Leginon

Rui,

Could you make snap shot of 3-way viewer to show how off this is?

Also, do two tests for me:

1. Create a duplicated preset of hl, call it hl2 or something
2. Use this preset in Exposure Node (Meaning remove the preset you had in its setting and add this one there)
3. targeting like you would normally to the high mag from an hole image in Exposure Targeting so that you will get images in exposure mode at the mag of hl.
4. Go to Scope node on the microscope pc, click the refresh button (green, recycling symbol) to get current microscope parameter.

What I want to see is whether the targeting is correct without change of magnification and to get from the scope the image shift it applied.

Next, repeat the test but use the en preset in Exposure Node.

There is a possibility that the deflector coil moves the image in opposite direction on your Krios at the mag in hl and en.

Anchi

Hi, Anchi,

I did the test you suggested, please see the attached images.
Step1: use hl3 in the exposure node
Step2: use en in the exposure node

Since I am using a room temperature grid, I have to manually mark the actual exposure position in pink (see target-vs-actual.png) for Step2.
I also attached another independent test (target-vs-actual-example2.png). I think there is a trend.

Again, if targeting at the center, no image shift applied, the hl and en are well aligned.

P.S. I notice there is no flip of orientation between low mag and high mag images for our titan microscope.

Rui

Sorry, I forgot one more information I need:

The image shift value for hl preset and hl3 preset since they are not at (0,0) like en.

I think you are suggesting that the when you want to move in x, it moves to y if that pink mark in your example is the actual location?

Anchi

Hi, Anchi,

Please see our hl and en presets settings. The hl3 is exactly the same as hl.

Yes, the pink mark is the actually en beam location that I saw on the microscope viewing screen after en exposure (without touching anything).
And I think you are right about the x-y direction thing.

Rui

Rui,

Could you confirm the image shift calibration matrix for hl and for en using Navigation node (direction and move scale) and then send us the matrix (You can read them off in Matrix node or Calibrations application in the Edit tool panel once the preset is sent to the scope) ?

The code for applying image shift does a transformation according to these. If your images orientations are the same for the two, these matrix should scale to each other roughly by their mag ratio in normal cases but it appears that it is not true in your scope.

Anchi

the image shift matrix for hl is:

-4.5791e-09 5.17145e-09
5.13622e-09 4.2449e-09

the image shift matrix for en is:
-1.20803e-10 -4.21528e-11
-3.79217e-11 1.21582e-10

I can use image shift in Navigation node to move to target in either hl or en presets. The targeting will only be off when leginon goes from hl to en.

I can't find anything wrong with the code.

We use the standard vector transformation in numpy to do this and applying them to your matrix does creates a significant rotation.

Since there is no other way to find the rotation of image between these two very different mags specific to your scope, we will need to assume that there is no rotation of the image, only the image shift.

I attach a changed leginon/calibrationclient.py file for you to test if this works out. I used the ratio of square root of determinant ratio between the two image-shift calibration to get the scale. Hopefully it is close enough.

Because the magnification ratio of your en/hl is unusually large, you should especially avoid targeting something far from the center. In physical unit, image shift is not expected to be accurate to be beyond 2 um on a magnification such as 27500x.

If what I have still give wrong image for targets that are close to the center, you have two options:

1. Try to find out the correct rotation of images between mags experimentally. Since image shift is decoupled from image orientation, there is nothing else to use since your stage movement isn't very good either at high mag.

2. Use iterative stage movement with navigator. This is the movement made by RCT and Tomography targeting. It is slower but good for larger distance.

http://emg.nysbc.org/projects/leginon/wiki/Iterative_Stage_Movement

Hi, Anchi,

Thanks! I will try that.

By looking at the image shift matrix for en and comparing with the one from our F20, we are concerned that the four numbers in the matrix are not at the same order of magnitude.
Do you think it is possible that our image shift / beam shift calibration is off for either LM or SA?
We are actually not sure if such calibrations have ever been done on our titan before.

We have asked the FEI engineer to do that tomorrow and we will let you know if it helps.

Rui

It is reasonable that the numbers in the matrix are not at the same order of magnitude between F20 with 15 um pixel camera and Titan with 6 um pixel camera. The ratio of the square-root of determinant of the two matrixes you sent was 53. In comparison, the ratio of the magnifications is 50. Therefore, I don't suspect a calibration problem.

While it is possible that your calibrations are off, image shift calibration I know of only deal with the scale, not the rotation. But Titan might be different with C3. I don't know enough about that.

Anchi

Hi Anchi & Rui,

I think we might have a similar problem on our Talos-Artica at UCSD - internally within a preset, the image shift and modelled stage position work, but when select exposure targets from Exposure Targeting, it seems that the collected images are 180 degrees off.

Was this problem ever resolved?

Thanks!
Mike

Michael,

I am working on a more permanent solution. If you are on 3.2, you can look for isLensSeriesChange function in leginon/presets.py and specialTransform in the same script for the place to make change that will hack it to work.

This condition statement will skip specialTransform. Therefore, if you change it so that it will proceed and add more rotation, then it will work right.

    if tem['hostname'] != 'jem-2100f' or not self.isLensSeriesChange(mag1,mag2):
      return pixelvect
 

What are the preset magnification ? You may be crossing lens series which is at the heart of the problem. Normally we avoid that but with small pixel camera and low camera mount, you may not have that choice.

Hi Anchi,

Thank you for your fast response! It is likely due to the relatively lower magnification (36000X) that we are using for exposure with our K2.

We are using:

hl: 4300X
en: 36000X

I will look into the code that you suggested to find the correct fix.

Thank you!
Mike

Just to document the changes in case future users have to address this problem, I updated the myami-3.2 code in the script:

$PYTHON_SITEPACKAGES/leginon/presets.py

Changing line 1998 from:

return (mag1 <=4000 and mag2 >=5000)

To

return (mag1 <=1000 and mag2 >=5000)

This fixed the targeting problem when selecting focus and exposure targets within exposure targeting.

Mike