Since a while we realise that the line-width (LWF#) of the peaks for the same experiment are different in version 2 or version 3. In version 2 you can choose the function but not in version 3. But it doesn’t seem to be related to this (see fig.). Which function is used in version 3?
If I manually measure the full width at half height (FWHM) it seems that apparently the LWF# in version 3 underestimates the FWHM.
Is this alright?
(all data in Hertz)
I think the default in Version 3 is to use a Gaussian fit. However, in the preferences (Peaks tab) you can switch between Gaussian and Parabolic fitting (Peak interpolation method) - and in fact the latest version (3.2.10) also includes a Lorentzian fit.
A couple of things to bear in mind: V3 has the option to view the Peak Symbols as half-heights - this is quite a nice way to check that the fitting is vaguely reasonable (sometimes, if there are overlap issues, then it isn’t!). Use PS to cycle through the different peak symbol types quickly, or else set it in the SpectrumDisplay Settings.
Also, peaks that are close together can be fitted together using RG (rather than RP which does individual fits).
As to how the fits are done: I don’t know to what extent the actual fitting code was changed between V2 and V3. I can check with @ejb. In V3 it uses a certain number of points around the peak position to do the fit (3 or 4 in each direction, I think) and so sometimes (I guess particularly for very broad peaks? Or if you have a high degree of point sampling?) you may find that the peak starting position affects the fit. I wonder whether we should make this a variable that users can set…
Vicky