A more general facility is also provided to associate specified FITS extensions with NDF components by means of entries in a file (see the EXTABLE parameter).

Details of the supported special formats and rules for processing them are given in topic "Special Formats"; the general-case processing rules are described in the "Notes".

FITS2NDF can also process both external and internal compressed FITS files. The external compression applies to the whole file and FITS2NDF recognises gzip (.gz) and UNIX compress (.Z) formats. Internal compressions are where a large image is tiled and each tile is compressed. The supported formats are Rice, the IRAF PLIO, and GZIP.

Both NDF and FITS use the term extension, and they mean different things. Thus to avoid confusion in the descriptions below, the term ‘sub-file’ is used to refer to a FITS IMAGE, TABLE or BINTABLE Header and Data Unit (HDU).

CONTAINER = _LOGICAL (Read)

ENCODINGS = LITERAL (Read)

EXTABLE = FILE (Read)

FMTCNV = LITERAL (Read)

IN = LITERAL (Read)

OUT = LITERAL (Write)

PROEXTS = _LOGICAL (Read)

PROFITS = _LOGICAL (Read)

TYPE = LITERAL (Read)

WCSATTRS = LITERAL (Read)

WCSCOMP = LITERAL (Read)

fits2ndf 256.fit f256 fmtcnv=f

fits2ndf 256.fit f256 fmtcnv=native type=_real

fits2ndf 256.fit f256 fmtcnv=t type=_real wcscomp=axis

fits2ndf 256.fit f256 noprofits noproexts

fits2ndf "$\ast$.fit,p$\ast$.fits" $\ast$

fits2ndf swp25000.mxlo mxlo25000

fits2ndf SWP19966.MXHI mxhi19966

fits2ndf data/$\ast$.silo silo$\ast$$|$swp$|$$|$ noprofits

fits2ndf "abc.fit,def.fts" "fgh,ijk" fmtcnv="F,T" noproexts

fits2ndf 256.fit f256 fmtcnv=f encodings=DSS

fits2ndf 256.fit f256 fmtcnv=f encodings="DSS,native"

fits2ndf "multifile.fit[extname=im3]" *

fits2ndf multifile.fit multifile extable=table1

Two other special cases are when a particular sub-file is specified by the IN parameter and when conversion is driven by an EXTABLE file.

The general rules for the conversion apply if the FITS file is not one of the "Special Formats" (including one defined by an EXTABLE) and Parameter CONTAINER is not TRUE.

The general rules are as follows:

• The primary data array of the FITS file becomes the NDF’s data array. There is an option using Parameter FMTCNV to convert integer data to floating point using the values of FITS keywords BSCALE and BZERO.
• Any integer array elements with value equal to the FITS keyword BLANK become bad values in the NDF data array. Likewise any floating-point data set to an IEEE not-a-number value also become bad values in the NDF’s data array. The BAD_PIXEL flag is set appropriately.
• NDF quality and variance arrays are not created.
• A verbatim copy of the FITS primary header is placed in the NDF’s FITS extension when Parameter PROFITS is TRUE.
• Here are details of the processing of standard items from the the FITS header, listed by FITS keyword.
  • CRVALn, CDELTn, CRPIXn, CTYPEn, CUNITn — define the NDF’s WCS and/or AXIS components (see Parameters ENCODINGS and WCSCOMP).
  • OBJECT, LABEL, BUNIT — if present are equated to the NDF’s TITLE, LABEL, and UNITS components respectively.
  • LBOUNDn — if present, this specifies the pixel origin for the $n^{\mathrm{\text{th}}}$ dimension.
• Additional sub-files within the FITS files are converted into extensions within the NDF if Parameter PROEXTS is TRUE. These extensions are named FITS_EXT_m for the $m^{\mathrm{\text{th}}}$ sub-file.
• An IMAGE sub-file is treated like the primary data array, and follows the rules give above. However, the resultant NDF is an extension of the main NDF.
• A BINTABLE or TABLE sub-file is converted into a structure of type TABLE. This has a NROWS component specifying the number of rows, and a COLUMNS structure containing a series of further structures, each of which takes its name from the label of the corresponding column in the FITS table. If there is no label for the $n$th column, the structure is called COLUMN$n$. These COLUMN structures contain a column of table data values in component DATA, preserving the original data type; and optional UNITS and COMMENT components which specify the column’s units and the meaning of the column. Thus for example, for the third sub-file of NDF called ABC, the data for column called RA would be located in ABC.MORE.FITS_EXT_3.COLUMNS.RA.DATA.
• A random-group FITS file creates an NDF for each group. As they are related observations the series of NDFs are stored in a single HDS container file whose name is still given by parameter OUT. Each group NDF has component name FITS_Gn, where n is the group number.

  Each group NDF contains the full header in the FITS extension, appended by the set of group parameters. The group parameters are evaluated using their scales and offsets, and made to look like FITS cards. The keywords of these FITS cards are derived from the values of PTYPEm in the main header, where m is the number of the group parameter.

• You can supply compressed FITS files, such as the Rice compression.
• NDF history recording is enabled in the output NDF.

• NDF2FITS
  • This is recognised by the presence of an HDUCLAS1 keyword set to ’NDF’. The conversion is similar to the general case, except the processing of FITS sub-files and HISTORY headers.
  • An IMAGE sub-file converts to an NDF VARIANCE component, provided the keyword HDUCLAS2 is present and has a value that is either ’VARIANCE’ or ’ERROR’.
  • An IMAGE sub-file converts to an NDF QUALITY component, provided the keyword HDUCLAS2 is present and has value ’QUALITY’.
  • FITS ASCII and binary tables become NDF extensions, however, the original structure path and data type are restored using the values of the EXTNAME and EXTTYPE keywords respectively. An extension may be an array of structures, the shape being stored in the EXTSHAPE keyword. The shapes of multi-dimensional arrays within the extensions are also restored.
  • HISTORY cards in a special format created by NDF2FITS are converted back into NDF history records. This will only work provided the HISTORY headers have not been tampered. Such headers are not transferred to the FITS airlock, when PROFITS=TRUE.
  • Any SMURF package’s ancillary IMAGE sub-files are restored to a SMURF extension, with the original names and structure contents. Thus the global HISTORY present in each sub-file is not duplicated in each SMURF-extension NDF.

• IUE Final Archive LILO, LIHI, SILO, SIHI

  • This converts an IUE LI or SI product stored as a FITS primary data array and IMAGE extension containing the quality into an NDF. Other FITS headers are used to create AXIS structures (SI products only), and character components.
  • Details of the conversion are:
    • The primary data array of the FITS file becomes NDF main data array. The value of Parameter FMTCNV controls whether keywords BSCALE and BZERO are applied to scale the data; FMTCNV along with the number of significant characters in the keywords decide the data type of the array. It is expected that this will be _REAL if FMTCNV is TRUE, and _WORD otherwise.
    • The quality array comes from the IMAGE extension of the FITS file. The twos complement values are divided by $-$128 to obtain the most-significant 8 bits of the 14 in use. There is no check that the dimension and axis-defining FITS headers in this extension match those of the main data array. The standard indicates that they will be the same.
    • The FILENAME header value becomes the NDF’s TITLE component.
    • The BUNIT header value becomes the NDF’s UNITS component.
    • The CDELTn, CRPIXn, and CRVALn define the axis centres. CTYPEn defines the axis labels. Axis information is only available for the SILO and SIHI products.
    • The primary headers may be written to the NDF’s FITS extension when Parameter PROFITS is TRUE.
• IUE Final Archive MXLO
  • This will usually be a single 1-dimensional NDF, however, if the binary table contains two rows, a pair of NDFs are stored in a single HDS container file whose name is specified by parameter OUT. The name of each NDF is either SMALL or LARGE depending on the size of the aperture used. Thus for OUT=ABC, the small-aperture observation will be in an NDF called ABC.SMALL.
  • Only the most-significant 8 bits of the quality flags are transferred to the NDF.
  • The primary headers may be written to the standard FITS airlock extension when PROFITS is TRUE.
  • The conversion from binary-table columns and headers to NDF objects is as follows:
    

    NPOINTS	Number of elements
    WAVELENGTH	Start wavelength, axis label and units
    DELTAW	Incremental wavelength
    FLUX	Data array, label, units, bad-pixel flag
    SIGMA	Data-error array
    QUALITY	Quality array
    remaining columns	Component in IUE_MX extension (NET and BACKGROUND are NDFs)

• IUE Final Archive MXHI
  • This creates a series of NDFs within a single HDS container file whose name is specified by Parameter OUT. Each NDF corresponds to a spectral order, and may be accessed individually. The name of each NDF is ORDER followed by the spectral-order number. For instance, when OUT=SWP, the 85${}^{th}$-order spectrum will be in an NDF called SWP.ORDER85.
  • Only the most-significant 8 bits of the quality flags are transferred to the NDF.
  • The primary headers may be written to the standard FITS airlock extension when PROFITS is TRUE. To save space, this appears once in the NDF specified by Parameter OUT.
  • The conversion from binary-table columns and headers to NDF objects is as follows:
    

    NPOINTS	Number of non-zero elements
    WAVELENGTH	Start wavelength of the non-zero elements, label, and units
    STARTPIX	Lower bound of the non-zero elements
    DELTAW	Incremental wavelength
    ABS_CAL	Data array, label, and units
    QUALITY	Quality array
    remaining columns (except 14-17)	Component in IUE_MH extension (NOISE, NET, BACKGROUND, and RIPPLE are NDFs each comprising a data array, label, units and wavelength axis)

  • It may be possible to evaluate an approximate error array for the absolutely calibrated data (ABS_CAL), by multiplying the NOISE by the ratio ABS_CAL / NET for each element.

  • The Chebyshev coefficients, limits, and scale factor in columns 14 to 17 are omitted as the evaluated background fit is propagated in BACKGROUND.
• IUE INES reduced spectra
  • This generates a single 1-dimensional NDF.
  • Only the most-significant 8 bits of the quality flags are transferred to the NDF.
  • The primary headers may be written to the standard FITS airlock extension when PROFITS is TRUE.
  • The conversion from binary-table columns and headers to NDF objects is as follows:
    

    WAVELENGTH	Start wavelength, axis label and units
    FLUX	Data array, label, units, bad-pixel flag
    SIGMA	Data-error array
    QUALITY	Quality array

• ISO CAM auto-analysis (CMAP, CMOS)
  • The CAM auto-analysis FITS products have a binary table using the ‘Green Bank’ convention, where rows of the table represent a series of observations, and each row is equivalent to a normal simple header and data unit. Thus most of the columns have the same names as the standard FITS keywords.
  • If there is only one observation, a normal NDF is produced; if there are more than one, the HDS container file of the supplied NDF is used to store a series of NDFs—one for each observation—called OBSn, where n is the observation number. Each observation comprises three rows in the binary table corresponding to the flux, the r.m.s. errors, and the integration times.
  • The conversion from binary-table columns to NDF objects is as follows:
    

    ARRAY	Data, error, exposure arrays depending on the value of column TYPE
    BLANK	Data blank (i.e. undefined value)
    BUNIT	Data units
    BSCALE	Data scale factor
    BZERO	Data offset
    CDELTn	Pixel increment along axis n
    CRPIXn	Axis n reference pixel
    CRVALn	Axis n co-ordinate of reference pixel
    CTYPEn	Label for axis n
    CUNITn	Units for axis n
    NAXIS	Number of dimensions
    NAXISn	Dimension of axis n
    remaining columns	Keyword in FITS extension

    
    Some of these remaining columns overwrite the (global) values in the primary headers. The integration times are stored as an NDF within an extension called EXPOSURE.
    The creation of axis information and extensions does not occur for the error array, as these are already generated when the data-array row in the binary table is processed.
    The BITPIX column is ignored as the data type is determined through the use the TFORMn keyword and the value of FMTCNV in conjunction with the BSCALE and BZERO columns.
• ISO LWS auto-analysis (LWS AN)
  • The conversion from binary-table columns to NDF objects is as follows:
    

    LSANFLX	Data array, label, and units
    LSANFLXU	Data errors, hence variance
    LSANDET	Quality (bits 1 to 4)
    LSANSDIR	Quality (bit 5)
    LSANRPID	Axis centres, labels, and units x-y positions— dimensions 1 and 2)
    LSANSCNT	Axis centre, label, and unit (scan (count index—dimension 4)
    LSANWAV	Axis centre, label, and unit (wavelength—dimension 3)
    LSANWAVU	Axis errors (wavelength—dimension 3)
    LSANFILL	not copied
    remaining columns	column name in LWSAN extension

• ISO SWS auto-analysis (SWS AA)
  • The conversion from binary-table columns to NDF objects is as follows:
    

    SWAAWAVE	Axis centres, label, and units
    SWAAFLUX	Data array, label, and units
    SWAASTDV	Data errors, hence variance
    SWAADETN	Quality
    SWAARPID	not copied
    SWAASPAR	not copied
    remaining columns	column name in SWSAA extension

• AAO 2dF
  • The conversion is restricted to a 2dF archive FITS file created by task NDF2FITS. FITS2NDF restores the original NDF. It creates the 2dF FIBRES extension and its constituent structures, and NDF_CLASS extension. In addition the variance, axes, and HISTORY records are converted.
  • The HISTORY propagation assumes that the FITS HISTORY headers have not been tampered.
  • Details of the conversion are:
    • The primary data array becomes the NDF’s data array. Any NaN values present become bad values in the NDF.
    • The keywords CRVALn, CDELTn, CRPIXn, CTYPEn, CUNITn are used to create the NDF axis centres, labels, and units.
    • The OBJECT, LABEL, BUNIT keywords define the NDF’s TITLE, LABEL, and UNITS components respectively, if they are defined.
    • HISTORY cards in a special format created by NDF2FITS are converted back into NDF history records.
    • The NDF variance is derived from the data array of an IMAGE extension (usually the first), if present, provided the IMAGE extension headers have an HDUCLAS2 keyword whose value is either ’VARIANCE’ or ’ERROR’.
    • The NDF_CLASS extension within the NDF is filled using the a FITS binary-table extension whose EXTNAME keyword’s value is NDF_CLASS. Note: no error is reported if this extension does not exist within the FITS file.
    • The FIBRES extension is created from another FITS binary table whose EXTNAME keyword’s value is FIBRES. The OBJECT substructure’s component names, data types, and values are taken from the binary-table columns themselves, and the components of the FIELD substructure are extracted from recognised keywords in the binary-table’s header. Note: no error is reported if this extension does not exist within the FITS file.
    • A FITS extension in the NDF may be written to store the primary data unit’s headers when Parameter PROFITS is TRUE. This FITS airlock will not contain any NDF-style HISTORY records.

Bailey, J.A. 1997, 2dF Software Report 14, version 0.5.

NASA Office of Standards and Technology, 1994, A User’s Guide for the Flexible Image Transport System (FITS), version 3.1.

NASA Office of Standards and Technology, 1995, Definition of the Flexible Image Transport System (FITS), version 1.1.