Re: Ruminations about time broadcast data formats

From: Markus Kuhn <>
Date: Mon, 07 Jul 2003 14:07:10 +0100

Steve Allen wrote on 2003-07-06 07:14 UTC:
> My scenario requires that the broadcast signals
> provide offset ( TI - UT ) as part of their format.
> This leaves open the question of at what accuracy level?
> I.e., what might happen with leaps of varying sizes?

As I said at Torino: If we start to redesign time signal formats, then
please get some computer scientists and communications people on board
and do it properly. The existing radio time-broadcast formats (MSF,
DCF77, WWVB, etc.) are extremely primitive constructions and represent
very inefficient 1950s hacks to piggyback semi human-audible data onto
existing standard-frequency transmissions. In a modern format, you would
have significantly more bandwidth and you would be able to add full
earth-orientation parameters (perhaps with several elements of a Taylor
approximation) as published in IERS bulletins to the data stream. This
discussion about how many bits to use to represent one particular
parameters is leading nowhere. In a modern narrow-band longwave
transmitter, you would drop the role of the carries as a
standard-frequency source and use ~200 Hz bandwidth QPSK to send a few
hundred bits per second in TDMA, such that time receivers can pick up TI
within a few seconds of activation. You would also use the modern format
specification tools developped, for instance, for digital TV (MPEG
data-type syntax) to specify the details of bitstream in compact
variable-width notations, not any crude BCD, pulse-counting, or
fixed-width ones-complement encodings. Parity bits would be replaced
with streaming message authentication codes, etc.

If you really want UT, then you usually want to have it with the best
possible accuracy and prediction available of course, and there is no
reason not to provide it.

Whether after a change to an atimic reference time (TI), it is actually
necessary to keep anything like UTC, which is coordinated to be
unambiguously predictable over some time interval and at the same time
tracks UT with some specified precision (Dennis McCarthy called this
UT1C in his talk), is doubtful.

You want to have:

  - one time to label events, and that would then be TI

  - the orientation of the earth to point your satellite dish
    and calculate sunrise, and that you extract this from the best available
    earth-orientation parameter Taylor approximations

I don't see a need for anything in between, and neither could anyone at
the Torino meeting when Dennis McCarthy brought up the subject of
whether there should be a UT1C.


Markus Kuhn, Computer Lab, Univ of Cambridge, GB | __oo_O..O_oo__
Received on Mon Jul 07 2003 - 06:07:28 PDT

This archive was generated by hypermail 2.3.0 : Sat Sep 04 2010 - 09:44:54 PDT