Shortt–Synchronome free pendulum clock
en.wikipedia.org
en.wikipedia.org
This is incredible. Its performance matched or exceeded quartz oscillators in modern electronic systems [0]. It's the ultimate (electro-)mechanical clock [1].
A good general-purpose quartz crystal oscillator is only stable to 20,000 ppb, a good temperature-compensated oscillator is stable to 1000 ppb, and a crystal oven (a quartz crystal in a sealed, temperature-controller oven, the best quartz oscillator you can ever made) is stable to 1 ppb. And this ultimate mechanical clock is stable to 2.31 ppb [0], I'd happy to use it as the time base for my SDR receiver or spectrum analyzer if I was given the chance... It's only measured for a month, so long-term stability can be worse (and crystal oven's short-term stability can be better), but you get the idea.
> Shortt clocks kept time with two pendulums, a master pendulum swinging in a vacuum tank and a slave pendulum in a separate clock, which was synchronized to the master by an electric circuit and electromagnets. The slave pendulum was attached to the timekeeping mechanisms of the clock, leaving the master pendulum virtually free of external disturbances.
The basic principle of accurate timekeeping - using a free-running primary oscillator as a reference, and lock a less-accurate secondary oscillator to the primary, is vital in electronic systems. Atomic clocks, radio/GPS time signal receivers, or NTP, all use this principle. This clock basically implemented a rudimentary form of PLL mechanically.
[0] https://en.wikipedia.org/wiki/Crystal_oven#Comparison_with_o...
[1] Now electro-mechanical clocks are making the comeback, in the form of MEMS oscillators in consumer electronics. But MEMS oscillators are often slightly worse than its equivalent crystal oscillators, and their main advantage is size (i.e. chip-scale integration), not accuracy.
Note that quartz crystals are also electro-mechanical devices, as the quartz crystal is physically vibrating. This is particularly easy to see in lower-frequency quartz oscillators, such as the common 32.768KHz ones, that physically look and act just like tuning forks: https://www.youtube.com/watch?v=_2By2ane2I4
This also means that mechanical vibration/shock can throw off quartz crystals: https://www.youtube.com/watch?v=M4Z7CjxfUU8
> This also means that mechanical vibration/shock can throw off quartz crystals
This is one strength of MEMS oscillators, they are more tolerate to shocks and vibrations. https://www.sitime.com/sites/default/files/gated/SiTime-MEMS...
https://www.taitien.com/wp-content/uploads/2016/01/NA.10M.20...
This OCXO for example starts off within 1ppb/day. But yearly is within 100ppb, and 10 year stability is within 600ppb.
In stock for $120 CAD on Digikey, for a single unit.
If I had to take a guess the reason is the only coupling through to the pendulum would be via the friction in its bearings or via movement of the air around it. Both of these are naturally minimised in most designs so normal vibrations are not an issue.
Long-term performance depends on:
– temperature (absolute, relative, gradient)
– atmospheric pressure, air currents
– humidity
– shock & vibration (air, ground, wall)
– material stability or other disturbances
– energy loss stability & energy gain stability
While I'm not trying to claim it has no effect, the effect must be slight enough as to be negligible, otherwise we'd have seen some kind of compensation as we see compensation for temperature, spring tension etc.
Perhaps I'm underestimating it though and the reason this is not compensated for is simply that these where the atomic clocks of their time. They would represent a significant investment, no idiot would install it in a building set back from a major truck road or expect to be able to run tap dancing classes in the room above. It's quite likely that the manufacturer set out standards for the building it was installed within, perhaps even down to the type of foundations it had to have.