Stage Reproducibility Correction

[John Donovan]

The latest version of PFE (11.2.2) now contains a stage reproducibility correction feature when combined with the Use Beam Deflection For Position option as seen here:



The idea being that when one has utilized the Digitize Image feature from the Automate! | Digitize window to digitize point acquisitions in more than one field of view, the stage needs to drive to each field of view before applying the beam deflection. Since no EPMA stage is perfectly reproducible, this feature drives the stage to each image center, retrieves the reference image utilized for the point digitization, acquires a new image based on the same imaging parameters as before, calculates the apparent image shift relative to the reference image and applies a correction using the image shift method. Then it utilizes beam deflection for each point digitized in that field of view.

Regarding the method details, our imaging programmer writes: "Image shift analysis employs Fast Fourier Transformed (FFT) based image correlation.  It finds the amount of shift of images based on the largest image correlation. If we use the usual spatial computation to find image correlation the running time would be very long. Using 2-dimensional FFT algorithm, we can cut off the computational speed thousand times faster while keeping the quality of the correctness of the image correlation."

I've tested the code on the SX100 and it appears to work (though in a very interesting manner which I will discuss in more detail once I understand it better!). It has not been tested on JEOL instruments, so it is possible that the image shift correction has the wrong polarity for the JEOL, but I do think it is correct as I now apply the image shift in a Cartesian orientation for both Cameca and JEOL instrument in micron units.

Please feel free to download and give it a try.

[John Donovan]

I promised some additional information on the instrument oddity I observed when testing the stage reproducibility correction on the SX100.

Basically I originally tried to test the stage reproducibility correction by apply a stage offset before the "current" image was acquired, which I hoped would induce a shift that the FFT imaging code would find, by comparing the "reference" or "base" image used in the beam deflection digitizing, with the "current" image acquired after the stage has ostensibly moved correctly to the image center.

However after manually moving the stage slightly to induce a "stage reproducibility" error, I observed  that the beam position was automatically deflected by the instrument to compensate for my manual stage movement. Apparently the instrument detected my manual efforts to induce a stage reproducibility error and it automatically compensated for it by deflecting the beam itself!  Note that the image shift parameters did not change for this internal adjustment so I am not sure how this instrument based correction is performed.

In a conversation with Karsten Goemann he suggested that I modify the stage "reference" position slightly to induce a pseudo stage error and indeed doing this, then allowed the new stage reproducibility correction code to perform as expected. 

Remember this stage reproducibility correction is designed for those situations in which the stage position is correct, but the sample position (for reasons of movement within the sample block or thermal drift, etc, or beam position drift for that matter!), is incorrect.

Attached below are two images from my SX100 on an MgO standard so I could see where the beam was deflected to.  And indeed the stage reproducibility correction feature worked excellently, placing the beam spot exactly where it was digitized on the reference image.

I should mention that this "internal" stage movement correction on the SX100 was only observed when the second image was acquired.  Just having the beam position deflected did not cause this interesting instrument behavior to occur. Maybe someone from Cameca can comment on this "internal" stage movement instrument compensation?