6.1 Beyond the Standard Calibration Pipeline

DrizzlePac is written in Python (with core drizzle algorithms written in C). Its interface is a departure from previously historically conventional IRAF usage. To learn more, please refer to the DrizzlePac Jupyter Notebook Tutorials for an introduction to using Python to run DrizzlePac tasks.

Instrument pipelines provide data calibrated to a level suitable for initial evaluation, but may not be suitable for scientific analysis. When users place a data request at the HST Archive, their data is processed using the best available software to calibrate data with the best available reference files. If it has been a long time since their data was retrieved from the archive, users are encouraged to re-download it to ensure that the data contains the most up-to-date header information and calibrations. 

There are presently two major steps in the processing:  (1) Calibration of individual datasets using instrument-specific calibration software, such as calacs for ACS and calwf3 for WFC3;  (2) Combining associated data with AstroDrizzle to produce a combined, distortion-corrected, and largely cosmic ray-free image. The second step cannot succeed without good results in the first.

There may be occasions when pipeline calibration of individual images require custom calibration by the user. Instances when automatic reprocessing is not ideal may include when a user has a preference for self-made calibration reference files, or the use of non-default calibration switches, or when using non-default software parameter values. Reason for these actions could be to improve hot pixels and cosmic ray removal or to deal with image persistence/other additional sources of noise.

For example, NICMOS data may require special attention: images from this camera often contain additional signal in the sky, persistence or pedestal effects (differing bias levels between quadrants in the chip) that require extra processing for removal. For more detailed information on recognizing and removing these effects in NICMOS data, please refer to Chapter 4, Anomalies and Error Sources, in the NICMOS Data Handbook.

Even when individual datasets from the archive appear well-calibrated, users should consider if reprocessing their images with AstroDrizzle on their home machines is beneficial. Drizzled pipeline data is created with conservative values which are stored in the MDRIZTAB reference file. More can be read about AstroDrizzle in the pipeline here. 

Some things to consider, depending on the type of data: drizzling with a finer output scale may produce better cosmic ray rejection. Using a smaller pixfrac will reduce correlated noise. Using both a smaller pixfrac and a smaller scale can produce a sharper PSF. In many cases, one can produce better images with a bit of effort. Further information can be found in this example notebook on optimizing image sampling.