Sunday, October 02, 2011

Neutrinos on the brain-7

OPERA and CERN were kept in sync. by using GPS in the "common view time transfer mode".   The synchronization was verified by using a clock that was moved.  Much is being made of the problem of relativistic corrections to the moving clock.  But perhaps that is not so important.

The common view time transfer mode is explained here by NIST.

An approach that builds and improves on the one-way technique is common-view time transfer. This technique allows the direct comparison of two clocks at remote locations. A generic common-view setup is illustrated in the figure. In this technique, two stations, A and B, receive a one-way signal simultaneously from a single transmitter and measure the time difference between this received signal and their own local clock. The data are then exchanged between stations A and B using any convenient method (email, FTP, etc.). {No moving clocks needed!}

The time difference between clocks A and B is calculated by taking the difference between simultaneous R - A and R - B clock difference measurements. If the travel times to the receivers are exactly equal, then the two receivers can synchronize their clocks with an accuracy that does not depend on the characteristics of the transmitter or the transmission medium. Fluctuations in the delays between the single transmitter and the two receivers also cancel exactly if they are completely correlated. This ideal situation cannot be realized in practice, but the method works well even if the two paths are not exactly equal, provided that they are nearly equal and that fluctuations in the two delays are highly correlated. Since the path delay is usually affected by various environmental parameters (such as ambient temperature), the common-view method generally works best if the distance between the receiver stations (the baseline) is small relative to the distance between either receiver and the transmitter. {The stations were about 730 km apart, the GPS satellites are 20,000km above the earth's surface.}
This geometry tends to ensure that the delay fluctuations (caused by the atmosphere for example) in the two paths will be highly correlated. One disadvantage of the common-view technique is that a means of exchanging data between the two stations must be available. {Not a problem here.}

NIST also says:

The accuracy of common-view GPS is better than for one-way since the stability of common-view is better, and also only a differential calibration is required between the two receiving stations. A differential calibration is made by carrying a portable GPS receiver between the two stations. This provides information on the difference between the delays of the receiving equipment at the two stations. An absolute calibration of the equipment delays at each station is not required. The accuracy of common-view time transfers is typically in the 1 to 10 ns range. The NIST Global Time Service provides traceable time calibrations using GPS common-view.

As far as I can tell, the "portable time-transfer device" does not actually involve a moving clock. If I understand the reference "ADVANCED GPS-BASED TIME LINK CALIBRATION WITH PTB ́S NEW GPS CALIBRATION SETUP" by Thorsten Feldmann, Andreas Bauch, Dirk Piester, Michael Rost, Elizabeth Goldberg, Stephen Mitchell, and Blair Fonville" findable on the web, the issue is as follows: 

Suppose you have two clocks A and B that keep time to a very high precision.  You have a common external time standard, the GPS signal, too. The problem in figuring out the offset between A and B, which arises because A and B are different equipment and so have different antenna, antenna cable, receiver delays, etc..  So to measure the offset between clocks A and B, you use the same device R at both places, transported carefully so as not to change its physical properties during transport.   This device R also works off the same  external standard - the GPS signal - and presumably has a common delay relative the GPS signal at both locations.  Moreover, the clock difference between this transported device R and the clocks A and B is measured using the same time-interval counter which is transported with R.   This way you have eliminated as much as you can the variations due to equipment.  If you eliminated them perfectly, you measure, then the difference between clocks A and B is simply the difference between A&R minus the difference between B&R - all the offsets due to R simply cancel.

So it is not the case that the syncing of CERN and OPERA clocks involved syncing the OPERA clock with a portable atomic clock, and then moving the atomic clock to CERN, and comparing with the CERN clock, with all the special and general relativistic effects that would affect the moving atomic clock in difficult-to-measure ways.