Focussing fit limit question
Added by Anonymous over 17 years ago
Due to a suggestion of Anchi, I'm trying two autofocussing stages in the focus sequence (to compensate for using a lower beam tilt itself in order to allow using a smaller objective aperture) - an initial autofocus at the same magnification as my current exposure preset (40k x) and a subsequent autofocus at a higher magnification preset (100k x). I'm also hoping that doing the final autofocus at a higher magnification will allow zooming in on the sometimes small areas of carbon on the lacey grids that I use. This seems to work okay based on the observed disappearance of Thon rings in the manual focus check. (The initial test was done with a Quantifoil grid.) However, I often get warning messages in the Focus node about exceeding the focus fit limit (currently set at the default starting setting of 1000, I think). What is the fit limit exactly and what units is it in? Is this a serious problem? I would have to check but I think the value (whatever it is) that exceeds the fit limit is typically about 1500. As I don't know the units, etc I don't have an idea if the fit limit is chosen to be a lot larger than the usual okay expected fit or of the same order,
William
Replies (1)
- Added by Jim Pulokas over 17 years ago
The autofocus routine acquires images at different beam tilt values and then cross correlates those images to measure the apparent image shift resulting from the beam tilt.
There is an of equation that relates this image shift to a value of defocus. We plug our measurements into this equation and then do a least squares fit to solve the defocus (and optionally the astigmatism). We always get a solution to this equation, even if the measurement or calibration is completely wrong. One way we found to determine if the defocus measurement is reliable is by looking at the sum of the squared residuals from the least squares fit (the value that is minimized by the solution). We call this value the "fit" and our threshold is the "fit limit". If the fit is not good, or higher than our threshold, then we declare the defocus measurement to be invalid.
We have not actually taken the time to figure out how to interpret the units of the fit limit value. We normally just set that value based on experience of what values seem to indicate a failure in the autofocus measurement. This value may change based on the magnification or defocus or other factors. I know most people here use a value between 500 and 5000, depending on how things are behaving.