On an RGB screen or a CMYK laser printer spot colors cannot be displayed directly and must be emulated by converting them into their process color equivalent. The Separation and DeviceN color spaces provide tint transform functions to do so. However, with NChannel colors spaces there exist separate transforms for the individual components and for the collection of the components. Which one to take?
Similar to the overprint preview feature this redundancy is somehow unfortunate, since some of the sub-standards, such as PDF/A, claim the fact that a unique reproduction is guaranteed.
Of course, I tried out some real world PDF documents to find out whether it makes a difference in appearance to use one or the other way to transform the colors. The not surprising answer is: it does! In order to understand this we have to go deeper into the mechanisms of the color processing pipeline.
If a rendering engine supports overprinting, then Separation, DeviceN and NChannel colors are rendered into separate channels without transforming them. At the very end when the final image is produced the channels are blended to receive the final color. So, in this mode it is quite clear that the tint transform functions of the individual colorants should be used. If a rendering engine doesn't support overprinting, it supports rendering into a surface which contains the process color channels only. In this mode it seems to be quite natural to use the collective tint transform function of the DeviceN color space.
Things get more complicated, however, if transparency is involved or if the target device supports one of the spot colors but not all of them. In this case you only get the correct result if you transform each colorant individually even if overprinting is not supported. But, what if an individual colorant doesn't provide a tint transform function? In this case you only have the option to use the collective tint transform function for the that colorant by setting the color values of the other colorants to 'no tint'.
Here's my conclusion: The tint transform of the DeviceN color space should only be used if the tint transforms of the individual colorants are not available. This produces a result which is independent of overprinting and transparency.
What is your take? Please let me know and post a comment.
