Discussion of Example 6: DIRBE FITS Headers

These primary and binary table headers are for an all sky map (OBJECT keyword) from the (INSTRUME keyword) Diffuse Infrared Background Explorer (DIRBE) on the (TELESCOP keyword) Cosmic Background Explorer (COBE) from the Cosmology Data Analysis Center (ORIGIN keyword). The primary header provides information that applies to all HDUs in the file. There are no primary data (NAXIS keyword); the EXTEND keyword is set to true, advising the reader that extensions may be present, as indeed they are. Since the primary HDU has no data, the value of the BITPIX keyword does not matter, as long as it is one of the permitted values. This particular file was created on September 28, 1994 (DATE keyword), but the original map file was created July 10, 1994 (DATE-MAP keyword). The map coordinates are for equinox 2000 (EQUINOX keyword).

COBE specific keywords (COMMENT) show that the data cover a period from December 11, 1989 (DATE-BEG) through September 21, 1990 (DATE-END) and give the size of the region of space projected on to one pixel (PIXRESOL). DIRBE-specific keywords (COMMENT) identify the (PRODUCT), the (VERSION) of the data reduction software, the nominal (intended) wavelength (WAVE9), and provide information about choices made in processing the data (SOLELONG, APPVEC, and ZLREMOV). The blank cards used as separators have an explicit COMMENT, in contrast with the use of blank lines in Example 5.

The COMMENT following the OBJECT keyword illustrates one approach to the problem of a keyword/comment string that is longer than the 80-columns of a card image.

The binary table (XTENSION keyword) has 393216 rows (NAXIS2) each of which has seven fields (TFIELDS), which take up a total of 27 bytes per row (NAXIS1). There is no heap for variable length arrays (PCOUNT). BITPIX and GCOUNT have the required values as described in section 3.6.1. Like the primary HDU, the extension header uses a COMMENT in the keyword field for blank lines.

Before the table description are a number of keywords describing the time of the observation. The COMMENT lines outline the meaning of the keywords and their values and provide a reference (TIMVERSN) with more details. MJDREFI and MJDREFF give the ephemeris Modified Julian Date corresponding to the zero point of the time system, also specified by TIMESYS. TSTART and TSTOP give the time of the measurements in seconds (TIMEUNIT) after the zero point.

The keywords describing the table follow. Detailed explanatory comments are provided at the top and after the keywords for each field. All fields have a short title (TTYPEn) keyword of eight characters or fewer. The titles are mixed case, but FITS software should not attribute any significance to the case; the value is to make it easier for human readers to understand the file. The COMMENTs furnish the full field titles and provide detailed explanations. Other COMMENTs are used to format the table for easier examination by human readers. The creators of the file have chosen to include a TUNITn keyword for every field, with a blank value if there are no units for the field. To do so is not required but this choice provides a design with a consistent set of keywords for every field. TFORMn keywords give the data type of each field. The header keywords describe the following fields:

1.

Pixel_no is a 32-bit integer giving the pixel number on the map.

2.

PSubPos is an 8-bit unsigned integer giving the pixel subposition, as explained in the comments.

3.

Time is a double precision floating point number giving the average time for the observations of a pixel. The comments note that the value is not physically meaningful for this data product. Including this item allows the designers to produce a set of data products with a a common format, simplifying the reading process.

4.

Photomet is short for ``Photometry'' and is a single precision floating point number giving a weighted average spectral intensity in MegaJanskys/steradian. The comments provide a precise definition and describe the weighting.

5.

StdDev is a single precision floating point number giving the standard deviation of the mean in the previous field, in MegaJanskys/steradian.

6.

WtNumObs is a 16-bit integer giving the sum of the weekly weighted number of observations used in deriving the annual averages.

7.

SumNRecs is a 32-bit integer giving the total number of observations.

Each row has one unsigned integer, of one byte, one two-byte integer, two four byte-integers, two single precision floating point numbers (four bytes each) and one double precision floating point number (eight bytes), for a total of 27 bytes, equal to the value of the NAXIS1 keyword.

A description of the DIRBE data products is available at http://space.gsfc.nasa.gov/astro/cobe/dirbe_products.html.