An important data collection strategy in previous MAD work [52] [53] [62] has been to collect each oscillation image at each X-ray energy before proceeding to collect the next image. This was to ensure that systematic errors arising from radiation damage in the sample, crystal slippage, changes in X-ray beam conditions would not effect the measurements of a particular unique reflection and its equivalent at other energies. Similarly, if the lattice symmetry allowed, the crystal would be oriented so that Bijvoet pairs of reflections could also be collected on the same image at the same time.
It was however decided not to follow this course of action for two reason. Firstly the calibration apparatus and beam-line control software had not been developed to a stage where the energy could be repeatedly changed, routinely and quickly for optimised and energy stabilised data collection. Secondly it was expected that the diffraction from the lysozyme crystal would not suffer seriously from radiation damage since the relatively high X-ray energy would reduced the extent of absorption by protein and solvent atoms in the sample and also lysozyme is known to survive long hours in X-ray beams with no serious loss of diffraction. Thus each data set was collected completely at one energy before moving on to the next.
Collection of diffraction data for all seven energies was performed on the same crystal of
the lysozyme derivative. The orientation of the crystal was found by searching for diffraction
representing a major zone of the crystal lattice. From such an image it was an easy task
to establish the orientation since the cell dimensions and space group were known.
It was found that the crystal lay with its 
reciprocal lattice vector 
away from the vertical
plane and its 
reciprocal lattice vector lying 
away from the rotation axis.
The crystal was left in this orientation so as to increase the possibility that the collection of
redundant reflections would `fill in' for reflections lost in the crystallographic blind region.
No attempt was made to measure Bijvoet mate reflections simultaneously.
The maximum resolution of the data collected was 
Å. This limit was chosen since it meant that
largest reflection intensities still did not exceed the dynamic range of the detector and allowed
complete data to be measured without requiring any extra measurement of the lower resolution
data. This would have necessitated additional scaling of the data and may have introduced
more systematic error.
The five sets of diffraction data
took fifteen hours to collect. For each set the experimental set up was essentially the same
except for the X-ray energy. A total of 
of data was collected per data set.
Data sets 3, 4 and 5 were collected using the calibration apparatus to monitor the incident
X-ray energy on X31 during each exposure. Baseline checks were made every half an hour as described in
section 
and corrections to the incident X-ray energy were
performed as found necessary but on average every five minutes. The two reflections from the silicon calibrator
crystal used to monitor the energy were the (
) and (
).
Fig. shows the reflections after being positioned about the white line
energy before the collection of data set 3.
The large asymmetry in the reflections in a consequence of the sub-optimal
mirror alignment on the X31 line. Attempts were made to improve upon the alignment but were generally
unsuccessful. This may have been because of deviations of the individual mirror segments on the
pre-aligned bench from the required toroidal shape. This asymmetry did not however affect the
stabilisation procedure since the line shape was found not to change with time.
Differences between the two calibration signals of less than 
were tolerated resulting in energy
stability to within 
. The () and () reflections were scattered into the
detector at a Bragg angle of 
The calibration apparatus was not used for data sets 1 and 2 because the anomalous scattering factors vary
slowly with energy away from the absorption edge and drifts in energy of a few 
are acceptable. This
meant that data collection could run automatically without intervention.
However it was discovered that during the first data set
the crystal had shifted with respect to the beam position either because it had not been correctly centred or
because the plasticine used to hold the capillary in place on the goniometer had become too soft. The crystal
was re-centred before the second data set was collected and a 
stream of cool air was used to
ensure that the plasticine did not soften.
Figure: Plot showing the signal obtained from the two calibration channels and the fluorescence
channel after the calibrator had been set up for data collection at the white line (set 3). The
reflections are the () and () (on the left and right respectively). The monochromator
was manually adjusted as required so as to keep the signals in both calibration channels equal thus
ensuring that the incident X-rays were always at the reference energy. The energy
was kept to within of the reference energy during data collection.
The X11 experiment was carried out ten days later. The normal operating energy for this
line of 
was found to be adequate for one of the measurements (
and 
) and meant that only one
energy change needed to be performed which was advantageous given that less than 12 hours had been
allocated for the experiment. The second energy was aimed at obtaining large contrast in
the 
scattering component and was set to be that used for data set 4.
The energy of the X-rays after realigning the apparatus was assessed
by inspection of a diffraction pattern from the lysozyme sample crystal. Since the cell parameters
had already been found from analysis of the X31 data at known energies this allowed a sensible
estimate of the X11 X-ray energy to be made relatively quickly. The energy change was completed
within 2 hours.
The higher dynamic range of the imaging plate detector on the X11 beam-line meant that diffracted intensities were on average four times larger than those measured on X31. The signal to noise ratio of the data was therefore expected to be improved by approximately a factor of two.
In addition to the derivative data an extra set of native lysozyme diffraction data was collected
on a conventional sealed tube source
using Mo characteristic radiation at 
or 
Å up to a maximum resolution of 2.2Å
to allow straightforward determination of the absolute heavy atom positions from isomorphous differences Fourier
techniques.