iraf.net

It is expected that,the RMS value in identify task should be somewhere around 0.01 for a good identification of reference arc. In identifying my arc ,i tried various things to minimize the RMS value but seems impossible to reduce it to around 0.01.can anybody tell me what would be the variables of interest in minimizing the RMS ?

Hello Amal,There is no particular RMS value that you should get. It depends on the resolution of the spectrum and the accuracy of the line list. Also it is not correct to simply increase the order of the fitting function just to minimize the RMS because then you can just be fitting the normal centering errors by reducing the degrees of freedom.The things that affect the accuracy of a dispersion fit are:- the resolution of the spectrograph
- the signal-to-noise of the arc spectrum
- the accuracy of the line list
- misidentifications
- use of many faint arc lines (related to S/N point)In your separate email, have a significant position error for a good line would most likely be that misidentified lines were used in the fit.Do dispersion solutions can be frustrating and sometimes a bit of an art. The best thing to do is look at the fit and see if the residuals appear random. At that point do not simply increase the order of the fit to reduce the residuals. However, you could delete the most deviant points which may be misidentifications or poor choices for fitting. And, of course, ultimately look at the wavelengths of the stronger lines to make sure you don't have a wrong solution, the wavelength should be within a fraction of the resolution (the arcseconds/pixel).I hope this helps.Frank Valdes

dear frank,
i have carried out calibrating the spectrums and the problem still persists.for higher wavelengths the values agree with theoretical results but moving towards lesser wavelengths,the deviation is around 3 to 5 angstroms.in my reference spectrums, peaks below 3800 A are unrecognizable though the spectrum exists for around 3650 A.my objective is to observe the balmer discontinuity point which occurs around 3650 A.however due to unrecognizability,the values below 3800 are extrapolated when identifying. could this be the reason for the deviation of values below 3800 A?the coordinate list at the place of research doesn't contain sufficient values.therefore i made a list by observing Fe I and Ne I emission wavelengths
with finer values.is something wrong with that?finally,i want to know whether IRAF identifies peaks and assigns nearby wavelength values when we mark some lines?it would be of great help if you could assist me in sorting out the above issues.
thank you.

[quote:4e7d98c141]
however due to unrecognizability,the values below 3800 are extrapolated when identifying. could this be the reason for the deviation of values below 3800 A?
[/quote:4e7d98c141]Yes, extrapolation of dispersion solution can quickly depart from the true wavelengths. How fast and how far depends on the order of the fitting function. So I do think this is your problem.[quote:4e7d98c141]
finally,i want to know whether IRAF identifies peaks and assigns nearby wavelength values when we mark some lines?
[/quote:4e7d98c141]Yes when you mark a line with 'm' and there is a dispersion function and a line list then to help you it assigns a wavelength that is sufficiently near to the predicted wavelength and then asks you whether to accept it. This can cause wrong wavelength assignments, particularly if all you have is weak lines and you don't have a reference spectrum as an identification map. Often one can allow some wrong assignments and delete them as bad outliers in the fitting. But if you don't get rid of wrong identifications then the solutions will not be very good.[quote:4e7d98c141]
the coordinate list at the place of research doesn't contain sufficient values.therefore i made a list by observing Fe I and Ne I emission wavelengths
with finer values.is something wrong with that?
[/quote:4e7d98c141]This should be fine.Frank

thanks a lot for the help.....