Hi Neil,
I extracted bits from your most informative post and labeled them for ease of reference, as these raise some questions:
Let's start from the premise that the key objective for colour management in printing photographic images is the reliability of perceptual results moving from display to paper (allowing for the inherent limitations due to differences of gamut and light between the two).
(1) So, turning to your comment "A" on the difference of the red/green balance between CM versus the comparator: which would be a more accurate rendition on paper? Will you be testing for that?
(2) Is there any contradiction between "A" and "C" insofar as in "A" the profile seems to produce less red in yellows and oranges?
(3) In "B" you say the curves are smoother with PM, but in "C" you say the greyscale step wedge is perfect. Is there any contradiction between these observations? When you print the image using the CM profile, is there visually less smooth tonal gradations on paper from a CM profile?
(4) In "D" where you say the profiles are "much more than acceptable", have you had a chance to rank them against say Epson's latest series of canned profiles for their recent crop of professional printers, or a custom profile generated by the HP Z3100?
(5) Re "E" I simply don't understand the meaning of this paragraph. Probably because I don't own a CM (yet) so I'm not familiar with its optimization routines and therefore I don't have an idea of what they are supposed to be doing. Grateful if you could elaborate these observations in case there are also other interested parties in my situation.
Mark
(Edited to change back from program-induced smileys to the actual letters I intended to show here!)
[a href=\"index.php?act=findpost&pid=222862\"][{POST_SNAPBACK}][/a]
I better start off clarifying ; this is the first set of profiles, both monitor and printer by no means extensive, nor exhaustive . Rather quickly done in between other things, I'll add.
1} Yes I will test everything the way I usually do. It is a bit too bad that I didn't beta test with Andrew and others. We all seem to find things and share our findings. Without that support I'm surely going to miss things that were shared on beta testing.
The colours I saw that were not right are on the hue angles pure zones as seen in ColorThink 3D graphing.
The colours I see in the print are more an overall perception of hues outside of say pure red, in between, same with the saturation of most colour.
The prints show things like skin tones that are a bit too red in the image as less red, and reflections in hair that should be blonde with some red, are towards cyan /green.
PAtches in the Fogra MW are lighter than they should be, I didn't measure the strip yet, didn't print to full scale yet either.
I would attribute this to both the LED spectrum of the CM, and the paper and ink being used. The paper is HP Advanced Glossy, with the HP pigments. A lot of OBA, and HP inks are quite prone to bronzing. So I think the bigger fault would be HPs inks and media with a spectro quite different than an i1 non LED. PErhaps the CM soft takes into account UV in a different way than PM.
2} Yes and no. Colorthink is good but the graphs only represent where the grid points create a volume within the hue angles (well a*b* plots to be more precise). The colours on the print are proof of application so I prefer to reference the print to monitor rather than assume too much from a 3D plot. I'll test more soon and really look at the colours and measure them too.
3} True the CM profile has a bumpier curve than PM5 for grey builds. Yet they are well handled in the profile, and seem to fit well in the LUTs. So less patches causing a few meanders in the grey build where they are aligning data samples to grid points hasn't seemed to create problems in greyscale printing. Again I'll have to print some large long tone prints to see for sure.
4} Hmmm. Form what I have, what I've heard, what I've read the Epson profiles are very good, and I think I saw that Epson USA made profiles with X-rite soft. HP use an internal custom program with some interesting quirks. their APS program is PM/i1Match hybrid with the Logo 6.x libraries so essentially PM5 with tweaks for the IsIs LED spectro. I made better profiles with PM and hand measuring a 918 and PM 5 than APS in the Z. But marginally so!
So no I don't think the 1st profile I made is as good as PM5, nor the quality of the recent Epson profiles, but close enough for professional use. The differences are not any greater with the CM than there is between other profile apps, and other spectros. Hence being thrilled with the quality, (sincerely I didn't expect it to be all that good) for a one off, first time profile that makes a very good screen to print match with 2 letter sizes pages, it is really incredible. A trained eye will pick out the differences but a vast majority of users wouldn't, nor should they as there seems to be little fault with my first try.
5} Yes I'm in the same boat. Actually anyone who uses a CM will be. X-Rite are doing themselves harm with oversimplification in the UI. IF they put the right options in and explained a bit about the options , users would step right up to the solution they need. UI design is still good, it's the missing options, and lack of details that take away from the max potential of the device.
Optimisation is adding a new page of patches (nodes if you like) to the first iteration of the profile. It is done by loading a jpg or tiff of any image, or a color palette already made in the color picker application. So if your greens are out, you load a landnscape and it builds a set of patches from the image and you print and measure.
The nodes are calculated and somehow incorporated into the profile. I don't know if they are added as grid points, or if they are moved , extrapolated, concatenated with previous existing points. Form what I did, the new prints were near the same as the old first profile. IT therefore looks like it simply over-samples and makes corrections to the first profile grid points but Andrew probably knows.
What is surprising, ; you print a page that you measure so fast you wonder if it really can work at all. It creates a subset, you print, measure , it calculates the profile and your done. You get a profile that is near large sample chart measured with much more expensive devices, and software costing much more. I'm impressed. I want to make some more. This little thing is fun. IT is what the Colortron should have been. Yet it couldn't have. I suppose with a bug fix update, possibly an upgrade this can only get better. A suivre.....