A new profile issue. Help please...

[Onslow]

Yep. Bad scan for sure.

Here's a few of the patches where the wrong patch was read in:

Row 15, Col. 4, Page 1  RGB values (255,42,23) (somewhere around a bright, saturated red) came out Lab (24, -12, 2), a very dark cyan.

Row 15, Col. 16, Page 1 RGB values(255, 191, 115) (a desaturated light orange) came out Lab (48, -46, -25) a strong cyan.

In examining the row it is clear that the patches from those locations in row 16, instead of row 15,  were read in by error.


There is a fairly good way to avoid these sorts of mistakes. Scan each set of prints 3 times saving the intermediate measurements. Then run "Data Measurement Analysis" in I1 profiler using the first and second file. This will show statistics on how well the two sets of scans match for each RGB patch. Any misalignment between the two will stand out. Do this for the first and second set, first and third set, and the second and third set. If two of the three match each other closer than about 4 dE worst case and better than 1.0 on average just average  the ones that are closest to each other. Xrite has a description on how to average measurement files in their FAQs.

Doug, thank you for your info. I am still learning my i1Profiler software and was not aware of this ability to analyse and average the data. I shall be looking into this.
I did have an issue on that row you mentioned. I had an error beep to rescan, I did so, it error beeped again and then said I should be measuring row 16.... Thank you for confirming that..

[Onslow]

I agree with Doug on doing duplicate (or triplicate if you are really compulsive).  I use ArgyllCMS to do my profiles and I import the two scans into an Excel spreadsheet and do the comparisons to see if there are any really bad reads.  It also shows the deviation from what the patch actually should be.  You can expect some minor variations but if something is really off then you know what steps to take.

Well, I'm not really compulsive but I do want to get to the bottom of this. I'll do a couple more scans to see the results.

[Mark D Segal]

Hi Onslow - whoever mentioned to scramble the patches (an option in i1Profiler) I agree you should do this. I recommend you use the XRite 2033 patch set, scrambled patches, and do a reprint of the profiling target using ACPU with the correct Media Type for the paper you are printing. Let it dry for several hours. Then re-read the target - if you are using an i1Pro2, use it in dual scan mode so you can profile for M0, M1 and M2. In case you aren't aware, right after scanning the target pages, first make sure to do a Save Workflow so if i1Profiler crashes (which it can) you don't need to start everything from scratch. Then make the first profile with the default M0, then go back to the Measurement Tab, select M1 as the measurement mode, and then skip over to the Profile tab and create a new profile in that mode with a slightly different name. Repeat for M2. Then examine the profiles in CTP to see whether the surfaces and contours are smooth - no dents or jaggies. If so, make test prints with each one of the three profiles and see whether one is just right for that printer/paper combination.

[Doug Gray]

Here's the RGB and scanned patch Lab values showing column 3 on the first page, rows 12 through 18

189   D12     185.00   42.00   23.00      45.36   36.88   31.82
190   D13     208.00   42.00   23.00      49.77   46.64   38.79
191   D14     231.00   42.00   23.00      53.54   55.85   44.96
192   D15    255.00   42.00   23.00    23.84  -12.48   -2.35
193   D16       0.00   63.00   23.00     23.83  -12.46   -2.40
194   D17      23.00   63.00   23.00      25.75   -9.89    2.13
195   D18      46.00   63.00   23.00      28.73   -6.59    7.05

You can see that the patches on rows 15 and 16 are essentially identical with extremely different RGB values.

As an aside, the most difficult errors to find are where a sliver of an adjacent row is scanned from a misalignment. These sorts of things can produce seemingly good profiles and the only way to really prevent them is to make the patches longer (an adjustment in I1Profiler) so alignment is less critical and/or scan the patches multiple times and compare them looking for significant dE's.

[Onslow]

Hi Onslow - whoever mentioned to scramble the patches (an option in i1Profiler) I agree you should do this. I recommend you use the XRite 2033 patch set, scrambled patches, and do a reprint of the profiling target using ACPU with the correct Media Type for the paper you are printing. Let it dry for several hours. Then re-read the target - if you are using an i1Pro2, use it in dual scan mode so you can profile for M0, M1 and M2. In case you aren't aware, right after scanning the target pages, first make sure to do a Save Workflow so if i1Profiler crashes (which it can) you don't need to start everything from scratch. Then make the first profile with the default M0, then go back to the Measurement Tab, select M1 as the measurement mode, and then skip over to the Profile tab and create a new profile in that mode with a slightly different name. Repeat for M2. Then examine the profiles in CTP to see whether the surfaces and contours are smooth - no dents or jaggies. If so, make test prints with each one of the three profiles and see whether one is just right for that printer/paper combination.

Thank you Mark. I have done a 2nd Rescan. The quality seems better. I was not aware that if there were issues, I could manually select the row and rescan, I did this several times on the 2nd Rescan after I realised the device was tipping whilst reading BUT not generating an error. So, I manually selected the row to rescan.
I have attached some comparisons. A is the original damaged scan, B is the first rescan, C is the 2nd Rescan.
I would attach a mov of the files in the 3d Grapher to illustrate what I see but for some reason, Color Think Pro 3 won't says there is an error OR the file is unavailable when I go to create the movie.
I think I shall do one more scan of the targets and do a compare. Is there away in the software to then read the three good scans and combine them into on profile? Or, is that not a good idea??

[Doug Gray]

Both rescans are close. Very close. Either one is fine. You could make a profile from each one and print reference images and I doubt you would see any difference at all under even close inspection.

As for how smooth the 3D gamut is, I have not found that useful at all other than identifying some sort of major error such as you initially encountered. Printers tend to be imperfect at the gamut edges and some of this has to do with the "cube" style granularity of the 3DLUTs in the profile. However, when inside the gamut volume, which is where 99+% of the image will be, the most important attribute is smoothness and lack of bumpiness in the printed colors.  This cannot be determined from the #D gamut plot, profile or software that generated the profile. What you have to do is print a separate set of color patches, read them, and compare the color printed with the color requested. There is a tool that can do this called PatchTool by Babelcolor that is excellent at this. Alternately, Matlab has some toolboxes that automate much of this too but it's a bit pricey.  Patchtool lets you print a set of color patches using the ICC profile then read the Lab color values and compare them.

[Mark D Segal]

Both rescans are close. Very close. Either one is fine. You could make a profile from each one and print reference images and I doubt you would see any difference at all under even close inspection.

As for how smooth the 3D gamut is, I have not found that useful at all other than identifying some sort of major error such as you initially encountered.

Yes, that is exactly what the 3D gamut view is good for. It is the first step for identifying a seriously defective profile, and often enough to incite one to start over without further ado. That said, using other tools will help refine the diagnosis.