PSD of a measured data set in millimeters

Hi, I am new to MATLAB as well as to doing spectral analysis, so I apologise if this seems a silly question.

I have a data set of a surface measurement, i.e. an x-y-z table of numbers. x and y are measured in mm-steps, z gives the surface position in µm (in fact, the z-values are the residuals of the measured values compared with the theoretical values, so they are small and can also be negative). What I want to do now is to get a PSD estimation of this surface, but not as a function of frequency, but as a function of the wavelength. So the x-axis should show the wavelength in mm, just like the following plot as an example:

Now, I've started reading up about all sorts of spectral analysis methods in Matlab, and so far I have tried the following:

Hs=spectrum.periodogram;
Hpsd=psd(Hs,data,'NormalizedFrequency',true)

And also the Welch method:

Hs=spectrum.welch;
Hpsd=psd(Hs,data,'NormalizedFrequency',true)

So these two methods give me two completely different results. I understand that the Welch method is using some kind of windowing and overlapping, whereas the periodogram is directly the FT of the autocorrelation function, is that correct? In any case both results look very different, and I am wondering how to interpret these differences. Also, I need to convert the results such that I get the wavelengths, not the frequencies, of the surface deviations.

I want to know which wavelength errors in my data contribute most to the surface defects and therefore want a plot comparable to the one given above.

Can anyone help me there please? Thanks a lot.

[datamonkey]

The data has been measured in a 2mm grid, so we have a surface profile measured in 2mm steps in both x and y directions.