Data Collection Hints (from sue)
Always:
Determine what data you need from the experiment. Select detector position and image time to:
- Make sure the dataset is complete - all possible reflections have been measured.
- Make sure the data is redundant - each reflection has been measured multiple times.
- Minimize detector overloads.
- Minimize radiation damage. At SSRL BL 9-2, a ten second exposure is a lot of x-rays.
Take a few minutes to run strategy to determine rotation ranges and maximum image widths for a complete data set (You can choose finer image widths than suggested if you wish). Position the detector so that the resolution limit of the crystal is near (but inside) the edge of the detector. Collect a test image to ensure there are no overloads on the detector face.
For a MAD Experiment:
- Collect the peak data set first so that you can try to solve using SAD if radiation damage is severe.
- It is important that the low resolution data be accurately measured and complete. This means no overloads.
- A good MAD data set will require collecting at least 6x the data as from a normal crystal (three wavelengths and enough redundancy to accurately determine the anomalous differences).
- Set the detector distance and image time so that there are barely no overloads.
- Try to collect the data in a single pass per wavelength (rather than high and low resolution passes).
- Collect a native high resolution data set from another crystal.
For very high resolution data:
Two passes will probably be needed to collect both the strong, low-resolution data and the very weak high resolution data. It has been recommended (Z. Dauter) that the low resolution pass ALWAYS be collected first. I disagree with this recommendation. The proper strategy will depend upon what is needed from the data set. For a known structure or an unknown structure I will attempt to solve using ACORN, I usually collect the high resolution pass first, without being too greedy, then the low resolution pass. This can give up to 0.2 A better resolution (0.9 vs. 1.1 A data, for instance).