iraf.net

I am using fxcor to determine the radial velocities of double lined spectroscopic binary targets. Unfortunately, some of the spectra are quite noisy. I would like to use filtpars to filter out some of the noise, but am not sure what values to use for the parameters cuton and cutoff. I would like to eliminate signals smaller than one resolution element (roughly 2 pixels) and larger than half the spectrum (1024 pixels). I would appreciate guidance as to what wavenumbers these values may correspond to.Thanks,
Sandy Liss

Firstly, you should be aware that filtering the data is not an exact science, and over-filtering can affect the results (especially width calculations) you get. In general, the task is pretty tolerant of even low S/N spectra.That said, you can use the 'f' command to get into Fourier Mode, the default plot is a power spectrum of the data where you can place the cursor to use the 'i' key to get the corresponding period trend in the data. In general, the left side of the plot is long-period trends such as continuum, however very broad lines in your data also contribute on this side. The right side of the plot is the higher-frequency trends such as the pixel-to-pixel noise. For a clean, high S/N spectrum the power plots will fall off gradually and then flatten out, that flat part is the noise you can usually filter without problems but of course in the real world the power spectrum is rarely that clean and determining the 'flat part' is tricky.Without seeing your data I can't really recommend specific values or whether you need to filter at all. The cuton value is usually less than about 5 and the cutoff depends on the power spectrum. See the FILTPARS help page for an explanation of the filter functions available.