the 220 millisecond period that the drive is energised. Because of the capacitor the current through the EMs decreases exponentially over the 220 millisecond drive period.
Without the capacitor, as the distance to the EM decrearses, the effect of the magnetic attraction would increase as the square law..However the magnetic field is decaying exponentially due to the discharging capacitor and (it is hoped) the two effects will produce a roughly steady accelerating force on the bob. The interference on the pendulum is now occurs only for 1/250th of its running time anyway.

The oscilloscope image below shows the falling voltage applied across the drive coils during the 220 milliseconds pendulum drive period. A side effect of this system is there is virtually no back EMF generated when the swicthing period ends, so protection of the switching transistor is not needed.

Due to inductance, the current waveform is slightly different..In particulat it has a slower rise.

The armature beneath the pendulum is of low remnance steel taken from a GPO relay. There are no fixed magnets in this design.

Here is the circuit diagram of this "capacitor drive" section.
(this is a scan of my own notes so the pin numbers etc are not relevant)

How it works. The Qbar output of ic2b in the Hipp toggle circuit , which is normally HIGH, is connected to the input (i/p) on the left. This turns on 2n3906 transistor which in turn turns on the 2n5193 pnp transistor above. TThe 3300uf capacitor is charged quickly to 5 volts exactly . The 5 volts comes from the 2940-5 low drop out voltage regulator (2940-5).

When the cHipp Toggle circuit calls for a pendulum impulse, Q of ic2b turns on the EM's drive transistor (lower right), At the same instant Qbar goes low , removing the 5 volt feed to the 3300uf capacitor. The capacitor is now the only source of current for the EMs during the drive pulse period. (220 Milliseconds)

* the 16ohm and 33ohm resistors in the electro-magnet circuit were adjusted to achieve a time between impulses of about 30 seconds.

The first version of this clock used a dial which required a drive pulse every 30 seconds. The description of the divide by 30 counter board which was previously shown here has been removed. The new scheme is described later on.

September 10th 2016 this page is being edited. The blog entries which follow are soon to be removed and only a description of the finished clock will remain..

The clock

Originally the pendulum rod was made up from 3 lengths of silver steel salvaged from old PC printers. But later I bought 1 mertre of 8mm Invar and replaced it. (£70 !)

The pendulum's semi arc is 1.28 degrees and is subject to a very small cyclical variation inherrent in a Hipp Toggle system. The average arc over the period between impulses is tightly controlled. The pendulum is about a "free" as a pendulum can be without being enclosed in a low pressure tank. (as in the Shortt / Synchronome clock)

The mahogany case came from a clock related junk shop in Brixham.(Now sadly gone !) It has been modified and varnished. It may have been easier to start the case from scratch.

The pendulum is suspended on an oak bracket and the top chops are similar to those used by Gents and the GPO No.36 clock – two penny washers and a length of threaded rod. A piece of .005” feeler gauge stock serves as the suspension spring.

Initially he pendulum bob was a brass shell case with a coil of 6” wide lead inside which was left over from the lead flashing when I built my workshop a few years ago. Before final assembly a length of steel pipe was added and polyester resin partly filled the space remaining to stabilise the whole. The final weight was 6Kg. But in May 2016 this was changed to a steel bob 8" tall x 2" diameter weighing 6 lbs.

The picture below (left) shows the pendulum stationary at BDC and the magnet drive assembly. The the optos interupters are behind the black plate, The electro magnets were taken from an old buzzer.

In spite of the small black shields, direct sunlight falling on the clock can upset the operation of the opto-isolators, therefore the clock needs to be positioned where the sun cannot directly shine on it.


Drive magnets and optos behind the plate
Oak pendulum support and chops

I bought a Racal-Dana 9904 counter/timer from eBay (£50) to help with this project. It has been useful in its "stop watch" mode and can measure the time between two pulse edge events to 4 places of decimal (1/10,000) A 30 second period of 30.0001 seconds would equate to .28 second error in 24 hours if my maths is correct. Later , when I changed to an alternating seconds slave movemnent, the 30 second divider board was removed and this timer could no longer be used.


Building a clock which incorporates ideas of your own and at the same time working in relative horological isolation throws up many issues. One can only wonder in amazement at early experimenters like John Harrison who would have needed great patience and dedication. Even checking the daily rate would have been a painstaking process.


During the summer I had noted occasional errors of 1 second or multiples thereof. I was well aware that opto-isolators are sensitive to prevailing light intensity and had mounted them behind a small black shield.

However I have now seen that where it is currently sited, as the evening sun sets, there are two short periods where shafts of sunlight, at different times from two different windows, can shine very brightly onto the clock. Not always of needs to be a good bright sunset and the positions of curtains also make a difference. The clock has now been running for the last two weeks with a large cardboard shield over the bottom half of the clock and there have been no unexplained jumps in the the apparent rate.


Various clock related projects were started over the Christmas and new year period; mainly to do with the making of hands and dials for new acquisitions. Close monitoring of the cHipp Toggle clock has been largely stopped and it's been left to carry on without any attention.

I've just finished re-reading the account of John Harrison's life story. At about the age of twenty he was already a clockmaker of great ability and had made two long case clocks, at least one of which he claimed was accurate to one second a month. This claim can be found repeated all over the internet when searching on JH's name for more information and it is never questioned. At this early stage of JH's life work he concentrated on the effects of temperature, friction/lubrication, improving the escapement and elimination of cycloidal error. Nowhere is the problem of barometric pressure mentioned until I found one reference to the making of a new pendulum clock to Harrison's original design that was finished in 2014. It was apparently tested in 2015 and was found to be equal to the one second a month rate but only after allowances were made for the changes in barometric pressure. I find this all very confusing and if any knowledgeable horologist ever reads this and can enlighten me please email me using the email address on my home page.

The cHipp Toggle clock has no friction other than that of the suspension spring and the air through which it passes. It has a very small arc which is tightly controlled and cycloidal error can be ignored. The is no escapement to interfere with the motion of the pendulum except for the precisely controlled impulse which are now every 50 seconds. Temperature control by Invar rod must equal that of the gridiron pendulum and, anyway, the temperature control in a centrally heated home does not allow the wild variations experienced in Harrison's home workshop. Therefore I am tempted to believe that an electrically impulsed clock of this type must be as good a pendulum clock as can be constructed whilst running in free air. Only the often mentioned quartz rod pendulum rods might offer a small improvement.

I have had one email from someone interested in the clock I've made and in my next blog I'll include a better description of how the electronics work. By the way I make no claim of this design being superior to other electronic pendulum clocks that you may come across on the internet. Some are indubitably better made ! But one thing I do think is that any design that makes use in its execution of timing circuits other than the pendulum I believe are "cheating" !


One of the advantages of the Synchronome system is that the current consumption of a Synchronome is miniscule whereas any electronic clock has a continuous drain. Of course, in a modern quartz clock this drain is reduced to a tiny current but the cHipp toggle clock is drawing 60 Ma for the three opto-isolators and more besides for the various LEDs that I've fitted.

This is not much of a problem as both systems have trickle charged batteries in case of a power cut. But I've had thoughts on reducing this in mine. So I'm thinking of stopping the cHipp Toggle and experimenting with a single laser diode beam and separate light sensors. I would also like to change the dial to indicate seconds so an additional sensor may be placed to get equally timed one second pulses. (One second pulses could be obtained from opto A or B but they would not be evenly spaced) I'll report on this again if I achieve anything useful !


The laser beam experiment was tried. I fixed a small reflecting surface at the suspension chops and projected a laser (from a laser pointer) upwards onto its surface and picked up the return beam below on a piece of card. The reflecting surface was polished stainless steel but the surface was imperfect and the return beam was imprecise. I decided it was an impractical idea and abandoned it.


It's time for an update. I have become suspicious about the effect of temperature variation s on the cHipp Toggle clock and so have made some changes. I wondered about the temperature stability of the lead filled brass cased bob. I've had a 7 pound mild steel one made and am monitoring the rate. I only bought 1 metre of Invar (expense) butI am hoping that the plain steel extension inside the bob (which includes the 4 ba rating thread) is approxiamate temp compensation for the upward expansion of the bob.


I have a 1 sec alternate polarity, silent impulse dial and have started to modify it to fit a dial plate in the cHipp toggle case. When this is completed and installed, the blue 30 second divider board will be dispensed with.


Work is in progress on a case for a new project. The case is the carcass of a GPO No36 clock in dark oak that I had but had no door. I am making a new one with a bar across for an integral dial. I will soon be collecting some parts of an unfinished clock project and the new clock will incorporate these parts.

Meantime I am remaking the dial plate and dial of the cHipp Toggle clock to accept a 1 second, alternate polarity silent Gents movement. This means the 30 second divider board can be dispensed with and a simple alternate pulse driver will be made incorporating an H-Bridge from eBay.


Just in case anyone does check in here occasionally I must report that not a lot has been done recently ..Holidays etc.! However the new dial is in position in the case and the electronic control boxes are "on the bench" awaiting the inclusion of two more ICs. The door carcass for the next clock has been made.

I have also bought two more Synchronomes one of which ( a late MK2) has two BO22439-3 printed circuit boards which send alternate polarity drives for half minute and one second dials. Work is in progress to trace out the circuit diagram. These boards must be rare as their is no published information on the internet - or a mention in the Synchronome bible by Miles. Please email me if you have any info on these boards.


The clock has now been fitted with the new Gents, silent, seconds indicating, alternate polarity slave clock mechanism . The dial motor resistance is 3.4 ohms and the required drive current is 220Ma. The dial was printed on silver card with a laser printer.

Clock dial, July 2016

The circuits of the cHipp Toggle unit and the pulse shaper unit shown at the start of this blog have been modified slightly and the circuits shown have been updated. The most significant change is the feed to the clock circuit. ic3a now gets an inverted feed of pulses from Optos A and B. A double pulse, centred closely on BDC, is now the feed to the new clock circuit.

Alternating polarity pulse driver for the Gents low voltage, silent dial mechanism.

The incoming double pulse is converted to a single pulse of about 220 milliseconds duration with a period of one second by 4013a . This is divided by 2 in 4013b. The two opposite phase (Q & Qbar) phase outputs are gated with the single pulse to produce positive going pulses of 220 milliseconds duration alternately, to drive the H-bridge. It was just luck that the spacing of Optos A & B were 220 Milliseconds apart and that the pulse resulting at 4013b i/p was a suitable length to drive the Gents dial movement.

H-bridge module with one pound coin.

The gem in this circuit is the tiny H-Bridge module sourced from ebay for a mere 99p post paid. Rated up to 10 volts and 1.5 amps its more than adequate to drive the dial motor. Pictured above with a £1 coin for scale, it features two H-bridges, so a couple of additional series dials could be run separately from the B output.

The circuit is "breadboarded" at present and it will be awhile before it is all tidied up and back in the case.

The steady current consumption is now 100ma - mostly due to the LEDs in the Opto-interupters and the other indicators.


A few days after the last blog entry I came down one morning to find the clock had stopped. A fault in the pendulum drive circuit. With summer visitors and other distractions it was only finally fixed yesterday. The problem was never exactly located but seemd to be due to a soldered joint on the strip-board it is made on.

So the 1 second alternate drive section is still bread-boarded.


Time for a small update. Still awaiting proper building of the new dial driver board. I want to re-use the 30 second divider board so that I can use the Racal timer again. It is so good at assisting with the job of getting the right close as quickly as possible. Waiting 24 hours between observation stretches my patience ! Recent fiddling has meant I need to regualte it again..


As the year come to a close I am trying to get to grips with microprosesors (Arduino) and mini computers (raspberry Pi) I have to admit to finding this very difficult at my time of life ! Knowing that 12 year olds in schools are lapping up this stuff does not help !
Meantime, some clock work goes on. The reason for the endeavors mentioned above is that I would like to replace manual monitoring of clock rate v baromentric pressure and temperature with an automatic system. A friend from the forum mentioned below has offered to help but I first need to gain some experience of doing basic things with them.
I shall start by monitoring a Synchronome ( I have two working) which create an impulse every 30 seconds. I've been trying various methods of extracting a clean, well defined pulse that is safe to feed into delicate electronics.

Almost everyone is aware that when an inductive circuit is abruptly broken a considerable spike of voltage of the opposite polarity to the battery is generated. In a Synchronome this spike is mildly tamed with a low value resistance, the main purpose of which is to prolong the life of the contacts.
In several of my attempts to extract an electrically isolated 5 volt pulse the spike generated by the break always got through.
The most recent system makes use of a 20 x 2.7mm glass reed relay element stuck to the long side of the Synchronome's electro-magnet (with tape of blutac) In spite of using twin screened wire and wrapping the reed in aluminiun foil earthe to the wire's screen, some of the spike was still getting through,
I eventually realised that the spike is actually a burst of radio frequency waves (seemingly from 1 to 1o Mhz) and are transmitted.then received by any nearby wiring. I found that this energy can be completely suppressed by connecting a .1ufd capacitor across the two main terminals on the Synchrome chassis.

If I had known this before, some of my other attempts might have worked but none was as convenient as the reed relay. It con be fitted in a trice without breaking into the working clock circuit. The logic voltage of the device on which the reed is connected can be used. In the test image below, a 6volt battery was used. The pulse illustrated has not been de-bounced or otherwise processed but looks good enough to be put straight to work in a logic circuit..

Pulse from a simple reed switch operated by the stray magnetic field of Synchronome EM's


In January the clock appeared to have gone faulty.. The second hand was twitching but not advancing. I substituted a different pilot dial and it worked OK. The problem with the 1 second alternate seconds "motor" could not really be identified, although I suspected the armature magnet might be weak.. However another strip down and some judicous bending of the soft iron pole pieces has got it going again. It has been put back in the case and my fingers remain crossed....

Meantime, first steps have been taken to get a graphical time monitoring system working on my Mk2 Synchronome..I have used the software written by the friend from the forum which runs on a Raspberry Pi.. I've ordered another Pi and hope to set up monitoring on the cHipp clock as well..More on that later...


A month has passed which (clock-wise) has been taken up entirely getting to grips with the raspberry Pi monitoring system. I have to acknowledge my thanks for the extreme patience of the creator/programmer of the Pi software (who, for brevity, will be referred to as Mike) who assisted me during this initial stage. At first I had to learn to get to grips with its operation - and to input command line commands EXACTLY as written. This did not come esaily to an old brain, unused to such things. It was first put to work monitoring a Synchronome clock but eventually transferred to the cHipp toggle clock which had been largely ignored while this learning process went on.
Mike's software is written to accept a pulse every 30 seconds but the cHipp clock is now outputting 1 pps. So the first task was to resurrect the old divide by 30 Unilab board and modify it to produce a short pulse every 30 seconds.
I won't relate all of the diifiulties that had to be dealt with but the final, and most significant, mod was removing the direct feed and changing to a galvanicaly isolated feed from /30 to the Pi. A miniature DIL reed relay was used to achieve this.
All of the problems related to mis-triggering of the Pi and/or the divider board by interference. The most signicant being the thermostat switch which controlled the small heating element (a resistor) in the cHipp clock case which regulated the temperature to 22 degrees. For the time being this has been disconnected. Finally, on the 2nd of March the monitoring has begun to work flawlessly. It is a remarkable system and the performance of the clock can be shown in graphical form and the data can be analysed.
In the next update I will upload a graph when I have one of reasonable length.


Data obtained from the computer monitoring is truely enlightening and within two days had shown errors which I may never have discovered without it. Basically, it showed up meanderings of the rate which had little correlation with observable changes of pressure and temperature. I stopped the clock and made the following improvements to the bob.
First, the bush at the top of the bob was not a good fit. It had merely been drilled with an 8mm drill and was slightly oversize. I re-drilled it to 9mm and soldered in a length of brass tube which was a perfect sliding fit.
Next, I fitted a projection to the rod, drilled to fit a pin projecting from the top of the bob. This was to prevent axial rotataion of the bob on the rod. This was important in this clock because the bob carries an armature underneath which could previously change its alignment with the impulsing electro-magnets.
Finally, I put a locknut under the rating nut to prevent it ever moving.

When I reasembled the clock and ran it again I had to re-rate it. With the rating nut locked of this was done with weights on the bob. It took about 100 grams but within 2 days I got it close.
Now it seems to be very much better and I am going to leave it a week to see if it will make the "within two seconds a week" target. I will post the result of the weeklong trial Monday 27th !


In 1771 Captain Cook had taken three years to travel from Plymouth to Deal; a distance of 250 miles.

The cHipp clock lost 0.164 seconds over the one week trial.

Both statements are true but they only tell a small part of the story. Here is the annotated Pi's computer record.

The spread is 0.55 seconds....Still not a bad result though !

Some relevant information. The clock case is insulated inside with 10mm of polystyrene. The temperature in the clock case is controlled with a small heater and thermostat set at 22 degrees C. A max/min thermometer showed that the stat. cut in at approxiamately 21.8 C and out at 23.1 C. Note the "saw tooth" variations in the blue "rate of change" record showing the effect of the cyclic variations due to the thermostatic control of temperature.

On Saturday 18th the clock was started and quickly rated. The last adjustment was the removal of a small weight from the bob on Saturday at about 10pm. The trial period then started at midnight, when the data shows that the clock was close to GMT, which for the purposes of the trial I've called zero (0). After a week, at midnight Saturday 26th the error was -0.164 seconds

The blue line on the graph shows the short term changes of rate, whereas the red line shows the average rate calculated at 5 minute intervals. During the day, when the temp is changing in" saw tooth" fashion, the effect of the consequent rate changes are clearly shown.

In the evenings when the central heating tends to swamp the clock's thermostatic control, the case temperature and room temperature stabilise at 22 degrees, the clock case control stops switching in and out and the the clock shows a loosing rate. This is the period from about 6pm to midnight.
On Wednesday there is a slight change. The pressure remained constant all day at 1028 day and the pattern is different. At 1028 Mb, while under the influence of the clock thermostatic control, the clock shows a very good rate.

Conclusion. The temperature compensation is not good enough. The errors are predominantly temperature related. The clock shows good stability and is capable of improvement.

As I write, the clock is aclimatising to thermostatic control at 24 C. I hope to see more clearly the effect of pressure changes and the phase of the moon. (lol)


The effect of the extra 2 degrees was unexpected. After the thermostat was reset to 24C the clock began to gain. 0.18 gram weight was taken off the bob, mid-day Tuesday when the next trial is deemed to have started. In spite of this unexpected result the clock shows the normal tendency to lose when the temperature increases everyday under the influence of the central heating and the warmer afternoons.

At this new temperature setting it can be seen that the ripples caused by the 1 C variations of the thermostat now occur at a greater frequency. By mid-day Saturday it seemed likely that the clock still had an overall gaining rate. This is shown by the yellow line and is estimated to be about 1 second in 11 days. At mid-day Saturday 0.1 grams was removed from the bob.

I decided to install a thermostat with a 0.1 degree differental and run a further test starting today, 03.04


The experiment with a more powerful heating resistor (36 watts) and a lower differential stat was a failure. All sorts of temp zones must have been caused within the case and graph of the resulting rates was more ragged and the rate was clearly adversely affected by the imposed heating.
A new trial was started without any heating at all and the results look promising after a week. But with a prevailing high pressure sitting over Cornwall I'm not sure yet what it tells me. Except that the original aim of getting a clock to keep time within 2 seconds per week has been met 1


My Homepage
what is sound mixing