Leginon

Magnification that you use to enter pixel size calibration is that at the film plate. Leginon only use that value to search in database the pixel size you entered during calibration. What matters is the pixel size value. In no place the post-magnification of the camera is considered since nor it needed to be considered when we do it this way.

The pixel size error is calculated based on the difference between value we give to Tecnai scripting in meters and the pixel size calibration you entered. The accuracy of the former relies on manufacture calibration. I have not seen it goes off that far in general, but I can not be sure whether you have performed beam shift calibration correctly as it is the trickiest of all (and optional. We didn't do it on our DD. The only consequence is that you need to shift the beam at the scope when it goes off). I would suggest that you try image shift matrix calibration first. If that gives you also large pixel size error, it would more likely be that pixel size you entered is off since from our experience that Tecnai calibration of image shift is usually better. If your pixel size error is large, and you are sure of the pixel size in the image acquire, you should increase the tolerance to accommodate it. We normally have to do that if defocus is set very far from zero. You just have to confirm that your calibration is valid by testing it in Navigation node.

One more caution note: You have a DE12 which pixel size is much smaller that of standard CCD. For example, at 50,000x microscope magnification, our Gatan 4k CCD has a pixel size of 2.2 A. Tietz F415 pixel size is 1.6 A, and DD pixel size is only 0.78 A. You will need to scale every magnification we recommended for different preset down if you want it to handle typical movement and autofocus error.

Thanks for your quick response. The question I have is how you get the pixel sizes at film plate because the images were taken on DDD (not on film) and the pixel sizes were calculated accordingly for DDD (not for film plate).

Another note: the magnification I was trying is 19,000x, not 50,000x. Besides, my calibrated pixel size on DDD for 50,000x is 0.83 Å, close to yours.

We don't get the pixel sizes at film plate. I am not sure we are talking about the same thing.

Because you would have sent a preset to the scope/camera pair and/or have acquired a test image, Leginon knows by the time you enter your pixel size that you are calibrating pixel size of DD. The values you enter, even though magnification is at the film plate, would be saved as values for the DD/Tecnai pair.

We use film plate magnification so that magnification becomes a microscope parameter, not a camera parameter. This makes it easier to set magnification to the microscope regardless of the camera installed since in some cases we have multiple cameras on one scope. The magnification sent to the scope would be the same in the way we do it.

Actually I did enter the pixel size on DDD for each scope magnification. One thing is that I put the presets for 19,000x, not 50,000x for fc, fa, en and ef. Does this matter? If not, I am not sure which step I did not do correctly.

My calibrated pixel size for 19,000x (screen up) is 2.09Å, what is yours?

I will try image shift calibration tomorrow before beam shift calibration.

2.1 A by extrapolation. I strongly recommend that you NOT to do beam shift calibration. DD can not do very short exposure, the beam-shift alignment tool would not work well any way. Use your time on required calibration instead.

My calibrated pixel size by imaging negatively-stained Catalase at 19kx is really close to yours.

I can try not to do beam shift calibration. But I think DDD should be able to do very short exposure. Do you know how short is for the beam shift calibration?

The only parameter that may need to be changed is the protection cover delay (change to let the beam pass or something, I don't remember the exact name). Usually there is a delay time (like 800ms) to minimize the vibration caused by the protection cover.

Maybe for the beam shift calibration in Leginon may require the beam on all the time? I called Dr. Jin Liang at DirectElectron about this, but he did not know if the beam-on all the time is required for this purpose.

Leginon's beam shift alignment tool reduces exposure time to avoid saturation of camera when it reduces the magnification to view the whole beam used in the higher mag preset. To give you a feel, we use as low as 7 ms exposure for beam shift alignment tool. DE12 can not go that low. It will round up to the next allowed number determined by the frame rate which in term is limited by clock frequency. Currently the shortest frame rate is around 25 ms. This has nothing to do with the protection cover.

Thanks for your clarification. Then why did I have such large pixel size error problem? Sorry I was a little confused. My calibrated pixel size is 2.09Å at 19kx, which is close to your 2.1Å. I tried to do beam shift alignment at 19kx. Maybe the exposure time I gave was too long? But it would not be related to the large pixel size error I guess.

The easy explanation would be that FEI calibration of beam shift from electronic unit to meters is off, or they simply decided to change the unit they are presented in. Don't worry about it unless you can't move to where you want after calibration.

The second easiest explanation is that you are focused very far from eucentric focus where all manufacture calibration was conducted at. Make sure you put the grid to eucentric height and reset defocus once you've focused the image to start with is always the first priority in good data collection.

If you can't move to where you want after calibration, then you should re-evaluate how you got the calibration. I've explained in the wiki how it works. The key is that correlation between the two images it used for getting the transformation matrix need to have clear peak.

Again, the concept is also easier when you watch image-shift and modeled stage-position calibration. Do them and you might realize what is wrong, if any.

Anchi