You may derive the normalized cross-correlation function for two spectra and calculate its maximum by either Gaussian fitting or finite differences. The location of the maximum represents the apparent shift of spectrum 1 with respect to spectrum 2.
|W1,F1||Wavelengths and fluxes for first spectrum.|
|W2,F2||Wavelengths and fluxes for second spectrum.|
|BEGL,ENDL||Beginning and ending wavelengths (respectively) denoting the region for cross-correlation, in the same units as W1 and W2.|
|delta||The apparent shift between the spectra, in the same units as W1 and W2. If multiple Gaussian components are specified, delta will be set equal to the first center value.|
|vinc||To use velocity space instead of linear wavelengths, set this keyword equal to the incremental velocity spacing desired (3-10 km/sec is recommended for IUE high dispersion data).|
|ccf||If desired, the normalized cross-correlation function can be output via this keyword.|
When running this program you will be shown a graph of the two spectra. Then, 31 points over the interval ± 15 × delw (or ± 15 × vinc if a value was entered via the VINC keyword). The program will be able to find the maximum (by finding the point of zero slope) if the maximum is well defined. The range can be extended if you think that the maximum exists outside the searched range. You may also elect to fit the maxima (there may be more than one) with gaussian functions (using GAUSSFITS.)
If you wish to work in wavelength space, you might want to use log( ) rather than for high dispersion:
High dispersion data should be corrected for heliocentric velocity shifts (see IUEVEL). It is recommended that no more than half the wavelength interval covered by three orders be correlated at one time. IUESPEC will tell you how many orders it has extracted; IUEHI3 extracts three.
Reasonably good correlation between the spectra results in a well-defined maximum. However, if the spectra have strong slopes, the cross-correlation function can be distorted. You may wish to normalize the continuum (e.g., with NORM, before running CRSCOR.