iraf.net

What does the "trace" part of apall do?In the manual I am using it says that the reason one needs to trace is the fact that the actual spectral line is not prfectlly parallel to the dispersion axis. That doesn't make much sense to me because when I look at my data, the spectral line is perfectly parallel to the dispersion axis. Does that mean for example that I don't need to "trace" my images?
Mathematically speaking what exactly is the tracing here?Thank you!

Does my question clear? Does it make sense?

Frank may have additional comments, however tracing is used to provide an optimal extraction of the spectra by fitting the data to match any tilt (i.e. dispersion not perfectly along rows) or curvature (e.g. introduced by the cross-disperser) in the data. If your data look pretty good already then a simple low-order trace might suffice and wouldn't hurt anything. Other instruments have quite noticeable curvature/tilt in the orders making this step more obviously needed. Hope this helps.

A simpler answer might be that the trace is what locates a spectrum in an image. If there are multiple spectra then there are multiple traces.As noted a spectrum is not always aligned with the columns/lines of an image. This is due to instrument alignment and also to optical distortion. The spectra will always be aligned with the dispersion but it is the dispersion that is not always exactly along an image axis. As a related point, it is possible for lines of constant dispersion to not be orthogonal with the dispersion which is another issue for spectral extraction and calibration.It is, of course, possible that a spectrum is aligned with the lines/columns of an image well enough for your purposes; though note that the trace location can also mean a subpixel position.As a note, most spectral analysis of 1D spectra extracted from 2D longslit/slitlet data goes through an extraction step (apall). However, there are features in the ONEDSPEC package that can plot or extract spectra based just on a set of columns.So now you have a longer exposition on your simple question Frank Valdes

Actually I still can't visualize cases when the trace step is needed and cases when the trace step is not needed.
Furthermore I was told the trace step in apall is only used when reducing data for stars not star clusters. So then if the clusters' data needs that alignment how is it going to be done without the trace step? That's what makes me uncertain about what the trace step in intended to do...

The simple answer is that anytime you extract a 1D spectrum from a 2D spectrum you need to trace. To do this you need something which has a profile with a shape that can be centroided. This applies to stars, galaxy cores, or even almost unresolved extragalactic globular clusters. I don't know what exactly your data is just from saying star cluster.I will be glad to continue this discussion but I will be gone next week on vacation.Regards,
Frank

The targets I observed are open star clusters that are spread over the field of observation.

OK, so when you put a slit across the cluster with one or more stars falling on the slit and disperse it across the detector you will get lines whose profile across the dispersion is that of the stars. Because the spectrograph will have lenses and the alignment of the disperser is never perfect, the lines will not be exactly along the lines or columns. By tracing, which means fitting the profile across the dispersion at a number of wavelength and then fitting a smooth curve to those points, you map out where the star spectrum falls on the detector. This trace may vary at different points on the slit. When you extract the light for the stars you sum across the profile whose position is defined by the trace.If you simply extracted by summing a set of lines or columns then as the profile moves a little across the dispersion you will not be getting the same fraction of light at all wavelengths which will cause the continuum or line indices to be relatively incorrect.The only time where the trace might be unimportant is if you use a very wide aperture to sum. The reason you don't want to do this, though, is that you then introduce a lot more noise than is necessary.Again, I am not clear if you mean the star clusters are extragalactic where they are not clearly resolved or nearby clusters where you resolve the individual stars. Since you say[quote:009c3f7dea]I was told the trace step in apall is only used when reducing data for stars not star clusters[/quote:009c3f7dea]you must be these are unresolved extragalactic clusters since otherwise you really are still taking spectra of individual stars. Note that if you are talking about unresolved star clusters then you can think of them as being like stars and the reasons for tracing still apply. Also, people take spectra of galaxies and trace the core so that when then use apertures offset by the core then they get the same distance from the core at every wavelength which would not be true if the tracing was not done.I hope I've explained this clearly enough.Frank Valdes

Yes, thank you so much.
You are right, I observed extragalactic star clusters. Could it be that the reason for not tracing is that while observing we were scanning the slit through the cluster? Does that matter?Thank you again for all the help!

I assume this means moving the clusters along the slit to broaden it. I suppose it could be the other way to produce some kind of composite spectrum. In either case it is still the case that you should trace. Note that there is a parameter "width" in apall that specifies a centering width. If the shape of the profile, that is the light as you cut across the spectrum at some wavelength, is broad or doesn't have a "peaky" shape you need to set the width to the full width of the profile so that is centers the whole flux rather than noise spikes.Frank Valdes

Thank you again for your patience in answering my questions, I have one more question if you don't mind (I want to make sure I am doing the right thing with my data reduction).
When I trace a standard star interactively I get a nice uniform looking "trace". But for the cluster when I do the trace interactively the points are all over the place. Is that Ok, or I should use higher orders for the function > 5 to get a better "fit"?