In the June issue of the Horological Journal, Colin Walsh reported on the Harrison Decoded conference, held at the National Maritime Museum (NMM), and the surprise announcement that Martin Burgess’s Clock B was awarded a Guinness World Record for being the ‘most accurate mechanical clock with a pendulum swinging in free air.’

For the benefit of those unable to attend the conference, this article outlines the conditions under which the clock was tested and summarises its behaviour during the recent 100-day trial. The trial was conducted in the Royal Observatory’s horological workshop, situated on the ground floor of the Great Equatorial Building. This room is not the perfect place to place a pendulum clock. The workshop is in constant use and, being inside a Victorian structure, is not greatly protected from the weather outside. However, the clock is mounted to a brick column that supports the Great Equatorial Telescope on the top floor of the building, which offers a good solid support. That said the building regularly reverberates with the energy imparted by enthusiastic (school age) visitors’ footsteps and the daily closing of the heavy doors to the adjoining gallery.

In an ideal world, the clock would have been placed in Flamsteed House basement to provide conditions conducive to good timekeeping performance: solid walls and near constant temperature. After all, the historic Shortt free pendulums were housed in this basement in the 1930s. But, as a Museum, any continuation of the tradition of testing precision instruments on site had to be an accessible part of the visitor experience and the decision was made because the workshop is visible to the public from the Time for the Navy gallery.


Burgess Clock B at the Royal Observatory, Greenwich

It is important to note that throughout the trial, the clock was kept within a Perspex case and that the purpose of the case was only to protect the clock from dust, spiders and humans; it did not provide a hermetic seal, so the pendulum was swinging in free air and therefore affected by any change in temperature, barometric pressure or humidity. The clock case was wired shut in April, 2014, and rendered tamper-proof by wax impressions that were kindly applied by the National Physical Laboratory (NPL) and the Worshipful Company of Clockmakers.

To ensure that there can never be any doubting that Clock B was fully exposed to atmospheric conditions throughout the trial, an environmental sensor was placed inside the case. This was connected to a Microset unit which provided a continuous log of air pressure, temperature and relative humidity alongside the clock’s rate and amplitude (arc of pendulum swing).

Fig. 1

Fig.1 Electronic logging of the atmospheric conditions inside the case and the clock’s rate and pendulum amplitude during the 100-day trial

The electronic data gives a very good overview of the clock’s characteristics and will serve as a useful reference to run through the environmental effects on the clock’s running. As can be seen in Fig. 1, there are two anomalous readings in the amplitude trace. These are not real events, but artefacts caused light interference to the optical sensor. With this weakness in mind, it was decided that the trial would be judged on results observed manually and that the electronic data would provide supportive evidence of the clock’s performance.


Barometric pressure fluctuated with some particularly large swings during the middle of the trial period, the largest of which was from 977 to 1038 millibars. From Fig.1, it can be seen there is no discernible effect on the clock’s rate, but when the raw data is put under a microscope, it becomes evident that there was a small effect and that around 96% of barometric influence was compensated for (see Fig.2).


Fig. 2 Scatter plot showing the effect of barometer on rate, from 2013-14, before compensation adjustment (left) and during the trial (right)


From the graph it is evident that the amplitude of the pendulum is affected by both barometric pressure and temperature. The clock has now been adjusted, as Harrison instructs, so that changes in amplitude caused by barometric pressure change (changing density of the air) affect the rate of the pendulum, using the suspension spring and cheeks, and compensate for floatation – the second effect of changing pressure. However, changing temperature also affects the amplitude and as the cheeks are not designed to correct for circular error, an unwanted change of rate occurs when the temperature goes up or down.


The most pronounced effect on the clock’s rate was caused by temperature change.  Looking at Fig.1, one can see a sharp dip in temperature between days 18 and 28. During this part of the trial the clock reacted positively to drop of around 4 degrees Celsius and the pendulum sped up by almost 0.000001 seconds per swing. As there are 86,400 seconds in a day, this change in rate equates to a gain of 0.086 seconds per day. This small temperature effect only became apparent after the barometric compensation had been adjusted as can be better seen in the scatter plots. The 2012/13 plot (Fig.3) shows that the clock was slightly over-compensated for temperature change (the clock ran faster in heat) before the adjustment for barometric compensation. The 2014/15 plot (Fig.3), shows that, without any change to the pendulum’s temperature compensation, the clock ran slower in heat owing to this entirely different class of temperature effect.


Fig3 Showing the clock’s different response to temperature before and after the barometric compensation

Here, it is reasonable to infer that around 94% of the physical changes in the components of the clock were compensated for and that this new characteristic was caused by changing density of air. The denser cold air causes more drag on the pendulum; therefore the pendulum loses more energy and reduces its arc of swing – causing the clock to run slightly faster.


The effect of relative humidity on the clock’s rate is minimal, if not zero. Any quantification of this effect is beyond the scope of the current data set.

Manual recording

Because the electronic data capture is not infallible, it was decided that a manual record should be taken of the clock’s going throughout the trial and should serve as the primary data. As agreed with the NPL, who acted as peer reviewer of the trial, an MSF radio-controlled clock was our main time standard, which was regularly checked against a computer time service corrected by network time protocol (NTP). At the beginning of the trial an eye and ear method was used to rate the clock. The telephone time signal was checked against the MSF clock and then used to gauge Clock B‘s rate. As the trial progressed, these observations were improved by using a smartphone to record a slow motion capture of the pendulum and an analogue MSF clock to precisely gauge the fractional part of the second using the beat scale.

The results

Fig.4 shows the both the electronic and manual plots.  The two lines are separated by around one quarter of a second because electronic data assumes a zero starting point and the manual record takes into account the fact that the Clock B was showing UTC -1/4 second at the beginning of the trial.


Fig4  Showing the manual record and electronic record for the duration of the trial

As can be seen, the electronic equipment failed for a few days towards the end of the trial and we lost a little of the electronic data, but the manual record continued and it was verified by a group of independent witnesses that on day 100 that the clock was well within the one-second margin. The final error after 100 days was -5/8 of a second with a maximum deviation from its starting point of 3/4 second, gaining Clock B its place in Guinness Book of World Records.

What next for Clock B?

At the time of writing, the clock is still running in the horological workshop at the Royal Observatory, Greenwich and plans are afoot to improve it by fitting an adjustable temperature compensation device. Once this is completed the clock is expected to comfortably achieve Harrison’s proposed “…nicety of 2 or 3 seconds in a year.” But for now, it remains locked inside the Perspex case and its performance is logged by new equipment running alongside the Microset system. The new equipment uses a picPET chip to record every oscillation with a much higher resolution than was previously possible. A sample of this new data can be seen in Fig.5, which clearly shows the changing energy imparted to the pendulum during the four-minute remontoire cycle.

Fig 5

Fig,5  A sample of data captured from Clock B using the new picPET chip

This trial and the preceding experiences with Clock B have shown us that the principles of John Harrison’s pendulum clock system were indeed correct. Work is ongoing to produce a book, based upon the two NMM conferences, that aims to collate this study of Harrison’s pendulum clock theory and enable clock-building projects to further explore this extraordinary alternative method of making a precision pendulum clocks.