From jimr@Maia.usno.navy.mil Wed May 20 15:26 EDT 1998
Received: by Maia.usno.navy.mil
(1.37.109.24/16.2) id AA197782354; Wed, 20 May 1998 15:25:54 -0400
From: Jim Ray (USNO 202-762-1444)
Return-Path:
Subject: Test of RINEX Option to Remove a Receiver Clock Offset
To: gpst@Maia.usno.navy.mil
Date: Wed, 20 May 1998 15:25:54 EDT
Full-Name: Jim Ray (USNO 202-762-1444)
X-Mailer: Elm [revision: 109.19]
Status: RO
TEST OF THE RINEX OPTION TO REMOVE A RECEIVER CLOCK OFFSET
==========================================================
M. Chin (National Geodetic Survey, NOAA)
W. Gurtner (Astronomical Institute, University of Berne)
Y. Mireault (Geodetic Survey Division, Natural Resources Canada)
J. Ray (U.S. Naval Observatory)
J. Rohde (U.S. Naval Observatory)
J. Zumberge (Jet Propulsion Laboratory)
SUMMARY
-------
W. Gurtner (AIUB) recently added a new option to the RGRINEXO program to
apply a user-specified overall apriori clock offset correction to RINEX
observation data. The option has been tested with a day of USNO data and
found to function as expected.
DISCUSSION OF RESULTS
---------------------
The new RGRINEXO option can be used to produce RINEX data for a station
with an overall clock offset removed. This option is only useful for those
receivers driven from a stable frequency standard where the clock offset is
well known apriori (at the nanosecond level). In such a case, the RINEX
observational data can be corrected to remove the offset as a convenience
for data users.
While this test demonstrates the validity of the clock correction option,
it remains to be determined whether it is wise to exercise this option
routinely. It can be argued that observational data should, as a matter
of principle, never be modified (although the correction value is reported
in the RINEX file which allows the orginal data to be recovered). Also,
the correction is limited to integral nanosecond values by the RINEX file
format specifications.
BACKGROUND
----------
Compared with GPS time or with some external timing standard, the internal
clock of any particular GPS receiver will have an arbitrary offset value.
In the case of TurboRogue receivers, the clock offset must not exceed 30 ms
compared with GPS time. If the offset is greater, the receiver will stop
tracking and reset its clock reference value to within 50 ns of GPS time.
If connected to a stable external frequency standard, the offset may remain
relatively constant for long periods of time. Normally, the RGRINEXO
program which converts TurboRogue binary data files to RINEX format stores
a single estimate of the receiver clock offset as a comment in the header
records. However, since this offset is estimated from the pseudorange
data (currently under SA conditions) its precision is limited to ~50 ns.
The RINEX specifications allow for the possibility of including a receiver-
derived clock offset correction. When invoked, the reported observables
should maintain their consistency by obeying the following relations:
Time(corr) = Time(r) - dT(r)
PR(corr) = PR(r) - dT(r)*c
phase(corr) = phase(r) - dT(r)*freq
where dT(r) is the clock offset. In this case, the clock offset correction
value should be reported on each observation epoch record according to the
format shown below so that it is possible to reconstruct the original
observations if necessary. As the output format for the receiver-derived
clock offset is limited to nanoseconds the offset should be rounded to
the nearest nanosecond before it is used to correct the observables in
order to guarantee correct reconstruction.
+-------------+-------------------------------------------------+------------+
| OBS. RECORD | DESCRIPTION | FORMAT |
+-------------+-------------------------------------------------+------------+
| EPOCH/SAT | - Epoch : | 5I3,F11.7, |
| or | year (2 digits), month,day,hour,min,sec | |
| EVENT FLAG | - Epoch flag 0: OK | I3, |
| | 1: power failure between | |
| | previous and current epoch | |
| | >1: Event flag | |
| | - Number of satellites in current epoch | I3, |
| | - List of PRNs (sat.numbers with system | 12(A1,I2), |
| | identifier, see 5.1) in current epoch | |
| | - receiver clock offset (seconds, optional) | F12.9 |
As an example, using a clock offset value of +234 ns will produce observation
records such as:
98 5 4 0 0 .0000000 0 8 13 06 23 17 26 24 27 10 .000000234
22414009.026 -8826591.467 9 -6877858.53745 22414008.2504
20906149.135 -18429992.301 9 -14361025.07446 20906150.6074
25036081.471 -2019865.994 6 -1573915.34443 25036085.2684
23114975.889 -4859986.034 7 -3786998.57145 23114979.7274
20930090.730 -18368012.814 9 -14312723.14246 20930092.5664
22504656.499 -13086408.400 8 -10197192.09545 22504659.9664
24770387.542 -3662656.331 6 -2854006.45944 24770390.5954
21514194.039 -17069325.299 9 -13300761.58845 21514195.9444
RGRINEXO OPTION
---------------
Until recently, the RGRINEXO program could apply only receiver-determined
clock offsets (or none at all). In January 1998, W. Gurtner (AIUB)
modified the program to permit a user-specified offset to be used instead.
To use this option, a new RGRINEXO.OPT file is needed with one line
containing the user-given clock offset (a zero for those stations that do
not have one) in all station-dependent input files that run with the same
option file. The program version should be V. 2.5.0 or later.
Thus, for the case of USNO, example input files are shown below.
Input File : RGRINEXO.OPT
APPLY/STORE REC.CLOCK OFFSET --> : 2 1: FROM RAW FILE, 2: USER-ENTERED
Input File :USNOO.HDR
USNO
40451S003
GL
GL
ROGUE SNR-12
T246
3.32.1.4
DORNE MARGOLIN T
309
0.0
0.0
0.000000234 (clock offset correction entered here)
usno1230.98o
y
TEST WITH USNO DATA
-------------------
The USNO IGS receiver uses an external 5 MHz reference signal from an
H-maser steered to UTC(USNO). Likewise, GPS time itself is steered to
UTC(USNO) by direct monitoring of the clocks of the GPS constellation at
USNO. While the two timescales have been kept within about 6.5 ns
(modulo 1 s) over the last two years (for 24-hour averages), the GPS time
steering algorithm has a "bang-bang" character that results in a saw-
tooth variation with a typical cycle of ~25 days. Even allowing for
thermal variations and other local instrumental effects, it is likely that
the clock of the USNO receiver is more stable than GPS time. There is,
however, an arbitrary clock offset for the USNO data that must be accounted
for. Specifically, in 42 days of results from March-May 1998 where USNO
was used as the reference clock station in solutions included in the IGS
Rapid combination, the satellite clocks were found to have an average offset
of 229.1 ns with respect to GPS time and a WRMS of 4.7 ns.
In principle, the USNO clock offset can be removed from the observational
data using the RGRINEXO option, at least to the nanosecond level. This
could be a convenience for users of USNO data seeking a clock reference
within a few nanoseconds of GPS time requiring no special user effort.
To test the RGRINEXO user option, USNO data were chosen for 04 May 1998 (1998
doy 124 = GPS wk 0956 day 1 = MJD 50937). The standard IGS Rapid report for
that day gave:
CENT STA| DX DY DZ RX RY RZ SCL RMS WRMS | TOFT TDRFT RMS
--------|------------------------------------------------|---------------------
usn 30 |-.02 .00 .01 .04 .04 -.07 .3 .09 .09 | 234.8 -1.2 .2
which indicates that the GPS satellite clocks (relative to the USNO receiver)
are offset from GPS time by 234.8 ns. In the test, therefore, the USNO data
from that day (and also 03 May since the USN analysis strategy uses a 27-hour
data arc) were corrected with a clock offset of 0.000000234 s (+234 ns). It
was expected, then, that the test solution should have an offset relative to
the IGS Rapid combination (TOFT) of 0.8 ns. The test RINEX data file(s) were
produced by M. Chin at NOAA.
RESULTS USING JPL'S PRECISE POINT POSITIONING TECHNIQUE
-------------------------------------------------------
J. Zumberge (JPL) analyzed both the standard and test RINEX data from USNO using
the precise point positioning (PPP) technique. He found the following results:
(1) a shift of -234.001 ns in the receiver clock estimate, with a standard
deviation of about 1.2 ps over the day;
(2) with respect to GPS time as derived from the broadcast ephemeris, the
average USNO clock offset estimate is 0.880 ns;
(3) less than 0.1-mm shift in estimates of horizontal coordinates and a shift
of -0.1 mm in the vertical for the USNO station position;
(4) same number of phase breaks;
(5) one additional phase outlier with the test data set.
Hence, this PPP test appears to validate the clock offset removal procedure.
RESULTS OF DUPLICATE IGS SOLUTION
---------------------------------
J. Rohde (USNO) reran the USN solution for 04 May 1998 (together with the last
3 hours of 03 May) exactly as in the standard case except replacing the USNO
data with the clock-corrected files. The satellite ephemerides and Earth
rotation files were sent to Y. Mireault (NRCanada) who repeated the standard
IGS Rapid combination for that day. All previous Analysis Center solutions
were the same. The new USN solution was labelled as "USO"; it was included
for comparison purposes only and was excluded from the actual orbit/clock
combination (similar to "brd"). The results from Mireault are shown in Tables
1 and 2 below. It can be seen that the USN and USO solutions are nearly
identical except for the removal of a clock offset of 234 ns from the USO
test solution.
Table 1: Daily transformation of each centre to the combined IGS orbit and
clocks.
WRMS - RMS weighted by the centres SP3 header accuracy codes.
STA - Number of stations in the daily solutions.
Units: meters, mas, ppb, nano-sec, nano-sec/day.
Table 1 GPS week: 0956 Day: 1 MJD: 50937.0
CENT STA| DX DY DZ RX RY RZ SCL RMS WRMS | TOFT TDRFT RMS
--------|------------------------------------------------|---------------------
brd n/a| .12 .11 .34 1.12 1.06 8.83 -2.6 3.81 3.81 | -1.3 -3.0 83.0
cod 94 | .01 .02 -.02 -.09 -.25 .01 -.5 .08 .06 | -1.0 -5.4 83.2
emr 24 | .01 .01 .01 .40 -.33 .48 -.3 .11 .11 | 615.4 12.2 .3
esa 40 | .00 .01 .00 -.22 .07 -.10 .2 .19 .15 | -48.8 -101.9 18.0
gfz 28 | .00 -.03 .00 .04 .12 -.10 .1 .07 .07 | .9 9.9 .6
jpl 21 | .01 .02 .01 -.20 .07 .04 .3 .08 .08 | -3.3 5.8 .9
ngs 55 | .00 -.04 -.01 .30 .07 -.32 .2 .15 .15 | -5.8 2.6 90.7
sio 51 | .01 .01 -.02 .04 .20 3.58 .4 .17 .13 | .0 .0 .0
usn 30 |-.02 .00 .01 .04 .04 -.07 .3 .09 .09 | 234.8 -1.2 .2
uso 30 |-.02 .00 .01 .05 .04 -.07 .2 .09 .09 | .8 -1.2 .2
Table 2: Daily fit for each satellite and each centre resulting from weighted
average combination. (GPS week: 0956 Day: 1 MJD: 50937.0)
The last column gives the IGS resulting satellite accuracy measures
found in the IGS SP3 header. Eclipsing satellites are flagged with
and "E" beside the PRN number.
"MEDI" stands for median of the centre's satellite RMS.
Units: cm.
Weighted Average
PRN | brd cod emr esa gfz jpl ngs sio usn uso IGS |
------|----------------------------------------------|
1 | 216 6 8 8 7 8 11 8 8 8 3 |
2 | 321 6 12 24 5 8 21 11 9 9 5 |
3 | 106 6 7 10 5 8 12 8 7 7 3 |
4E | 223 11 10 14 9 13 18 16 12 12 5 |
5 | 174 7 9 9 4 5 13 9 6 6 3 |
6 | 260 7 9 4 6 7 8 11 7 7 3 |
7 | >>> 4 6 18 5 5 13 8 7 7 3 |
8 | 383 6 11 7 7 8 12 9 5 6 3 |
9 | 119 8 7 10 5 4 8 8 10 10 3 |
10 | 202 5 7 12 7 6 15 19 7 7 3 |
13 | 133 9 9 15 14 10 10 28 12 12 5 |
14 | 255 8 12 25 11 8 9 9 7 7 4 |
15E | 337 7 10 26 7 8 21 8 7 7 5 |
16 | >>> 13 25 54 8 12 8 55 9 9 7 |
17E | 198 6 16 18 6 11 16 12 11 11 5 |
18 | 172 5 7 34 5 8 13 12 10 10 5 |
19 | 558 7 9 9 7 6 20 11 5 5 4 |
21 | 102 8 18 18 5 8 14 11 8 8 4 |
22 | 112 5 11 8 7 7 8 9 11 11 3 |
23 | 209 15 14 36 11 14 26 39 16 16 7 |
24E | 483 12 7 14 5 4 20 16 7 6 4 |
25 | 177 5 9 9 5 6 8 6 8 8 3 |
26 | 196 7 10 11 8 5 24 8 8 8 4 |
27 | 312 8 11 10 5 8 11 16 7 7 3 |
29 | 165 8 6 10 5 9 13 12 5 5 3 |
30 | 151 7 11 8 8 6 12 9 7 7 3 |
31 | 255 6 8 6 8 6 12 6 6 6 3 |
------|----------------------------------------------|
RMS | 381 8 11 19 7 8 15 17 9 9 |
WRMS | 381 6 11 15 7 8 15 13 9 9 |
MEDI | 209 7 9 11 7 8 13 11 7 7 |