From jimr@maia.usno.navy.mil Fri Mar  5 15:54:38 EST 1999
Received: (from jimr@localhost)
	by maia.usno.navy.mil (8.8.6 (PHNE_14041)/8.8.6) id PAA23468
	for gpst@maia; Fri, 5 Mar 1999 15:52:29 -0500 (EST)
From: Jim Ray (USNO 202-762-1444) 
Message-Id: <199903052052.PAA23468@maia.usno.navy.mil>
Subject: Colocated common view calibration method
To: gpst@maia.usno.navy.mil
Date: Fri, 05 Mar 1999 15:52:29 EST
X-Mailer: Elm [revision: 212.4]
Status: RO

CALIBRATION BY COLOCATED COMMON VIEW OF SATELLITE CLOCKS
========================================================
   J.R. Ray   (U.S. Naval Observatory)


SUMMARY
-------
Calibration of the timing delays in a geodetic GPS receiver system
should be feasible relative to a GPS timing receiver system which 
has been calibrated "absolutely" by comparing satellite clock values
observed simultaneously.  The two receiver systems must be located in
close proximity and all cable delays must be accurately known.  If
care is taken to apply equivalent analysis methods to both data sets,
it should be possible to determine the end-to-end calibration delay
through the geodetic system to the 1-ns level or better.

CCV CALIBRATION DESIGN
----------------------
A schematic for calibration by the colocated common view (CCV) method
is shown below.  A geodetic GPS receiver/antenna system must be
installed close to an operational timing receiver.  In order to transfer
"absolute" calibration to the geodetic system, it is necessary that the
absolute delays through the timing system be accurately known (or
accepted).  A common frequency standard should be used for both
systems.  In order to preserve the geodetic calibration through
subsequent power cyclings and internal clock resets, it is also necessary
for receivers such as the AOA TurboRogue to monitor the 1 pps output
continuously.

                          U-O-U
                           . .
                          .   .
                         .     .
                        .       .
                       .         .
                      .           .
                     .             .
                    .               .
                   .                 .
                  .                   .
                 .                     .
                .                       .
               .                         .
              A                           A
             ===                         ===
              |                           |
              | __________                | __________
              |_|  Rx-G  |  1pps          |_|  Rx-T  |
   Geodetic     |        |-----             |        |    Timing
   system       ----------    |             ----------    system
                     ^        |                  ^
                     |        -->(counter)       |
                     |              ^            |
                     |  5 MHz       |     5 MHz  |
                     -----------(clock)-----------


Given the large timing errors due to selective availability (S/A),
it is essential that the two receivers be synchronized to make
observations of the same satellites at common epochs.  This can be
done using an approach similar to the BIPM common view schedules
for single channel timing receivers.  Multi-channel timing receivers
are obviously preferable.  It should be noted that geodetic receivers
collect and time tag observations according to broadcast GPS time.

ANALYSIS CONSIDERATIONS
-----------------------
Data from each receiver system can be analyzed to determine estimates
for the observed satellite clocks relative to the receiver clocks.
That is, the two estimated clock differences at a specific epoch can
be represented as

    Obs_G = Clk_sat - (Clk_ref + cable_G + calib_G)

for the geodetic system, and as

    Obs_T = Clk_sat - (Clk_ref + cable_T + calib_T)

for the timing system.  The notation is

    Obs_G = estimated satellite clock difference from geodetic system
    Obs_T = estimated satellite clock difference from timing system
    Clk_sat = true satellite clock reading
    Clk_ref = true station clock reading
    cable_G = cable delay for 5 MHz input to geodetic receiver
    cable_T = cable delay for 5 MHz input to timing receiver
    calib_G = calibration bias for geodetic system
    calib_T = calibration bias for timing system

It is assumed that the timing receiver calibration bias has been
determined independently (presumably to the few-ns level).  The cable
delays must be accurately measured.  Then the calibration bias of the
geodetic system can be determined by

    calib_G = (Obs_G - Obs_T) - (cable_G - cable_T) - calib_T

This calibration quantity accounts for the entire delay of the geodetic
system from the effective phase center of the antenna to the point 
within the receiver where the data are time-tagged.

While this approach is straightforward and seemingly simple, considerable
care must be taken to ensure that the observed satellite clock differences
are comparable from the geodetic and timing systems.  Otherwise, the
inferred calibration bias will be corrupted.  The most important
components that must be considered are:

1) antenna locations -- The geocentric coordinates must be in the same
high-accuracy terrestrial reference frame and must account for the same
deformational effects (e.g., solid Earth tides).  Few cm (~0.1 ns) errors
in antenna locations are readily attainable but much larger errors are
common; time-varying tidal effects can be as large as ~0.5 m (~1.5 ns).

2) satellite orbits -- The same ephemerides must be used for both
analyses or corrections must be applied for the effect of different
ephemerides.  Broadcast and IGS Final orbits differ at the few-meter
level (5 ns or greater).

3) ionospheric propagation -- Normally, observed L1-L2 phase differences
are used to correct for ionospheric delays in geodetic analyses.  Delay
differences with ionosphere models can be 10 ns or larger, sometimes
much larger.

4) tropospheric propagation -- The total delay due to the neutral
atmosphere is about 7 ns at the zenith and about three times larger
at 20 degrees elevation.

Assuming that the geodetic and timing analysis procedures are properly
matched, the estimated satellite clocks (Obs_G and Obs_T) should be
determined with a precision of roughly 0.2 ns.  It is likely, therefore,
that the calibration error will be dominated by uncertainties in the
cable delays (cable_G - cable_T) and the timing receiver calibration
(calib_T).

DISCUSSION
----------
Instrumental methods for delay calibration are also possible and
desirable.  Only in this way can the individual contributors to the
delay be identified and characterized.  The CCV analysis method, on
the other hand, is independent (except for cable delays) and end-to-end.
Thus, it is necessary to verify instrumental methods.  It is also
possible that the CCV method provides a more accurate determination, at
least relative to an accepted timing receiver.  However, it can only be
applied at facilities equipped with calibrated timing receivers.