[Tool] ChromIQ – a macOS and Windows GUI for ArgyllCMS printer profiling (v4.3.0)

[Im-back-Arno]

Hi all,

Following progress while I have a moment but still traveling.

@itsab1989 : I wouldn’t worry about currently 40 users. It’s still under development and changing too rapidly to be spread widely. There will be a large user-base for it once it’s more stable. You could possibly even consider eventually licensing it for commercial use to the x-rites, calibrites and paper manufacturers out there for many coffees . Retain free for non-commercial use only. That way, it will spread (because it’s free) and you might eventually make some money for all your efforts, while us hobbyists will still be able to enjoy the fruits of your work. This is the way some software companies bring in new paying users. For example, I’m using Autodesk Fusion for free for my hobbies, but when I would use it for my company, I would need to, and would, pay. ChromIQ seems to quite generously beat the options that i1Profiler is providing and software development in a corporate environment is silly expensive, so there could be commercial interest. They might want to add a shell with their logo on it, but do we care?

Anyway, with that said, I have yet another question: I recall reading about ink limit calibration, which is hidden but can be activated in the preferences (?). Do I want to do that? I could imagine that when switching from dense Epson paper to a spongy, say, fancy Hahnemühle photo rag, that the amount of ink that gets absorbed could differ. Hence, it seems to make sense to first determine what amount of ink creates 100% density for each cartridge, no? It could be enough to tell my R1900 whether it’s printing on glossy or matt, and it might be sufficient, but for off-brand papers it might matter more(?).

Arno

 

[itsab1989]

Hi Arno!

Actually I checked again yesterday and it seemed that I got the numbers wrong. The latest version at that point had a little over 100 downloads. That’s progress!

Today we released 4.3.0 which focused mainly on the report feature that Knut designed and as always should squeeze some bugs. Also as always once we decided to drop the beta tag immediatley another little issue appears and the first updates follow…
But my AI subscription runs out tomorrow I think and I will not re-subscribe for another month. Instead I will give it some time and collect some feedback.
There might be more things one could add but I am pretty confident that it gets its basic job done pretty well (and actually a lot more).
When a few things added up I might work on it again to fix and add things but for now I finally want to take the break that I planned to take so many times by now. And I have a feeling that the same is true for Knut to a degree. Such a project can be fun at times but also mentally draining. And as a side effect I could not use my computer for anything else for the last few weeks because it was running automated tests day and night.

I will not be able to license this app though because it relies on ArgyllCMS which is free and does not allow this. Which I think is fair. I mean although we built „just“ a GUI around this, extended the functions of it in some cases and had the help of Claude artificial intelligence and Knut real intelligence this was quite a lot of work. Argyll itself is… like 20 years old? And I might be wrong but I think Graeme Gill seems to have created this completely on his own with no such help. So it would feel weird to sell this to people which was never my plan from the beginning.
I had issues using the command line and Knuts script finally enabled me to create mir first own printer profile with Argyll and improved my prints a lot. His script seriously did not get enough praise for how good it is. But I think people who were comfortable with the command line had mostly already found a way to work with it and to most others even the script was too „scary“.
That’s why I wanted to make this GUI. And I am more than happy with it.

But since I have started I paid at least around 100€ for this project every month. In September I subscribed to the highest tier Claude account which was 200€ even and I bought the CR30 spectro which cost me another 60€. Of course this is in part a hobby and if I could not afford it I would not have spent all of this money. But I also drags you in somehow because once you have paid you want to use as much of it as you possibly can and you have to follow the rules of the AI company with their usage limits. This time we are at a good point to make a cut for now and take a break as I said.

When you are back from travelling and had some time to play with it I would ask you for a favour though (just if you want of course - please don’t fell pressured): You have a very nice blog on your website and if you are looking for a topic for a new post in that blog maybe you could share some of your experiences on how to improve your black and white printing. This would make me very happy!

Regarding your other question: the calibration function is mainly useful for true CMYK (+N) printers and not for consumer printers. Of course they are using cyan, magenta and yellow as well but are considered RGB printers. I have added those options for completeness but I had no real way to test them and you probably would not really benefit from it. But more experienced users here might know more about this.

 

[Im-back-Arno]

Hi Arno!

Actually I checked again yesterday and it seemed that I got the numbers wrong. The latest version at that point had a little over 100 downloads. That’s progress!

View attachment 19323

Today we released 4.3.0 which focused mainly on the report feature that Knut designed and as always should squeeze some bugs. Also as always once we decided to drop the beta tag immediatley another little issue appears and the first updates follow…
But my AI subscription runs out tomorrow I think and I will not re-subscribe for another month. Instead I will give it some time and collect some feedback.
There might be more things one could add but I am pretty confident that it gets its basic job done pretty well (and actually a lot more).
When a few things added up I might work on it again to fix and add things but for now I finally want to take the break that I planned to take so many times by now. And I have a feeling that the same is true for Knut to a degree. Such a project can be fun at times but also mentally draining. And as a side effect I could not use my computer for anything else for the last few weeks because it was running automated tests day and night.

I will not be able to license this app though because it relies on ArgyllCMS which is free and does not allow this. Which I think is fair. I mean although we built „just“ a GUI around this, extended the functions of it in some cases and had the help of Claude artificial intelligence and Knut real intelligence this was quite a lot of work. Argyll itself is… like 20 years old? And I might be wrong but I think Graeme Gill seems to have created this completely on his own with no such help. So it would feel weird to sell this to people which was never my plan from the beginning.
I had issues using the command line and Knuts script finally enabled me to create mir first own printer profile with Argyll and improved my prints a lot. His script seriously did not get enough praise for how good it is. But I think people who were comfortable with the command line had mostly already found a way to work with it and to most others even the script was too „scary“.
That’s why I wanted to make this GUI. And I am more than happy with it.

But since I have started I paid at least around 100€ for this project every month. In September I subscribed to the highest tier Claude account which was 200€ even and I bought the CR30 spectro which cost me another 60€. Of course this is in part a hobby and if I could not afford it I would not have spent all of this money. But I also drags you in somehow because once you have paid you want to use as much of it as you possibly can and you have to follow the rules of the AI company with their usage limits. This time we are at a good point to make a cut for now and take a break as I said.

When you are back from travelling and had some time to play with it I would ask you for a favour though (just if you want of course - please don’t fell pressured): You have a very nice blog on your website and if you are looking for a topic for a new post in that blog maybe you could share some of your experiences on how to improve your black and white printing. This would make me very happy!

Regarding your other question: the calibration function is mainly useful for true CMYK (+N) printers and not for consumer printers. Of course they are using cyan, magenta and yellow as well but are considered RGB printers. I have added those options for completeness but I had no real way to test them and you probably would not really benefit from it. But more experienced users here might know more about this.

Hi Sebastian,

I didn’t know the history of Argyll, nor its limitations regarding use. In my ignorant mind I thought that also i1Profiler must have been built on it. Alas, no ROI then, so I’ll buy you a coffee . I’ve done a decent amount of programming in my time (Assembler, Basic, Visual Basic, Pascal and a bit in C) so tend to think that I understand what it takes, but I never had the luxury of AI to help with that. What you (with the help of others) pulled together is truly impressive though, AI or not, and I really do appreciate it very much that it is available to me and just at a time when I decided to turn some of my work into printed form. Respect and hats off to you and everyone who has helped!

Of course I can, and will, write about it on my blog, even when I don’t think my website has a lot of reach. I’m also well connected to MrLeica (google that) who has much more reach, as well as Alain Oguse, who seems to be well connected to the QTR world. I’ll try to do some PR where I can. ChromIQ deserves it! Perhaps you can consider publishing an article about ChromIQ in a suitable magazine or journal as well(?).

Anyway, I’m itching to be back home and find time to properly dig in. For sure I’ll provide feedback here about my findings from an end-user viewpoint.

Thanks again and, yes, take some quality time with your family!

Arno

 

[know]

Hi All !

Now that the community, and the world, has the awesome capabilities of ChromIQ available, with a myriad of pre-created chart presets, I would like to challenge the community to help out to verify the quality of the most important presets and their differences, so that everyone can learn from how certain differences in focus in the patch sets in a chart affects the result on paper, and that we all become confident in what gives the best results without overdoing number of pages or overdoing certain traits of a patch set.

For example, how many steps of Near-neutral double-rings are good enough, so that patches can be used on other things than filling unnecessary number of patches into the grey-steps (which demands a lot of patches).
I have an example here: Previously, in other posts on the forum, also associated with my old Argyll_Printer_Profiler script, several people reported great results using the TC9.18 target with Spyderprint greys using around 1260 patches (one example in this post done by pharmacist: post-148565). That target had 16 steps of near-neutral greys with two rings.
The preset in ChromIQ called "i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist", which has ben verified to give great results by both @pharmacist and Arno (@Im-back-Arno), if I remember correctly, uses 31 steps of near-neutral greys with two rings. I would argue, though not proven yet, that 22 to 24 steps of near-neutral greys with two rings is more than enough to give acceptable results without wasting patches on grey-steps that could be used on other parts of the chart patch set, especially given the good results from the TC9.18 target with Spyderprint greys with 1260 patches.

That was my theory when making several of the patch sets included in ChromIQ, but not verified, and begs to be answered through verification tests.

Specifically looking at the differences between
"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist" (left in the comparison image below)
and the other similar preset supplied
"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-Uniform 11x11x11-Edge Emphasis" (right in the comparison image below),
several questions would be really nice to get answers to, which I go into below...

Here is a comparison of the two presets patch distribution, using the "Show patch distribution 3D" tool in ChromIQ:

The key differences:

Preset Name (Target) / Characteristics	i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist	i1Pro · A4-1944p-3pages-Portrait-w7.5mm-Uniform 11x11x11-Edge Emphasis
No. of patches	1944	1944
Main shape	10x10x10 Cube (1000 patches)	Uniform 11x11x11 Cube (1331 patches)
Grey steps	32 steps near-neutral grey with 2 rings + 32 neutral greys (572+32=604 patches)	24 steps near-neutral grey with 2 rings + 37 neutral greys (432+37=469 patches)
Edges, corners and extremes	Only Corner focus with 48+18=66 patches, not for all edges.	Strong distributed edge emphasis (144 patches)
Additional features	- Strong addition of Skin tones (144 patches) - Additional Highlights and Shadows (12 patches) - Rest of patches distributed across various generators (Oceans, Foliage, Sunrises, Flamingos, Pastels): 24+24+24+24+18=114 patches	- Cube with uniform distribution of patches
Summary:	The reduced cube dimension of 10x10x10 and less edge definition, compared to the other preset, gives 331+78=409 patches more to use for additional focus on grey steps, skin tones and corner extremes (including black and white), and distributes the rest unevenly using the focused generators. Possibly giving an advantage for images with people in them.	The one higher cube dimension and full coverage of edges gives an evenly distributed patch set and good edge-definition for general use, giving no special consideration to a particular feature of an image (such as skin colors). The grey steps are lower than the other preset, but still significantly higher than the well tested TC9.18+spyderprint greys (24 steps is smack in the middle between 16 and 32 steps).

My questions are:
1. How much better, if at all visible, is 32 steps near-neutral grey with 2 rings compared to 24? How much is good enough for color prints versus black and white prints? Can one see a difference between them, or is 24 steps sufficient for expert quality prints?

2. Is there a clear visible advantage for images with people in them using the 10x10x10 cube with added focus on skin-tone-patches, compared to the uniformly distributed 11x11x11 cube?

3. Is there a visible difference having stronger edge definition for the Uniform 11x11x11 cube preset?

4. With some extra Highlights and Shadows for the "Real World Target-by Pharmacist", is that visible on print comparing the two?


I would be really interested to see this type of comparison for several patch sets, especially those most interesting for users, so that everyone could get a better understanding of what is good enough for their use when selecting profiles in ChromIQ, and also enable users to better build chart presets that suit there needs, knowing how certain changes in the patch editor will affect the resulting prints...

I hope the enthusiastic users of this forum can help out and make such comparisons, providing the scientific results for all of us to benefit.


Other comparisons that could be done too:

"i1Pro · A4-1296p-2pages-Portrait-w7.5mm-(medium quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-1296p-2pages-Portrait-w7.5mm-Uniform 9x9x9-Edge Emphasis"


"i1Pro · A4-648p-1page-Portrait-w7.5mm-(standard quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-648p-1page-Portrait-w7.5mm-Uniform 6x6x6-Edge Emphasis"


"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-3888p-6pages-Portrait-w7.5mm-14x14x14-Skintones-Edge Emphasis-Plus"


"ColorMunki · A4-300p-1page-Landscape-w9.0mm-TC3.00 Target-by Pharmacist"
vs.
"ColorMunki · A4-300p-1page-Landscape-w9.0mm-TC3.00 Equivalent Target (ChromIQ Editor)"
vs.
"ColorMunki · A4-306p-1page-Portrait-w10.0mm-Slow Reading Speed"

Cheers,
Knut

 

[Im-back-Arno]

Hi All !

Now that the community, and the world, has the awesome capabilities of ChromIQ available, with a myriad of pre-created chart presets, I would like to challenge the community to help out to verify the quality of the most important presets and their differences, so that everyone can learn from how certain differences in focus in the patch sets in a chart affects the result on paper, and that we all become confident in what gives the best results without overdoing number of pages or overdoing certain traits of a patch set.

For example, how many steps of Near-neutral double-rings are good enough, so that patches can be used on other things than filling unnecessary number of patches into the grey-steps (which demands a lot of patches).
I have an example here: Previously, in other posts on the forum, also associated with my old Argyll_Printer_Profiler script, several people reported great results using the TC9.18 target with Spyderprint greys using around 1260 patches (one example in this post done by pharmacist: post-148565). That target had 16 steps of near-neutral greys with two rings.
The preset in ChromIQ called "i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist", which has ben verified to give great results by both @pharmacist and Arno (@Im-back-Arno), if I remember correctly, uses 31 steps of near-neutral greys with two rings. I would argue, though not proven yet, that 22 to 24 steps of near-neutral greys with two rings is more than enough to give acceptable results without wasting patches on grey-steps that could be used on other parts of the chart patch set, especially given the good results from the TC9.18 target with Spyderprint greys with 1260 patches.

That was my theory when making several of the patch sets included in ChromIQ, but not verified, and begs to be answered through verification tests.

Specifically looking at the differences between
"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist" (left in the comparison image below)
and the other similar preset supplied
"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-Uniform 11x11x11-Edge Emphasis" (right in the comparison image below),
several questions would be really nice to get answers to, which I go into below...

Here is a comparison of the two presets patch distribution, using the "Show patch distribution 3D" tool in ChromIQ:
View attachment 19327
View attachment 19328

The key differences:

Preset Name (Target) / Characteristics	i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist	i1Pro · A4-1944p-3pages-Portrait-w7.5mm-Uniform 11x11x11-Edge Emphasis
No. of patches	1944	1944
Main shape	10x10x10 Cube (1000 patches)	Uniform 11x11x11 Cube (1331 patches)
Grey steps	32 steps near-neutral grey with 2 rings + 32 neutral greys (572+32=604 patches)	24 steps near-neutral grey with 2 rings + 37 neutral greys (432+37=469 patches)
Edges, corners and extremes	Only Corner focus with 48+18=66 patches, not for all edges.	Strong distributed edge emphasis (144 patches)
Additional features	- Strong addition of Skin tones (144 patches) - Additional Highlights and Shadows (12 patches) - Rest of patches distributed across various generators (Oceans, Foliage, Sunrises, Flamingos, Pastels): 24+24+24+24+18=114 patches	- Cube with uniform distribution of patches
Summary:	The reduced cube dimension of 10x10x10 and less edge definition, compared to the other preset, gives 331+78=409 patches more to use for additional focus on grey steps, skin tones and corner extremes (including black and white), and distributes the rest unevenly using the focused generators. Possibly giving an advantage for images with people in them.	The one higher cube dimension and full coverage of edges gives an evenly distributed patch set and good edge-definition for general use, giving no special consideration to a particular feature of an image (such as skin colors). The grey steps are lower than the other preset, but still significantly higher than the well tested TC9.18+spyderprint greys (24 steps is smack in the middle between 16 and 32 steps).

My questions are:
1. How much better, if at all visible, is 32 steps near-neutral grey with 2 rings compared to 24? How much is good enough for color prints versus black and white prints? Can one see a difference between them, or is 24 steps sufficient for expert quality prints?

2. Is there a clear visible advantage for images with people in them using the 10x10x10 cube with added focus on skin-tone-patches, compared to the uniformly distributed 11x11x11 cube?

3. Is there a visible difference having stronger edge definition for the Uniform 11x11x11 cube preset?

4. With some extra Highlights and Shadows for the "Real World Target-by Pharmacist", is that visible on print comparing the two?


I would be really interested to see this type of comparison for several patch sets, especially those most interesting for users, so that everyone could get a better understanding of what is good enough for their use when selecting profiles in ChromIQ, and also enable users to better build chart presets that suit there needs, knowing how certain changes in the patch editor will affect the resulting prints...

I hope the enthusiastic users of this forum can help out and make such comparisons, providing the scientific results for all of us to benefit.


Other comparisons that could be done too:

"i1Pro · A4-1296p-2pages-Portrait-w7.5mm-(medium quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-1296p-2pages-Portrait-w7.5mm-Uniform 9x9x9-Edge Emphasis"


"i1Pro · A4-648p-1page-Portrait-w7.5mm-(standard quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-648p-1page-Portrait-w7.5mm-Uniform 6x6x6-Edge Emphasis"


"i1Pro · A4-1944p-3pages-Portrait-w7.5mm-(expert quality)-Real World Target-by Pharmacist"
vs.
"i1Pro · A4-3888p-6pages-Portrait-w7.5mm-14x14x14-Skintones-Edge Emphasis-Plus"


"ColorMunki · A4-300p-1page-Landscape-w9.0mm-TC3.00 Target-by Pharmacist"
vs.
"ColorMunki · A4-300p-1page-Landscape-w9.0mm-TC3.00 Equivalent Target (ChromIQ Editor)"
vs.
"ColorMunki · A4-306p-1page-Portrait-w10.0mm-Slow Reading Speed"

Cheers,
Knut

Hi Knut,

Small correction: The only presets I tried so far was a 3 pager for 10x15 from Pharmacist, which I couldn’t get to work, and a one pager for A4 that immediately gave me a quite neutral B&W print with good tonality (which promises something when I find time to do a 3 pager A4…). Based on the ongoing discussion on experiences from others, I concluded that around 2000 patches should suffice and that the 1944 from Pharmacist should likely be a sweet spot for B&W due to the emphasis on neutral greys. I didn’t test anything further yet though, but I ordered a 50 page pack of Epson Archival Matt (similar to Epson Enhanced Matt), because I didn’t want to blow through fancy Photo Rag until I knew what I was doing.

Anyway, when approaching this scientifically, unless there is some non-linearity in how various colors respond to calibration, I seems to me that giving more preference around neutral grey could perhaps be overkill and might actually introduce errors. I also don’t think that color calibration cares whether it’s skin tone, or grey, or something else.

What could happen with giving emphasis on a certain color range is that the other ranges suffer and longer interpolations have to be done there, which in turn could provide incorrect slopes towards the ranges that are emphasized in the patches. Not sure if I make sense, it’s mostly intuitive what I try to describe, but there is some math and rationale behind it: I assume that curve fitting is done with polynomials, and less smooth data require higher order polynomials, which don’t extrapolate well and can become unstable outside of the region where most emphasis is given on the fit accuracy. Hence, I wonder whether a simple equidistant distribution without local emphasis of more patch densities in preferred regions would not be better. The extreme example is my initial test: Full focus around neutral grey and never mind what the colors are outside of this since I want to print B&W anyway. This did not work at all.

For this reason, I think it would be interesting to compare a fully equidistant distribution across the entire 3D cube with a patch distribution that gives clear emphasis on a certain range, neutral grey being an obvious candidate. I’m perfectly fine to test this, but need to find time, which I’m not exactly drowning in.

And yes, I fully agree that ideally the user-base should test which profiles work best and report back so we can all benefit collectively.

Cheers,

Arno

 

[Im-back-Arno]

Hi Knut,

Small correction: The only presets I tried so far was a 3 pager for 10x15 from Pharmacist, which I couldn’t get to work, and a one pager for A4 that immediately gave me a quite neutral B&W print with good tonality (which promises something when I find time to do a 3 pager A4…). Based on the ongoing discussion on experiences from others, I concluded that around 2000 patches should suffice and that the 1944 from Pharmacist should likely be a sweet spot for B&W due to the emphasis on neutral greys. I didn’t test anything further yet though, but I ordered a 50 page pack of Epson Archival Matt (similar to Epson Enhanced Matt), because I didn’t want to blow through fancy Photo Rag until I knew what I was doing.

Anyway, when approaching this scientifically, unless there is some non-linearity in how various colors respond to calibration, I seems to me that giving more preference around neutral grey could perhaps be overkill and might actually introduce errors. I also don’t think that color calibration cares whether it’s skin tone, or grey, or something else.

What could happen with giving emphasis on a certain color range is that the other ranges suffer and longer interpolations have to be done there, which in turn could provide incorrect slopes towards the ranges that are emphasized in the patches. Not sure if I make sense, it’s mostly intuitive what I try to describe, but there is some math and rationale behind it: I assume that curve fitting is done with polynomials, and less smooth data require higher order polynomials, which don’t extrapolate well and can become unstable outside of the region where most emphasis is given on the fit accuracy. Hence, I wonder whether a simple equidistant distribution without local emphasis of more patch densities in preferred regions would not be better. The extreme example is my initial test: Full focus around neutral grey and never mind what the colors are outside of this since I want to print B&W anyway. This did not work at all.

For this reason, I think it would be interesting to compare a fully equidistant distribution across the entire 3D cube with a patch distribution that gives clear emphasis on a certain range, neutral grey being an obvious candidate. I’m perfectly fine to test this, but need to find time, which I’m not exactly drowning in.

And yes, I fully agree that ideally the user-base should test which profiles work best and report back so we can all benefit collectively.

Cheers,

Arno

Some further thoughts: It all depends on what Argyll is using to interpolate the data points. I know from experience that if one fits a high density data range with higher order polynomials in order to get an accurate representation in that region, and then also want to fit a few points outside of this region of interest, the fit in the region of interest will actually become less accurate. If Argyll interpolates linearly, or limits to a second order maximum, then the fit will be smoother, but at that point creating a higher density in a region of interest could be in vain.

Do we know how Argyll fits the measured data?

 

[know]

Hi Knut,

Small correction: The only presets I tried so far was a 3 pager for 10x15 from Pharmacist, which I couldn’t get to work, and a one pager for A4 that immediately gave me a quite neutral B&W print with good tonality (which promises something when I find time to do a 3 pager A4…). Based on the ongoing discussion on experiences from others, I concluded that around 2000 patches should suffice and that the 1944 from Pharmacist should likely be a sweet spot for B&W due to the emphasis on neutral greys. I didn’t test anything further yet though, but I ordered a 50 page pack of Epson Archival Matt (similar to Epson Enhanced Matt), because I didn’t want to blow through fancy Photo Rag until I knew what I was doing.

Anyway, when approaching this scientifically, unless there is some non-linearity in how various colors respond to calibration, I seems to me that giving more preference around neutral grey could perhaps be overkill and might actually introduce errors. I also don’t think that color calibration cares whether it’s skin tone, or grey, or something else.

What could happen with giving emphasis on a certain color range is that the other ranges suffer and longer interpolations have to be done there, which in turn could provide incorrect slopes towards the ranges that are emphasized in the patches. Not sure if I make sense, it’s mostly intuitive what I try to describe, but there is some math and rationale behind it: I assume that curve fitting is done with polynomials, and less smooth data require higher order polynomials, which don’t extrapolate well and can become unstable outside of the region where most emphasis is given on the fit accuracy. Hence, I wonder whether a simple equidistant distribution without local emphasis of more patch densities in preferred regions would not be better. The extreme example is my initial test: Full focus around neutral grey and never mind what the colors are outside of this since I want to print B&W anyway. This did not work at all.

For this reason, I think it would be interesting to compare a fully equidistant distribution across the entire 3D cube with a patch distribution that gives clear emphasis on a certain range, neutral grey being an obvious candidate. I’m perfectly fine to test this, but need to find time, which I’m not exactly drowning in.

And yes, I fully agree that ideally the user-base should test which profiles work best and report back so we can all benefit collectively.

Cheers,

Arno

Thanks Arno,
Sorry, I must have mixed you with another user

I agree with your reasoning from a logical viewpoint. However, just so you and other readers understand what I meant... Trying one of the profiles is good, but it is the specific comparison of two same-size chart presets that is the most valuable, as we can learn how the differences between them affects the results in both color and black and white, which we can learn from. Profiling only one of them, then we learn little in comparison. Profiling both, on same type of paper and on same printer, that gives us more data. If you are up for it, I think everyone would love to see the results of this posted here. I certainly would find it interesting.

Of course, whoever sees the potential of testing some of the comparisons I mentioned, and is willing to do it for the community, that would be very welcome and appreciated. What do we have a good tool for if not to use it to improve our understanding

Cheers,
Knut

 

[know]

Some further thoughts: It all depends on what Argyll is using to interpolate the data points. I know from experience that if one fits a high density data range with higher order polynomials in order to get an accurate representation in that region, and then also want to fit a few points outside of this region of interest, the fit in the region of interest will actually become less accurate. If Argyll interpolates linearly, or limits to a second order maximum, then the fit will be smoother, but at that point creating a higher density in a region of interest could be in vain.

Do we know how Argyll fits the measured data?

When I ask Google this is what I get as answer:

ArgyllCMS fits measured color data into a printer ICC profile using a specialized mathematical approach called Regularized Linear Splines, implemented through its internal RSPL (Regular Spline Library). It deliberately avoids using standard global polynomial equations or traditional high-order splines, which are prone to unwanted oscillations ("wiggles" and "overshoots") when modeling complex multi-dimensional color behavior. [1, 2]
According to the official ArgyllCMS Source Code Documentation, the math and implementation function through specific mechanisms:

1. Scattered Data to Regular Grid Interpolation
Printer test charts (.ti3 files) consist of scattered data points unevenly distributed across the device space. The ArgyllCMS colprof tool must map this scattered information onto the rigid, multi-dimensional grid structures required by standard ICC Color Lookup Tables (cLUTs). [1, 2]

• The Algorithm: The method is inspired by D.J. Bone’s technical paper, “Adaptive Multi-Dimensional Interpolation Using Regularized Linear Splines”. [1]
• Smoothness and Control: Regularized linear splines combine the flexibility of local linear interpolation with global smoothing constraints. This prevents the profile from over-fitting to instrument measurement noise while ensuring first-derivative continuity. [1]

2. Spline Smoothing and Error Tolerance
Instead of slavishly forcing the surface through every single imperfect measurement, colprof balances fitting accuracy with mathematical smoothness. [1]

• Handling Uncertainty: The tool employs the -r parameter (average deviation/uncertainty percentage) to guide the spline fitting.
• Higher uncertainty values force the algorithm to rely less on individual scattered points and smooth the collective grid response further, suppressing localized device variations or instrument errors. [1]

3. Multi-Dimensional Inversion (Reverse Interpolation)
Creating the inverse B2A tables (mapping the Profile Connection Space back to device values like CMYK) requires finding a pathway back through the forward model. [1, 2]

• The RSPL system handles this via reverse multi-dimensional spline interpolation.
• It natively solves complex boundary and multi-valued problems during this inversion, such as managing Total Area Coverage (TAC ink limits via -l), calculating the black ink generation locus (-k/-K), and executing gamut boundary clipping. [1]

 

[ajaytanna]

I’m using ChromIQ with a Chnspec CR30 to create a CMYK printer profile for PrintFab. I successfully created a 4-channel CMYK calibration .cal file, and with Preset = None ChromIQ successfully generates a 450-patch CMYK chart.
However, selecting Apply & Embed (-K) fails with:
[ERROR] ChromIQ layout engine: calibration has 4 channels, target has 3
[ERROR] Chart generation failed.
None and -I generate the chart successfully. Is this a bug in the -K implementation, or am I missing a setting required to make the target 4-channel CMYK?