Hi, Klaus Karcher wrote:
some good news: with cctiff -p, at least the ordinary transformations are ok -- but the deviceLink transformations are still different.
Ah, cool - I've tried that, with the "synthetic" profiles I created earlier, and the discrepancies in transforming from a 1.0-gamma profile to a real profile pretty much vanish if the -p option is passed to cctiff.
I also compared with the lcms tifficc output - lcms seems to be affected by this issue too, but not as severely.
What's interesting is that with a pathologically high gamma (out of curiosity I tested with 4.0 as well as 1.0) there are discrepancies, but they're far more subtle than with the 1.0 gamma profile.
As for building the device link - do your results improve if you use the -n flag to icclink? It reduces the error here, but certainly doesn't eliminate it.
Oh, BTW, Graeme, the documentation for icclink's -n option seems to be missing.
All the best, -- Alastair M. Robinson