MSI-T focus errors
Added by Cheri Hampton almost 14 years ago
Hi, after submitting targets we keep getting the same focus node errors, although data acquisition occurs anyway. The first is focus measurement failed:fit=10616. while the limit is 1000. The second is Process tartget failed: missing calibration matrix. The calibrations were just done in calibrations application. Where else do these calibrations need to be done?
Cheri
| focus_error_snap.png (1.26 MB) focus_error_snap.png |
Replies (5)
RE: MSI-T focus errors - Added by Anchi Cheng almost 14 years ago
First, if the correlation peak is visible and where Leginon displays the cross as the peak found by fitting is on top of the peak, and you feel that you can trust its result, go ahead and increase fit limit in the Focus Sequence settings so that it can include your larger value. Fit limit is not normalized for different pixel size, so our default might not be large enough for you.
Second, I see that your logging window says it received 5 targets. I hope it was not 5 focus targets :)
Third, the calibration error is harder to figure out without other information. If you compare Focus Settings dialog, and Presets Manager with the preset of the image the targets come from (normally hl) selected so that you can see if the move type in Focus node is calibrated for hl. From 2.1 onward, you will need image shift calibration for your Focus/Exposure preset mag as well for image shift to work. By doing so, the range in which image shift can be used is increased and more accurate when the mag difference between hl and fa/fc/en/ef are larger. If this is not the problem, you might need to give use more details on the your focus sequence in case you have asked the node to do a task that you didn't calibrate for.
Anchi
RE: MSI-T focus errors - Added by Bob Grassucci almost 14 years ago
First point: How can I view the individual images that are correlated so I can determine if the peak is reasonable or not?
The other 2 points were resolved.
RE: MSI-T focus errors - Added by Bob Grassucci almost 14 years ago
I suspect that the problem lies with the beam tilt/Defocus calibration. I do the calibrations for hl 5000X and Fc 80000X and get the image to move 10-20 percent and yet the peak is at an extreme corner. I am attaching screen captures the beam induced image shift is about half of a hole and tet the peak is at the corner. Any thoughts? Thanks.
Bob
The attachments are png files
RE: MSI-T focus errors - Added by Anchi Cheng almost 14 years ago
O.K. it took me the whole morning to write this new wiki page and you've found the button before I finished it :)
The correlation peak by intensity is fine at 1/10th of the image from the corner as I explain in there and judging from the calibration images you show, that is about half a hole. Now, when you say peak is at an extreme corner, did you mean that peak finder find its peak at extreme corner? If it is at the extreme corner, you may have an image gain correction problem that created an unusually high, and sharp, origin peak. Retake bright and dark images may help. You can also save an corrected image of an empty area with the preset used for your calibration (should be the same dimension, binning and exposure time as your new bright and dark images). Then you can Fourier transform the resulting image to see if there are still feature uncorrected in there.
These images you show are at 5000x? what is the binned pixel size? My documentation mentioned above give an estimate of what pixel size is needed for a given focusing accuracy.
Now, back to the original question about seeing large fit error but acquisition went o.k. Did you see that mainly in Z Focus node? If so, don't worry. The default sequence used in Z Focus node uses stage Z to correct defocus offset determined by beam tilt. For Leginon to do this, it set the defocus to eucentric focus first during the measurement so that the defocus measured is relative to the eucentric height. As a result, there is little contrast in the images it acquired during the measurement if the grid is already close to eucentric height. The failure of fitting keep the node from making the correction in moving the stage z, exactly what we want.
Anchi
RE: MSI-T focus errors - Added by Bob Grassucci almost 14 years ago
Hi Anchi,
Sorry I hit the button before you posted the explanation. As you can see from the image the peak is not in the extreme corner but near it as you show in your new documentation so probably there is no peak in the corrected image. The calibration shows an offset of 358,-292 pixels on a 512X512 image.
With a 8X binning will be a pixel size of 168 Angstroems. I'll study your new document to get a more complete understanding.
Yes the error was reported in the z-focus which is done at 5000X or hl mode so we should be good to go.
Your new documentation with the pictures is much appreciated. Thanks again.
Bob