There probably were not a lot of us that noticed it during the countdown to midnight, and the New Year, last Saturday, but this year, time needed a tweak. At 23:59:59 on December 31, 2016, an additional second was added to UTC (Universal Time Coordinated, the international time standard) so that, for exactly one second, UTC time was 23:59:60.Â
This might sound a little “who cares” for most of us, but managing the Leap Second is, among other things, essential for little things like running the Internet, and ensuring GPS doesn’t think you’re halfway to the Moon when you’re just trying to find your mother-in-law’s house (literally).
Needless to say, since accurate GPS and hey, a working Internet, are nice things to have, and since managing time for both is already a complicated business, why do something like add a Leap Second? The answer is that UTC isn’t based on astronomical observations â at least, not anymore.Â
UTC used to be based on the rotation of the Earth around its axis, as observed at Greenwich. Once upon a time â a simpler, happier time â the second was exactly 1/86,400 of a day. By the mid-1950s, however, clocks had gotten accurate enough that we’d figured out that the Earth’s rotation on its own axis was irregular, so in 1952, the International Union Of Astronomers decided to define the second as a fraction of one orbit of the Earth around the Sun: a second would now be 1/31,556,925.9747 of a tropical year.
However, the year turned out to have the same basic problem as the day; it’s irregular, changing slightly in length from one year to the next. (This is different, by the way, from the problem that requires the insertion of an extra day in a Leap Year; the Leap Year is inserted to keep the Gregorian Calendar in sync with the seasons, but the reason for the Leap Year, is that there isn’t a whole number of days in a year, not that an astronomical year varies slightly in length from one year to the next.) The search, therefore, was on for a definition of the second that didn’t rely on irregular astronomical phenomena. And, by the 1960s, the atomic clock had become accurate enough to offer a better definition â a second, it was decreed, would now be, “the duration of 9,192,631,770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the cesium-133 atom.” Â
In plain English, atoms vibrate at certain frequencies depending on their energy level, and by basing the second on an atomic frequency, you get a definition of the second based on something that’s always the same, for every cesium atom, no matter where, no matter when. Forget pendulum swings, forget balance wheels, forget quartz crystals: atoms are the ultimate stable oscillator (and they’re everywhere). Today, we still use this as the official international definition of a standard second.
And this is where the trouble starts. As it turns out, atomic clocks are much more stable than the Earth’s rotation around its axis, or its orbit around the Sun, and it soon became clear that while an atomic clock-based time standard (UTC) was great to have, it meant that there was going to be a cumulative difference between UTC, and observed mean solar time. While both the astronomical day, and year, are irregular, the day overall has been getting slightly longer for at least the last few centuries. To keep UTC and mean solar time in sync, a Leap Second is occasionally added to UTC. Exactly when to add a Leap Second depends on how much the Earth’s rotation is slowing (which is happening for several reasons, including drag caused by the tides) and it’s up to the good folks at theÂ International Earth Rotation and Reference Systems Service to say when it’s going to happen. If they deem a Leap Second necessary, they give notice about six months ahead of time.
So what’s that mean for GPS accuracy?
GPS works thanks to a system of satellites positioned in orbits about 20,000 km up (there are currently 32 satellites in orbit). It’s run by the U.S. military. When you use a GPS receiver, you’re receiving a signal from (at minimum) four satellites to get a fix: the signal from three satellites is used to triangulate on your position, and the signal of a fourth satellite, to provide a time correction. Where the satellites are relative to you, is determined by how long it takes a signal to travel between you and the satellites, and for the whole thing to work, the system has to use extremely precise clocks.
GPS can accurately determine position to around 30 centimeters, anywhere on Earth (barring physical barriers to radio signals, or electronic interference) but that’s only if the satellite atomic clocks, and the more precise atomic clocks on the ground that correct them, are providing accurate time. The whole system is simple in principle, but timekeeping accuracy is everything. A nanosecond (one billionth of a second) error means a position error of about a foot, which means a one second error puts you off by a billion feet: 189,394 miles, which is around 5/8 of the way to the Moon. At that level of sensitivity to clock precision, GPS has to compensate for effects described by Einstein’s theory of General Relativity â clocks moving with respect to each other, will see each other’s clocks as ticking at different rates, and clocks experiencing different forces of gravity will have the same problem.Â
Thanks to relativistic effects, to a clock on the ground, GPS satellite clocks look like they’re running 38 microseconds faster, which produces a cumulative error of 10km per day, so if you do find your mother-in-law’s house accurately with GPS, you can thank Albert Einstein â and the ultra-precise atomic clocks that keep the whole system in sync.
How does GPS handle Leap Seconds? Basically, it doesn’t â there’s no moment where the clocks on GPS satellites read 23:59:60, or a moment when the clock is frozen for one second. Instead, the GPS system transmits GPS time, while also embedding in the signal the current number of seconds difference between GPS and UTC. Your GPS receiver is responsible for doing the conversion.Â
Leap seconds don’t need to be inserted very often â since 1972, it’s happened a total of 27 times. Managing them, or rather mismanaging them, has created some major problems in the past. The 2015 Leap Second was widely publicized but it still caused issues â certain types of Internet routers turned out to be vulnerable, which caused service outages atÂ Twitter, Instagram, Pinterest, Netflix, Amazon, and Apple’s music streaming series Beats 1 (so if your ability to watch Santa Claus Conquers The Martians at 12:30 AM was impaired on January 1, 2016, now you know who to blame). And the 2016 Leap Seconds caused problems here and there as well.Â
The question naturally arises for owners of watches that use GPS receivers for accurate time: does my watch account for the Leap Seconds? For Seiko, Casio, and Citizen, the answer in all three cases is yes; the Seiko Astron, for example, will automatically look for Leap Seconds corrections to GPS time embedded in the GPS signal, on the first occasion it syncs to GPS, after June 1 and December 1 of every year.
The Leap Seconds system is therefore, somewhat controversial; because they’re inserted as a correction to an error that’s irregular (unlike the Leap Year, which we need once every four years like, well, clockwork) it’s not possible to build a Leap Seconds correction into any kind of clock, whether electronic or mechanical; the necessity for a Leap Seconds is based on comparing astronomical observations of inherently non-periodic variations in the Earth’s rotation and orbit, with atomic clocks.Â
Because the difference between UTC and non-Leap Seconds corrected time (like GPS time) is pretty minute, and because Leap Seconds can cause serious network and navigation issues, some people think we should just forget the whole idea. At the World Radiocommunication Conference in 2015 (which took place, appropriately enough, in Geneva, under the auspices of the UN) it was decided by participating nations to put off deciding whether to abolish the Leap Seconds correction until 2023. The difference between mean solar time and UTC isn’t huge, mind you â it takes about a thousand years for a one hour difference to accumulate â but if we do ditch the Leap Second someone’s going to have to cope with the offset, sooner or later.