Spitzer Documentation & Tools
MIPS Instrument Handbook

Chapter 6              Data Products

6.1  Data Products Overview

The nature of the SSC pipeline data processing of MIPS data and the resultant data products are briefly described here and in the following section. 


In order to fully understand and assess the characteristics of MIPS data, the data processing capabilities, and limitations for the Ge:Ga data in particular, it is important for the observer to carefully examine and understand the photoconductor detector behavior (section 2.3), calibration strategy (sections 4.1.1 and 4.1.2), data acquisition methods (Chapter 3), and how these relate to each other.  Gordon et al. (2004, PASP, 117, 503) has more details about the data processing algorithms than are included here.


MIPS data are delivered to observers as FITS image files through the archive interface.  Because of the complexity and redundant nature of the MIPS AORs, three distinct types of data are generated by the pipeline processing.  Observers will be able to select which kinds of data they would like to download through the archive interface.


Level 1: Basic Calibrated Data products (BCD) for MIPS are data derived from a single data collection event (a DCE, or a single frame exposure).  The Basic Calibrated Data are planned to be the most reliable product achievable by automated processing.  BCDs for all AOTs are in the form of FITS images; the SED BCDs are dispersion-corrected and calibrated two-dimensional spectral FITS images.  Ge BCD images are corrected for the effects of cosmic ray impacts; cosmic rays are flagged (but not removed) for Si.  The total integration time for the individual BCD frames is the observer-specified exposure time (i.e., 3, 4, 10, or 30 seconds), not the total integration time accumulated during the multiple on-source exposures that result from an AOR.  No background or sky subtraction is made.  BCD products are composed of 2-dimensional FITS image files and a full set of header information keywords, including distortion coefficients, epoch, pointing information, and pipeline and calibration versions.  The BCD products are calibrated in MJy/ster.


Level 2: Post-BCD products (post-BCD) are derived from a full AOR (e.g., a scan map or a dithered photometry observation).  These data are the result of combining all individual BCD frames from a single AOR; in the case of photometry mode, the product is an averaged and registered single image suitable for photometric measurements, and in the case of a scan map, the product is a registered mosaic, with first order removal of seams between the component images.  MIPS post-BCD products are calibrated in MJy/ster.  The post-BCD products are delivered to the observer in the form of a single FITS image file per AOR, and include a full complement of header keywords.  The BCD products are, for the most part, what you should start with when working with MIPS data.  For Ge data, in particular, the post-BCD data products are a good way to get an overview of your data; however, to do science, you should revert to the BCDs and make your own post-BCD products.


Level 0: Raw Data products are the unprocessed array images (in unprocessed counts per pixel).  These FITS image files allow the observer direct access to the data, but still contain the difficult-to-calibrate detector behavior inherent to Ge:Ga detector technology for the 70 micron and 160 micron frames.  No cosmic ray removal will have been performed.


Calibrated Data Units

The Level 1 BCD product, which is the primary data returned to observers after pipeline processing, consists of individual frames where the pixel values are in units of MJy/sr.  Jansky is a flux density unit defined as:

1 Jansky = 1 Jy = 10-26 W m-2 Hz-1 =


The conversion between Jansky and flux density in W m-2 per unit wavelength is accomplished via

 x 10-26 x


where the wavelength bin-width is specified in the same length units as  and c.  For example, if c is taken as 3x1014 micron s-1 and  is specified in micron, the above equation results in  being in units of W m-2 micron -1.


MIPS Data Processing

Gordon et al. (2005, PASP, 117, 503) has more details about the data processing algorithms than are included here.  Basic Calibrated Data (BCD) images are accompanied by additional FITS files.  These include, but are not limited to: traceable uncertainties per pixel (an error image), calibration and pipeline reduction pedigree, and data masks for pixels flagged for various data quality criteria.


Production of the Level 1 Basic Calibrated Data includes removal of electronic signatures (as applicable) such as a ''droop'' correction in the 24 micron array data (see also chapter Chapter 7), dark current subtraction, non-linearity of sample ramp slopes (Ge:Ga), a robust fitting of ramp slopes to determine total counts per second (Ge:Ga), and the flagging and removal of single or multiple cosmic ray hits using robust Bayesian techniques for the fully sampled Ge:Ga exposure ramps sent to the ground.  When possible, corrected ramp segments are adjusted and averaged as required.  The inherent redundancy of the MIPS data acquisition adds additional compensation for cosmic ray effects in the combined data.  Bad pixels are identified and masked as appropriate. 


For a uniform diffuse background, the pixels in a BCD image report equal values (within noise statistics; a.k.a. flat fielding of the arrays is performed).  The flat field of the Ge:Ga array data is determined and maintained temporally for the Ge:Ga arrays as described in section 4.1.4.  Correction of raw counts to flux density per pixel is done based on the detector behavior described elsewhere in this chapter.  The flow, or order, of pipeline processing steps appears in Figure 5.1 (the BCD Si pipeline), Figure 5.6 (the BCD Ge pipeline), and Figure 5.7 (the post-BCD pipeline).  These pipelines will continue to be refined during operations.


The Final BCD products for 24 micron data include (but are not limited to) the following items: slope image, difference image (all zeroes except for pixels saturated in the slope image), replaced image (identical to slope image, except non-zero difference pixels are replaced), error images, and mask images.  Note that the 24 micron pipeline does not correct for latent images, and although it flags cosmic rays, it does not correct for them.  The Final BCD products for Ge (70 and 160 micron) data similarly include the following items: fully calibrated slope images, the errors associated with the slope images, and mask images.


Whether Level 2 post-BCD  products are sufficient for a particular program depends upon the specific scientific needs of that program.  For example, the MIPS pipeline does not provide the type of refined data processing needed to achieve ''super-resolution'' imaging that can be achieved with the available MIPS data acquisition modes.  Observers requiring this type of processing in the combining of BCD images will need to apply standard techniques, or develop new ones as required by the demands of the data and/or the science goals.  The software used for the Post-BCD pipeline is available (along with copious documentation) in the Data Analysis section of this website.


The MIPS post-BCD products are derived per AOR, and do not extend to multiple AORs in an Observer's program.  It is composed of FITS image files of size MxN (as defined by the specific observation or map) that are the result of optimally mapping the constituent frames into the post-BCD data frame with minimal loss in resolution.  There is one post-BCD FITS file per band per selected wavelength per AOR, except for scan map AORs where there are always 3 files per AOR.


The pointing offsets used in mapping of individual frames into the post-BCD products are derived from an optimal offset estimation method, e.g., point source extraction.  If insufficient point sources are available, then the data are aligned using spacecraft and CSMM pointing knowledge only.  Individual frames are adjusted such that the overlap regions are statistically similar, within the limitations of the large-scale background and slow response characteristics of the Ge:Ga detectors.  This implies that some edges might be visible for certain types of varying backgrounds.  This will especially be true for adjacent legs of a scan map over a region with significant and varying background.  The absolute level to which edge effects and scan leg baseline offsets will be removed in the post-BCD products will not be known with certainty until more flight data are obtained.  For observers requiring very uniform maps of extended source regions, there should be some expectation of additional required data processing to correct for edge effects and scan leg baseline offsets in the Ge:Ga data of a scan map.  Such effects are minimized in the other MIPS observing modes.  The zodiacal backgrounds are maintained only within an individual AOR.


Observers who wish to produce maps of areas larger than can be imaged in a single AOR could merge either the individual BCDs or the post-BCD mosaics themselves, in order to obtain the combined dataset.  Characterizable Ge:Ga long-term detector behavior are included as part of the processing as needed.  No background or sky subtraction is made as part of the post-BCD products processing.  Ancillary data images similar to those included with Basic Calibrated Data are also provided with the post-BCD products.


The level of MIPS pipeline performance is contingent on proper observation design (one that takes full advantage of the standard MIPS calibration), and a suitable target selection.  Suitable target selection includes an assessment of the impact of the general background to be observed during the observation, the MIPS saturation limits, and any bright sources that come within the MIPS fields of view.


Absolute pointing knowledge of the focal plane array pixel centers based on the Pointing Control System is better than:

         24 micron : about 1.4'', 1 sigma radial

         70 micron : about 1.7'', 1 sigma radial

         160 micron : about 3.9'', 1 sigma radial           


From the science data frames themselves, refined pointing knowledge of sources must be based on centroiding of objects visible in a MIPS image.  The uncertainties will depend on the signal-to-noise ratio of those source detections and the related uncertainty in the centroids, and on the accuracy of prior knowledge of the positions of some or all of the sources.  Uncertainty in relative positions between sources (source offsets) can be much lower than those associated with absolute pointing knowledge, for detections that can support this type of analysis.