iraf.net

[b:2a97495062]From: [/b:2a97495062]Frank Valdes <valdes@noao.edu ([email]valdes@noao.edu[/email])>
[b:2a97495062]Date: [/b:2a97495062]November 4, 2005 12:06:19 PM MST
[b:2a97495062]To: [/b:2a97495062]Paola Rodriguez <paola@astro.ufl.edu ([email]paola@astro.ufl.edu[/email])>
[b:2a97495062]Subject: [/b:2a97495062][b:2a97495062]Re: Question about spectra reduction[/b:2a97495062]
Hello Paola,
The effect you are seeing is caused by a combination of the hot pixel and the variance weighting.  If you use either variance weighting or clean then the algorithm creates a smoothed profile of the 2D spectrum and the extracted value is based on the fitting of the profile to the data at each column.  The fitting, which is basically the normalization of the observed data, is greatly affected by hot pixels.  Cleaning often helps to iteratively remove the hot pixel so it doesn't affect the normalization.  However, if the hot pixel is very strong then the iterations actually remove the good data in preference to the hot pixel because both the good and hot data are then seen as deviant from the profile normalized initially by all the data including the hot pixel.  Note that when this happens it causes the bad things you mention both with and without cleaning.
So, you need to either not use the variance extraction (which you already have as the second "band" of the extractions) or you need to use the saturation parameter to chop off the really hot pixels a priori before the normalization, etc.  I recommend you always use the saturation parameter for this data.  The benefit of the variance weighting is sometimes overrated but you can use it with the clipping of the hot pixels.  I also recommend the cleaning but only if you can minimize the effects of the hot pixels, again using the saturation parameter.
Yours,
Frank Valdes
On Oct 21, 2005, at 9:34 AM, Paola Rodriguez wrote:
[quote:2a97495062]Good morning,
 I am Paola Rodriguez Hidalgo, a graduate student from UF. I am experimenting problems when using the task "apall" to extract spectra from 2D images. Sometimes a weird "bump" shows up in the spectra. It happens sort of randomly, although I have the suspicion that it might be related to a hot pixel you can see in the 3 images at physical= (2048,280) or image=(1549,102). However, we cannot explain how a single point created such a huge and wide bump. Unexpected bumps have actually  shown up in other spectra too.
I am sending you:
- 3 different spectra extracted from 3 different images (apfBb.0041.0001,apfBb.0042.0001,apfBb.0043.0001)
- the 3 images (fBb,0041,fBb,0042,fBb,0043)
- a text file with the parameters I used by default when running "apall" (apallParam.txt)
- other 2 spectra extracted with different parameters (see below)
The data was taking on September 29th with the GoldCam at the 2.1 m telescope (Kitt Peak).
(Since I was running "apall" interactively, the final images had a trace order of 7, instead of 4.)
We had set the parameter "clean" off because, for previous experience in our group, clean on has created changes in the continuum flux or troughs in the spectrum. Just for this spectrum we have tried with clean=yes or setting the saturation to a number lower than the value of the hot pixel. Both have removed the bump (see apredo42b.0001 for the clean=yes and apredo42.0001 for saturation=6500). Still, we would like to understand the problem and be sure that this hot pixel is really the responsible of this bumps before attacking the problem.
Thank you very much,
Paola
[/quote:2a97495062]