Previous: Integration,
Now we are ready to scale the data. Press the Tab "Scaling". The following window will appear.
The tick boxes and values are the default values. Press "Scale sets" to scale the sets together. After some time the results will show:
The statistics are not great and the percentage of deviating reflections is quite high. The plot of scale factors shows some peculiarities:
Pointing the mouse at the lowest red dot, and clicking the first mouse button shows the number of the frame in the Display Frame field. Here, it is image 35 (To create this example case we have replaced the real image by a completely black image as an illustration).
This causes the strange behaviour of scale factors. If we look at the scaling options we can "untick" the Scale restrain. This will make the program calculate individual scale factors for each image, without restraint.
Now a second run of scaling will show the following plot of Scale vs. image number:
The scale factor of image 35 is now equal to zero. Therefore it is better to reject this image from scaling. "Splitting" the first set in two and leaving out the bad frame can do this. Go back to the "Main" level and select the first set.
Now go to the controls and press the button: "Split set".
This will pop up the following window:
Move the slide bars until the split is at image 34, the last good one of the first part. Then select the second set and change the Starting frame number to 36.
Now image 35 is eliminated and we have three sets: One from 1 to 34, one from 36 to 90 and the third set from 1 to 90. Since all the images have been processed, we can move on to scaling straightaway by pressing the Tab "Scaling".
Now a scaling run will give much better results:
and the graph of scale factor vs. frame number has no gaps anymore.
Ticking the box : Use rejections on next run, will kick out the reflections that are flagged for rejection.
The resulting R-factor has now dropped by another tenth of a percent.
The graph of Chi2 vs. resolution is still a bit high on the low-resolution side. This can be changed by playing with the "scale factor" (normally 1.3) and the "Error model". Ideally the Chi2 should be one over the whole range of resolution. Changes of Chi2 will have no effect on the intensities, but only on the sigmas of the intensities. The level can be changed by the scale factor. Lowering it to e.g. 1.1 or 1.2 will increase the average level of Chi2. The shape of the curve can be changed with the "error model".
The button "Adjust error model" can be found under "Controls".
To adjust this you need the R-factors as a function of resolution. These can be found at the end of the log file, by clicking "Show Log file". The editor can be selected in optionsà editor.
The important curve is the "Linear R-factor":
These values should be used as input for the "error model".
The result of scaling after these changes will be a much smoother Chi2 graph.
Pressing the button: "Show Redundancies" shows the redundancy of the data. The resolution bin can be selected. The window will look as follows:
By pressing the button "Reciprocal Lattice" you can get a pseudo precession picture in the display window. You have to select whether you want to see the h0l, hk0 or 0kl plane. Pressing the button hkl+++ or hkl--- will allow you to pan through different levels of hk, kl or hl planes.
If all your data have been properly scaled this way, you can run cifin to convert the data to CIF format as input to the maXus software.