| Re: wavelength calibration [message #69939] |
Tue, 23 February 2010 09:07  |
ameigs
Messages: 12
Registered: March 2009
|
Junior Member |
|
|
On 20 Feb, 05:38, sid <gunvicsi...@gmail.com> wrote:
> Hi,
> Please give some tips and suggesions on wavelength calibration for
> spectral data
> regards
> sid
Where is the spectra from? What type of instrument and do you have
measurements of the dispersion as a function of instrument wavelength
(and probably how it varies across your detector for a given setting)
or a theoretical prediction for the dispersion of the instrument.
Steps:
1) Measure the dispersion accurately; ideally across the entire range
for which you wish to use the instrument. In the visible (assuming),
using various light sources (Ne, He, Hg, Ar, Xe pen-ray lamps are
convenient), take as many different wavelength settings as you can
that have at least 1 pair of lines from the lamps (more pairs equals
better result for the variation across the detector at a given
setting). Fit the line pairs (Gaussian, centroid, whatever, to
determine the line centers on the detector in "pixel" space. Then the
sometimes hard part; identify these lines. Once you are reasonably
sure of the identifications, plot wavelength of the lines (from the
wavelength table you id'ed the line with) versus "pixel" for a
particular instrument setting (wavelength). If this is a straight line
(and you have a few pairs) then you're lucky and have a non-varying
dispersion across the detector for that setting. Assuming near linear
curve at all settings, then plot the slope of the line (nm/pixel) of
each setting versus the instrumental wavelength setting. This will
give you a good central (ie center of the detector) dispersion versus
instrument setting curve which you can then either use appropriate
interpolation, polynomial/spline fitting or best, fit the theoretical
dispersion formula to this (nice for Czerney-Turner spectrometers).
With this curve, then you can go on to step two.
2) Apply the dispersion along with the central dispersion to your
pixel data ( wave[p] = wave0[p0] + (p-p0)*dispersion[wave0] ). This is
your first guess. Now try to match up your line peaks to the
wavelength tables (NIST, etc). p0 is the center pixel of your
detector, wave0 is the instrument setting, dispersion[wave0] is the
dispersion from you nice curve from 1 at wave0. You may find that you
need to adjust p0.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
