Does anyone know how LabView determines standard deviation on their Gaussian Peak Fit VI?

[LaserPro2010]

The VI reports standard deviation that is quite high. I am wondering if they go out to 5-sigma or something along those lines.

If anyone has had this same issue, please let me know

Thanks in advance!

[altenbach]

Did you graph the data and best fit on the same graph? Do they agree?

I just played a little bit with it and, in my limited experience, the gaussian fitting VI is often not very good (especially for data with small values). I probably would use levenberg marquardt instead.

Can you attach your VI containing some typical data?

[LaserPro2010]

Interesting you mention the robustness of the algorithm. I have heard in the past that it is not very good. In fact, the data you feed into it must be VERY Gaussian otherwise the fit is just garbage. Luckily for me, my data is quite clean, see the attached picture.

You also mention Levenberg-Marquardt. Is a VI available somewhere or would I need to write my own?

Thanks for your prompt reply!

[altenbach]

What is your LabVIEW version?

Do you give any initial estimates?

You can use "nonlinear curve fit", but you need to make a model subVI, which is quite trivial. You also need to provide initial estimates.

[LaserPro2010]

LabVIEW 2010 SP1

I have not added initial estimates; working on that right now.

Let me just mention that for the sample data I attached (which is very nicely fitted) I get a standard deviation of ~4.5, but by simple examination of the graph, you can see that the profile width is ~1.9. However, I am interested in finding the deviation from the center at approximately 13% of the maximum amplitude (common in my line of work). My problem is that I don't know how LabVIEW decides how far to go out from the center to report the standard deviation. They may go out to 5% or even 1%.

[altenbach]

Make sure your x-value array is scaled correctly.

The definition of stddev is not open to interpretation. The formula is given on the help page and seems correct.

[altenbach]

You did no attach any sample data, just an image. Can you attach an actual data file?