XCOR N1 N2 N3 N4 N5 SLO SHI TAPER FILE OPTION [F1] [F2] SET NAME -- Cross-correlates spectra. See below for more details. Parameters: N1 -- First object spectrum N2 -- Last object spectrum N3 -- First template spectrum N4 -- Last template spectrum N5 -- First slot for cross-correlations. They are stored as spectra with 5000 A corresponding to 0 km/s. SLO -- First shift in pixels to compute. SHI -- Last shift in pixels to compute. A search will be made for a REAL header parameter called 'XCOR_SHIFT' and if found SLO and SHI will be assumed relative to this. If an arc is available, it is assumed that 'XCOR_SHIFT' has units of 'km/s', otherwise pixels. This should help pick out the right peak in cases of low signal-to-noise especially when binary motion is significant. The value of XCOR_SHIFT is converted to the nearest equivalent integer pixel shift and then cross-correlation is carried out from SHIFT+SLO to SHIFT+SHI. The range SLO to SHI can then be reduced to an amount consistent with the need to bracket the maximum. TAPER - Fraction to taper at ends of spectra. This reduces end effects. FILE - ASCII file to store results (blank to ignore) OPTION - 1 -- Bayesian computation 2 -- Normal (but with weights) If OPTION=1 F1 -- Lower limit of veiling factor from prior information. Normally this will be set equal to 0. F2 -- Upper limit of veiling factor from prior information. If the spectra have been correctly normalised, F2 should in theory be no more than 1, however, it does correlate with Vsin i and the spectral type. Thus if you use a very broadened template, it may want to reach F > 1, and so there may well be reason to set F2 > 1. It is unlikely that F2 should ever be < 1. SET - Yes to set mask of pixels in object (not template) NAME - Name of header item to store radial velocity in. (uncertainty will go to 'NAME_err') Carries out standard computation of cross-correlation, interpolating over regions masked out of analysis. Maximum cross-correlation located by parabolic approx to three points at maximum. Purely statistical uncertainty computed. Option (1) Cross-correlation if stored is actually equal to (sum W S T)**2/(2 sum W T**2) where S is spectrum and T template plus a correction for the known range of the fraction of the spectrum between 0 and 1. (2) The same without the correction (3) Stores sum W S T in one block and then sum W T**2 in the next. If the spectra were normalised to have unity continuum which was then subtracted then these cross-correlations can be back-projected to form a `skew map'. The routine does not account for uncertainties in the template spectra. The Bayesian correction factor reduces probability for negative cross-correlation using prior knowledge that the template can only make up a factor f between 0 and 1 of the spectrum. Spectra and templates must be correctly scaled. Scaling is probably best done by first fitting a constant to a region of the spectra of both templates and targets and then dividing this through. Note that a spline or higher order poly is not desireable because it might change the relative line strengths at different wavelengths. Following normalisation then one should subtract a fit to the continuum. Another possible method is to apply a band-pass filter with 'bfilt', especially to filter out longer term variations. The velocities that xcor produces are only accurate if prior to using it you have rebinned both targets and templates onto an identical velocity scale with 'vbin'.Related commands: vbin , bfilt and back .
This command belongs to the classes: measurement .