Dear Gerd,
I can assure you that I have no shares in Dectris nor any
commecial connections with them. What I do have is a lot of still
vivid memories of CCD images, with their wooly point-spread function
that was affected by fine-grained spatial variability as well as by
irredicible inaccuracies in the geometric corrections required to try
and undo the distortions introduced by the fiber-optic taper. By
comparison the pixel-array detectors have a very regular structure, so
that slight deviations from exact registering of the modules can be
calibrated with high accuracy, making it possible to get very small
residuals between calculated and observed spot positions. That, I
certainly never saw with CCD images.
I do think that asking for the image width was a highly pertinent
question in this case, that had not been asked. As a specialist you
might know how to use a CCD to good effect in fine-slicing mode, but
it is amazing how many people there are still out there who are told
to use 0.5 or even 1.0 degree image widths.
Compensating the poor PSF of a CCD by fine slicing in the angular
dimension is a tall order. With a Pilatus at 350mm from the crystal,
the angular separation between 174-micron pixels is 0.5 milliradian.
To achieve that separation in the angular (rotation) dimension, the
equivalent image width would have to be 0.03 degree. For an EIGER the
numbers become 75 microns, hence 0.21 milliradian i.e. 0.012 degree.
Hence my advice, untainted by any commercial agenda :-) .
With best wishes,
Gerard.
--
On Thu, Jul 13, 2017 at 01:25:08PM -0500, Gerd Rosenbaum wrote:
> Dear Gerard,
>
> you sound like a sales person for Dectris. Fine slicing is perfectly fine
> with CCD detectors - it takes a bit longer because of the step scan instead
> of continuous scan. The read noise issue is often overstated compared to the
> sample induced scatter background. If for fine slicing at 0.05 degree or
> less the diffraction peaks go too close to the read noise make a longer
> exposure - signal goes up, ratio signal to sample-induced-BG less, as for
> any fine slicing, same read noise.
>
> It would be helpful to analyze the dense spot packing along layer lines if
> we knew the wavelength and the sample-to-detector distance (assuming this is
> a 300 mm detector) and the rotation width - as you pointed out. That would
> help to distinguish between multiple crystals (my guess) and lattice
> translocation disorder. Fine slicing is definitely needed to figure out what
> the diffraction pattern at 120 degree could tell you in terms of strong
> anisotropy .
>
> Best regard.
>
> Gerd
>
> On 13.07.2017 08:20, Gerard Bricogne wrote:
> >Dear Tang,
> >
> > I noticed that your diffraction images seem to have been recorded
> >on a 3x3 CCD detector. With this type of detector, fine slicing is
> >often discouraged (because of the readout noise), and yet with the two
> >long cell axes you have, any form of thick (or only semi-fine) slicing
> >would result in spot overlaps.
> >
> > What, then, was your image width? Would you have access to a
> >beamline with a Pilatus detector so that you could collect fine-sliced
> >data?
> >
> > I would tend to agree with Herman that your crystals might be
> >cursed with lattice translocation disorder (LTD), but you might as
> >well try and put every chance of surviving this on your side by making
> >sure that you collect fine-sliced data. LTD plus thick slicing would
> >give you random data along the streaky direction. Use an image width
> >of at most 0.1 degree (0.05 would be better) on a Pilatus, and use XDS
> >to process your images.
> >
> >
> > Good luck!
> > Gerard
> >
> >--
> >On Thu, Jul 13, 2017 at 01:21:02PM +0100, Tang Chenjun wrote:
> >>Hi David,
> >>Thanks for your comments. Although the spots become streaky in certain
> >>directions, I have processed the data in HKL3000 and imosflm, which
> >>suggested the C2221 space group (66.59, 246.95 and 210.17). The
> >>Rmerge(0.14), completeness(94.8%), redundancy(4.6) are OK. When I tried to
> >>run Balbes with the solved native structure, the molecular replacement
> >>solution was poor. So I ran Balbes with the split domains of the native
> >>structure. Although the solutions were also poor, I found the MR score of
> >>one solution above 35. On the basis of this solution, I tried to run
> >>Buccaneer and the Rfree could be 0.46. Unfortunately, there are four
> >>molecules in the asymmetric unit and it is to hard for me to reduce the
> >>Rfree further.
> >>
> >>All best,
> >>
> >>Chenjun Tang
