[time-nuts] Temps Atomique International

Magnus Danielson magnus at rubidium.dyndns.org
Thu Oct 22 20:14:41 UTC 2009

Dear Martyn,

Martyn Smith wrote:
> Hello,
> I have a question about TAI (Temps Atomique International) time.
> As I understand it TAI is UTC time without the leap seconds.
> Is that correct?

Not really, TAI is calculated out of EAL and UTC is calculated out of
UTC. EAL is calculated out of comparative measures between UTC
laboratories. That's the short story.

> So is that the same as GPS time?

No, GPS time is offset by 19 seconds from TAI, to match UTC at the GPS
epoch at 1980-1-6T00:00:00Z. GPS time is maintained such that it is
steered towards UTC(USNO) modulus 1 s. The effective differens is a few
ns. UTC(USNO) is steered towards UTC.

> I need to be able to generate accurate TAI time as opposed to UTC time.

Essentially the same problem as producing UTC, as you still want the
long term corrections between GPS time and TAI/UTC time, which is
announced in the same message as the leapsecond message. However, GPS
time could act as a replacement, even if it technically isn't directly TAI.

Then people say you can't get (the) TAI, but can't get (the) UTC or
(the) GPS-time, but you can get approximations of them all.

> I'm a bit confused as to what the difference between the two is.

Confusion is the name of the game.


More information about the time-nuts mailing list