When I do a DF calculation for one specific room compared to doing DF calculation for a whole planfloor (with that one specific room included) I get different resultats for that specific room? shouldn't one get the same results regardless of changing any data, model settings or basically anything. The only difference between the to calculations is that I calculate more than one room in the second calculation.
Sometimes they only change is perhaps 0.01% which I consider OK. But when it comes to 0.1-0.5 % something tells me there is a difference in doing a calculation with one or many rooms at the same time.
There are two reasons results may differ between the two runs:
1) Stochasticity: There is (pseudo)randomness in the way Radiance samples the irradiance hemisphere at each diffuse bounce, so variation between runs is expected. One would expect this variation to be smaller at higher values of the -ad (ambient divisions) parameter.
2) Irradiance Caching: To save time during a raytrace, Radiance caches and reuses irradiance values on surfaces in the scene. (This is why running two sensors doesn't take twice as long as running one.) If the -aa and -ar parameters imply a coarsely-resolved cache, it is possible to see light leaks between spaces. To test if this explains your variance, you could try turning the cache off (by setting -aa 0). Note that this may significantly increase run times...
3% variance is unusual. It may be that if your model is very, very large or in units of mm, the ambient parameters (specifically -ar) need to be increased. Also if your geometry is very complex (shading screen with tiny holes, venetian blinds) you may need to increase the -ad value. Its difficult to say more without seeing the model.
I always import a dxf.file from ArchiCAD where it's saved in "meter" in ArchiCAD before saved as a dxf.file.
The dxf.file is normally around a size of 200 000 kB. The Rhino file then normally gets around the double, and yes it's fairly complex because of a direct import from the architect drawings.
I have found out that by saving the dxf.file in meter the time for a single run becomes much faster.
Is this a correct way of doing this?