"Thomas A. Gardiner" <[EMAIL PROTECTED]>:
>It should help, because the manuals say that changing the order of the
>deltas changes which coordinate DX thinks varies fastest...  I also tried
Yes, that's why I chose the delta sequence posted here: z (last index)
varies fastest in my case.
>then it can be done.  The gridpositions and deltas syntax that you are
>using is short for something dx refers to as a product array; look into
>the user guide appendix B for details.
No, it's not a product array, but a regular grid as in example 1 of
App.B; this example was the one I used for the grid description because
my grid is regular: delta x, y and z are fixed, although different for
different directions.

It turned out that the actual problem seemed to be a different one. I
kept using my original grid now, but the point is that in the FORTRAN
binary writeout, 1 machine word (4 bytes in my case) is prefixed to the
array and 1 machine word is postfixed; although I had noted that there
was a 8 byte-excess in the record length compared to the raw length of
the data, I had thought that the 8 bytes were both postfixed. This lead
to a 4 byte shift of my whole data, which explains a number of problems,
some of which I didn't mention because I originally did not recognize
their relation to the coordinate issue; in particular, it explains the
shift in the vector components by one item. To read the data, one must
skip the first machine word of every record (step).
I don't know if the problem is compiler- or architecture dependent, or
if it is a FORTRAN issue and possibly does not appear in C (I seem to
recall that records are separated in a different way by C, though).
However, maybe this helps other unexperienced users like me.

Thanks to Tom Gardiner for discussing this issue with me.
-- 
------------------------------------------------------------------------
Thomas Ruedas
Institute of Meteorology and Geophysics, J.W.Goethe University Frankfurt
e-mail: [EMAIL PROTECTED]
http://www.geophysik.uni-frankfurt.de/~ruedas/
------------------------------------------------------------------------

Reply via email to