[Proj] Problems using grid shift files

[Paul Kelly]

Hello,
I think it would help if you give an example of one of the transformations 
that gave the incorrect results, with the exact cs2cs command-line you 
used and test points.

On Thu, 2 Apr 2009, Ziegler Stefan wrote:

> Mmmmh, I think this is beyond my mental skills ;-) So I start with some
> questions: what is a pivot datum? And how can I chain datum
> transformation (without piping in the shell)? How does proj does a
> re-projection? I assumed something like: projection A -> ellipsoid A
> ----> here comes the datum transformation ----> ellipsoid B ->
> projection B. So if I know the accurate relation between ellipsoid A and
> B (like the +towgs84 params) I'm able to reproject more or less exact?

No, there is generally an extra stage:

Source Datum --> datum transformation A --> Pivot datum --> datum
transformation B --> Target Datum

Only when either your source or target datum is the same as the pivot 
datum is there only one datum transformation. For all other cases you need 
two sets of datum transformation parameters, for the source and target 
datum to some intermediate pivot datum. Parameters for use of WGS84 as a 
pivot datum are well defined for most datums. But in your case when you 
want to use CH1903+ as a pivot datum instead you need to calculate a new 
set of datum transformation parameters.

As of yet I'm not sure exactly how to calculate the new parameters but if 
you post some details of the conversion you're trying to do with some 
sample points, maybe someone will be able to work it out for you.

Paul