IJC linearization demonstration
2003-07-05 by Antonis Ricos
To demonstrate the effects of linearization in IJC, I have uploaded an Excel graph that shows three plots: - the "aims", which are the ideal densities for the given dmin and dmax (.05 to 1.69 in this case). These are generated internally by IJC once the readings of the unlinearized print are entered. - the unlinearized plot which shows how far from the aims the initial ink curves are. The idea is to show that a simple, approximate "throwing together" of some curves, is a good enough starting point (yes, within limits, but without worrying about accuracy across the scale). - the linearized plot which shows the effect of the internal calculations of the program once the readings for each of the 26 gray patches was entered. The near perfect plot means that all the initial "sloppiness" is removed automatically in one step by the linearization engine of IJC. Some caveats and limitations: - This particular excercise was done with a profile that uses small amounts of LightCyan and LightMagenta in the 2200 to neutralize the extra warmth of the gray. In theory, I should have turned off these tinting inks while linearizing, and then turned them back on - sort of cheating the linearization a bit. Because in this case they add very little density, the effect is inconsequential. However, when it comes to major color shifts (such as in my earlier sepia profiles), I abandon linearization alltogether so I can compensate for density independently of color. The linearization engine presently does not distinguish between the two. - This demonstration was based on readings of relatively "fresh" prints: They were dried with a hair dryer and read within approximately the same time after they each came out of the printer. Any dry-down effects are thus equalized. It so happens that the OEM inks of the 2200 dry to a pretty stable density relatively quickly. But that is not the case with all inks. The FS black, for example, takes almost a day to reach final density on Photo Rag. This throws a "curve" (pun intended) to the linearization engine. I find it more reliable to linearize manually (by precise moves on the curves of the inks) and plot the results in Excel anticipating the final drydown. Most of the difficulty comes from the very short and steep curve required for the black to kick in at the end of the scale. Of course, that's assuming I want to chase down every little + or - 0.01 deviation from the aims. In most cases, you can get close enough to print - but it's a caveat worth bearing in mind. In general, it's a good idea to have a sense of what a given deviation from the aim densities does to a real image: If you have the subtlest of gradations across a sky (or a skin tone), the smallest deviation will break it. If you have a rough grainy image with many abrupt variations of grays, you'll hardly see a +- 0.04 error. The other little dark secret in all this chasing down of the perfect gray scale is that all those "cheap densitometers" we often talk about here (like the XRite 810) have a whole lot more accuracy in the shadows than they do in the highlights. Nothing wrong with the instruments, they are just missing an extra decimal point. When it comes to densities less than 0.10, there is only one digit to work with. Shadows get the benefit of 3 digits. Given the log nature of the scale, that results in a fairly rough read of the highlights. Going after that extra decimal, however, could cost thousands. So, we after "good enough" here - a reality check to keep in mind! Since it's been a while since IJC has been released, I hope these observations are useful to those making profiles with it. Antonis