Problems with Gauss package
greensilverfish wrote on Aug 21, 2008
I am using gauss to smooth some spectra. I have a intel Mac running OSX, and the latest version of iraf. The parameters I have set are :
sigma = 2.2 Sigma of Gaussian along major axis of ellipse
(ratio = 1.) Ratio of sigma in y to x
(theta = 0.) Position angle of ellipse
(nsigma = 4.) Extent of Gaussian kernel in sigma
(bilinea= yes) Use bilinear approximation to Gaussian kernel
(boundar= nearest) Boundary (constant,nearest,reflect,wrap)
(constan= 0.) Constant for boundary extension
(mode = ql)
When I run gauss, lines are being added to my spectra. (possibly real lines are being divided up). If I run (on another computer and older version of iraf) with the same parameters set they get a correctly smoothed spectra. What is going on? How do I fix gauss on my computer? Is gauss calling some other function, that I need to reset?
Thanks
sigma = 2.2 Sigma of Gaussian along major axis of ellipse
(ratio = 1.) Ratio of sigma in y to x
(theta = 0.) Position angle of ellipse
(nsigma = 4.) Extent of Gaussian kernel in sigma
(bilinea= yes) Use bilinear approximation to Gaussian kernel
(boundar= nearest) Boundary (constant,nearest,reflect,wrap)
(constan= 0.) Constant for boundary extension
(mode = ql)
When I run gauss, lines are being added to my spectra. (possibly real lines are being divided up). If I run (on another computer and older version of iraf) with the same parameters set they get a correctly smoothed spectra. What is going on? How do I fix gauss on my computer? Is gauss calling some other function, that I need to reset?
Thanks
Mike Fitzpatrick wrote on Aug 21, 2008
Because the task uses the interpolator, my first suggestion is to be sure you're using the Jan14 patch release and not the original v2.14 binaries that contained a bug:
If the file dates from before Jan14 you'll need to update the binaries to pick up the change needed.
Otherwise, when you say 'lines' do you mean blank columns in the image or actual spectral-looking lines? GAUSS is not a script task so you should only need to unlearn that task to reset the params. If you could post a URL to the image (or put it in ftp://iraf.noao.edu/pub) I'll try to reproduce it.
-Mike
cl> dir bin$x_images.e long+If the file dates from before Jan14 you'll need to update the binaries to pick up the change needed.
Otherwise, when you say 'lines' do you mean blank columns in the image or actual spectral-looking lines? GAUSS is not a script task so you should only need to unlearn that task to reset the params. If you could post a URL to the image (or put it in ftp://iraf.noao.edu/pub) I'll try to reproduce it.
-Mike
greensilverfish wrote on Aug 21, 2008
By adding lines I mean actual spectral-looking lines. I have uploaded both the original image (serp1.norm.fits) and the smoothed image (serp.smooth.fits ) to the link you sent me.
Thanks
Thanks
Francisco Valdes wrote on Aug 21, 2008
This result is a consequence of two things: 1) using a 2D task which works in pixel space on 1D data stored in the 2D multispec format and 2) the spectra having different wavelength zero-points. What this means is that you are using a 2D gaussian which is convolving both along the dispersion but also across the multiple spectra which are stored as lines in a 2D image. The strong lines, when plotted in pixel space (the '$' in splot or the 'f' key in specplot, are in different places in different images lines of the multispec format so they end up leaking into the other spectra at different places and appearing like new spectral lines.
However, it is still fine to do what you are doing. The only change you need is to set the gauss.ratio parameter to 0. so the 2D gaussian is effectively 1D along rows. The other method would be to use splot's smoothing function, probably with the non-interactive version called bplot.
Yours,
Frank Valdes
However, it is still fine to do what you are doing. The only change you need is to set the gauss.ratio parameter to 0. so the 2D gaussian is effectively 1D along rows. The other method would be to use splot's smoothing function, probably with the non-interactive version called bplot.
Yours,
Frank Valdes
Last post on Aug 21, 2008