Automated Pipeline Processing for AXAF Ground Calibration

J. DePonte, M. Conroy, K. R. Manning

Smithsonian Astrophysical Observatory, MS 81, 60 Garden St., Cambridge, MA 02138 USA

K. Presley

TRW, MS 81, 60 Garden St., Cambridge, MA 02138 USA

Abstract:

1. Standard Data Processing Overview

For the XRCF ground calibration, named pipelines have been defined to accommodate the automatic reduction and analysis of data produced from tests of the AXAF Science Instruments. Standard pipeline processing is partitioned into 4 levels. At each level data is archived. In Level 0, the telemetry substreams are isolated, the science and engineering streams are decommutated, and data gaps and detected errors are recorded. In Level 1, the raw events are processed to remove instrumental effects, and standard screens are applied to the mission time line---a data store for housekeeping, engineering, and gap information to build temporal filters. Standard masks (spatial,pulse height amplitude) and temporal filters are then applied to the data to identify the standard event list. Level 2 processing is test specific for XRCF and is identified with a particular test at a particular source energy. The data is analyzed and test-specific products are output. Test data is compared to the predictions from simulation runs of the same test, and residuals are determined. In Level 3, data from individual tests are stacked into a test-suite. A theoretical model is fit across multiple test points, and a model is derived to predict performance at any arbitrary point spanned by the test.

2. Key Components and Interaction

Pipeline processing is achieved through the interaction and execution of key components in the ASCDS. Prior to XRCF, a Test Database is generated with an entity for each test to be performed during calibration.

Data processing is controlled via the test status tracker (TST)---a component in the system that monitors and invokes applications by tracking status fields in the Test Database. The database fields are updated by software components as different phases of the data processing are completed. The TST invokes a pipeline by sending the Pipeline Executive a Pipe ID and Test ID. The Pipe ID identifies the preconfigured pipeline and corresponds to a pipeline profile in the Profile database. Pipeline Profiles are ASCII templates which identify the tools, the run sequence, and default parameters for the programs that compose the pipeline. The Test ID defines the set of data to be processed. Pipelines are run under Pipeline Control (Mistry 1995) by the Pipeline Executive. The Pipeline Executive (PEX) component provides pipeline queuing and management and orchestrates the execution of multiple pipelines. The PEX invokes a search and retrieve client to retrieve data from the archive. Profiles are submitted to the Executive which places them in a queue for execution by Pipeline control. Pipeline control (PCTR) provides the ASCDS the ability to create a data processing stream from information captured in a profile. The PCTR also allows the user to monitor the data processing for integrity, and provides error handling and error recovery features. Upon completion of the processing, the Pipeline Executive invokes an archive client to store data products, the processing log, and bookkeeping information in the XRCF Calibration Database.

Figure 1 illustrates an example of the interaction of the components in the Level 2 HRC pipeline for the Point Response Function (PSF) test.

  
Figure 1: Example of Pipeline Component Architecture.
Figure 1: PS 21 Kb

  
Figure 2: ASCDS Software Architecture of Pipeline Components.
Figure 2: PS 15 Kb

3. Layered Architecture

The layered software design of the ASCDS partitions components and allows for the migration of the Release 1 data system which supports XRCF, to future data system releases which will support the flight system. Each component is isolated and communicates to other components via a defined API as illustrated in Figure 2. Each layer hides the details of its internal structure and mechanism from the other layers. Modularity and extensibility are key to the design (Conroy 1996).

  
Figure 3: ASC Predictor/Test Pipelines.
Figure 3: PS 20 Kb

4. XRCF Predictor/Test Pipeline Runs

The intent of the predictor pipeline is to use the current best representation of telescope models to predict performance. Comparison of predictor and actual pipeline results occur at several stages in the calibration of AXAF and is a key component to the calibration of the telescope at XRCF. An example of a predictor/test pipeline interaction and the data dependencies are shown in Figure 3. Comparisons of results from the two pipelines allow the refinement of models and data processing.

Prior to XRCF, the predictor pipeline is run using calibration models of the individual telescope components, the High Resolution Mirror Assembly (HRMA) and the Science Instruments (SIs) to simulate test data. Simulation runs will coincide with planned XRCF tests and standard data analysis is run to predict the performance of the integrated telescope. These predictions are used as input to the test pipelines run at XRCF on actual test data. During XRCF, predictions are continually updated as information is gained from the analysis of test results.

After completion of XRCF calibration, predictor pipelines for all tests will be run to fine tune the models, resulting in telescope and component models at 1-G. Final predictions for flight performance will be made using these models and a 0-G prediction of the HRMA. The 0-G predictions are verified during the orbital verification phase of the mission to determine flight models. At each phase of the calibration effort, automated pipelines will be utilized to facilitate the processing.

Acknowledgments:

This work is supported by the NASA contract NAS8-39073 with the ASC.

References:

Conroy, M., Doe, S., & Herrero, J. 1996, this volume