The Guardian has an article on the verification of a 250-year-old claim by clock maker John Harrison over the accuracy of his designs:
After a 100-day trial, the timepiece known as Clock B – which had been sealed in a clear plastic box to prevent tampering – was officially declared, by Guinness, to be the world's "most accurate mechanical clock with a pendulum swinging in free air".
It was an intriguing enough award. But what is really astonishing is that the clock was designed more than 250 years ago by a man who was derided at the time for "an incoherence and absurdity that was little short of the symptoms of insanity", and whose plans for the clock lay ignored for two centuries.
[...] At a conference, Harrison Decoded: Towards a Perfect Pendulum Clock, held at Greenwich yesterday, observatory scientists revealed that a clock that had been built to the clockmaker's exact specifications had run for 100 days during official tests and had lost only five-eighths of a second in that period.
The same story is also covered at The Telegraph and The Independent.
Wikipedia has more background on John Harrison. NOVA's Lost At Sea: The Search For Longitude (transcript) may also be of interest.
(Score: 1, Informative) by Anonymous Coward on Monday April 20 2015, @10:34PM
More like 2 seconds per year. So just under 10 minutes in 250 years (5/8 * 365.24/100 = 2.283 s/yr *250yr = 571s = 9m31s).
(Score: 4, Informative) by martyb on Tuesday April 21 2015, @12:33PM
Looking at in another way, I am even more amazed: the clock's accuracy was within 20 parts per billion. That, in a mechanical device that was designed 250 years ago... before there even was a USA!
(60 * 60 * 24 * 365) / (5/8) == (60 * 60 * 24 * 365) * (8 / 5) == 1 part in 50457600 == 20 parts in 1.009e9
Wit is intellect, dancing.
(Score: 2) by martyb on Wednesday April 22 2015, @02:06AM
Blegh! Something was gnawing at me as I wrote that early this morning... I left out the part that it was off by 5/8 of a second over the course of 100 days! Please mod parent down. Even though it is quite late in the evening, I hope this time (pun intended!) I can get the calculations right! Revised calculation follows:
The clock was found to be off by 5/8 of a second over the course of 100 days.
How long would it take to be off by one full second?
It would take 160 days for the clock to be in error by 1 full second.
How many seconds are there in 160 days?
There are 13,824,000 seconds in 160 days.
==> So, the clock was accurate to 1 part in 13,824,000.
How many parts per billion is that?
The clock was accurate to 72 parts per billion.
Wit is intellect, dancing.