IGS Electronic Mail 17 Aug 04:28:19 PDT 2005 Message Number 5189
Author: Gerd Gendt and Ralf Schmid
This message is to alert users of pending changes that will profoundly affect
IGS products. The changes concern the conventions for handling phase
corrections due to antenna effects. The related activities go back to the 2002
IGS Workshop in Ottawa [Rothacher and Mader, 2003] and were put into concrete
recommendations during the Workshop in Berne last year [Schmid et al., 2005a].
Even data analysts who do not use IGS products directly should be aware of
these developments and should consider what changes, if any, are appropriate in
their own procedures.
Early in the era of GPS geodesy it was appreciated that the phase response of
GPS tracking antennas does not correspond to an ideal point source. That is,
with respect to a fixed reference point within the antenna, the signal
wavefronts of constant phase do not form perfectly hemispherical shells.
Instead, the phase can vary depending on viewing direction (elevation and
azimuth angles toward the satellite being observed). Since most high-quality
antennas are manufactured to tight physical specifications, the variations
between different models are generally much larger than among different
replicas of the same model. For the highest quality of geodetic results, it is
therefore necessary to account for these antenna effects in the data analysis
based on antenna type. Otherwise errors of up to several cm can occur in
estimates of the antenna position.
Relative antenna corrections
Despite this realization, the application of antenna phase center corrections
measured in anechoic chambers [e.g., Schupler et al., 1994] was found to give a
global GPS frame differing from VLBI and SLR in scale by about 15 ppb,
equivalent to a global height shift of ~10 cm. (Note that a +1 ppb change of
the terrestrial scale corresponds to a uniform height shift of +6.4 mm.) This
was considered an unreasonably large discrepancy, considering the good
agreement between VLBI and SLR, so the chamber measurements were not adopted
for general use.
Nevertheless as new antenna types joined the IGS network, which was initially
dominated by the AOA Dorne-Margolin choke ring models, it became necessary to
somehow account for differences in the phase responses. After extensive
discussions, it was agreed at the IGS Analysis Center Workshop held at Silver
Spring (19-21 March 1996) that relative phase center corrections would be
applied by 30 June 1996. The corrections were to be determined using
short-baseline field measurements wherein the "AOAD/M_T" antenna was taken to
be the defining standard [Mader, 1999]. For each antenna model, a NEU offset
value was adopted for the mean location of the antenna electrical reference
center compared to the external, physical, antenna reference point (ARP).
Relative to this mean phase center, the antenna phase center variations (PCVs)
were then measured as a function of elevation angle. Using this method the
elevation range for PCVs has been limited to 10 degrees due to ground noise.
The results have been maintained since that time for IGS and general use in the
As new antenna models have become available, new calibration measurements have
been added to the file before the antenna could be included in the IGS network.
It was generally recognized that the approach of using relative antenna
calibrations was an expedient made necessary to avoid cm-level errors, but it
was not a complete or permanent solution. For one thing, the relative
calibration measurements are really valid only for short baselines. Even for
the same antenna model, on long baselines a satellite is viewed from markedly
different elevation angles so the relative PCVs are not adequate [Mader, 1999].
Furthermore, the discrepancy in the GPS frame scale must ultimately be
understood and addressed. Note that even with the relative PCVs the current IGS
terrestrial frame is smaller than ITRF2000 (before rescaling) by nearly 3 ppb
(see the weekly IGS SINEX combination reports by R. Ferland) and that the scale
of the GPS frame has varied with time by more than 1 ppb in 2000 [Ge et. al.,
Absolute antenna corrections
It was suggested by several people that neglect of any non-ideal effects in the
satellite antennas might explain the apparent failure of chamber PCVs. Since it
has not been feasible to measure independently the satellite antenna
characteristics very accurately [Mader and Czopek, 2002], those properties must
be determined from the GPS data together with other usual geodetic parameters.
Springer  was the first to demonstrate quantitatively the difficulties of
the problem due to very high correlations among tracking antenna offsets and
PCVs, satellite antenna offsets and PCVs, estimated station heights, and
estimated tropospheric parameters. Due to these correlations, the general
problem is singular. A solution is possible if the terrestrial frame scale is
fixed by adopting a set of fiducial coordinates for the tracking network and if
the "absolute" phase center corrections (offsets and PCVs) for the tracking
antennas are known from external calibration measurements.
Meanwhile there exist absolute offsets and PCVs determined by a robotic system
developed by the University of Hannover and the company Geo++ [Menge et al.,
1998], which include azimuthal values and elevations down to 0 degrees. PCVs
for those tracking antennas not measured by the robotic system have been
inferred using the prior relative PCVs together with the absolute patterns for
the AOAD/M_T antenna. These inferred patterns are still only valid to elevation
angles of 10 degrees.
Schmid and Rothacher  demonstrated that it is possible to determine
satellite antenna offsets and PCVs if the absolute tracking antenna models and
the station coordinates in ITRF2000 are fixed. The results have been
independently validated by GFZ Potsdam [Ge and Gendt, 2005].
Recently, for the derivation of the 'official' IGS satellite antenna models
(PCVs and offsets) 11 years of data were reprocessed by TUM and GFZ while
aligning the solutions to IGb00. The solutions for the satellite offsets have a
trend, which is caused by an error in the vertical rate of the IGb00 (~1 mm/y).
Referencing the offsets to a given epoch (2000.0 in the antenna model) will
stabilize the scale in the GPS network solutions. Tests with the new model have
shown that the IGb00 has a scale rate of ~0.15 ppb/y compared to the new
solutions, which is consistent with the reported vertical rate error of IGb00.
So, a complete and consistent set of absolute PCVs for both tracking and
satellite antennas has been assembled. A test version of this set is available
in the file (see IGS Mails #5149, #5187):
The file contains in total:
- 106 antenna calibrations, where
- 14 are from Geo++ (elevation and azimuth, robot calibrated)
- 16 with identical construction are copied from the above 14
- 76 are from NGS (elevation only, converted form relative model)
- 45 antenna and radome calibrations
- 10 are from Geo++ (elevation and azimuth, robot calibrated)
- 35 are from NGS (elevation only, converted form relative model)
Among the 40 antennas possessing full calibration models (azimuthal PCVs and
elevations down to 0 degrees) from robot calibrations are most of the antennas
dominating the IGS tracking network. Converted relative calibrations have
mainly been added for reasons of completeness and in order to facilitate the
use of the file outside the IGS.
Note that there are several important differences involved in implementing the
new absolute patterns:
* The organization of the calibration information now uses the "ANTEX" format,
which is documented at:
The old format did not allow for satellite antenna corrections nor azimuthal
* Absolute PCVs can be reported as functions of azimuth as well as elevation.
* Robotic PCVs are measured down to 0 degrees elevation, whereas the relative
PCVs (and absolute PCVs derived from them) usually extend only to 10 degrees.
* Absolute PCVs and phase offsets are reported for individual satellites. The
PCV values are the same for all satellites within each block type, which are
tabulated in the file:
However, the z-offsets (in the direction from the satellite center of mass
toward the Earth's center) are satellite-specific.
* The IGS satellite block designations generally match those of the GPS
operators, except note that the Block IIR group is divided into IIR-A (first
eight satellites) and IIR-B (the next four launches: SVN 47, 59-61). There
was apparently a redesign of the antenna array. It is expected that the next
launch, of the first "modernized" IIR-M satellite, will start a new IIR
Under no circumstances should users mix absolute and relative PCVs! Moreover,
absolute PCVs require corrections for both satellites and tracking antennas
Implementation issues for absolute PCVs and handling radomes
Based on the work summarized above, sessions at the IGS 2002 "Towards
Real-Time" Workshop, held in Ottawa (8-11 April 2002), and the IGS 2004
Workshop and Symposium, held in Berne (1-5 March 2004), were devoted to issues
surrounding implementation of absolute PCVs. Important background information
is summarized in the position papers from these two meetings: see Rothacher and
Mader  available in:
and Schmid et al. [2005a] available in:
The IGS Analysis Centers (ACs) are currently in a test phase of using the
absolute antenna PCV corrections in parallel, unofficial solutions.
Associated with adoption of the absolute PCVs, the handling of antenna radomes has also been modified:
* PCVs for antenna + radome pairs will be used, where measured. The effect of
radomes was more or less ignored in IGS relative PCVs (some ACs are using the
NGS relative antenna+radome calibrations presented in igs_01.pcv). This
change can cause apparent station height differences up to several cm. In
many cases, suitable radome measurements do not exist; for those, the radome
effects will continue to be ignored for the remaining life of that pair.
* Any new antenna + radome pair must have calibration measurements before it
can be used in the IGS network.
* Position information for a station not having a calibrated antenna + radome
pair should be used with care. Vectors to nearby physical points, including
co-located techniques, can have errors up to several cm. Replacement of
antenna equipment can lead to discontinuities at a similar level.
* All station operators are urged to avoid using radomes unless absolutely
essential. If used in the future, radomes must be of a type suitable for
calibration and calibration measurements must be performed before IGS use;
Schedule for IGS changes
If current tests are successful, the IGS plans to implement operationally the
absolute antenna PCVs for tracking and satellite antennas on 1 January 2006
[the final date will be decided after the evaluation of the AC parallel tests].
The appropriate ANTEX file will be announced beforehand, but it will be quite
similar to igs_05.atx.
Summary of expected effects of IGS PCV changes
* TRF scale -- Because the IGS terrestrial frame products have always been
rescaled to match ITRF, there should be minimal impact in this respect except
that the unscaled GPS frame solutions should be much closer to ITRF and
should be more stable over time [Ge et al., 2005].
* TRF distortions -- The relative positions of the stations within the IGS
frame will change, in general. The changes will probably be largest for those
stations with calibrated antenna + radome combinations. Because of this, such
stations will probably have to be removed from the IGS set of reference frame
stations. Smaller shifts will be seen due to using azimuthal PCV corrections
in some cases.
* New IGS00 realization -- For the reasons above, the IGb00 realization of
ITRF2000 will no longer be suitable. A new version and revised set of
reference stations will be needed.
* Troposphere estimates -- Due to the correlations between antenna PCVs and the
zenith troposphere delay estimates, the latter values will likely change by
noticeable amounts. Biases between GPS estimates and those of VLBI or water
vapor radiometers should be considerably reduced [Schmid et al., 2005b].
* Orbits -- Because the scale of the satellite orbits is not very sensitive to
the TRF scale (being largely determined through Kepler's 3rd law), there
should be no major effects on the satellite orbits. Some positional changes,
at the quoted noise level, will likely occur however. (Comment: Some
satellites such as PRN28 and PRN03 will be more affected, because they have
large Z-offsets from the nominal values. For instance, the MIT absolute
model orbits for these satellites are much closer to the current IGS orbits
than are the (official) relative model solutions, MIT uses a loose radiation
model which allows these offsets to have a large effect due to high
* Clocks -- Any effects on satellite or station clocks should probably appear
minor compared to the quoted noise levels. However, the systematic effect of
the TRF scale change will be inherent in the new clocks.
* Long-term continuity -- A discontinuity should be expected for all IGS
product time series on 1 January 2006 (TBC), when the new absolute PCVs are
implemented. For this reason (and others), the IGS plans a complete
reanalysis of all historic GPS data; see IGS Mail #5174. Only when the
reanalysis results are available and incorporated into a future version of
ITRF will the fullest level of consistency between frames be achieved.
* Mixing products -- As a general rule, users should avoid mixing results from
solutions using different PCV conventions. When mixed results are used, a
thorough consideration should be given to possible systematic differences.
Considerations for regional networks
In general, analyses of data from regional networks can probably continue using
the relative PCVs with little or no impact, even with IGS orbits fixed.
However, this will depend on the size of the network and the effects should be
However, SINEX solutions submitted for combination (e.g. solutions from EUREF,
SIRGAS and NAREF which enter into the Regional Network Associate Analysis
Center combinations) should all be based on an agreed upon antenna model,
ideally the one used for all other IGS official products.
On the other hand, precise point positioning (PPP) results using IGS products
will not be self-consistent, before and after the PCV change, due to the
inherent shift in TRF scale.
Ge, M., and G. Gendt. Estimation and validation of IGS absolute antenna phase
center variations. in Proc. IGS 2004 Workshop and Symposium, ed. M. Meindl,
pp 209-219, Berne, Switzerland,2005.
Ge, M., G. Gendt, G. Dick, F. P. Zhang, Ch. Reigber. Impact of GPS satellite
antenna offsets on scale changes in global network solutions, Geophysical
Research Letters, 32(6), 2005, L06310, doi: 10.1029/2004GL022224, 2005,
Mader, G.L., GPS antenna calibration at the National Geodetic Survey, GPS
Solutions, 3(1), 50-58, 1999.
Mader, G.L., and F.M. Czopek, The Block IIA satellite - Calibrating antenna
phase centers, GPS World, 13(5), 40-46, 2002.
Menge, F., G. Seeber, C. Voelksen, G. Wuebbena, and M. Schmitz, Results of
absolute field calibration of GPS antenna PCV, in Proc. ION GPS-98, Nashville,
TN, pp 31-38, 1998.
Rothacher, M., and G. Mader, Receiver and satellite antenna phase center
offsets and variations, in Proc. IGS 2002 Network, Data and Analysis Centre
Workshop, eds. P. Tetreault, R. Neilan, and K. Gowey, pp 141-152, Ottawa,
Schmid, R., and M. Rothacher, Estimation of elevation-dependent satellite
antenna phase center variations of GPS satellites, J. Geodesy, 77(7-8),
Schmid, R., G. Mader, and T. Herring, From relative to absolute antenna phase
center corrections, in Proc. IGS 2004 Workshop and Symposium, ed. M. Meindl,
pp 209-219, Berne, Switzerland, 2005a.
Schmid, R., M. Rothacher, D. Thaller, and P. Steigenberger, Absolute phase
center corrections of satellite and receiver antennas: Impact on global GPS
solutions and estimation of azimuthal phase center variations of the satellite
antenna, GPS Solutions, DOI: 10.1007/s10291-005-0134-x, 2005b.
Schupler, B.R., R.L. Allshouse, and T.A. Clark, Signal characteristics of GPS
user antennas, J. Inst. Navigation, 41, 277-295, 1994.
Springer, T.A., Common interests of the IGS and the IVS, in Proc. IVS 2000
General Meeting, eds. N.R. Vandenberg and K.D. Baver, Koetzting, Germany,
pp 296-305, 2000.