Using Gatan K2 Summit in Leginon¶

The main use of K2 Summit is in Counted or Super-resolution mode in combination with dose fractionation (frame saving).

If you want to use DM's dose fractionation, activate the "save frames" check box in the particular preset camera configuration. You can also use DM's frame alignment algorithm by activating "align frames" check box, but the latter is not recommended since the alignment in DM will delay the return of the image to Leginon and therefore affects throughput greatly.

Configuration¶

1. Make sure your camera configuration is set to Leginon image orientation and instruments.cfg matches that resulting orientation and dimension.
   • This is likely 270 degree rotated with a flip around vertical axis. New version of DM can acquire its internal darkware gain references in this configuration.
2. K2's dose fractionation (frame saving) only works with full camera dimension without binning.
3. K2 interface in DM restricts the exposure time to multiple of some set number, depending on the mode. When you enter in DM an invalid number, it automatically change it to the valid one when you click the acquire button. pyscope/dmsem.py has hard-coded values for these precision. Compare them to the precision of the DM version you have since it may change with DM version.
4. K2 frame exposure time is adjustable in Leginon gui. However, you need to check if the number is acceptable by DM.
5. Frames are saved as unsigned 16-bit mrc image stack in K2 computer under D:\frames by default as set in the code inside pyscope/dmsem.py in the function calculateFileSavingParams. You must create the frames directory first. Leginon will not do that for you. You may change where the frames are saved here but will need to make corresponding changes when setting up raw frame file transfer

Recommended dose rate and exposure time for different K2 modes¶

This recommendation is based on our experience.

As a general rule, the dose rate corresponds to the value used during DM gain reference acquisition. You can put down the small viewing screen once the intensity is adjusted at the beginning of DM procedure to get the values. You will notice that Linear mode calibration uses much higher beam intensity than the calibration for counted/super-resolution modes. For our Tecnai F20 at 200 kV, the former reading of exposure time by the small viewing screen is 0.25 s while the value for latter is 1.6 s. You should use similar intensity during Leginon calibration and data collection.

At 1.2 A/pixel, we use the following for final exposure that has binning of 1: |mode|small viewing screen exp time (s)|detector dose rate (e/physical pixel)|acquisition exposure time for ~ 20 e/A^2 specimen dose with 1.2 A/pixel pixel size (s)| |Linear|0.25|64|0.5| |Counted/Super-resolution*|1.6|10|5.0|
*Most literature recommend at even lower dose in the range of 6-8 e/s/pixel. This is just a number we have had success with.

Preset Recommendation¶

See Pre-MSI_Set-up specific for K2 camera using counted/super-resolution mode.

Calibrations¶

Most calibration and operation are similar to that of a typical digital camera.

Leginon treats linear/counted/super-resolution modes of K2 Summit as three cameras. Therefore, each needs its own calibrations.
A python script is available to make the copying of the calibrations easier. See RE: Super resolution format unavailable to calibrate matr.... You may need to modify it for different modes you are copying calibrations to.

Gain/Dark Correction Handling¶

Counted/Super-Resolution mode¶

K2 Summit in counted/super-resolution mode has two layers of gain reference correction, known as hardware (HW) correction and software correction.

Hardware dark correction is essential for electron counting, without it, its processor can not find individual electrons.
Hardware gain correction is a rough gain correction. It is only done to make sure the data from different ADC boards are reasonably normalized before counting.

After these hardware correction and counting, the images are integers, representing the number of counts received at each pixel. We call this type of data "raw".

Software dark correction for counted/super-resolution mode is fake. The image taken in dark condition is set to 0 internally and output as such if a dark exposure is requested externally by SerialEM and Leginon.
Software gain reference taken in DM is equivalent to the bright image taken in Leginon. The software gain correction, as well as Leginon's correction use this image of the blank area to normalize the more subtle differences from pixel to pixel.

After this level of gain normalization, the image data type is float.

Since Leginon duplicates the DM software gain normalization, we decided to receive images from DM without software gain normalization. This is set in the code, and means gain correction is still needed in Leginon.

1. The hardware correction has been applied in DM before counting. Therefore, the "raw" image Leginon get are integers and roughly gain corrected.
2. The dark image in Leginon should always be 0. Therefore, acquiring dark image only need to be done once per camera configuration. There is also no need to average several images in the making.
3. The bright image acquired in Correction node should be taken at the dose rate to be used. It should also be taken so that accumulated exposure time is at least 100 second (DM reference accumulates 500 seconds of exposure by taking 50 of 10 second exposure). If you don't want to wait this long, a reasonably good reference can normally be obtained with averaging 20 of 5 second exposure. Alternatively, take a single very long exposure (100 s).

The bright image will display variation in sensitivity of the pixels. There is likely a general gradient across the detector known as growth zone like this attached image:

Frame Saving¶

When frames are saved:

• The image returned to Leginon gui and saved as usual IS gain normalized by Leginon as usual. This is an integration over all the frames. With the gain correction, the image is 32-float mrc image.

• The movie frames recorded are "raw" as defined above. This saves some time and keeps the file at 1/2 the size. The format is 16-bit integer MRC image stack. We plan to save in the future even smaller number of bits. Appion will use the bright/dark/correction plan of the image transferred to Leginon to make corrected frame during frame stack making.

Linear mode (Not recommended as we found the quality is unstable)¶

1. The dark image will have a large mean and the value changes with exposure time. Ideally, the dark images should therefore be taken at the same exposure time as the later images to be corrected. Leginon can account for small differences but not a full range.
2. To avoid saturation by the accumulated dark current, long exposure in linear mode is not recommended. 0.5 s is typical.
3. Do not use frame saving with linear mode. It is not worth while.

Preparation in DM¶

Hardware dark and gain correction need to be done by DM to count electrons. You should therefore obtain dark and gain references and upload them to the hardware (HW). You can do these in Leginon image orientation in newer version of DM.

Once a while, but not very often (such as once a week or longer), all references should be retaken. This takes at least 15 minutes.

Gatan recommends frequent hardware dark reference update. This does not require beam. You can do it everyday while getting the scope ready.

1. Make sure the temperature of the camera is stable at where it should be cooled to.
2. Set the Direct Detection mode to Counted (just in case of DM gui bug).
3. Click on "Update HW Dark Reference". This takes about 4 minutes.

Operation¶

1. Typical frame saving preset parameters for counted/super-resolution: 0.2 s frame exposure time, 4-5 s total exposure time at 8-10 e/physical pixel/s (Some argues that 5 e/physical pixel/s is better). You can use DM's reading of dose rate while it is in counted mode and UNBINNed to get a reasonable estimate.
2. Because of its small size, we have a modified Quick-start procedure we use regularly at NRAMM FEI F20.
3. /myami/rawtransfer.py should be run on the file server to transfer the raw frames and give them names that are synced with the integrated image as it is produced. See DDD_raw_frame_file_transfer. K2 frames are saved as mrc stacks.
4. See Direct_Detector_Frame_Processing for tools used for processing the frames.
5. To gain correct and align the frames, start parallel instances of Appion makeDDRawFrameStack.py See GainDark correction of the raw frame with or without drift correction.

Warning¶

• Because of a delay in communication with DM when Leginon is started, camera configuration shown in Correction Node settings may not corresponds to what is used last time when you quit Leginon. Make sure you check this.

Leginon functions that do not work with K2 camera¶

General functions

1. The dose matching tool in Preset Manager. K2's exposure time is not fine grained enough to make small adjustment according to the small dose deviation. You will have to use fixed exposure time and adjust beam intensity by hand to achieve the right dose. We find out the equivalent screen reading from the microscope for the acceptable 8-10 e/physical pixel/s dose rate for the detector and make adjustment there.
2. Tomography correlation algorithm only takes square images. Until that is changed, Frame saving is not possible since only a full size unbinned image can be paired with frame saving.

Frame saving camera functions

1. The image returned to Leginon display is always integrated over all the frames with K2 summit camera. Selection of frames to use as in DE series does not work here.
2. Readout delay is not adjustable on K2. Leave it at 0.

Clean up¶

Frame saving generates a lot of files during acquisition and during processing.

1. python script in leginon directory "cleanddraw.py" can help you clean up the transferred raw frames when you start to run out of disk space. See Feature #1784 for more details.
2. Selective removal of aligned frame stack in Appion runs is still to be written.