# How do we accurately measure time?January 7, 2010 5:47 PM   Subscribe

How did people make clocks more accurate before accurate clocks?

I was listening to a podcast recently that mentioned making clocks in the 1700s more accurate, so they could be used for astronomical observations and I was wondering how you make a clock accurate if there is no other accurate time keeping device to use to judge accuracy? What is accuracy judged against for time? In my personal experience of time, it is something that is based on broader astronomical systems, but the detailed level (seconds, even minutes) are based on how clocks record the time.

I suspect this reflects some basic misunderstanding of clocks and time, so I need someone to spell it out for me please.
posted by eclecticlibrary to Science & Nature (24 answers total) 12 users marked this as a favorite

I can't remember all the details but this is pretty much the central theme of the book Longitude: The True Story of a Lone Genius Who Solved the Greatest Scientific Problem of His Time, the derivation of Longitude for accurate navigation requiring extremely accurate clocks. It's a highly readable and interesting book, I found.
posted by Rumple at 5:51 PM on January 7, 2010 [4 favorites]

Actually, the big drive to improve clock accuracy was for use in navigation. On the ocean, it's possible to work out your latitude using astronomical observations, and it would also be possible to work out your longitude if you know exactly what time it is.

The whole story is told in the book Longitude, appropriately enough. Highly recommended.

Anyhow, back to your question. If you know that a certain star is going to rise (or any astronomical event) every 24 hours exactly, you can use that to calibrate your clock. Astronomical observations were the yardstick against which clock accuracy was judged. Now we define a second as 9,192,631,770 oscillations of cesium 133, and need to add leap seconds to deal with the slowdown of our planet.
posted by adamrice at 5:57 PM on January 7, 2010 [1 favorite]

At noon, set your clock to 12:00. The next day at noon, what time does the clock show?

(This means you need to be able to define what "noon" is. And as your clock gets more accurate you might have to start comparing the drift after a week, or after a month, etc.)
posted by phliar at 5:59 PM on January 7, 2010

The astronomical event that was actually used is "High Noon". Your clock should show "24 hours" (well, "23-hours 56-minutes-and-change", but, whatever) from noon to noon.
posted by AsYouKnow Bob at 6:02 PM on January 7, 2010

There were very accurate clocks long ago -- pendulum clocks. How you measured their accuracy was by timing durations between astronomical events at night e.g. the exact time that a particular star appeared at the horizon. That duration wasn't exactly 24 hours, but astronomers by then did know what the duration was, and that could be used to compare against what a clock said to see how accurate that clock was.

The big problem for longitude calculation was that you can't use a pendulum clock on a rocking ship. The solution was figuring out how to make high quality and very consistent spring steel for clock springs, which couldn't be done until someone figured out how to melt steel.

In between those two historical events there was an intermediate one. Astronomers learned that the movement of the four main satellites of Jupiter were extremely regular and certain observable events could be predicted months ahead with an accuracy of a second or two. If the night was clear, and Jupiter was up, and you had a decent telescope (and it didn't take much of one) you could observe the movement of Jupiter's moons and use that to check your clock's accuracy. You could also use it to set your clock, if the clock had gone out for some reason (like trying to operate a pendulum clock during a storm at sea).

I've seen pictures of books which were printed hundreds of years ago which consisted only of tables of predictions of observable events in the Jupiter moon system.
posted by Chocolate Pickle at 6:10 PM on January 7, 2010 [1 favorite]

appears that adjusting the swing width and the pendulum length on pendulum clocks were some of the early improvements to mechanical clocks...
posted by SueDenim at 6:12 PM on January 7, 2010

Astronomical observations, as everyone said. The instrument most often used for making such observations is a transit circle, or meridian circle. These were vitally important instruments for timing; the Prime Meridian was originally defined as "the meridian passing through the transit instrument at the Observatory of Greenwich".

There's a great bit in Twenty Thousand Leagues Under the Sea about Captain Nemo making a related sort of split-second observation to verify that he had reached the South Pole. (Like everyone else at the time, Verne wasn't aware that there was land underneath it. Science marches on.)
posted by tellumo at 6:12 PM on January 7, 2010

Sundials
posted by phredgreen at 6:15 PM on January 7, 2010

Response by poster: thanks everyone for these answers, they are all useful, so I won't mark a best one. I will have to do some reading and track down a copy of Longitude.
posted by eclecticlibrary at 6:16 PM on January 7, 2010

This is only sort of on-topic, but moving this into modern times, the way to set many clocks accurately is by having standard time servers. Computers do this automatically now, but before this was possible, people would travel the world with super precise clocks just to set the time in the satellite observatories. I know because my Dad was one of them.
posted by jessamyn at 6:18 PM on January 7, 2010 [2 favorites]

If you know that a certain star is going to rise (or any astronomical event) every 24 hours exactly, you can use that to calibrate your clock. Astronomical observations were the yardstick against which clock accuracy was judged.

There are no celestial events which take place at exactly 24 hour intervals. There are events which are exactly one sidereal day apart, but a sidereal day is 23 hours, 56 minutes, 4.091 seconds.
posted by Chocolate Pickle at 6:19 PM on January 7, 2010 [2 favorites]

Good answers here and I love to see chocolate pickle serving our precision needs 23 hours, 56 minutes, 4.091 seconds a day

Longitude is indeed a good read, but you should also check out the first section of Danile Boorstin's The Discoverers which is a quick survey of time measurement tools and concepts from Babylon on up. Highly recommended.
posted by shothotbot at 6:37 PM on January 7, 2010 [1 favorite]

There are events which are exactly one sidereal day apart, but a sidereal day is 23 hours, 56 minutes, 4.091 seconds.

Did they know that was how long it took? Because you don't really need 24 hours, you just need to know how far apart two things are.
posted by smackfu at 7:16 PM on January 7, 2010

Did they know that was how long it took?

They didn't know to millisecond accuracy, but they did know that it was a sidereal day, and they did know that there's one more sidereal day in a year than solar day, and they knew how long a year was supposed to be. So they could calculate the length of a sidereal day. It was as accurate as any time keeping they were capable of.
posted by Chocolate Pickle at 7:41 PM on January 7, 2010

Thirding Longitude. Great book!
posted by intermod at 8:09 PM on January 7, 2010

Chocolate Pickle: There are no celestial events which take place at exactly 24 hour intervals. There are events which are exactly one sidereal day apart, but a sidereal day is 23 hours, 56 minutes, 4.091 seconds.

I think you are confusing the sidereal day and the solar day. The solar day is indeed exactly 24 hours within 2 milliseconds. Of course the solar day varies through the year by about +/- 15 minutes due to the eccentricity of the earth's orbit, but averaged over the year, the time from high noon to high noon is 86400.002 seconds. That fact is the basis of our time system and the reason we use a 24-hour clock as our timebase.

The daily deviation from the exact 24 hour solar day is shown by the the analemma, that funny figure eight thing you see on some globes or sundials. Twice a year the solar day, noon to noon, is indeed exactly 24 hours.
posted by JackFlash at 8:44 PM on January 7, 2010

Actually I explained that poorly. The solar day varies by about +/- 30 seconds throughout the year but cumulatively the mean can vary by +/- 15 minutes. So a solar day is a celestial event that is accurate within less than 30 seconds that can be used to calibrate a clock and exactly accurate twice a year.
posted by JackFlash at 9:02 PM on January 7, 2010

Twice a year the solar day, noon to noon, is indeed exactly 24 hours.

This isn't really true. There are points in time at which the instantaneous angular motion of the sun is the same as its mean motion. (There are actually four such points per year, corresponding to the local minima/maxima of this graph: two for the equinoxes, and two for perihelion/aphelion.) But even near those points, successive noons will only approximately line up; between one day and the next, the sun's motion will have slowed down or sped up slightly, depending on exactly what time of day the equinox or apsis occurred.

Distant stars, on the other hand, will transit exactly once per sidereal day (to within a few milliseconds) no matter where the earth is in its orbit, so they're much more suitable for calibrating clocks. And the length of a sidereal day is easy to calculate from the length of a solar day, as Chocolate Pickle said.

Sorry if this is being too pedantic.
posted by teraflop at 9:17 PM on January 7, 2010

Did they know that was how long [a sidereal day] took?

They knew the length of the year at least as precisely as 365.25 days (that's what the Julian calendar works out to; the Gregorian, to 365.2425 days; IIRC, the Mayans knew it to 365.2422). From that you can easily compute the ratio between the length of a solar and sidereal day, and if you define the mean solar day to be 24h, then the mean sidereal day has to be a smidge less than 23h 56m 4.1s.

Of course this depends on knowing the relation between the year and the day lengths; this is obvious from a Copernican perspective, but I suspect they understood it well enough in the geocentric model as well.
posted by hattifattener at 10:21 PM on January 7, 2010

Teraflop, yes, much better.
posted by JackFlash at 10:31 PM on January 7, 2010

...and if you define the mean solar day to be 24h, then the mean sidereal day has to be a smidge less than 23h 56m 4.1s.

It's even easier than that. The reason for "mean" solar day is that the length of the day changes as Earth moves through its elliptical orbit. But the sidereal day isn't affected by that. You don't need a "mean" sidereal day because all sidereal days are identical. They don't depend on Earth's orbit; they only depend on Earth's rotation, and on human scales that is very consistent.
posted by Chocolate Pickle at 10:45 PM on January 7, 2010

Sure, but the amount of difference is negligible within the context of people 250 years ago trying to determine how accurate their pendulum clocks were.

The sidereal day changes as a result of tidal slowing from the moon and sun. It also is affected by the precession of the equinox. But none of those effects are very great.
posted by Chocolate Pickle at 10:58 PM on January 7, 2010

Good point, CP. I was careful to write "mean" because of the wobbliness of the solar day, but I guess the regularity of the sidereal day is the whole reason we're talking about it in the context of timekeeping, huh?

But you still need to refer things to the mean solar day, because that's where your definition of an hour comes from. Came from. Whatever.
posted by hattifattener at 12:04 AM on January 8, 2010

Simple: they would wind and set the clock every day when the stick in the front yard had no shadow at noooooon.... my2cents
posted by RENNER8592 at 1:07 AM on January 8, 2010

« Older What must I do to teach in the US?   |   Is there such a thing as head attacks? Newer »