12.07.2015 Views

GP-B Post-Flight Analysis—Final Report - Gravity Probe B - Stanford ...

GP-B Post-Flight Analysis—Final Report - Gravity Probe B - Stanford ...

GP-B Post-Flight Analysis—Final Report - Gravity Probe B - Stanford ...

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

14.2.2 Independent Data Analysis TeamsThe <strong>Stanford</strong> team has formed two independent, internal teams to attack the data analysis task. This was doneprimarily for internal quality control: two teams, processing the exact same data, should produce very similarresults. The results are compared formally on a regular basis where discrepancies are noted and investigated.These comparisons will unearth subtle software coding errors, which may go undetected if only a single analysispackage is used. Each team, while working under the constraint of using the same source data, has considerablelatitude in what methods of analysis to apply during each of the data analysis phases. This has the beneficialeffect of better understanding the nature of the source data and what signal processing techniques are mosteffective in estimating the performance of the four science gyroscopes.To provide a benchmark for both of the teams, sets of “truth model data” are constructed and are run on the twodata analysis packages. In these truth model cases, the signal structure, instrument calibrations, and noises areprecisely known and can be analyzed in a relatively straightforward way without sophisticated signal processingtools. Both data analysis packages should return identical results when run on such truth data; if they do not,this test reveals a fundamental difference in the algorithms that may be rectified quickly. Once the algorithmsagree on “truth”, they will be re-benchmarked regularly to ensure that evolutionary changes in the codes do notcompromise their ability to correctly analyze straightforward data sets.14.2.3 Data GradingThe <strong>Gravity</strong> <strong>Probe</strong> B science data consists of approximately 5300 guide star valid intervals during which theinstrument is able to measure the orientation of each of the four gyroscopes with respect to the line of sight tothe guide star. Due to vehicle anomalies or environmental effects, not all of these intervals contain valid sciencemeasurement data. If these data are used in the analysis algorithms, significant errors in the overall results maybe introduced.Consequently, a data grading scheme is used to separate the compromised data from the clean science data. Thegrading system does not remove data from the set, but simply flags segments of the data that do not meet anestablished grading criteria. Roughly 20 criteria have been set to flag sub-standard data on a gyro-by-gyro basis.The criteria chosen are primarily a function of vehicle state, such as “guide star out of view,” “vehicle in eclipse”,“loss of drag-free control,” “SQUID unlocked;” these all directly affect the validity of the data being sent fromthe spin axis readout system.During data analysis runs, the analyst selects the grading criteria that will apply during the run. Then, databaseextraction tools will then return the data that meet the grading criteria. Immediately, the same algorithm can berun with different grading criteria to assess the sensitivity of the result to various types of low quality data.420 March 2007 Chapter 14 — Data Collection, Processing & Analysis

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!