[LEAPSECS] preprint about timekeeping for neutrino experiment
zefram at fysh.org
Mon Oct 3 20:28:03 EDT 2011
Rob Seaman wrote:
>An interesting discussion of a difficult measurement:
I've found a description of the Time-Transfer Device that is the
subject of that paper. The original OPERA paper doesn't actually
say that a TTD was used for synchronisation, it says that the
synchronisation was "independently verified", by PTB, using a TTD.
<http://operaweb.lngs.infn.it/Opera/publicnotes/note134.pdf> is a note
from PTB describing their part in the affair, and it's a very limited
part. Their job was only to compare corresponding parts of the timing
gear at the two labs, checking the delay between GPS signals and the
resulting PPS signal at the input to a timestamping unit.
PTB's note explains why they used a portable device, and it's got nothing
to do with actual time transfer. There's a problem that units of the same
species have individual variations, so measurements made with different
units are not directly comparable at the finest precision. So they
transfer one unit between the labs in order to perform corresponding
measurements with the same unit, so that the unknown biases in the unit
cancel themselves out. The unit does not maintain its own time scale, so
the path it took between the labs is irrelevant. Contaldi's assumption
that "this device [is] a transportable atomic clock" turns out to be
wide of the mark. (He redeems himself by the footnote deploring the
lack of sources on this point.)
The synchronisation may still have gone awry in other areas not covered by
PTB's work, which are not described by anything I've seen yet. There's
also a clear error in that the OPERA paper treats time as Newtonian:
the two labs are synched to GPS time, hence to TAI, and there's no
discussion of the difference between this time scale (SI seconds on the
geoid) and time along the neutrino path (varying between 1 km above and
30 km below the geoid).
<http://operaweb.lngs.infn.it/Opera/publicnotes/note132.pdf> discusses the
geodesy in reasonable detail. Down at the bottom of it, the endpoints
are ascribed coordinates in ETRF2000, and Pythagoras's theorem is used
to determine the path length. At the precision stated, gravitational
length contraction must make this calculation invalid.
Do the gravitational time dilation and the gravitational length
contraction cancel each other out, when viewed from a suitable reference
frame? I'm out of my depth here.
More information about the LEAPSECS