September 12, 2016 at 3:50 pm #1362
What is your technique to correct for the effects of slit tilt in your spectra?
One way that I have used is to cross-correlate each row in the accompanying arc spectrum with its central row and apply the measured lags to the corresponding rows in the science frames using linear interpolation. Do you have an alternative method, or use the science frames directly? If so, could you describe it and why you prefer it?
Note: an example of the effect on spectra of slit tilt was presented at the South American Data Workshop in Brazil in October, 2011. The relevant images are on on slide 44 with an example of how to use the IRAF pipeline to correct for it: http://www.lna.br/SAGDWorkshop/Apres/SAGDW_RSchiavon_GMOS.pdfSeptember 12, 2016 at 4:20 pm #1363
This should be taken out in the reduction automatically? I have been planning (for some time) to try a method similar to what you describe to help constrain the wavelength calibration a bit more robustly than matching N spectra separately against a line list, but I believe the current procedure should nevertheless take tilts out when it works properly. Of course the differences between rows are not really linear shifts but will be close (maybe indistinguishable for MOS) and I haven’t developed this enough to be able to describe an optimal algorithm. I expect it would normally be best to stick to using the arc for tilts and just use the lower-S/N science frames for correcting any zero-point shift with respect to the arc (if taken in the day). Sometimes you do get more bright sky lines than arc lines at longer wavelengths though.
September 13, 2016 at 12:57 am #1365
- This reply was modified 3 years ago by jturner.
I use a method similar to what you described, but with several small tweaks to make the code robust. You can find the routine that I’m using, called “align_trace3.pro” (it’s in IDL) in the Red Flamingos pipeline for Flamingos 2 here:
(look in /pro/general_pro/)
It is really the same principle with GMOS, I’ve been using the same routine to reduce GMOS data.September 14, 2016 at 9:01 pm #1366
@jturner: Thanks for the reply. Yes, that’s true that is is taken out in the reduction. I was mostly asking to stimulate discussion about people’s practices, or for people who may be doing their own reductions.
So far I’ve found that the differences between rows in the arcs were linear in my data but then the effect was very small overall (< 1 pixel shift over all ~200 rows) and so my experience probably isn’t representative. I haven’t experimented with higher-order functions but I would be interested to see at what level of tilt the effects become non-linear.
I agree with using arcs for tilt and science spectra for correcting zeropoint shift – this is what I do as well.September 14, 2016 at 9:02 pm #1367
@jgagne: Thanks for sharing your code! This does look similar to what I did although with a lot more options. I’m curious – in what circumstance would you use the option to smooth? I can see how it would reduce noise but would it also reduce the significance of the cross-correlation peak?
September 15, 2016 at 11:01 pm #1369
- This reply was modified 3 years ago by Catherine Huitson.
Yes, where I would expect linear shifts not to work so well is for long slit data, since you can see strong line curvature on those scales that varies from one end of the spectrum to the other. Thanks for the numbers.September 20, 2016 at 9:50 pm #1375
Yes, that makes sense – our slits are only short. Thanks for the reply.
You must be logged in to reply to this topic.