Positioning by GPS can be derived today with a precision of a few millimeters. A GPS receiver determines the distance between the electrical phase center of its antenna and the phase center of a GPS satellite’s transmitting antenna as a pseudorange or a carrier phase observation. It is well known that the phase center of the antenna is not a stable point.

For each GPS receiver's antenna the phase center changes with respect to the elevation angle and azimuth of the incoming satellite signal (Rothacher et.al., 1995). Different antenna calibration models are given in general by a set of antenna offsets with respect to the mechanical reference point and a list of variation values that help a user to refer to the ground points under determination.

Knowing Phase Center Variations (PCVs) is especially important in case different antenna types are used at the end points of a baseline. Mixing antennas usually happens in GPS regional and permanent networks like national networks, IGS and EPN. An uncalibrated antenna will certainly introduce errors that combined with other error sources result in significant erroneous point estimations, growing on long baselines (Mader 1999, Fotiou and Pikridas 2006). Even if identical antenna types are used the effects of PCV values do not cancel out for long GPS baselines. This is due to the earth curvature that causes elevation differences and therefore common satellites are seen at different elevation angles by the end points of a baseline.

In order to overcome the above problems various calibration models have been generated and used by the IGS community, namely the relative and more recently the absolute antenna calibration models (Schmid et al., 2005).

The relative phase center variation models are based on the assumption that the Alan Osborne antenna type AOAD/M_T has been

Fig 1: Differences in the horizontal component caused by the transition from relative to absolute PCVS.

approved of being the “Zero” antenna. This antenna type forms a standard with elevation dependent variations set to zero referring to a mean fixed offset. PCVs for a calibrating antenna can be determined using short-baseline field measurements (Rothacher et.al., 1995). Thus for each antenna type correction values were adopted relative to the external or mechanical antenna reference point (ARP, MRP). A database of relative calibrated antenna types has been generated with free access to everyone. The drawback is that the corrections are dependent on the zero/reference antenna and that PCVs at low elevations are not reliable due to the increment of noise and multipath in measurements below 10 degrees (Mader 1999).

Combining GPS with other space-geodetic techniques becomes difficult in case of unmodeled systematic errors due to improper GPS antenna calibration models. As a consequence scale differences have been seen in GPS reference frames. Due to the above mentioned disadvantages relative models can no longer satisfy the increasing accuracy requirements.

Until November 2006 relative elevationdependent PCVs were applied within the IGS and EPN. After that date (GPS week 1400) the IGS has adopted the absolute PCVs for its routine generation of precise orbits and station coordinates. In IGS-Mail 5438 (2006) a new file with absolute antenna models, named igs05_1390.atx, was made available to the GPS community. Ever since several updates of this file were followed. The EPN started to use absolute antenna models at the same time with IGS.