Antenna Issues

The observed range between a receiver antenna and the transmitting satellite antenna is the distance between the electrical phase centres of both antennas. However, the geometric antenna centres of both the transmitting and receiving antennas and their electrical phase centres do not coincide, and can differ for receiver antennas by centimetres and for satellite antennas by up to one metre (Mader and Czopek,2001). The offset between the geometric antenna centre and the electrical phase centre of the receiver antenna can be described using a constant and varying component. The constant offset arises from the difference between the mean phase centre and an antenna reference point (ARP), whereas the variable offset depends on the elevation angle, azimuth and intensity of the incoming signal and differs for L1, L2 and the ionospherically free linear combination (Schupler and Clark,1991;Rothacher et al.,1995).

For transmitting satellite antennas, the offset between the electrical phase centre and the geometric reference point has been assumed to be constant for a specific satellite block type. The IGS therefore uses offsets for Block I, II, IIA and IIR, which are referenced to the centre of mass of the GPS satellite. Currently there are no elevation–dependent antenna phase centre variations applied for satellite antennas. Mader and Czopek(2001) however showed that the IGS satellite phase centre offsets differ significantly from offsets determined during a satellite antenna phase centre offset calibration carried out on a Block IIA satellite antenna.

Receiver antenna phase centres also vary between different antenna types, but are consistent between antennas of the same model. For high precision geodetic choke ring antennas using the Dorne Margolin antenna element, the antenna phase pattern was shown to have hardly any azimuthal phase centre variations (W¨ubbena et al., 2000). For short baselines, this means that by using identical antennas and orientating them in the same direction, the effect of antenna phase centre variations can be mitigated. If baselines are longer, and satellites do not appear in the same elevation angle anymore, then the bias does not difference out and antenna phase centre models must be applied.

Chapter 3. The Global Positioning System (GPS) for Monitoring Tide Gauge Sites 47

If antenna types are mixed, then Rothacher (2002) stated that even with application of antenna phase centre models, there would be no guarantee of sub–centimetre height determinations, especially for longer baselines, where the ionospherically free linear com- bination is used, and where tropospheric delay parameters are estimated. Furthermore,

Rothacher (2002) concluded that if precise height determination was the objective, the same equipment (antenna, receiver, monument, etc.) should be used at all stations and any changes to the equipment avoided.

There are currently three different methodologies in use for antenna phase centre cali- brations, which can be grouped into absolute and relative calibration procedures (Rothacher and Mader,2002).

The first method for absolute calibrations of GPS antenna phase centre variations was reported bySchupler and Clark(1991). This method involves measurements in an anechoic chamber1 where the antenna to be tested is mounted on a positioner, which enables rotations around two independent axes and shifts in three directions. A transmitting antenna is kept fixed, while the receiving antenna is rotated through zenith angles from -90◦ _{to +90}◦ _{in various azimuths.}

In relative antenna phase centre calibrations the phase centre variations of a test antenna are determined with respect to a reference antenna over a small baseline (<10 m) (Mader and MacKay,1996). The antenna phase centre variations of the reference antenna are assumed as known, and if known absolutely, absolute antenna phase centre variations for the test antenna can be derived.

In recent years, an in field automatic absolute antenna phase centre calibration method was developed by W¨ubbena et al. (1998). In this case, the antenna to be calibrated is mounted onto a precisely calibrated robot, which rotates the antenna around a fixed point, i.e. the nominal phase centre. By carrying out the experiments over two siderial days the multipath effect, which is a problem of the relative field calibration, can be largely reduced. Furthermore, the antenna elevation mask is dynamically adjusted depending on the inclination of the antenna in order to guarantee an elevation mask of 18◦_{, further} 1_{An anechoic chamber is a room of which all interior surfaces are lined with radio frequency absorbent}

Chapter 3. The Global Positioning System (GPS) for Monitoring Tide Gauge Sites 48

improving the multipath situation. As this procedure is carried out fully automatically, the antenna can be rotated and inclined into several thousand precise positions, resulting in a high resolution and precise phase centre variation model (W¨ubbena et al.,2000).

By investigating the geometric relation between the satellite and receiver antenna elevation–dependent phase centre variations, Rothacher and Mader (2002) found that these phase centre variations were not separable. The elevation–dependent phase centre pattern of the receiver antenna may thus be interpreted as a phase centre pattern of the satellite antenna and vice versa. They concluded that this was the reason for the inconsistencies seen between the absolute receiver antenna phase patterns determined in the anechoic chambers and the relative phase patterns. The analysis of global IGS data and different phase centre patterns recently obtained for Block II/IIA and Block IIR satellites, in combination with absolute phase centre variations from anechoic chamber and in–field absolute calibrations showed that a large part of the change of scale in the GPS, as reported by Herring (2001); W¨ubbena et al. (2000); Heflin et al. (2002) and Herring (2002), was due to the mis–modelling of antenna phase centre variations. Based on this, Rothacher and Mader(2002) suggested that the IGS should adopt these recently determined satellite antenna offsets (Mader and Czopek, 2001), together with the consistent satellite antenna phase centre variations (Rothacher and Mader,2002) and the absolute receiver antenna phase centre variations determined byW¨ubbena et al.(2000), in order to produce a more consistent ephemeris.

Recently,Rothacher (2003) announced the availability of a test set of absolute antenna phase centre offsets and phase centre variations for receivers and satellites. The satellite antenna phase centre variations have to be used together with the absolute receiver antenna phase centre variations in order to avoid scaling of the GPS results. However, to date these have not been applied in the generation of IGS products.

In order to protect GPS antennas at remote sites from snow and ice accumulations or weathering, antennas can be equipped with radar domes; covers that are transparent at GPS frequencies. However, there is evidence that radomes affect the incoming GPS signals and directly bias station coordinate estimates. For local networks, the effect does not exceed the several millimeter level, however for regional networks, the effect increases

Chapter 3. The Global Positioning System (GPS) for Monitoring Tide Gauge Sites 49

remarkably as soon as local troposphere parameters are to be estimated in the adjustment (Kaniuth and Stuber,2002). The resulting height errors have been reported to be as large as 3 cm for a cut–off elevation angle of 10◦ _{and tend to increase with increasing cut–off} angle. Furthermore, the error depends on the antenna and radome type and there are suggestions for a dependence on the local antenna environment. Although it has been reported that the addition of a radome does not affect phase and amplitude patterns noticeably (Schupler and Clark, 2001), a systematic lowering of the mean phase centre position by 2 to 3 mm is observed, and visually confirmed by the height time series shown in Johansson et al. (2002). If both antennas are fitted with identical radomes, it can be assumed that the effect cancels completely on short baselines and partly on longer baselines (Braun et al.,1997;Kaniuth and Stuber,2002;Schupler and Clark,2001).