[Proj] How to convert a sphere to ellipsoid with correct datum?

[Hermann Peifer]

On 02/09/2010 17:17, Jan Hartmann wrote:
> On 09/02/10 12:27, OvV_HN wrote:
>> In my opinion it would be simpler or even better to do a conversion from
>> lat-lon to cartesian XYZ starting with the old ellipsoid and to convert back
>> from cartesian to lat-lon with the parameters for the new ellipsoid. Regular
>> datum transformation routines do that, but a Helmert transform is done in
>> between (in cartesian space). So inactivate the Helmert transformation by
>> giving dx=dy=dz=0, rx=ry=rz=0, ds=1.0.
>>
> Yes, that's logical but it just doesn't work. Doing a transformation
> with Helmert transformation parameters set to 0 and 1 just gives results
> that are hundreds of meters off.
>

As an exercise, I just transformed lat-lon from one ellipsoid to another 
one by ussing the "Abridged Molodensky transformation" (EPSG dataset 
coordinate operation method code 9605) which doesn't go the long route 
(geographic 3D to geocentric + geocentric to geocentric + geocentric to 
geographic 3D) but uses some sort of shortcut. Using the 3 parameters dX 
= dY = dZ = O, I ended up with results which are pretty close to those 
from cs2cs. 1 meter away, perhaps. So I guess your hundreds of meters 
offset must have other reasons.

Hermann