From jimr@maia.usno.navy.mil Thu Aug 27 09:37:33 EDT 1998
Received: (from jimr@localhost)
	by maia.usno.navy.mil (8.8.6 (PHNE_14041)/8.8.6) id JAA01176
	for gpst@maia; Thu, 27 Aug 1998 09:34:46 -0400 (EDT)
From: Jim Ray (USNO 202-762-1444) 
Message-Id: <199808271334.JAA01176@maia.usno.navy.mil>
Subject: Evaluation of 12 August changes at USNO
To: gpst@maia.usno.navy.mil
Date: Thu, 27 Aug 1998 9:34:46 EDT
X-Mailer: Elm [revision: 212.4]
Status: RO

EVALUATION OF CONFIGURATION CHANGES AT USNO, 12 AUGUST 1998
===========================================================

On 12 August 1998, several changes were made to the USNO IGS station
configuration which were intended to reduce its sensitivity to environmental
variations.  The AOA TurboRogue SNR12 receiver was moved to an
environmentally controlled chamber within the USNO clock vault.  More
significantly, the antenna cable (about 90 m in physical length, rather
than the 60 m previously reported) was relocated to run almost entirely
within the building rather than exposed on the rooftop.  The antenna cable
type was also changed to reduce the electrical sensitivity to temperature
variations by an order of magnitude (to ~0.02 ps/m/C).

Data collected after the station changes have been examined and appear to
verify that the changes were highly successful.  There is now no evidence
of any diurnal variation at all, within the formal error of the clock
estimates (< about 100 ps).

Data from 18 August 1998
------------------------
As a laboratory exercise on the last day of the GIPSY class last week,
Lara Schmidt, Merri Sue Carter, Lisa Nelson (NIST), and I wrote a shell
script intended to estimate clocks (and other things) for a small GPS
network using fixed IGS orbits and fixing phase cycle ambiguities.  There
were a number of problems encountered along the way but basically it
seemed to work reasonably well thanks to much help from Ken Hurst (JPL).

For the test solution, we used data from ALGO (Algonquin Park, Ontario,
Canada), AMCT (Colorado Springs, CO, USA), GODE (Goddard Optical Facility,
Greenbelt, MD, USA), NLIB (North Liberty VLBA site, Iowa, USA), and NRC1
(Ottawa, Ontario, Canada) in addition to USNO (connected to USNO Master
Clock 3, Washington, DC, USA) which was used as the clock reference.  The
test day was Tuesday, 18 August 1998.  All of these stations use 5 MHz
external frequency inputs from H-masers.

Plots of the clock differences (wrt USNO), in postscript format, can be
viewed at: http://maia.usno.navy.mil/gpsclocks/18aug98/ .

Unfortunately, AMCT had data for less than half the day.  ALGO and NRC1
appear to be the most useful for assessment of clock performances.  Each
has a very nearly linear trend of roughly +/- 2 ns/d.  Residuals from a
linear fit are also available at the Web site for ALGO, GODE, and NRC1.
Even though the (GODE - USNO) differences looks rather linear, the GODE
residuals are considerably larger and more systematic than for ALGO and
NRC1, probably due to the less stable NR maser there (which incidentally
failed a few days later).

Assuming that the ALGO, NRC1, and USNO "clocks" are independent, then their
highly linear differences -- free of any diurnal trends to within <100 ps --
are strong evidence for the effectiveness of the 12 August changes at USNO.
The peak-to-peak temperature range was about 21 to 32 C.  A plot of the
outside temperature variation at USNO for GPS week 0971 (16-22 August) is
available at: http://maia.usno.navy.mil/gpsclocks/0971/USNO-temp.971.ps .

The RMS of the (ALGO - USNO) residuals is 41.5 ps while the (NRC1 - USNO)
RMS is 49.0 ps (with 286 epochs for each).  The formal error of each
differential clock estimate is ~112 ps.

Comparison of GPS weeks 0969 and 0971
-------------------------------------
It could be argued that diurnal variations at ALGO, NRC1, and USNO should
be nearly in phase with one another (due to the sites being at nearly the
same longitude) and therefore would be removed in these clock differences.
However, that is clearly not the case when the USNO Rapid clock estimates
for GPS week 0969 (02-08 August, the week before the USNO configuration
changes) are compared with those for GPS week 0971 (16-22 August, the 
week after the changes).  These results can be viewed at
http://maia.usno.navy.mil/gpsclocks/ for all the stations used in the
daily USNO Rapid analyses.  Most revealing is the comparison of the ALGO
residuals:

http://maia.usno.navy.mil/gpsclocks/0969/ALGO.969.res.ps  
http://maia.usno.navy.mil/gpsclocks/0971/ALGO.971.res.ps  

Diurnal variations are very evident in the former but undetectable in
the latter.


Paul Wheeler and his colleagues in the USNO Time Service Dept. are to be
congratulated for planning and implementing the USNO changes.

=========================================================================
Jim Ray                                   e-mail: jimr@maia.usno.navy.mil
U.S. Naval Observatory, EO Dept.          phone : 202-762-1444
3450 Massachusetts Avenue, NW             fax   : 202-762-1563
Washington, DC  20392-5420
=========================================================================