Yes, unfortunately an "orientation matrix" is completely useless without 
a frame of reference.  All it does is tell you where the cell axes lie 
relative to "X", "Y" and "Z", but "X" may be the x-ray beam, or 
sometimes that is "Z" and "X" is the spindle axis. Different people have 
different and strongly-held opinions about this.  The "fast" and "slow" 
axes on the detector are also a matter of widely-varying convention.  
The origin of the detector could be the first pixel in the image file, 
the upper-left corner, or it could be in "cameraman view" vs "sample 
view" and the beam center is also only meaningful relative to whatever 
the "origin" is.   Also, if you are at one of the four beamlines in the 
world that rotates their spindle the opposite direction of everyone 
else, then you may have to reverse the sign of one axis as well, usually 
"Y", or there may be an option such as "REVERSEPHI".  XDS lets you 
define all these things manually and there are formally no "default" 
settings.

To be fair, it is difficult for developers to keep track of what all 
~150 beamlines around the world, plus commercial "home lab" equipment 
manufacturers are doing.  I've been trying to some time to assemble a 
"library" of recent lysozyme datasets from as many beamlines as 
possible, but my collection is still far from complete:
http://bl831.als.lbl.gov/example_data_sets/

That said, I have written a few jiffies for inter-converting XDS and 
mosflm conventions, and posted them here:
http://bl831.als.lbl.gov/~jamesh/orientation_matrix/

Instructions are at the top of each awk script.  For example, 
mosflm2xds.awk takes the first three lines to be the *.mat file from 
mosflm, but also needs the mosflm input script after that to figure out 
how to convert it into XDS conventions.  Use it like this:
cat auto.mat mosflm.inp | ./mosflm2xds.awk >> XDS.INP

You can convert the resulting XDS keywords into an XPARM.XDS file using 
xds2xparm.awk, which can be useful for feeding into some pipelines.  If 
you already have an XPARM.XDS, you can convert it back into XDS keywords 
using xparm2xds.awk, and these keywords can, in turn, be fed into 
xds2mosflm.awk to get a mosflm *.mat file again.  Note, however, that 
there can be some information loss here.  The *.mat file does NOT 
contain the full description of the detector and spindle geometry that 
XDS may be using.  I make some guesses in these scripts if not all the 
relevant XDS keywords are provided.  It works with my setup at ALS 
8.3.1, but I have not tested it thoroughly with other cameras.  I would 
be interested if you find any bugs!

As for what you're missing, could it perhaps be a reversed phi 
direction?  That might look like a 180-degree rotation.

Good luck!

-James Holton
MAD Scientist

On 12/28/2014 3:12 PM, Igor Petrik wrote:
>
> Takanori and Pierre,
>
> Thank you for your quick responses. If you read my post you will see 
> that I used the xds2mos.py script to convert the xds orientation to 
> Mosflm format, but this gives me a result that is inconsistent with 
> the matrix calculated directly by moslfm or DENZO. For orthorombic 
> space group P222 there are only a hand full of equivalent matrices 
> related by changing sign of one of the vectors. The matrices generated 
> by XDS and Mosflm do not seem equivalent. Or maybe I'm missing something.
>
> Thanks,
> - Igor
>