iraf.net

Dear iraf users and support, after fiddling a lot and reading the documentation for doecslit as well as for echelle data reduction, I still do not know how to retain control (ie, interactivity) of the comp *re*identification task within doecslit.I reduced a set of 3 object files, obj2, obj4 and obj6, which are
sandwitched between 4 comp files, comp1, comp3, comp5, and comp7.
I first extracted a long exposure arc taken at the beginning of the night,
comp0, and used ecidentify to identify lines and do the dispersion
solution. The RMS of the fit was 0.0033A as I expect from the resolution
of the echelle. I then used the long exposure comp as a template to
identify and fit arc lines in subsequent comps, within the task doecslit.
All went well, except that the last 2 comp files (comp5 and comp7) were
reidentified and the RMS was 0.049, which is about 10 times worse than I
want.I wanted to know whether I could get a better RMS from comp5 and comp7. I
therefore extracted these two arcs with apall (creating files comp5.ec.fits and comp7.ec.fits) and did the dispersion solution manually, using the following command:> ecid comp0.ec,comp5.ec,comp7.ecThe first comp to come up, comp0.ec, was previously identified (the file
eccomp7.ec is in directory database) and therefore many arc lines are
already identified. I fit (RMS=0.033A), write and quit, and automatically
get comp5.ec on the graphic display, with the same identified lines. I do
an 'a', 'c' (for center all lines) and I then fit again. The RMS of
comp5.ec is calculated as 0.049A, but I see a few lines which have a
larger than average residual. I delete those lines by hand and refit. Now the RMS is 0.0035A. The same happens for comp7. So there is nothing inherently wrong with comp5 and comp7, but I NEED to be able to go into the fitting
interactively during the doecslit routine to delete those lines. Right
now, doecslit uses ecreidentify and does not let me do the fitting
interactively.I tried to change the ecreid paramters to make sure that the bad lines did
not get picked up, but it did not work. I also tried to identify and fit
the two arcs comp5 and comp7 by hand before doing doecslit, but the
problem is that the names do not match (to do it by hand I do an apall and
call comp7.fits, comp7.ec.fits with the corresponding solution in
database called eccomp7.ec, but the doecslit routine seems to extract the
comps giving them a name like obj6comp7.ec with corresponding wavelgnth
solution in database called ecobj6comp7.ec.)So, I am not sure how to gain some control to improve the wavelength
solution of comp5 and comp7 beyond what doecslit does for you using
ecreident.Just before posting this, I managed to improve the performance of doecslit
ecreidentify by altering the paramter sparams.cradius (NOTE I had NOT been
able to improve the performce of the reidentification by altering the
ecreidentify.cradius paramter!). However I only managed to get comp7
right, but comp5 is still fit with RMS about 10 times worse than it
should.Do you know how I can gain interactivity of the reidentification and
fitting process while in doecslit?Thanks! Orsola

---

Orsola De Marco, Ph.D. Astrophysics Department American Museum of Natural History New York, NY 10024, USA 212 496 3444

Hi Orsola,You were clear in the description of the problem and concern. I wish I could be more helpful. What you describe is precisely why identify/reidentify had the feature added to be able to jump into the interactive fitting. But the echelle track based on ecid/ecreid did not get so updated. So there is no feature, other than possibly reworking the doecslit scripts in some way, to get the behavior you want. What you are doing by trying to find the parameter settings that might handle the misidentifications is a good, though limited, approach.As I periodically remind people, the DOxxx tasks are canned tools to integrate data reductions for common well-behaved data but if these can't handle specific problems then people will have to resort to doing reductions with the lower level tools such as you were doing. The other part of the answer is that at this point it is unlikely that there will be much new development on the ECHELLE package and the DOECSLIT task.I wish I could give you a more useful answer.Yours,
Frank Valdes

Hi Frank, thanks for the reply. You are right. The only work around is to use APALL to extract objects and comps and then use only ECIDENTIFY which has the interactive mode.Darryl Willmarth had told me to use DOECSLIT because it has the good feature that it extracts the comps using the aperture set of the nearest object. This is indeed an important feature for accurate wavelength calibration, but it can be mimiced by APALL by giving the appropriate reference object aperture file to the comp being extracted.So all is well.Thanks again, Orsola

---

Orsola De Marco, Ph.D. Astrophysics Department American Museum of Natural History New York, NY 10024, USA 212 496 3444