spectro-aras.com

Christian,

I noticed on your last spectrum of AX Per regions of enhanced noise in each order (shown magnified here)
I do not think this is real but is instead a problem with ISIS processing. I have been experiencing this same phenomenon quite often and have devised a workaround (described below).
I believe quite strongly that it has something to do with your optimal binning routine. Isis now provides a flag to turn optimal binning on or off but I don't think the flag is working on the object spectrum as I get exactly the same result if it is checked or unchecked. HOWEVER turning off optimal binning does remove the noise on the calculation of the flats and blaze functions.

- Workaround (and why I think the optimal binning is the culprit)
Simply divide the object images (and the offset and dark frames) by 10 in floating point format (sometimes only a division by 5 is needed) and reprocess The noise features go away. I suspect if you do the same with your latest AX Per spectrum you will see the same.

Here is an example of mine. The first plot processed without division shows the same sort of noise. In the second plot the images were all divided by 10 before processing. Noise gone!
I have had to rely on this workaround quite often but not always. We need a proper fix!

Tim

Tim, the origin of your problem is the spectrum sampling, less then 2 pixels per FWHL in your data (a typical Nyquist theorem violation).

If I consider your setup (spectrograph on the telescope, excellent optical performance, type of camera (pixel size)...) I recommend really to work in bin 1x1 mode.

See also and download the latest version of ISIS (I can process your spectrum with these version - many thanks for sending the data, very useful for me) :

http://www.astrosurf.com/buil/isis/eshe ... index.html

Christian