Subject: Re: [ngs.noaa.gov cors_req #4096] OPUS: Date: Fri, 18 Oct 2002 16:30:23 -0400 (EDT) From: Mike Cline Reply-To: cors@proton.ngs.noaa.gov To: ray@gwinn.com CC: cors@proton.ngs.noaa.gov, cors@proton.ngs.noaa.gov Ray Gwinn, You are worse than the little kid who asks why strawberries are red. I am reminded of a former colleague who once ranted, "I can respond to questions about how to solve a problem the right way, but I'm not about to respond to every hairbrained attempt to solve a problem the wrong way!" Ray Gwinn, you are fast trying my patience with your questions about why you fail to get the right answer when you try to solve problems any hairbrained wrong way. When I submit your file bkly2710.02o to OPUS with correct identification of the antenna, the NAD83 ellipsoid height of the ARP is 692.593 meters with peak-to-peak uncertainty of .025 meters. In contrast, the published NAD83 ellipsoid height of the ARP is 692.580 meters. The discrepancy between the one-day OPUS height and the multi-day published height is 692.593 m OPUS-derived ARP - 692.580 m Published ARP ------- .013 m discrepancy +/- .025 m OPUS peak-to-peak uncertainty The answer looks pretty good to me when we calculate the height the correct way--in other words, by correctly identifying the antenna. The antenna calibration for antenna model NOV600 states that the ARP-to-L1PC offset is 0.0905 meters (see http://www.ngs.noaa.gov/ANTCAL/Models/NOV600.shtml ) The NAD83 ellipsoid height of the L1PC for the OPUS-derived solution of bkly2710.02o is therefore 692.593 m OPUS-derived ARP + .091 m ARP-to-L1PC offset ------- 692.684 m OPUS-derived L1PC Now you want to solve the problem the wrong way by failing to identify the antenna. When I submit the same file to OPUS without identifying the antenna (NONE), the NAD83 ellipsoid height of the default nominal L1PC is 692.719 meters. The discrepancy between the right answer and the hairbrained answer is: 692.719 m Hairbrained OPUS-derived L1PC - 692.684 m Correctly determined OPUS-derived L1PC ------- .035 m discrepancy +/- .021 m Hairbrained OPUS-derived peak-to-peak uncertainty The discrepancy is outside the peak-to-peak uncertainty by 14mm. Moreover, the difference between the hairbrained one-day OPUS solution and the published solution is: 692.719 m Hairbrained OPUS-derived L1PC - 692.671 m Published L1PC ------- .048 m difference Note that this nearly 5cm difference contains two sources of error: first, the comparison is between one day of data versus the published mean of 10 days of data; second; one value is calculated in a hairbrained manner and the other is calculated correctly. Why does the hairbrained method give such a large 5cm discrepancy? First, the ARP-to-L1PC offset of 0.0905 m is only half the calculation. The remaining half of the calculations have neglected the variations in the phase centers due to the elevation angles of the arriving GPS signals. Note from the antenna calibration table that when the signal arrives from elevation angle 85 degrees (i.e., almost overhead) the L1PC is another 0.7mm higher than the nominal value. When satellite signals arrive from 45 and 40 degrees above the horizon, the L1PC is another 8.1mm higher than the nominal value. These elevation-angle dependent terms are totally neglected from the hairbrained method of L1PC calculation. Where do the neglected terms go? They spill into the tropospheric corrections for atmospheric water vapor that directly affect the height calculations, thereby giving unpredictable and hairbrained heights. Moral of story: The correct answer depends on identifying the antenna, so that one form of calculation (phase center variations) doesn't corrupt another form of calculations (tropospheric corrections). If the antenna model were unimportant, why would we even bother with antenna calibrations and ARP coordinates? Everyone could always choose "NONE" for the antenna, regardless of make and model, and life would be easy--except the heights would be wrong! M Cline