Optimization of a Voigt profile

[TheInfamousYano]

A Voigt profile is achieved by fitting the convolution of a gaussian and a lorentzian to a set of data. Here is what I'm hoping to accomplish in LabView:

1. Acquire data (done)
2. Calculate Voigt profile based on guess coefficients (done)
3. Calculate least squares error from Voigt fit (done)
4. Minimize error by changing the guess coefficients and repeating steps 2 & 3 (NOT done)

I can't use the built-in LabView optimization VIs like Downhill-Simplex because they require an analytic formula. The convolution also prevents me from using a nonlinear fitting scheme like Levenburg-Marquardt. What can I do?

Thanks for your time

[altenbach]

Sure, the stock Lev-mar cannot be used because they calculate the function one point at a time. However, they can easily be modified to do operate on fuctions that return the entire array at once. Over the years, I have posted plenty of examples (e.g. 1DFit.llb posted here).

I have written much more complex fitting algoritms that also incorporate convolutions with multiple lineshapes. (For example to simulate experimental EPR spectra). You can contact me privately for a more detailed discussion (see my e-mail address on the FP of the quoted VI).

Also have a look at my pre-publication article  for an example (see e.g. figure 3). Here I determined excess Lorentzian broadening using Lev-Mar fitting of the comvolution.

Your problem should work equally well, I've done it. 🙂

Message Edited by altenbach on 07-29-2005 10:40 AM

[AlfredoT]

Would you mind posting your code so that I may take a look at it. I am actually trying to design a vi in order to do steps steps 2 and 3 from your list. I'd like to use something similar to yours, if you don't mind, so that I may not be recreating the wheel here. Thank you in advance.

[altenbach]

You replied to a post that is 4 years old and many things have changed since then. Try searching for newer posts on the topic.