Good Profile vs Bad Profile

[Bill W]

Greetings,

I have run a few profile runs on presses in our Texas plant. The last run produced a strange looking profile. Until this last run I have been using a custom chart with 992 patches - I am a narrow web flexo printer and "regular" charts do not fit on my press web. This last run I built a reduced IT8.7-4 which fits on my web, but unfortunately can only be read in patch mode using an iO.

The IT8.7-4 chart I used to reduce also had the P2P with it. I built two reference files from the chart so I could scan them separately, for testing purposes. While the 2 new references files, I made, in some places have values such as 69.9%, where the original had a value 70%, I felt that they represented the IT8.7-4 and the P2P properly.

The profiles I made from the last run using the reduced IT8.7-4 are not as smooth as the previous profiles made from my 992 chart. I have attached two screen shots of each, labeled Good Profile 1, Good Profile 2, (from 992 chart), and Bad Profile 1 and Bad Profile 2 (from the reduced IT8.-4 chart).

Anyone have a clue as to why Bad Profile looks like it does. Not only does it have a strange gamut projection, but it also appears to have some strange dips in the lines between values as the color goes from dark to light.

The only thing we have changed between the two runs represented here, besides the chart, is the anilox rollers. They are brand new and not much different in configuration from the ones used on the first run.

-Bill-

 

[Bill W]

Good Profile vs Bad Profile 2

Good Profile vs Bad Profile 2

Greetings,

I attempted to attached pictures of the two profiles, but apparently I am "attachment challenged". So, I will try again.

-Bill-

 

[Bill W]

Good Profile vs Bad Profile 3

Good Profile vs Bad Profile 3

Greetings,

Sorry, I have fixed my "attachment challenges", now I just have to make sure I submit the correct attachment.

In my last post, Bad Profile 1 is incorrect, I have attached the correct Bad Profile 1 to this post.

-Bill-

 

[dherder]

Hi Bill,
When you say you "reduced" the chart, what do you mean? Was it scaled at all, or was the patch count just reduced?

Also, a few questions:
What profiling software was used?
Can you email the measurement data? I'd be interested to see if this happens with other profiling packages.
The IT8/7.4 should have 1617 patches. I am assuming that there was some difference in the magenta area of the gamut between press runs, and that because there are fewer patches (maybe in the saturated a* region?), the gamut that was computed was a "best guess" at device behaviour in that region.
It may make more sense to actually create two forms and run one with half the patches (~808) and then run another form after (the other 809) and perform averaging on several sheets to get a more accurate representation of the device condition.

regards,
dave

 

[Bill W]

Greetings Dave,

I started with an IT8/7.4 that I received from CGS that included the P2P. It has a total of 1960 patches. Misus the 275 for the P2P, and the "extra" white patches in the last row, the target has 1642 color patches - a few more than the standard IT8/7.4. This may have something to do with the CGS needs for certain patches for use in it iteration process - I got my self in a bit of a pickle earlier with my custom target not containing those certain patches.

I separated the P2P from the rest of the patches in Photoshop, removed the gaps between the IT8/7.4 patches and moved some patches around replacing most of the extra white patches so the new target had 55 rows, instead of the original 56 rows. I then presented this file to ColorLab and created a reference file. I also created the separate P2P reference file so I could scan both parts of the target separately. The P2P was also massaged to remove the gaps and a few patches moved to make 55 rows.

My next step was to recombine the P2P and the IT8/7.4 (with extra patches) in to one file and format it to fit my press sheet by reducing it (not deleting any patches).

I then printed my new target through my CGS rip and measured the print for uniformity. It's gamut projection mapped smooth, like the good sample attached to my original post.

While the "bad profile" magenta was right at the bottom of the density I have discovered I need to run our inks, it was still 3.26 dE to the G7 solid magenta Lab spec. In my previous runs, I have keep the dE's of the solid inks under 2, as I find the G7 spec of 5 to be a little to high.

I used Profilemaker to make my profiles, and I have attached my measurement file for you to work with.

A few years ago when I started to play with color management I only had a DTP 41 to use. Being naive about the variations of a flexo press, I did create split targets that I ran back to back - not a good idea - way too many variations to average out. My experience is that it is best to fit the target on one set of plates.

Thanks for your interest.

-Bill-

 

[rich apollo]

ProfileMaker has the characteristic of being very precise - it doesn't perform as much smoothing to the data as some other profiling packages. This may be desirable to some folks, and not to others. As you've introduced more data points, you've introduced more opportunities for variance, or deviation.

That's my theory.

 

[dherder]

...format it to fit my press sheet by reducing it (not deleting any patches).

I think that is where your problem is. By reducing the target, you are introducing a huge measurement error into the .cgt file. If you are reading the chart with an i1iO in patch mode (hope it doesn't overheat!!!), then it looks like the device is reading into the adjacent patches, thereby giving erroneous measurement data.

If you broke the form into two runs (and did not scale the chart), you would avoid this problem. However, you have noted that you were unsuccessful with this in the past. Quite a dilemma here. If the patch order is randomized, I would imagine that if you did two runs, the variance may be avoided some; if you pulled several sheets from each run and averaged across the forms, you would get an even better result.

 

[Louis Dery]

Hi Bill,

I’ve made a profile (with PerfX CM Proâ„¢) with your dataset and I got a quite good profile, even if your data reported a peak of 14 DeltaE (patche L=32.95, a=-31.58, b=-2.09) and average deltaE of 2.4.

I would like to know what black generation you’ve use to create yur profile (TAC, K start, etc.) in order to compare with the results you got with PM.

Thanks!

Louis

 

[Bill W]

Thanks to All,

Rich, I agree that in my case too many patches might be more detrimental than helpful.

Dave - I would not have attempted such a thing as reducing a target, except I just had great success at a Cal Poly FIRST seminar doing it with a target made by one of their instructors that was reduced even smaller. It made quite a good profile and I used the same iO as I used for my "bad" profile. Yes, it took my iO 1 hour and a few minutes to read my target. Do you think the water I sprayed on it to keep it cool could have played any part . I have, and still do, averaged targets, anywhere from 5 to 20 from two runs with the plates flipped 180 degress, and even without splitting the target, I still get very wavy TVI curves. A challenge I have had discussions about before.

Louis - I made two profiles with different settings:
1 - GCR3; start 0; total ink 320
2 - GCR3; start 20; total ink 320
3 - I thought I made a third one using UCR, but I apparently dumped it when it do not look any better.
A little education please - I thought that the different settings in a profile did not affect the total gamut of a press, only how color would be converted if the profile was used to covert, lets say an RGB image to a CMYK image. I have made profiles with GCR, UCR, different settings, etc, and when viewed in ColorThink they appear to have the same volume and gamut projection. Am I missing something? Also, how to you evaluate whether a profile is good or bad?

Finally, I just finished a series of press tests using my old custom target on the same press. Bbefore doing the final runs, I measured one sheet of the setup run and made a profile. It looked really good, smooth with a clean gamut projection. Until I read the post from Louis in which he states he make a good profile from my "bad" measurement file, I was leaning towards a bad set of plate on the last run as the reason for failure. Now I am scratching my head again. I of course will be using this new press run for my profile, but still will be curious what caused the last the last press run caused the "bad" profile.

-Bill-

 

[Louis Dery]

Louis - I made two profiles with different settings:
1 - GCR3; start 0; total ink 320
2 - GCR3; start 20; total ink 320
3 - I thought I made a third one using UCR, but I apparently dumped it when it do not look any better.
A little education please - I thought that the different settings in a profile did not affect the total gamut of a press, only how color would be converted if the profile was used to covert, lets say an RGB image to a CMYK image. I have made profiles with GCR, UCR, different settings, etc, and when viewed in ColorThink they appear to have the same volume and gamut projection. Am I missing something? Also, how to you evaluate whether a profile is good or bad?

Finally, I just finished a series of press tests using my old custom target on the same press. Bbefore doing the final runs, I measured one sheet of the setup run and made a profile. It looked really good, smooth with a clean gamut projection. Until I read the post from Louis in which he states he make a good profile from my "bad" measurement file, I was leaning towards a bad set of plate on the last run as the reason for failure. Now I am scratching my head again. I of course will be using this new press run for my profile, but still will be curious what caused the last the last press run caused the "bad" profile.

-Bill-

Hi Bill,
My question about the black generation was not related to gamut but just to replicate the same type of profile in order to compare the conversion results. If you can send me your "bad profile", I would like to compare with the results I got with mine.

Louis

 

[Bill W]

Greetings Louis,

Be glad to send you the profile - just need you email address.

-Bill-