[Tool] ChromIQ – a macOS and Windows GUI for ArgyllCMS printer profiling (v3.9.0)

[itsab1989]

Well, I would call it unusable, similar to the other failures. The odd thing is, they are either very good or very bad. I've been looking at the profiles I've made with the LittleArgyllGUI and none of them are as good as the two successful ones with ChromIQ. I can send the ColorMunki files if they would help.

Send them and I‘ll let Claude check. Maybe it finds a hint.

 

[pkk]

Send them and I‘ll let Claude check. Maybe it finds a hint.

Okay, i just read Claude's report. He has mostly good news but I'm going to have to be led through by the hand to understand exactly what happened to avoid it happening again. Claude's explanation does not jive with the workflow I've seen.

The NatMatteV2 profile that you sent was not built from the measurement next to it. Your reading of that chart is actually very good — when I built a profile straight from your measurement data, it came out excellent too. The problem is that the .icc profile file you had was an older, leftover profile from before; it was never rebuilt from your final reading, so the profile and the measurement didn't match. A mismatched profile like that behaves badly even when the underlying measurement is perfect — which is exactly what you saw.

Why this can happen: in ChromIQ, measuring a chart and building the profile are two separate steps. After you finish reading all the strips, the app guides you to the Build Profile tab — but it stops there on purpose, because that tab is where you can adjust the profile settings before creating the ICC. You then have to click Build Profile on that tab to actually generate (or regenerate) the profile. If that final "Build" click doesn't happen — for example after re-reading a chart — the folder keeps the previous profile, and that older one is what gets used.
▎
▎ So the simple rule going forward: whenever you measure or re-measure a chart, finish by clicking "Build Profile" on the Build Profile tab to create a fresh ICC from that reading. That guarantees the profile always matches the measurement it came from.

I went back & tried to build another profile from some of my failures without success. There is only one T3 file to be used so i don't see what he is doing. I've attached two more sets of files if Claude would like to try them. In both of these I'm sure there is only one measurement file. Any chance my Argyll is corrupted? Thanks again.

 

[itsab1989]

Here is what Claude concluded:

▎ Hi,
▎
▎ Thanks for sending the two extra sets — they actually told me exactly what's going on, and the news is good: there's nothing wrong with your i1Pro, your ColorMunki, or your copy of the software. The same setup made your two excellent profiles, generated these charts correctly, and built every profile I tested without a single complaint. A broken instrument or a corrupted install can't give you perfect results part of the time — so we can take "corrupted Argyll" off the table completely. You're in good shape; we just need to pin down one step.
▎
▎ Here's the key thing I learned, and it explains why your rebuild attempts didn't help: your bad profiles aren't all failing for the same reason. There are actually two different problems mixed together — and they need two different fixes. That's almost certainly why my earlier advice didn't match what you were seeing on screen.
▎
▎ Problem #1 — the profile was never rebuilt from the final reading.
▎ This was the NatMatteV2 one I looked at first. Your measurement of that chart was genuinely great. The trouble was that the profile file sitting next to it was an older, left-over profile that didn't match that reading — it was never rebuilt after the final measurement. So the profile and the measurement disagreed, and a mismatched profile prints badly even when the reading underneath it is perfect. For that type, simply rebuilding the profile from the measurement fixes it — which is exactly what I did, and it came out excellent.
▎
▎ Problem #2 — the measurement itself has a few bad readings in it.
▎ This is what's happening with the two new charts (the Epson SemiMatte and the Pro1000 Gloss). And this is the important part: for these, rebuilding can't fix them — and that's not your fault at all. When I rebuild a profile, I'm building it from the measurement file. If that measurement has mistakes baked into it, every profile I build from it inherits those same mistakes. So you did everything right when you tried to rebuild — there was just nothing good to rebuild from. That's why it felt like hitting a wall.
▎
▎ What "bad readings" actually means here — and why it's so clearly a reading issue and not a paper or printer issue: in the Epson chart, a patch that was printed as a dark grey got recorded by the meter as pure paper-white. Another dark-grey patch was recorded as near-black. A dark blue came back logged as a light green. None of those are physically possible for a real patch on paper — the meter simply captured the wrong spot during the scan. That's a classic strip-reading hiccup, where the instrument briefly loses its place as you draw a strip through it. (It also explains why your ColorMunki struggled too — both meters read by dragging strips, so they're both prone to the same kind of slip. And it's why X-Rite's diagnostic passed: that test checks the optics, not how a strip glides across them.)
▎
▎ The good news is how contained it is:
▎
▎ - The Pro1000 / Gloss chart is almost perfect. Out of the whole chart, only a handful of patches are off — and five of them are in a single strip. Carefully re-reading mainly that one strip should rescue the whole profile. The software lets you re-read individual strips, so you don't have to start over.
▎ - The Epson / SemiMatte chart had a rougher scan — the slips are sprinkled across many strips, so this one is best re-measured from the top. It's not that the chart is bad; that particular reading session just didn't go cleanly.
▎
▎ So here's the simple game plan going forward:
▎
▎ 1. Re-measure the two charts above (full re-scan for the Epson; mostly just the one strip for the Pro1000). Re-measuring — not rebuilding — is the fix whenever the reading itself has slips.
▎ 2. Then build the profile from that fresh reading. Reading a chart and building the profile are two separate steps on purpose (so you can set your profile options first), so the last thing to do after any measurement is click Build Profile to make a new profile from that reading. That guarantees the profile always matches what you just measured — and avoids the Problem #1 situation entirely.
▎
▎ One more thing from the earlier round: I already rebuilt a corrected version of your NatMatteV2 profile from your (good) measurement of it — it self-checks beautifully, far better than the leftover one you had. Make sure you're using that corrected file rather than the original, and you should see that paper come right in line with your FineArtMatte result.
▎
▎ A few tips to keep strip-reads clean (these alone usually end the random failures):
▎ - Re-calibrate the i1Pro on its white tile right before you start reading a chart.
▎ - Pull each strip through at a slow, steady pace — rushing or uneven speed is the #1 cause of a slip.
▎ - Keep the chart flat against the backing board and make sure the ruler/guide is seated firmly so it can't shift mid-strip.
▎ - On satin or semi-matte papers, watch for sheen — it can occasionally confuse the meter's sense of where one patch ends and the next begins.
▎ - If a strip beeps oddly, re-reads, or just feels off, re-do that one strip on the spot. A few seconds then saves a whole bad profile later.
▎
▎ Bottom line: your gear is fine, and the "either great or terrible, nothing in between" pattern makes total sense now. The great ones were built fresh from a clean reading. The bad ones were either a leftover profile that never got rebuilt, or a scan with a few slips in it. Get one clean reading, build from that, and you'll get the great result every time.
▎
▎ Happy to look at any future ones if you're ever unsure — but I think you've got it from here.
▎
▎ Thanks again for the detailed files; they made this easy to track down.

 

[pharmacist]

@itsab1989 : the latest beta version for the patch generator has lost the shuffle possibility

 

[itsab1989]

@itsab1989 : the latest beta version for the patch generator has lost the shuffle possibility

View attachment 18943

The create chart tab now takes care of the layout. As long as you activate the randomize option there it should shuffle them on clicking apply

 

[pharmacist]

Ohh I did not see the shuffle button. Maybe i am overlooking it.

 

[itsab1989]

It’s quite a lot of new options and changes to the old behaviour so it takes a while to get used to. But I just tried and it worked at least on my side.

Just made a small update so the clip border content is not restricted by the margins, the overwrite option for targen settings is moved out of the section so it is visible even when the options are collapsed and added an option to auto update the preview when layout settings are changed to match what the editor had before

 

[Alan G]

@pharmacist - I used your 1944 patch set to do a profile for Moab Entrada. I was a little skeptical that the smaller patches would read OK with my i1 Pro but it was not a problem other than the occasional 'not enough patches read' error. The profile came out quite well and the gamut was pretty much the same as I got using the precondition method. The peak error was 1.86 and the avg error was 0.32. The high errors were all black or near black patches which is what I observe with matte papers. It might be a little high and go down if I rescan that row but am satisfied with the results. Soft proof was fine and I will do a couple of real prints to see how this works compared to the regular profile I have been using.

 

[pharmacist]

@pharmacist - I used your 1944 patch set to do a profile for Moab Entrada. I was a little skeptical that the smaller patches would read OK with my i1 Pro but it was not a problem other than the occasional 'not enough patches read' error. The profile came out quite well and the gamut was pretty much the same as I got using the precondition method. The peak error was 1.86 and the avg error was 0.32. The high errors were all black or near black patches which is what I observe with matte papers. It might be a little high and go down if I rescan that row but am satisfied with the results. Soft proof was fine and I will do a couple of real prints to see how this works compared to the regular profile I have been using.

Thanks for your update. @hackensacker seems to have good results too (compared to the preconditioning profile method). I hope you will get the same results as @hackensacker has obtained with this 1944 target. The "secret" is just a massive amount of sampling points on and around the neutral axis so any deviation of given printer/ink/paper combination from true neutral grey is estimated as good as possible without the need of a preconditioning profile. In most cases the behaviour is not straight forward but with a serpentine curvature from pure white to pure black. By having 2 rings around the pure grey axis the deviation from the pure grey axis can thus be accurately determined and corrected accordingly to compensate from the off-axis behaviour for a given R=G=B point. Thanks to @itsab1989 work and my suggestion to have that injection possibility of pure greys + near neutrals we now can produce targets that maximally emphasize that very important area of the RGB colour space. Not possible with i1Profiler and even not with targen of ArgyllCMS as you will need a secondary profile (with a preconditioning profile to determine the neutral axis) to describe this area. This way the profiling is faster (no 2 steps needed) to get a similar result. Actually the Spyderprint Extended greys target is based on this principle. Now we have AI that helps with an even better extended greys target injected in the target.

 

[itsab1989]

It is so good to hear that all of this work is useful for someone. I know Knut is following this conversation here as well and is probably happy about this, too.

Someone requested that I should add a GUI for collink as an additional tool. This was a pretty easy task so it is included in the latest beta. Maybe some of you know what this is really good for