Dear Bryan,
The quick answer is no. As David Waterman
mentioned, it has a default value for the gain for each type of
detector that it can deal with.
A more detailed answer. An incorrect value for the gain can be
indicated by values of the BGRATIO which differ significantly from
unity (1.0). BGRATIO is the ratio of the rms variation in the
background to the variation expected on the basis of Poisson
statistics, using the gain to convert from digitised values in the
image to X-ray photons. This is calculated for all measured spots, and
binned as a function of intensity for each image measured (and printed
in the full logfile). Mosflm prints a warning message if this differs
from 1.0 by more than 10% and will suggest an "improved" value for the
gain that should be used.
There are a host of caveats in this procedure. For example, if the
images contains significant diffuse scatter around the Bragg spots,
the BGRATIO may be above 1.0 ... this is probably the commonest
effect, but does not mean the gain is wrong. If for any reason the
mask definition (defining the boundary between background and spot)
has not worked correctly so the spot extends into the background, this
will also give a BGRATIO of one (in this case, the BGRATIO will tend
to be close to 1.0 for weak spots but greater than 1.0 for strong
ones). The boundary is controlled by the "Profile tolerance"
parameters, which are sometimes set artificially high to help process
images where the spots are not fully resolved.
This is why Mosflm does not automatically update its default value for
the gain based on the BGRATIO.
As David has mentioned, this procedure also assumes that adjacent
pixels are independent, which they most certainly are not (except
possibly for some pixel detectors), due to the point spread function
of the detectors and corrections that are applied to the raw images.
Does it matter ? The gain is used to identify outliers in the
background plane determination (eg due to zingers, shadows, ice spots,
hot pixels etc) which are rejected from the calculation, so if it is
significantly in error this will introduce systematic errors in the
integrated intensities. This can show up in the cumulative intensity
distribution in Truncate if the gain is a very long way off. I have
not done a proper study of this, but I think it would need to be out
by more than 20% to have a significant effect. The gain is also used
to calculate sig(I), however, the sig(I) values from Mosflm are
adjusted in SCALA to reflect the true variation between symmetry
related reflections so that providing the multiplicity is high enough
for this to work correctly this will not have any real effect on the
final merged data.
The bottom line is that the estimates for sig(I) that emerge for this
procedure seem to be quite good, in that the correction factors that
are subsequently applied in SCALA for cases where other systematic
errors are small (ie no radiation damage, absorption etc etc) are very
close to 1.0.
Best wishes,
Andrew
On 3 Mar 2011, at 20:34, Bryan Lepore wrote:
wondering if mosflm can automatically estimate the gain.
i.e. i gather it is still estimated the usual way.
-Bryan
