Leginon

This error is showing up because you have set up the Hole and Z Focus nodes to use modeled stage position as the move type, but it cannot locate the calibrations for this in the database. The tricky part is that you need to have this calibrated at the magnification you used for the square image, not the hole magnification, because square is where the targets were selected. It sounds like maybe you have already done a stage model calibration, but not for the square magnification. Since you already have a model, you only need to do the magnification adjustment calibration for the square magnification. In the current user manual, this is described in Chap 8. You can also test out your calibration in the Navigation node by first going to the square preset, acquiring an image, select "modeled stage position" as the move type and then click on the image to see if it will move to your clicked position.

sorry, I was "guest"

Maybe we can figure out how to prevent guest access...

I carried out the magnification adjustment calibration described in the manual. I retried the Square Targetting stage in MSI setup and things seem to get further after I set up the hole finder. During this I have to focus in the Z Focus node, I think. I get to Exposure Targetting and have to setup the hole finder for that. I am prompted to focus after that (which I do) - in Focus node I think. Eventually this gets stuck with Square node waiting for 719 to be processed, Hole node waiting for 728 to be processed and Exposure node taking moe than 15 minutes to acquire an image, eventually giving an error message "Process target failed: (104, 'Connection reset by peer'). "Busy wheels" still spin and also Exposure Node continues. I noticed that Leginon gave the following errors:

Exception in thread Thread-139:

Traceback (most recent call last):

File "/usr/lib/python2.3/threading.py", line 436, in __bootstrap

self.run()

File "/usr/lib/python2.3/threading.py", line 416, in run

self.__target(*self.__args, **self.__kwargs)

File "/usr/lib/python2.3/site-packages/Leginon/targetmaker.py", line 82, in calculateAtlas

self.panel.atlasCalculated()

File "/usr/lib/python2.3/site-packages/wx/_core.py", line 10548, in getattr

raise PyDeadObjectError(self.attrStr % self._name)

PyDeadObjectError: The C++ part of the Panel object has been deleted, attribute

access no longer allowed.

Also, I discovered the Leginon client crashed (or rather that it was no longer running for some reason, presumably a crash - I've reverted to starting the Leginon client in a DOS window, rather than from the shortcut, so I will still have a window with error messages if it crashes again). I'm not sure if the above problems are due to some temporary network problem or some other difficulty. I did notice that in the Focus node I get the error "Process target failed: defocus at HT: 200000 mag: 50000x". This may be a separate problem. I attempted to re-run after doing a Beam Tilt node focussing related calibration (although the calibration was done poorly). I think this re-run got stuck in a similar manner to what happened earlier. I'm not quite sure how to do the autofocus calibration with the Beam Tilt Node in the Calibration application (defocus and stigmator calibrations). The manual says in step 6 of the defocus calibration (and step 4 of stigmator) to go to the Image Viewer under the branch labeled "imv". I can't see a branch or anything labeled "imv" in the Leginon GUI. I tried using the web-based Image Viewer but that doesn't show the correct image (not even after pressing the acquire button in the Beam Tilt Node). So how do I get to the Image Viewer that I need to use with the Beam Tilt node calibrations?

William

"Anonymous" wrote: This error is showing up because you have set up the Hole and Z Focus nodes to use modeled stage position as the move type, but it cannot locate the calibrations for this in the database. The tricky part is that you need to have this calibrated at the magnification you used for the square image, not the hole magnification, because square is where the targets were selected. It sounds like maybe you have already done a stage model calibration, but not for the square magnification. Since you already have a model, you only need to do the magnification adjustment calibration for the square magnification. In the current user manual, this is described in Chap 8. You can also test out your calibration in the Navigation node by first going to the square preset, acquiring an image, select "modeled stage position" as the move type and then click on the image to see if it will move to your clicked position.

Sorry, my fault. The documentation refers to a node that we no longer use. I will make the correction. You will have to use manual focus function to do defocus check. Here is how:

1. Create a fc preset in Presets Manager if you have not done so.See the table for example preset for the configuration. The main thing is to have camera bin of 1 and a small area so that it is not too slow.

2. Check settings in Focus node, you need to make sure that "fc" is the preset used in the top portion of the node, i.e., the acquisition part.

2. After you "Correct Defocus" in Beam Tilt node, Click "Reset Defocus" in the same window.

3. Goto Presets Manager and send "fc" to scope. This will reset the camera so that it has bin of 1 and will set defocus to whatever the preset specified.

4. In Focus Node, click on the Toolbar where Manual Focus is (The icon says MF).

5. You will get a pop-up window that shows the power spectrum of the image taken at "fc" preset.

6. Manual Focus/Toolbar>check the number on the left of m. It should be 0. Then click "Set Instrument" to send that defocus to scope.

7. If the Thon rings in the power spectrum disappears, your calibration is good.

Note that the camera configuration is now that of fc. You will need to send the fa preset if you are to redo or continue Beam Tilt calibration.

You should not get the message as long as you have the complete autofocus calibration for the HT and mag you use for autofocusing. It does not matter whether the calibration is a good or bad one, as long as it is complete, i.e., has both defocus and stigmator calibration. If you can "measure" in Beam Tilt node, it should be completed.

You may want to start using Simulated Target tool for testing your focus node or to start a partial test of your MSI sequence. Just click on it where it is available. Leginon treats the event as if it recieved a target for the current position. I would suggest that you try out your Focus node, and then try Exposure node, and then try to start from Hole node. If none of them fail, you may have run into some event that is not your fault. We have seen that error before but has not seen that since before your workshop. Jim may have more input.

What are suitable values for the Beam Tilt node parameter settings (for higher magnifications)? The manual starts to say what the parameter settings are in step 2 of the section on defocus calibration for the Beam Tilt node; but the actual values are not there. (Later in step 7 of the section it says what the suitable values are for 5000x and other low magnifications and mentions that the two defocuses used in the calibration need to be higher and further apart for low magnifications),

William

The default values are for 50,000 x to 62k x.

beam tilt=0.01 (You should use the 100 um objective aperture for it). If

you use a smaller aperture, you should experiment on it. You would

pick the largest beam tilt that does not cause the tilt image to change

contrast ( an indication of beam being clipped by the aperture).

You will find the rest of parameters in the new on-line manual in the Calibrations chapter under "Autofocus Calibration with Beam Tilt Node"

"anchi" wrote: The default values are for 50,000 x to 62k x.
beam tilt=0.01 (You should use the 100 um objective aperture for it). If
you use a smaller aperture, you should experiment on it. You would
pick the largest beam tilt that does not cause the tilt image to change
contrast ( an indication of beam being clipped by the aperture).

You will find the rest of parameters in the new on-line manual in the Calibrations chapter under "Autofocus Calibration with Beam Tilt Node"

By the way, do you normally use the 100 um objective aperture (which is the no. 4 objective aperture, I think) on the F20 at Scripps? We don't seem to be achieving anything like the 0.1 um accuracy in the defocus when autofocussing that is suggested in one of talks about Leginon. We currently use the larger no. 4 objective aperture but I'm considering changing to the no. 3 objective aperture (40 um, I think) to improve contrast and hopefully give a stronger correlation signal for autofocussing to work with. The no. 3 objective aperture was not used in the past; because when the objective aperture is aligned in diffraction mode it was found to be poorly centred at low magnification in the sq preset so it may be only possible to get an image from the CCD of a small part of the area available through the objective aperture in the sq preset (although this does not seem to be so much of a problem any more - at least not at present). Do you know of any alignments that would get round this problem (so the objective aperture appears centred both in diffraction mode at high magnification and in "normal mode" - i.e. not in diffraction mode - at low magnification)?

Also, I'm wondering what the best way is to calibrate the Beam Tilt node - is working with a negative stain specimen and using smaller apertures where possible going to be suitable (in order to carry out the calibration with the higher contrast)? Or is it better to carry out the calibration with a setup to normal running conditions (cryo specimen with a larger objective aperture)?

William

The autofocus calibration for us can either be done with a negative stain grid or a cryo grid to get about 0.1 um accuracy at 50,000x. If your contrast is in general so low that false correlation peaks are found, you will more likely have failed autofocus (high fitting residual) than inaccurate measurements because the measurements over fit the equation. Therefore you need to first separate the two effects.

(1) failed autofocus- The normal cause is false correlation peaks. This happens more often if the contrast is low or your CCD image is noisy. When you display the correlation, does the peaks stand out well and correspond to the shifts of the images?

(2) inaccurate autofocus- There are many causes. One possibility is that the image shift induced by beam tilt at a given defocus is too small on the CCD pixel scale. This is why the accuracy at 5000x is much lower than at 50000x using the same camera configuration. The other common one is that the rotation center which defines the untilted beam is not truly untilted, either during the calibration or during the measurements. A third possibility is that the image shift applied during calibration and measurements are very different such as many microns. We also expect that the measurements are not as good if the defocus is off too much as it deviates from the linear assumption of the equation. In our setup, the error for the measurement is greater with greater defocus. Do you use "Z focus" at lower mag for encentering the grid at each square?

With a smaller objective apperture, you lose the ability of tilting the beam higher, therefore will reduce the accuracy of the measurement although you will indeed increase the contrast. You can always compensate it by going to higher magnification so that the shift is obvious but then you will lose on the range of defocus the measurement is accurate. Again you can get arround that problem by adding an autofocus step at lower magnification to bridget the gap, then you will lose some time. Anyway, the options are endless.

I suggest that you send us a few of your typical images. It would be useful to have fa, fc, and their parent images so that we can compare. The magnification, effective pixel size, and your focus sequence and autofocus settings will also be useful. In addition, send me an image taken under the condition you calibrate the autofocus in Beam Tilt node. In other words, at the same area, defocus, magnification, and camera configuration.

Anchi

"anchi" wrote: The autofocus calibration for us can either be done with a negative stain grid or a cryo grid to get about 0.1 um accuracy at 50,000x. If your contrast is in general so low that false correlation peaks are found, you will more likely have failed autofocus (high fitting residual) than inaccurate measurements because the measurements over fit the equation. Therefore you need to first separate the two effects.
(1) failed autofocus- The normal cause is false correlation peaks. This happens more often if the contrast is low or your CCD image is noisy. When you display the correlation, does the peaks stand out well and correspond to the shifts of the images?
(2) inaccurate autofocus- There are many causes. One possibility is that the image shift induced by beam tilt at a given defocus is too small on the CCD pixel scale. This is why the accuracy at 5000x is much lower than at 50000x using the same camera configuration. The other common one is that the rotation center which defines the untilted beam is not truly untilted, either during the calibration or during the measurements. A third possibility is that the image shift applied during calibration and measurements are very different such as many microns. We also expect that the measurements are not as good if the defocus is off too much as it deviates from the linear assumption of the equation. In our setup, the error for the measurement is greater with greater defocus. Do you use "Z focus" at lower mag for encentering the grid at each square?

With a smaller objective apperture, you lose the ability of tilting the beam higher, therefore will reduce the accuracy of the measurement although you will indeed increase the contrast. You can always compensate it by going to higher magnification so that the shift is obvious but then you will lose on the range of defocus the measurement is accurate. Again you can get arround that problem by adding an autofocus step at lower magnification to bridget the gap, then you will lose some time. Anyway, the options are endless.

I suggest that you send us a few of your typical images. It would be useful to have fa, fc, and their parent images so that we can compare. The magnification, effective pixel size, and your focus sequence and autofocus settings will also be useful. In addition, send me an image taken under the condition you calibrate the autofocus in Beam Tilt node. In other words, at the same area, defocus, magnification, and camera configuration.

Anchi

How would you like me to send the images? Should I just e-mail them to the Leginon support e-mail address?

William

That would be fine, or send it directly to me through e-mail. Try sending just one image first. The other option is put it on a website where Jim and I can pick it up.

"anchi" wrote: That would be fine, or send it directly to me through e-mail. Try sending just one image first. The other option is put it on a website where Jim and I can pick it up.

I have some images to send but not including the requested image in the Beam Tilt node. I'm not sure how to get the Beam Tilt node image as this would be done in the Calibration application which has nodes which can acquire images but no nodes which can manually acquire images and then save them in the data directory, as far as I know (unless it's possible to force the Focus node to do this somehow). I'm thinking about finding the area for use in the Beam Tilt node, then sending the preset used in the Beam Tilt, then quitting the Calibration application and acquiring the image of the area with the correct conditions in the Manual application, and afterwards returning to the Calibration application and carrying out the calibration in the Beam Tilt node. This will require sending the preset again, I think, after getting back to the Calibration application (because the application won't remember the preset it sent before it quit earlier). Is that going to be suitable or is there a more direct way to get the image taken under the conditions used to calibrate in the Beam Tilt node?

William

I don't need the beam tilt images, yet. Just the parameters you used. To get

leginon to record an fa image, you just need to include the preset in the acquisition sequence. For example, you can add fa after fc in the Focus node. This way, when Leginon acquire the final image after autofocusing, it will not only acquire and save an fc image but also an fa image, without the beam tilts, of course.

Another hint, Calibration (1.1) has an Focus node, so does any MSI application. To run only the node, you can (1) use Presets Manager to send

fa preset to the scope. This resets any image shift left from your previous manipulation. (2) Click on the Simulated Target tool in the tool bar of the

Focus node. What this tool does is to pretend that it recieves an target at the current condition and perform the function of the node at the target. It is very useful in debuging problems and testing.