[time-nuts] 15 Seconds error...??

[Magnus Danielson]

On 01/18/2012 11:26 PM, Tom Van Baak wrote:
>> Use of TAI is fine, as long as you won't tell anyone near the
>> timelords, as obvious TAI is supposed to be "hands off".
>
> Magnus, Mike,
>
> The issue, I suspect, is that "TAI" still, mostly, in print and in
> people's minds, refers to the abstract paper clock, maintained
> by the BIPM, rather than some physical realization.
>
> For physical realizations we have all the UTC(k) clocks, one or
> sometimes more than one per country. The "UTC" you get from
> GPS is a slightly noisy version of UTC(USNO) which itself is a
> slightly noisy version of UTC. Depending on your technical or
> legal needs you can apply corrections to shift your native GPS
> tick to better match the UTC(k) of your choice. For example, I
> know NIST uses a GPS CV setup in such a way so that the tick
> becomes a UTC(NIST) tick rather than a UTC(USNO) tick.
>
> All this sort of phase manipulation occurs with flavors of UTC
> and we're all used to it. So it's convenient when we talk about
> "TAI" that it's just the paper clock we're talking about.
>
> On the other hand, when nanoseconds and issues of virtual
> or physical clocks don't matter, you can add 34 seconds to
> a wrist watch can call it TAI if you want. It's what I do here:
>
> http://www.leapsecond.com/java/gpsclock.htm

You are not entirely correct. There are physical replicas of TAI in a 
few labs, give or take biases from correct TAI. It's being measured and 
you find these measures in the BIPM files. So, there are a few TA(k) 
clocks out there... alongside the UTC(k) clocks.

Cheers,
Magnus