Creating & mosaicking data cubes ******************************** * Each stack of reduced fibre spectra is resampled onto a 3D x-y-λ grid (with reference to the MDF extension) using ``gfcube``: .. code-block:: none gfcube csteqbpxprgS20140919S0059 fl_atm+ fl_flux+ fl_var+ fl_dq+ gfcube csteqbpxprgS20140919S0062 fl_atm+ fl_flux+ fl_var+ fl_dq+ You should normally enable all of the options. The atmospheric dispersion correction automatically straightens the spectra to about 0.1" or better. * To inspect the cubes, you can give any IRAF task a 1D or 2D image section, eg. to see a single image plane: .. code-block:: none display dcsteqbpxprgS20140919S0059[sci][*,*,650] 1 There are also various programs around that can display projections of a 3D cube directly, or you can load one from the DS9 GUI and step through the planes. * Attempt to measure any spatial offsets between the cubes and shift their World co-ordinate systems to reflect those differences (so that features will have the same World co-ordinates in each one). Such offsets can be due to intentional dithering, flexure or re-acquisition on different nights. Any spectral offsets should already have been dealt with during :ref:`wavecal` or :ref:`flexure`. In this particular example, there is no significant difference between the exposures and the following step actually makes things slightly worse for some reason, so can be left commented out: .. code-block:: none # pyfalign dcsteqbpxprgS20140919S0059,dcsteqbpxprgS20140919S0062 For nebulous sources, you should specify ``method=correlate``. If the registration doesn't work reliably, you can use any available method (eg. ``imexam``) to determine pixel offsets, multiply them by the pixel scale (usually 0.1) and subtract them from the header keywords CRVAL1 & CRVAL2. * Resample all the cubes onto the same spatial grid and co-add them: .. code-block:: none pyfmosaic dcsteqbpxprgS20140919S0059,dcsteqbpxprgS20140919S0062 dcsteqbpxprgS20140919S0059_add var+ To check the alignment, you can run this command with the ``separate+`` option, which resamples the cubes onto the same grid but in separate FITS extensions, instead of co-adding them (this can also be used to align them prior to more sophisticated co-addition with another program, such as ``imcombine``). .. figure:: std_cube.png :scale: 32% :align: center Image of the target star, from a single wavelength plane near the middle of the co-added data cube (left) and the spectrum from the central pixel (right).