[LONG] ECIDENTIFY: another dispersion relations ?

Petr Skoda wrote on Jun 21, 2002

I have been devoting considerable time to the echelle data reductions for
more than two years both in IRAF and MIDAS. Having this experience I
compared the same raw data reduced in MIDAS HEROS instrument pipeline and
IRAF dofoe and doecslit  packages. The order extraction  and background
subtraction seems to be working better in IRAF but the most difficult part
is the wavelength calibration.

The ecidentify approach using the bivariate Chebyshev and Legendre
polynomials  gives to much freedom for higher powers
(especially at the edges of orders) and lower power polynomials are not
fitting well the dependency.  So I would like to know if it is possible to
add another type of fit to the ecidentify (or rather where to modify the
source of the task).



In fact I would like to implement the fitting formula of Hensberge and
Verschuern (ESO Messanger 58,p51) in form of

lambda(x,m')=1/(1+c1*m') [a0+(a1+b1*m')*x+a2*x^2+a3*x^3]

where m'=offset-m    is the relative order (the same as aparture)
and a0,a1,a2,a3,c1 coefficients where c1 is about 0.01

I see (by overplotting echelle overlap regions) from the HEROS package
that this fit is very accurate for several typical echelle spectrographs
like FEROS, HEROS, CASPEC and it is enough to identify only about 10 lines
all over the chip. With the same lines the IRAF ecidentify is completely
mislead even with the higher values of xord and yord - and more than 80
lines must be used in IRAF to give simillar precission. Moreover, the IRAF
must begin with identification of several lines in each order but for this
global fit is better to use one line in about each  5-th order.


I have seen in release notes for IRAF 2.12 this sentence :

IMRED.ECHELLE.ECIDENTIFY

    * Help page describes how to externally evaluate the dispersion fcns.

but I cannot find in the help the  how to evaluate the disp. fit
externaly - the only improvement is in detailed description of the
algorithm but I would expect some recipe how to replace the default
fitting (chebyshev legendre) by some user defined function.

Would it be possible to do this ?

Could you describe here your experince with using ecidentify on you
particullar data ? I am afraid the cookbook "A user Guide to Reducing
Echelle Spectra With IRAF" is probably not fully  applicable to the modern
echelle (even fiber-based) spectrographs with typically 50 or more orders.
Especially the wavelength calibration is quite tricky and magic task and I
have spent a lot of time to get good estimates before the automatic
searching of lines (by command "l") can really find all lines well.

Can somebody comment on this ?



*************************************************************************
*  Petr Skoda                         Phone : +420-204-649201, ext. 361 *
*  Stellar Department                         +420-204-620361,620136    *
*  Astronomical Institute CAS         Fax   : +420-204-620250           *
*  251 65 Ondrejov                    e-mail: skoda@sunstel.asu.cas.cz  *
*  Czech Republic                             pskoda@mbox.cesnet.cz     *
*************************************************************************

---

Last post on Jun 21, 2002