The following keywords are used (*i* is the number of the axis):

- NAXIS
*i* - the number of pixels along the axis
- CTYPE
*i* - the name of the axis. The axis name is uniquely related to the physical coordinates of the pixels along the axis. This name consists of several parts, separated by hyphens. The first part denotes the real axis name: the physical coordinate system. The second part is the type of projection (meaning that there are more ways to go from grid position to physical coordinates). At the moment only projections for spatial axes (like RA and DEC) and spectral axes (FREQ) are defined.
- CRPIX
*i* - the location of the reference pixel on the axis in file coordinates.
The grid position of the reference pixel is always zero.
For this position the physical coordinate is precisely defined in
CRVAL
*i*, and therefore we call this position the projection centre. - CRVAL
*i* - holds the physical coordinate at the reference pixel location in
some units. The units are defined in the non-standard FITS
descriptor CUNIT
*i*. - CUNIT
*i* - holds the units of the physical coordinate at the reference pixel
location, i.e. the units of CRVAL
*i*. When this descriptor item is not defined, the standard FITS units will be assumed. - CDELT
*i* - holds the pixel separation in units specified in CUNIT
*i*along the axis. Note that CDELT*i*may be negative. - CROTA
*i* - holds the rotation angle in degrees of the
actual axis from the stated coordinate type in CTYPE
*i*. This descriptor makes only sense for spatial axes. In GIPSY it is assumed that the spatial latitude axis carries the rotation information.

When writing, the items NAXIS*i*, CTYPE*i* and CRPIX*i*
may **never** be modified using GDS write routines.
This may only be done using the routine
GDS_EXTEND.
This is because this information is referenced so frequently that it
has to be maintained separately for efficiency reasons.

- Equatorial sky coordinates: RA, DEC
- Galactic sky coordinates: GLON, GLAT
- Ecliptic sky coordinates: ELON, ELAT
- Supergalactic sky coordinates: SLON, SLAT
- Spectral frequency coordinates, equidistant in frequency: FREQ
- Spectral velocity coordinates, equidistant in velocity: VELO

- Equatorial (1950.0)
- Galactic
- Ecliptic (1950.0)
- Supergalactic
- Equatorial (2000.0)

For sky systems the rotation angle of the sky plane with
respect to the axes is stored in CROTA*n*, where *n* is the number
of the latitude axis. The epoch of the coordinate system in stored in the FITS
descriptor EPOCH.

There following **projection systems** are implemented:

- Aitoff Equal Area projection, AIT
- Equivalent Cylindrical projection, CYL
- Flat projection, FLT
- Gnomonic projection, TAN
- Orthographic projection, SIN
- Rectangular projection, ARC
- Transversal projection, GLS
- North Celestial Pole projection, NCP
- Stereographic projection, STG
- Mercator projection, MER

V(radio) = DRVAL*i* - *g*.CDELT*i*.c / f0

V(optical) = DRVAL*i* + (1 / (CRVAL*i* + g.CDELT*i*) -
1 / CRVAL*i*).f0.c

Where c is the speed of light and f0 the rest frequency of the spectral line. GIPSY uses for the grid separation in velocity the following:

DDELT*i*(radio) = -CDELT*i*.c / f0

DDELT*i*(optical) ~= -CDELT*i*. c/f0 . (f0)^2/(CRVAL*i*)^2

The rest frequency is stored in descriptor FREQ0 in units of Hz.

The different **velocity systems** are encoded in the following
way:

VELSYS CTYPEi Meaning ------ --------- ----------------------------- 1 FREQ-OHEL Optical definition w.r.t. SUN 1 FREQ-OLSR Optical definition w.r.t. LSR 2 FREQ-RHEL Radio definition, w.r.t. SUN 2 FREQ-RLSR Radio definition, w.r.t. LSRThe function VELPRO does the velocity - grid transformations.

- Inverse Wavelength, INVLAM
- Wavelength, LAMBDA
- Logarithmic Wavelength, LOGLAM
- Parameter axis, PARAM
- Polarisation, POLN
- Time, TIME

*p* = CRVAL*i* + *g* . CDELT*i*

*g* = *p* - CRVAL*i* / CDELT*i*

For determining the axis type, the function
AXTYPE can be used.
Note that natural units are usually **S.I.** units.
In GIPSY the routines involved in coordinate transformations use these
units! In case other units are stated in CUNIT*i* or BUNIT*i* the
conversion factor can be obtained with function
FACTOR.