On Aug 26, 2011, at 11:16 AM, Andrew Rodney wrote:

It would be interesting to bring the iteration into a discussion. I found it did absolutely nothing if I started with a sufficiently large patch sample from the get go (for RGB, 1728 patches). IOW, testing I did showed that adding another 3000 patches resulted in a fraction of a dE difference in the two profiles. Chris Murphy found similar results testing a CMYK profile.

I also think we need to be clear what iteration does and doesn’t do (so far, an open question if you will). My understanding is unlike say Oris, this iteration process is not designed to account for device drift or changes. Its designed for an improved resulting profile. In my tests that didn’t happen. What I was told by X-rite is that iteration is useful for those who start with an initial small patch process (ala ColorMunki which uses initially only 50 patches, then iterates the 2nd 50). That technology works amazingly well IMOH. But if an advanced user starts with a sufficiently large patch sample, and that number is open to debate, then iteration isn’t at all useful. That’s been my experience.

AR is right on...the iteration feature in i1Profiler is not to be confused with what ORIS Color Tuner or GMG ColorProof will do during either a calibration or profile iteration process. In a GMG world, *calibration* iteration is basically iterating the "full" gamut back to it's original reference or starting point to account/correct for device drift....this is more like the typical profile of an inkjet printer. GMG *profile* iteration is a bit different in that you're iterating a particular "color match" profile such as GRACoL, ISO, SWOP, etc. on your inkjet device. i1Profiler's "iteration" should more properly be called profile "optimization" in my opinion. The way I would describe the process is that the initial profile is built from a patch set that has no bearing on the actual device it's profiling (it has no clue as to device behavior/linearity/gray balance at this point). So the initial patch set is rather random in nature although it uses some sort of patch set generator algorithm one would assume. Once this initial profile is created, NOW it actually has some information about the device that was just profiled...if you proceed to iteration after you create the initial profile, a 2nd patch set can be generated that sort of "tightens up" the original profile in terms of gray balance and other perceived deficiencies in the original profile. I think of it as "filling holes" in the initial profile that the first patch set didn't adequately cover. How well does it work? In my world of CMYK proofing RIPs, I'm already starting with fairly good device behavior as a result of tools such as in-RIP ink limiting (per-channel and total) and linearization....plus I tend to start with upwards of 3,000 patches (I've been using a 2-page iSis chart that pushes close to the 6,000 patch limit of i1Profiler). In all my comparisons profiling different inkjet printers on various media, I have consistently seen an improvement between the original and the optimized profile...but the improvement is on the order of TENTHS or HUNDRETHS of a dE...IOW, if not statistically insignificant, it's at least visually insignificant. It's nothing like the improvement you might see iterating a GMG profile where the original profile starts out in the 1dE avg./5 dE peak range and iterates down to perhaps .25 dE avg. with no more than 2 dE peak values. In my experience with most Epson inkjets and various RIPs that use ICC profiles, I'll typically see on the order of .70-1.0 dE average matching to the GRACoL data set using i1Profiler. Generally that's much better in my experience than ProfileMaker and perhaps slightly better than MonacoPROFILER. That's NOT saying it's of no value....but it's pretty clear that if you start with an initial patch count that's high enough, iteration/optimization is of questionable value. With a fast/automated device such as the iSis, measuring upwards of 2,000 patches is trivial...but on a manual device such as an i1Pro, there MAY be some benefit in terms of time and media usage with measuring a few hundred patches and then improving it further using optimization. Myself, I've never tested the theory that a small number of patches would show a substantial improvement using the iteration feature....I simply start with enough patches to begin with that optimization is generally not necessary. :-) Regards, Terry Wyse ______________________________________ Terence Wyse, WyseConsul Color Management Consulting G7 Certified Expert FIRST Level II Implementation Specialist