4.1 Antenna Testing

Overall, the goal of antenna calibration studies has been to determine the best general phase corrections for a given antenna model with the understanding that there are specific site influences which can modify the pattern and introduce additional measurement errors. There are principally two ways to measure the phase patterns of GPS antennas. The first is in an anechoic chamber (e.g. Schupler and Clark, 1991; Meertens et al., 1995) and the second is in the field using actual GPS signals. For the field measurements there have been two types of calibrations, one using antenna rotations (e.g., Meertens et al., 1995) and the other using relative differences with respect to a reference antenna (e.g., Mader and MacKay (1996) and Rothacher et al., (1995)). While a more sophisticated in-situ field reference calibration system has been proposed by the UNAVCO Community, rotation tests remain an important element in individual antenna calibration and testing. Although the different calibrations are not all in agreement, the results are starting to converge.

Anechoic Chamber Measurements

Over the last few years a series of chamber tests have been conducted by Goddard/Bendix (Schupler et al., 1995) and recently one set of tests by UNAVCO/Ball Aerospace at the Ball anechoic chamber located in Broomfield, Colorado (Meertens, et al., 1995). With these tests, a full three-dimensional absolute phase and amplitude pattern is measured. The patterns resulting from the UNAVCO/Ball tests are shown in Figure 4-2. In Figure 4-3 the corresponding horizontal phase offsets for six GPS antennas used by the UNAVCO Community are plotted. For these antennas, the offset of the L3 combination (typically used for baselines longer than a few kms) is up to 8 mm and there is considerable variability between designs in both magnitude and direction. In Figure 4-4 the Ball chamber results are compared with the Goddard results (Schupler et al., 1994, 1995) for specific antennas showing a large difference between the offsets determined from the two chambers. The Ball chamber tests were run using the antenna and low-noise-amplifier (LNA) combinations provided by the manufacturers. The chamber source transmitted at nine frequencies near L1 and L2 to simulate GPS spread spectrum modulation. Observations were at five degree intervals over all azimuths and over ±120 degrees of elevation. Thus, more than 60,000 digital phase measurements were recorded for each antenna. The centers of rotation of the antennas with respect to the chamber mount were determined using a laser. The enhanced capabilities of this state-of-the-art chamber may account for any disagreement between the results presented here and previous results (Schupler and Clark, 1991, 1994).

Figure 4-2. L1 Phase Center Patterns for Several Antennas. (10 degrees of L1 phase is approximately 5 mm. Each of the "sombrero" plots show zenith values in the center and 5 degree steps outward ending at 10 degrees above the horizon.)

Figure 4-3. Horizontal Phase Center Offsets (labeled L1C, L2C, and L3C) Determined from Chamber Tests. (The antenna axis of rotation is the zero point. Error estimates are ±0.5 mm for the measured L1 and L2 offsets.)

Figure 4-4. Horizontal Phase Center Offsets Derived from UNAVCO Field Antenna Rotation Tests.

Field Measurements

Two types of field tests have been conducted on Table Mountain near Boulder, Colorado, to validate antenna calibration parameters determined by the chamber tests. First, antenna rotation tests were conducted using antennas of the same type. In these tests, antennas are aligned to north on one mount and to the south on another mount, and then each antenna is rotated by 180 degrees. The observed difference in baseline length is equal to four times the average horizontal phase center offset from the rotation axis of the antenna. UNAVCO has conducted antenna rotation tests since 1989 on available antennas including the Trimble 4000SX MICRO, 4000SLD L1/L2, SST and GEOD L1/L2 GP, the Ashtech GEODETIC L1/L2 L, TI-4100, FRPA, and the AOA Dorne-Margolin T. The results are available in a series of UNAVCO technical reports. In the second field test, calibration corrections determined by the chamber tests are used in surveys between mixed antenna types on known baselines.

Chamber measurements showed horizontal phase center offsets as large as 3 mm (L1) and 4 mm (L2) with an uncertainty of 0.5 mm, yielding as large as 8 mm horizontal offsets for L3 (Figure 4-3). These offsets agree within 1 mm for L1 and L2 with offsets determined by field rotation tests for the AOA Dorne Margolin T and Trimble GEOD L1/L2 GP antennas (Figure 4-4) as well as for the Trimble SST (not shown). The UNAVCO/Ball results offer for the first time a confirmation with chamber measurements of the horizontal offsets observed in the field antenna rotation tests.

The consistency of the phase center offsets of GPS antennas is important when trying to achieve repeatable geodetic measurements accurate to the millimeter level. UNAVCO has previously tested the NASA-UNAVCO pool of Allen Osborne Associates Dorne Margolin (DM) choke ring antennas manufactured in 1992. These results showed an L3 horizontal baseline scatter of 1.5 mm between antennas of the same model number. The nominal phase center offset for these antennas is zero. Full rotation tests conducted by Trimble in early 1996 on their prototype choke ring antenna showed an L3 systematic horizontal offset of up to 3 mm, significantly larger than previously found and attributable to the particular pair of DM elements used. Random tests of DM elements subsequently delivered to Trimble showed sub-mm offsets in their tests. Full rotation field tests were conducted on 11 Trimble choke ring antennas from the new ARI equipment purchase prior to delivery to their owners. Using three or more antennas allows for absolute calibration of the horizontal phase center offsets. Application of the absolute calibration results show sub-mm L1 and L2 offsets and L3 offsets of less than 1.8 mm in both horizontal components (Figure 4-5). The effect of changing the processing elevation cutoff from 20 degrees above the horizon to 10 degrees is less than 10 percent. The relative vertical offset derived from swapping of antennas shows a scatter of less than 1 mm. The magnitude of the horizontal phase center variability of the choke ring antennas, up to 1.8 mm for the L3, is comparable to the offset in the Ball chamber results of Figure 4-3. It is therefore not clear whether the observed chamber-derived offset is due to measurement errors or actual offset which might be expected from antenna variability.

Figure 4-5. East/West Baseline Component Scatter for Trimble Choke ring Antenna Rotation/calibration Tests. (Red "+" symbols represent scatter without individual antenna phase offsets applied. Green "*" symbols represent scatter of position residuals after estimation of individual phase offsets.)

Antenna Mixing

ARI field tests showed that L3 mixed antenna, Ball chamber-derived mean offset corrected measurements with no tropospheric estimations were in error by 20 mm or less in the vertical (Figure 4-6, left panel) and by 1 mm or less in the horizontal. If hourly tropospheric estimations are included in the L3 solutions, the vertical error increases to as much as 87 mm for the Trimble GEOD L1/L2 GP (SSi) to Ashtech Dorne Margolin (Z12) (Figure 4-6, right panel).

Application of mean offset and pattern corrections reduces the vertical error for troposphere corrected Trimble GEOD L1/L2 GP to AOA Rascal and Ashtech Dorne Margolin antenna mixes to 13 mm or less. The least successful application of the offset and phase center pattern corrections has been with the Trimble 4000ST L1/L2 GEOD antenna where the residual error is as large as 50 mm[1].

Figure 4-6. Summary of mixed-antenna vertical solutions with offset corrections only (+ symbols) and offset plus phase center pattern corrections (o symbols). L3 solutions are shown with tropospheric estimation (right panel) and without tropospheric estimation (left panel). The columns, separated by vertical lines, show 10 different antenna mixes. Ground truth is indicated by the dotted line.

In general, the ARI mixed antenna baseline solutions show vertical errors of 12 mm or less without tropospheric estimation, and up to 50 mm with tropospheric estimation. This compares with 1 mm errors achieved with unmixed antennas on short baselines with no tropospheric estimations. The residual antenna mixing error has several possible sources. First, an anechoic chamber provides an ideal low multipath environment whereas observations in the field are influenced by local site conditions including multipath and monument effects. Second, differences in phase patterns between mixed antennas can be easily confused with tropospheric delay, particularly at low elevation angles. The wealth of mixed antenna data collected during the ARI will permit independent checking of future general antenna phase variation models expected from the IGS as provided by Gerry Mader and Markus Rothacher without having to perform further field tests for these antennas.

[Next] [Previous] [Up] [Contents]

Last modified Thursday, 19-Jan-2006 17:17:18 UTC

Home | About Us | Contact Us | Support | Search | Facility | PBO | Education & Outreach

Comments: webmasterunavco.org
© 2009 UNAVCO, Inc.