Leginon

About the first problem of focus tracking problem. It is likely that your negative tilt model is no longer good. Could you attach here two typical tomography tracking graph in x,y directions like Figure 1 in the wiki documentation Fitting_the_tilt_axis_model so that we may spot some differences in goniometer behavior?

About the second problem:
What kind of drift are you talking about? Is it a blurred image, i.e., image moved during the time of the exposure, meaning the drift is in the order of several pixels at the resolution of your imaging in the exposure time. Or is it a slower one so that one image and the next looked moved? Any estimate of the drift rate, Angstroms per seconds or nm per seconds, etc? If you don't see drift when the shutter is out, there might be some problem in the stability of the image when the shutter (a beam deflector) goes in and out. We have out DM shutter setting always at shutter one pre-specimen, do you normally do it the other way? When you say the apparent drift returned the next session, has the shutter settings the same as at the end of the previous session or did you or the software default it back? If it is related to the shutter, you ought to see it through Leginon with the following test:

Use a specimen that you know is stable, no-charging etc. Select Drift Monitor node. Go to advanced settings and select your camera other setting for drift monitor testing. save that by 'ok' the changes and then click on the "play" tool in the main panel. In low dose mode, this would take images at the given interval. This will measure your drift rate. When you reduce the interval, the shutter will need to get in and out more often (one per exposure), you might see a change of drift rate it measured. If the value you get are very different for short from long interval, you will be more sure that it comes from the shutter. Try this when tilted negatively might be able to tell you if it is slipping

Indeed that if you have severe drift problem (say coming from a loose specimen holding mechanism when it is tilted in the negative direction) the model-based tracking will not be able to do well because it would use the drifting between images as a sign that the specimen is away from the alpha rotation axis and will over-correct the defocus.

Anchi

Hi Anchi,

Thanks for the quick reply.

I've attached the X,Y, and Z-axis plots for a representative tilt series, that went considerably over-focus on the negative side the the tilt series.

I'm not able to run the drift monitor node as we're trying a couple of other things at the moment. In answer to your questions, though, it is indeed a blurred image, which seems to nevertheless be tracked moderately well throughout the tilt series. So one image and the next don't look moved. I'd estimate (off the top of my head) that it can approach ~100nm over a 0.1second exposure. Our DM shutter one is now set to pre-specimen, and still says that -- so unless there's an internal contradictory setting, I believe this is currently the case. I don't know what the setting was like before.

So, we're very confident it's not a loose specimen or other physical movement within the column.

I'll try your suggestions with the drift monitor when I get access to the scope next.

Thanks!

-Morgan

Check Low_dose_shutter_configuration_for_Gatan_camera_in_Digital_Micrograph_program page for our normal settings and how to figure out the correspondance at your scope.

The quick drift points to the possibility that more than one shutter is activated when you take images. Watch the LED on Gatan camera box and make sure that it is not blinking more than it should.

Regarding the defocus tracking,your x-axis is shaped like cosine and with very large (The axis on the left shows that your prediction and actual position are off 3 um when it was tilted to +/- 60 degrees. The curve is also symmetrical in positive and negative tilt direction. This curve is typical of the case when model axis is far away from the center of the detector but is not sloppy. In fact, I can estimate how much off using this diagram:

(a-3) / a = sin(60 degrees)

Therefore, a = 6 um. This is quite large. You should check your graph before the scope went down to see if it was much different. Also check the first four images of your tilt series to see if defocii for them are drifting. When the model axis is far from the imaging axis, the defocus change very quickly at low tilts. If you are at high enough mag, you can even see the defocus change by eye.

When the axis is off, you can change the model and image shift your final exposure area so that the ccd is imaging closer to the physical tilt axis, then the assumption behind the modeling will more likely to hold so that the focus is tracked. However, if your axis is indeed 6 um off, it would be better to have the engineers adjust it back.

Anchi

Anchi,

Many thanks for the response.

Your shutter settings worked a treat.

We just collected a tilt series without loss of focus on the second half of the tilt series. We'll keep monitoring, and talk to the engineers on Monday.

Have a good weekend,

-Morgan