On 8 Dec, 2011, at 11:43 , Zefram wrote:
> Steve Allen wrote:
>> http://www.ucolick.org/~sla/leapsecs/draftTF460-7.html
> 
> "TAI is not physically realized and consequently is not suitable for
> time dissemination." -- what on Earth is that supposed to mean?

It's a technical argument.  I think a physical realization of UTC is UTC(k),
so if you want a real-time estimate of UTC you'll be getting it by tracking
UTC(k), for some value of `k'.  Here's how some of those UTC(k)'s did in the
November BIPM Circular T:

   1 - Coordinated Universal Time UTC and its local realizations UTC(k). 
Computed values of [UTC-UTC(k)]
       and uncertainties valid for the period of this Circular.
       From 2009 January 1, 0h UTC, TAI-UTC = 34 s.

   Date 2011    0h UTC         SEP 30   OCT  5   OCT 10   OCT 15   OCT 20   OCT 
25   OCT 30    Uncertainty/ns Notes
          MJD                   55834    55839    55844    55849    55854    
55859    55864     uA    uB    u
   Laboratory k                                         [UTC-UTC(k)]/ns

   CH   (Bern)                   -0.5     -1.2     -2.0     -3.6     -4.5     
-6.0     -8.0    0.3   1.7   1.8
   IT   (Torino)                 -0.8     -3.1     -3.1     -3.6     -5.0     
-5.9     -6.8    0.7   1.9   2.0
   JATC (Lintong)                17.4      7.6     -0.2     -0.8     -7.9     
-9.5    -14.3    1.4   5.0   5.2
   KRIS (Daejeon)                 7.2     13.9     10.2      7.2      8.1      
6.9      6.1    0.3   5.1   5.1
   NICT (Tokyo)                  19.0     18.5     18.5     19.2     21.1     
22.7     21.8    0.3   4.7   4.7
   NIST (Boulder)                10.2     10.8     11.2      9.6      8.9      
8.5      7.4    0.3   5.0   5.0
   NRC  (Ottawa)                -19.9    -22.5    -28.0    -33.1    -29.0    
-29.0    -28.2    0.3   5.2   5.2
   NTSC (Lintong)                 1.4     -4.6     -4.8      0.2      4.5      
6.6      7.6    1.4   4.9   5.1
   ONRJ (Rio de Janeiro)          6.2      0.2     14.2     21.1     10.3      
6.9      9.3    3.9  19.6  20.0
   PL   (Warszawa)               18.5     15.2     10.1     -2.7    -15.6    
-19.1    -23.2    1.5   5.1   5.3
   PTB  (Braunschweig)            4.0      4.6      5.1      3.3      2.8      
2.2      1.5    0.2   1.4   1.4
   SG   (Singapore)              16.6     19.3     24.5     26.2     29.4     
27.8     28.7    0.3   5.1   5.1
   SU   (Moskva)                  7.0      9.0      9.6     10.6      9.0     
10.4      8.1    1.2   5.2   5.3
   TL   (Chung-Li)               28.6     29.4     29.6     27.5     28.5     
27.9     26.9    0.3   4.9   4.9
   USNO (Washington DC)           6.2      6.2      8.5      7.6      6.7      
5.8      4.9    0.3   4.1   4.1

So any of those UTC(k)'s would be pretty good.  The worst are off by a few 10's 
of
nanoseconds, the best by 10 ns or so.

Analogously, the physical realization of TAI would be TA(k), so I guess a 
real-time
estimate of TAI would be some lab's TA(k).  Here's how you would have done 
using the
above labs' TA(k)'s in the same month:

  2 - International Atomic Time TAI and Local atomic time scales TA(k). 
Computed values of [TAI-TA(k)].

   Date 2011    0h UTC       SEP 30      OCT  5      OCT 10      OCT 15      
OCT 20      OCT 25      OCT 30
          MJD                 55834       55839       55844       55849       
55854       55859       55864
   Laboratory k                                             [TAI-TA(k)]/ns

   CH   (Bern)              40261.2     40192.7     40125.0     40056.7     
39988.1     39919.7     39851.2
   F    (Paris)            167705.4    167704.2    167703.1    167697.6    
167696.9    167695.8    167693.8
   IT   (Torino)           106735.2    106835.3    104974.1    103876.8    
102400.3    101043.6     99654.0
   JATC (Lintong)          -50417.8    -50441.2    -50469.5    -50490.8    
-50519.3    -50544.8    -50570.2
   KRIS (Daejeon)           38299.3     38368.4     38420.0     38473.4     
38532.5     38587.7     38643.3
   NICT (Tokyo)              -126.7      -124.2      -118.4      -114.8      
-107.3      -101.9       -96.7
   NIST (Boulder)       -45367935.4 -45368124.8 -45368314.4 -45368505.8 
-45368695.5 -45368884.9 -45369075.0
   NRC  (Ottawa)            26053.9     26030.2     26003.5     25977.4     
25960.6     25939.5     25919.2
   NTSC (Lintong)           12928.3     12957.5     12982.3     13014.0     
13038.9     13066.6     13094.0
   ONRJ (Rio de Janeiro)    -9973.7    -10013.8    -10038.7    -10059.4    
-10087.4    -10107.9    -10135.7
   PL   (Warszawa)          -9392.6     -9414.5     -9440.9     -9471.9     
-9496.6     -9519.0     -9547.6
   PTB  (Braunschweig)    -328739.5   -328727.5   -328714.3   -328698.5   
-328685.7   -328681.1   -328673.1
   SG   (Singapore)          8084.6      8108.3      8134.5      8158.2      
8185.4      8207.8      8232.7
   SU   (Moskva)         27277354.3  27277501.9  27277648.1  27277795.6  
27277940.4  27278088.3  27278233.7   (1)
   TL   (Chung-Li)           -473.7      -476.2      -478.7      -480.1      
-480.6      -484.9      -485.8
   USNO (Washington DC) -35094388.5 -35094680.6 -35094972.9 -35095267.2 
-35095560.1 -35095852.7 -35096145.4
   - Note on section 2:
   (1) SU  : Listed values are TAI-TA(SU) - 2.80 seconds.

So instead of 10's of nanoseconds with UTC(k), with TA(k) you get microseconds 
if
you're lucky and milliseconds if you aren't (and seconds with TA(SU)).

The reason the values of UTC(k) are so close together is that they are sometimes
steered to make that so.  UTC(k)'s are the times which are disseminated, so they
try to keep them all as close to each other as possible.  The value of an 
individual
TA(k), on the other hand, is never steered; they don't really give a crap how 
far off
individual TA(k)'s are from each other since they don't disseminate those 
values,
so the error accumulated in each TA(k) is the frequency error of that lab's 
clock
integrated over the entire time that lab has had a clock.

UTC, in the real-time form of UTC(k), is maintained for dissemination.  TAI, in 
the
real-time form of TA(k), is not.  If you want realtime TAI you are way better 
off getting
a UTC(k) and adding 34 seconds.

Dennis Ferguson
_______________________________________________
LEAPSECS mailing list
[email protected]
http://six.pairlist.net/mailman/listinfo/leapsecs

Reply via email to