Printer profile testing patch chart

[rasworth]

I'm in the middle (targets are drying) of a 2025 patch printer profile generation.  My question is how should one construct a patch set to test the profile?  It's easy enough to test the neutrals, Doug had previously posted a 101 group convertible to i1Profiler scan format.  I could crank out a small printer profile patch set for the colors and add it on, but is that the correct way to exercise the profile?

I want to end up with a maximum of 405 patches (single letter size sheet with i1Pro patch size).

Thanks for input,

Richard Southworth

[Mark D Segal]

Just saw this - the notification went to my spam folder - very unusual; moreover I'm having an inconvenience with the notification system from LuLa - the threads are not hyperlinked in the notification emails as they used to be, so I need to copy-paste the references to a new browser tab. Don't know if others are also experiencing this issue.

OK, on topic, Richard, you don't need many patches to verify whether a printer/paper/profile combination is performing with adequate accuracy. A small sample suffices provided that all patches are within the gamut of the profile's gamut volume and there is adequate representation of hue and saturation. What's "adequate": I take inspiration from the various commercial producers/users of industry-strength process control targets that range from 24 to 77 patches depending on whose it is. If patches are OOG, two variables are being examined: (1) accuracy of ink laydown and (2) how Rendering Intent is being handled. If the main interest is (1), best that all the patches be within gamut, and that gamut volume of course varies considerably between gloss/luster and matte papers. I'm not sure it's too useful getting into sampling theory about what size of a sample is sufficient to represent the population, but the adequacy of these small samples becomes readily apparent upon using them and seeing the impact on print quality when using profiles that turn up differing dE test results that are large enough to impact our visual perception.

Perhaps the most convincing approach would be to experiment: construct a small set and a reasonably consistent large set, test a profile with both and see how much difference it makes to the results, in terms of whether the indicators are sufficient to show what you want to know.

[rasworth]

Mark,

Thank you.  I re-read your stress article, had pretty much come to the same conclusion, i.e. experiment with small and (somewhat) larger sets.

Now to scan in 2025 patches with an i1Pro...

Richard Southworth

[Alan Goldhammer]

Mark,

Thank you.  I re-read your stress article, had pretty much come to the same conclusion, i.e. experiment with small and (somewhat) larger sets.

Now to scan in 2025 patches with an i1Pro...

Richard Southworth

That's how I've spent my day!  Profiling my new Canon Pro-1000 and did the first four papers.  For those interested, Dry Creek Photo has a nifty replacement for the Adobe Color Print Utility that actually prints out the right size patches!!!!   https://www.drycreekphoto.com/tools/ChartPrinter/   The Adobe utility's patches are about 5% smaller than they should be under WinOS

[Mark D Segal]

No support, user's risk, Windows only. Good of them to put that up for Windows users. Being on OSX I'll continue with ACPU.

[Alan Goldhammer]

No support, user's risk, Windows only. Good of them to put that up for Windows users. Being on OSX I'll continue with ACPU.

Quite right.  ACPU works ine on MacOS.  A lot of us badgered Adobe about the WinOS problem and they pretty much didn't care and hence no fix of the problem.  One can still read patches using the manual ruler that ships with i1 instruments but positioning needs to be carefully done.  I used the Dry Creek program for yesterday (and today!!) reads and it was a pleasure seeing the 'real' size patches.

[rasworth]

I've been printing targets thru i1Profiler with no issue, on W10.  As a slight convenience it handles multi-page targets automatically, don't have to feed the pages in one by one.

Richard Southworth

[Alan Goldhammer]

I've been printing targets thru i1Profiler with no issue, on W10.  As a slight convenience it handles multi-page targets automatically, don't have to feed the pages in one by one.

Richard Southworth

I take the cheap route and use ArgyllCMS so I have to find an app to print targets.

[rasworth]

Alan,

You can use i1Profiler, the target generation and print is in the free section.  It's an useful program to have on your system even without a license, can measure targets, output in cgats, etc.  The only time I plug in my dongle is to actually generate a profile.  I am assuming the same holds true for a Mac.

https://www.xrite.com/service-support/downloads/i/i1profiler-i1publish_v1_8_2

Richard Southworth

[Mark D Segal]

I am assuming the same holds true for a Mac.

Richard Southworth

Yes.

[rasworth]

Ok, back to the main topic.

I created the 2025 patch target using Doug's latest generator, with parameters -m 10 -l 5 -n 263 -s 2 -r, translates to outside grid dimension 10, double density dark grid dimension 5, 263 neutrals/near neutrals, with a previously generated training printer profile.  The 263 neutrals number was chosen to exactly total out at 2025, which fills 5 letter sheets with iPro1 size patches.  Attachment 1 shows the patch set in the training profile space.  Generated the profile with i1Profiler, input with i1Pro uv cut.

I'm starting profile testing with color patches, will do neutrals later.  Being basically lazy I looked for an automated way to generate a reasonable set of colors, ended up using BabelColor.  It has a clever patch generation tool, that iterates thru L*a*b* space, with the option of retaining only those values that are inside an arbitrary profile gamut.  After playing with the "knobs" I ended up with 214 L*a*b* values, distributed thru the profile space.  The actual list is the second attachment, and the graph of the colors distribution in the printer profile space is the third.

BabelColor also provides the ability to print out a scannable 16 bit tif of the patch list (thanks to Doug for pointing this out), which I then brought into Photoshop and printed thru the profile, absolute rendering intent.  After drying I scanned the image with the same i1Pro uv cut, again using BabelColor, translating the spectrum data directly to L*a*b*.  Compared the values to the input list generating 214 dE2000 values, average was .57 with 10 values above 1.0 and less than 1.6.

So far I'm happy with my profile, I'll see how well it does with neutrals.

Richard Southworth

Added by edit - the BabelColor program is PatchTool.

Second edit - oops, n = 163, not 263.

[rasworth]

I should have stated before, I would like to hear comments on the adequacy, or lack of it, of this approach.

Thanks,

Richard Southworth

[Mark D Segal]

It's probably more than adequate.

Taking inspiration from the process control industry (and it is an "industry"), the IDEALLIANCE process control strip has 84 patches including neutrals and non-neutrals. It's a CMYK construct, but I supposed the same colours could be useful for RGB printers that also spray flavours of CMYK inks. About 12 of its patches are OOG for Ilford Gold Fibre Silk in an Epson SCP5000 (very wide gamut combination), and about 17 or more patches OOG for matte papers having volumes in the 500K-575K range. So using this strip for measurements of in-gamut colours alone would require adjustments.

The Barbieri control strip for their (really expensive) colour management systems is 48 patches, which they obviously have determined to be adequate for colour control verification purposes.

Absent a basis in sampling theory specific to this purpose, it's really hard to say how many patches one really needs to satisfy oneself that a profile is performing at an acceptable level of accuracy. Perhaps the more the merrier, but at what point does the marginal utility of more patches top-out? Could be a low number. Like 24 well thought-out values such as we have in the venerable ColorChecker that has been used for decades to do this kind of thing. So your generating over 200 can't be harmful. I'm only raising a second-order question of how necessary for the purpose.

I think you are very well covered here. The distribution relative to the gamut space you used looks eminently coherent and the methodology seems fine to me.

[rasworth]

Mark,

Appreciate the reassurance.  I just printed out 101 neutrals, generated with the same methodology in PatchTool.  I guess if I had been a little more clever I would have stitched it and the color patches together, would still fit on a single sheet.

I agree it's probably overkill, but the difference between scanning 80 patches vs. 300 is just a few minutes.  I'm really impressed with the BabelColor program, although one could say I suffer from SAS (software acquisition syndrome), quick to pull the trigger to buy once I saw a reference.

Richard Southworth

[Mark D Segal]

I have Patchtool also (same syndrome). It's a useful application, though I have to say from my periodic usage, I find its operation with OSX can be quirky.

[rasworth]

I printed and measured 101 neutrals, test image attached (also useful for monitor visual neutrals test).  Discarded the first and last six patches as out of range for the paper, average dE2000 on the remaining 89 was .49, three patches over 1, all at the low end.  Here are the first 15 dE2000 values:

L0   3.4733588
L1   2.5086567
L2   2.4167779
L3   1.5070952
L4   1.0084492
L5   1.3736149
L6   1.07466
L7   0.9987372
L8   0.8995739
L9   0.374725
L10   1.0301797
L11   0.3951753
L12   1.6917682
L13   0.3903298
L14   0.6377794

L* = 12 was the obvious worst at dE2000 = 1.69.  I'm curious as to why the "bumpiness" at the low end, do profilers generally have trouble in this region?

I believe the profile should be good for printing anything, probably overkill for my humble Pro-100 but justifiable as this whole process has been educational for me.  Thanks to Doug for his patch generator, and to both Doug and Mark for their helpful comments.

Richard Southworth

[Mark D Segal]

Is this for matte or glossy/luster paper? If the latter, by L12 the results should be quite linear. If matte, L12 would generally still be below the distinguishable threshold for an Absolute Rendering, as for most Matte papers maximum black usually falls in a range of L14-L18. Have you tried remeasuring that one patch? Sometimes these outlier results occur because of a measurement glitch or the spot measured in that patch has some kind of irregularity not shared by the rest of the patch.

[rasworth]

Mark,

It's Canon semi-gloss, minimum l* approx. 5.  I also saw this when I did a Black and White emphasis profile, with 256 neutrals thru Doug's 1st program.  I'll go back and check measurements.

Richard Southworth

[rasworth]

Re-measured the test print, same results.  Things look better if I choose dE76 instead of dE2000, but still a little bump on that patch.  I am using an i1Pro, which I believe is less accurate at the low end than some of the newer instruments.

Richard Southworth

[Mark D Segal]

Re-measured the test print, same results.  Things look better if I choose dE76 instead of dE2000, but still a little bump on that patch.  I am using an i1Pro, which I believe is less accurate at the low end than some of the newer instruments.

Richard Southworth

I'd be surprised if it were the instrument, Richard, as if it were, I would expect other erratic readings elsewhere along the scale. The main difference between dE76 and dE2000 is that the latter attempts to correct for the non-uniformity of human visual perception. One does find that sometimes the dE76 readings return lower numbers than the dE00 readings, but more often than not it's the reverse. But the fact that both readings still produce an irregularity in that one patch could mean that the printer didn't print that particular value in as linear a manner as it was supposed to, for who knows what reason. I'd bet that if you reprinted the target you may get a different outcome. Happens.