CPC - Color Perception Confusion (1 Viewer)

[Ray-Col]

CPC = Color Perception Confusion


My own color samples arrived this week. I was seriously confused at first and will share some of my (more obvious) observations below:

- most colors are as before seen, some are quite different
- all Dayglos (except orange) are waaay more vivid/intense than the samples I got from @PatTheCat a while ago, barely sitting in AdobeRGB (maybe) but definately outside of sRGB.
- The whole line frome Blurple to Retro Lavender is more differentiated - just bigger color-gaps between them
- Maroon is a dark but rich brownish red, not nearly as dark as before
- The Bright White is much lighter as seen before - really clean and bright
- The brass flakes in weighted colors are more coarse than in the first samples I've seen - seems like I have to overhaul my material definitions

Quick notes to my mold samples

- I have six different shades of white in my hands - seven if I count the one from the color samples. Photos don't do justice to the slighter variations but from the snap you should clearly distinguish three.
- The diameters of the various molds ranging from 39,3 mm to 39,8 mm, I was thinking they're the same?!

Two extras were included (ha! thanks!), they look like promo-chips.

#1 Base color is a variant of Mandarin Red (differs slightly from the color sample) with spots in green (match), yellow (match) and purple (clearly less saturated than my sample but a match in luma)

#2 Base Imperial Blue (darker and more greenish than sample) and a green I've never seen before. It doesn't even come close to any of my sample colors: a very warm, rich middle green. Dunno what to make of this. Could be that it is a severely degraded Dayglo Green?!

Onto some tests:
I put my samples on a decent scanner and snapped them with a dslr as well (whitebalance from graycard), averaged the 'measurements' and put the new colors into my renderpipeline.
The results are (mostly) ok when setting my monitor to D5000 or D5500 which is roughly my lighting situation indoors before noon with mostly overcast sky. D65 is too cold/bluish compared to real-world light (and there is no technical lightsource emitting D65 with a full wavelength spectrum in the first place, afaik).
I discounted all dayglo colors from the scans as they look plain wrong (in general much too bright and undersaturated). It seems that the combination of scanner lighting/CCDs and the conversion of light waves into other visible wavelengths these dayglo pigments do is just beyond the gamut of the device.

Taking all observations into account I would assume that there is a mild color variation from the factory coupled with a significant degradation of the Dayglo pigments and a slight degradation of some other pigments over time (unless you keep them in the dark basement with your potatoes).
I'm kinda lost now and the whole approach to get good color representation from my renders feels quite questionable. And it explains some of my observations when looking at photos from actual chips and samples that always threw me off in the past (I'm no slouch when it comes to compensating lighting situations and typical camera tendencies when viewing pictures).

My temporary conclusion and approach after these tests boiled down to this:

• I go with my personal visual interpretation an do my best in directly comparing/adjusting monitor colors with real chip colors in (roughly) D50 lighting and calibration-setting - got no light-box and it's more likely I'm buying chips instead of one. I do a warm-up of 20 - 30 min to complete human chromatic adaption and my monitor does warm up as well, so both of us are at 'working-temperature'.

• I will then shift these colors in order to compensate for the shift from D50(D55) to D65 (widely used when calibrating with standard tools in the ~$100 range).

• devil-may-care

I did exactly this today from scratch with the colors for my metronome set. Funny thing is, I ended with colors pretty close to the colors I already had. But it will take some time to update the other colors as well and do a full recap of all things involved in render setups.

What I problably won't do is this:

• let my (still questionable) samples get measured by a costly pro-service and convert these colors from *L *a *b to a gamut like AdobeRGB or any other colorspace applicable to RGB-monitors (I assume that a fair amount of Dayglos would still be out of these).

• then go with these measurements and postulate them as 'ultimate real' colors.

• secretly adjusting render lighting until getting 'pleasant' results.

Am I just a nit picking german, annoying, grumpy and macula-degenerated? I'm aware of simplyfiying, leaving out nuances and ignoring some things in the color-finding process but I have no reason to believe that the majority of viewing devices out there would profit from colors 'closer' to the real chips (whatever this may be) and I'm far from being confident at this point.
Can anyone confirm or counter my observations? Any reasonable (not overly expensive/sophisticated/third-party) methods to improve from where I am? Would like to hear some other points of view or possible solutions/suggestions to to my dilemma.

 

[Psypher1000]

The diameters of the various molds ranging from 39,3 mm to 39,8 mm, I was thinking they're the same?!

I'm surprised that's all it is - from the smallest to the largest, I wouldn't have been shocked if there was a full 1mm difference. But yes, they definitely vary, and I think all are between 39-40. MD-50 is one of the smallest; FDL is one of the largest.

Taking all observations into account I would assume that there is a mild color variation from the factory coupled with a significant degradation of the Dayglo pigments and a slight degradation of some other pigments over time (unless you keep them in the dark basement with your potatoes).

It might also have to do with the mix of dye and clay for each print run. While I'm certain there's a general ratio they use, I'd guess the exact amount of ingredients mixed each time for each color varies slightly - this is a human endeavor, after all. But yes, clearly there are variations of colors from print run to print run. The same is true with Paulsons as well - particularly with lavender, although if you look at the Empress sets the reds, greens, and pinks (and, to a lesser degree, the purples) all varied from print run to print run even within the same rack. You can also see in hot-stamped Paulson color samples that there are variations from run to run.

Any reasonable (not overly expensive/sophisticated/third-party) methods to improve from where I am?

I think you said it best yourself earlier with the "devil-may-care". You can't be any more exact than the materials presented, and the chip factory itself doesn't even maintain high levels of color consistency from batch to batch. There's only so much you can do, mate. Take your best stab, and present your renders w/the caveat of "actual results may vary".

 

[BGinGA]

a green I've never seen before.

Probably the old dayglo green before the formula was changed (previous ingredients no longer available or too costly).

 

[Mental Nomad]

Am I just a nit picking german, annoying, grumpy and macula-degenerated?

No, you have over-developed, hyper-sensitive cones in your retinas. (But you are a nit-picking German.)

I have no reason to believe that the majority of viewing devices out there would profit from colors 'closer' to the real chips

Agreed - the variance in how most devices display colors is so much larger than the fine tweaks you'll be making that it's a pointless exercise - it would be a bit like pouring water on your pennies to save them while your dollar bills are on fire.

On the variations in the chips, themselves - yes. And yes. The molds, in many cases, came from different original manufacturers. The colors are hand-mixed per batch when production runs are coming up. Older sample chips coming from an acquaintance may well have been out on display in sunny windows for years before coming to you. Almost anything is possible.

All that being said, the nit-picker in me intensely appreciates all the hard work you're doing to produce the most accurate possible renders and images. You have my respect and my gratitude, and when my time for CPC customs comes (and it will) I'll be even more grateful... but I'll also do that thing that we hear so often, here on the forums: I'll get me the color samples to see in person!

 

[Ray-Col]

MD-50 is one of the smallest; FDL is one of the largest.

perfectly right - spot on (y) :thumbsup:

I think you said it best yourself earlier with the "devil-may-care". You can't be any more exact than the materials presented, and the chip factory itself doesn't even maintain high levels of color consistency from batch to batch. There's only so much you can do, mate. Take your best stab, and present your renders w/the caveat of "actual results may vary".

On the variations in the chips, themselves - yes. And yes.

I assumed this but to get some confirmation from seasoned chippers at least gives me some relief.

Probably the old dayglo green before the formula was changed

makes sense, another nugget from the past I'm not aware of ...

No, you have over-developed, hyper-sensitive cones in your retinas.

I've worked in automotive coatings four years back then and colors sometimes have to be tweaked to Delta E < 0,3 where my boundary was around 0,9

(But you are a nit-picking German.)

Glad you confirmed this too.

All that being said, the nit-picker in me intensely appreciates all the hard work you're doing to produce the most accurate possible renders and images. You have my respect and my gratitude, and when my time for CPC customs comes (and it will) I'll be even more grateful... but I'll also do that thing that we hear so often, here on the forums: I'll get me the color samples to see in person!

What I found most useful from my renders so far is seeing (better than chiptool) colors in context. Splash pot (on custom cloth), chip case and some other scenes to be generated when getting suggestions in the future - a good way to evaluate sets beyond isolated single stacks. Don't hesitate to drop a PM on me whenever you have any kind of render request I can help with (especially things I didn't do before).

What eyes are capable of

And where we can't trust them (get close and stare at the black point for a while and the gray cloud completely disappears)

 

[CraigT78]

And where we can't trust them (get close and stare at the black point for a while and the gray cloud completely disappears)

Whoa.

 

[Mental Nomad]

What eyes are capable of

It wasn't obvious, to me, what colors were in the boxes... all I knew was that they probably weren't Blue and Yellow. So I did a little basic Paint editing:

Ah. So the one on the left is grey:

But the one on the right is grey:

 

[Psypher1000]

But the one on the right is grey:

But what color is the damn dress!?!?

 

[Ray-Col]

It wasn't obvious, to me, what colors were in the boxes... all I knew was that they probably weren't Blue and Yellow. So I did a little basic Paint editing:

Obviously you're the ultimate color-degen! The gradients are pretty sick. (y) :thumbsup:

------------

Bad lighting today over here and metronome progess delayed for another week - I just dressed up some chips and toyed around with fake bumps. And guess what - I can throw hotstamps into the CPC-Mix soon. It's only kinda fake (real modeling would definately break the mold [yup, lamer pun]) but could come in handy and any design is implementable for a decent preview.

 

[onerand]

Another excellent thread. Lots of knowledge

 

[tommythecat]

Bad lighting today over here and metronome progess delayed for another week - I just dressed up some chips and toyed around with fake bumps. And guess what - I can throw hotstamps into the CPC-Mix soon. It's only kinda fake (real modeling would definately break the mold [yup, lamer pun]) but could come in handy and any design is implementable for a decent preview.

Great...now I want clay chips with edges buffed and polished to a high shine!

 

[kmccormick100]

This is really fascinating, with a background in design and photography, color has always really interested me.

Not to fuel your fire, but we do have a spectrophotometer at work that I could probably bribe one of the techs to get LAB values off a few of the chips for... haha.

I could at least get Pantone Matches for quick reference as a standard, although those aren't that precise of values, and nothing will take into effect the batch to batch slight color variance. Oh well.

 

[Ray-Col]

Not to fuel your fire, but we do have a spectrophotometer at work that I could probably bribe one of the techs to get LAB values off a few of the chips for... haha.

WHAT?! This would be outstanding!
Can anybody hack this forum software soon to implement a bigger like-button for things like that?

At least we have reason to believe that our samples are
a) up to date
and
b) there's a good chance they came from the same batch

I'd be particular interested in the following colors for 'reference grid':
Gray / Light Blue
Imperial Blue / Blue
Purple / Retro Lavender
Retro Red / Mandarin Red
Light Green / Green

Pairings shouldn't suggest that I'm interested in Delta E between them (most common measurement in Color Labs), I'm after the absolute measurements (D65/10° if possible). The transformation to other spaces and gamuts I can do myself later.

Dayglo Tiger or Dayglo Pink as sidekick - really curious how far outside of sRGB they are ...

I could at least get Pantone Matches for quick reference as a standard

Pantone matches (solid uncoated preferably) are very welcome as well. Even statements like 'Canary is inbetween 460U and 609U' should be pretty clear and of valuable significance!
If these color guides were not as costly as 50 mid/high-level customs, I'd get me some, too. :

although those aren't that precise of values, and nothing will take into effect the batch to batch slight color variance. Oh well.

Ha - welcome to the club!

Great...now I want clay chips with edges buffed and polished to a high shine!

May I encourage you to get friends with some car-body painter?

 

[kmccormick100]

Here's some ballpark Pantone PMS values, some are closer than others.
DG pink 1915c
Light blue 7458/630C ish
Canary ~ 129C
DG peacock ~ 2925c
DG tiger 171c
Charcoal 425c
Arc yellow 1565c
DG peach 1635c
DG Saturn 611/612/583c tough color
DG yellow 610c
Yellow 117c
Lavender 2655/2665
Retro lavender 258c ish
Red 188c
Mandarin red 1807c
Retro red 207c

 

[Ray-Col]

Here's some ballpark Pantone PMS values, some are closer than others.

Wow - thank you so much for the PMS-values and the time you donated to my project!

First conversion tests from several internet pages lead to mixed results (as expected) in all checked spaces/gamuts (CMYK, RGB, HSB, LAB) and therefore resulted in initial confusion. The differences were so big sometimes that I seriously doubt any reasonable/scientific background of some sites ...

Eventually I found a page that gave some insight to their own measurements and guess what - their conversions are the ones that matched my perception the closest.
That's what they said regarding to their values:
'The L* a* b* (together with the corresponding HLC, RGB and CYMK) values are based on the average of various measurements using various spectrophotometers using D65 light with a standard observer according to CIE 1976 which may be updated or modified by any other relevant available information.'
This was the first (and only so far) correlation to a specific lightsource (and the one we use the most, too) in measurements I stumbled upon and I lean towards trusting their colors atm.

I did only some samples yet, so I'm still in the process of evaluating your references.
One important question arises: Which part(s) of the sample chips did you use to compare with the PMS patches? The main crosshatched surface or the flat indents of the mold? Both? A mix?

With these references real measurements with a spectrophotometer should be obsolete, don't bother your colleague and keep the bribe-money for chips.
Again thanks a buch for doing this!

DG Saturn 611/612/583c tough color

Would you share my impression that this color magically changes everytime you look at it?

I'm back with more observations, soon. Foreshadowing: Some colors perfectly match my visually generated ones and others are so far off, I couldn't believe it. Every single answer in this realm seems to generate ten new questions.

 

[72o]

Here's some ballpark Pantone PMS values, some are closer than others.
DG pink 1915c
Light blue 7458/630C ish
Canary ~ 129C
DG peacock ~ 2925c
DG tiger 171c
Charcoal 425c
Arc yellow 1565c
DG peach 1635c
DG Saturn 611/612/583c tough color
DG yellow 610c
Yellow 117c
Lavender 2655/2665
Retro lavender 258c ish
Red 188c
Mandarin red 1807c
Retro red 207c

DG saturn always seemed a little 'different' to me. Like it just didn't fit in with the standard or the DG colors.

 

[Ray-Col]

Some additionally comparisons/observations to new samples and pantone matches

• Red, Retro Red, Mandarin Red: I had these on the lighter, less saturated side but not by much

• Charcoal: I was a tad cooler and darker but apart from that - close

• DG Pink: We're together on this one, almost retina burning nuclear. Seems to fit sRGB barely, close to the boundary. The old samples must have been seriously degraded.

• DG Peacock is out of sRGB by a fair amount, I can't get enough saturation pumped in it for the luma level I percieve

• DG Tiger/DG Peach/DG Arc Yellow: wow, these dayglos are poppin' now, never saw that in the old samples

• Yellows: All tricky ones imho. I tend to see them lesser saturated than the pantones overall, esp. Canary.

• Lavender/Retro Lavender: Did you accidentally mixed up numbers, names or something else? Nothing fits here, hue, luma, chroma - all so much more vivid and off hue-wise compared to any of the samples in this color range. Really weird regarding that all the others were quite close.

These are the current input colors for the renderengine:

Now I'm on to fix the flake size and amount in the weighted materials and use them as visible 'dirt' in lighter colors simultaneously ...

 

[kmccormick100]

Some additionally comparisons/observations to new samples and pantone matches

• Red, Retro Red, Mandarin Red: I had these on the lighter, less saturated side but not by much

• Charcoal: I was a tad cooler and darker but apart from that - close

• DG Pink: We're together on this one, almost retina burning nuclear. Seems to fit sRGB barely, close to the boundary. The old samples must have been seriously degraded.

• DG Peacock is out of sRGB by a fair amount, I can't get enough saturation pumped in it for the luma level I percieve

• DG Tiger/DG Peach/DG Arc Yellow: wow, these dayglos are poppin' now, never saw that in the old samples

• Yellows: All tricky ones imho. I tend to see them lesser saturated than the pantones overall, esp. Canary.

• Lavender/Retro Lavender: Did you accidentally mixed up numbers, names or something else? Nothing fits here, hue, luma, chroma - all so much more vivid and off hue-wise compared to any of the samples in this color range. Really weird regarding that all the others were quite close.

These are the current input colors for the renderengine:

Now I'm on to fix the flake size and amount in the weighted materials and use them as visible 'dirt' in lighter colors simultaneously ...

I'll take another look at them today in daylight ad bright white and double check. A couple of them were really tough to come up with values, nothing super close.

 

[Ray-Col]

I'll take another look at them today in daylight ad bright white and double check.

Take your time. I've got plenty of things left to work on. Many thanks for your help!


I think I'll go with this flake variation from now on.

Bringing the whole table outside when the sky is cloudy (completely in diffuse daylight) the flakes should disappear and leave some dirt in lighter colors when there is no strong directional lightsource to reflect.
And they do.

I wish I had an assistent to distribute all these surface changes to the other 120 predefined objects waiting to spring to life in custom sets with other molds ...

 

[Mental Nomad]

Lest you think the effort of doing what you're doing, and of posting comparisons like the last one, is unappreciated or wasted... it's not.

I totally love, appreciate, and respect the attention to detail you're putting into this.

Edit: totally had to think about why the sparkle from the brass flakes disappears in purely diffuse light... but they should still show up with their brass color, regardless, no?

 

[Ray-Col]

Sorry, my bad. I threw you off and rightfully so because it's plain wrong wording.

I used a physically based term (diffuse light) describing the visual outcome that is solely caused by the specific geometry of the scene (camera position, object position, light position).

I'll try to set it right by clarifying/differentiating what's really going on:
The surrounding sky can't be reflected and thus the flakes 'disappear' because the reflected color (including tint) gets buried in most of the chip colors. They show up as darker speckles in the lighter ones only. That's what I was after as I can observe it in the real chips.
Here are two stacks in pure brass to show what's actually reflected and you may get the idea why the not so bright reflections are masked by the chip color when they are sparse and lacking contrast.

So the effect has nothing to do with 'diffuse lightning' physically, it was just the wording I coose for the situation where a whole sky is evenly illuminating instead of distinct lightboxes I used in the first render where mirroring flecks were visible. These are the two hdr-images (reduced in size and not hdr anymore, obviously) in use:

It will look completely different if I tilt a whole stack so that it can reflect the evenly bright sky which is illuminating the scene.

Although I try to use PBR as much as I can, not all of the visible effects are directly correlated to pure material characteristics. As far as metal surfaces are involved, they are really absorptive (which comes from the fact that the nearly free electrons in the metal follow the oscillations of the radiation, thereby depleting its energy), but some only in part of the visible range and this is where the 'tint' in metallic reflections comes from.

The reflectivity of the brass I use is given by the Fresnel equations in terms of the index of refraction. They describe the angle dependency and further tell that the higher the difference in the index of refraction the more light will be reflected at the interface. This simulation is done by a special 'conductor node' I use in the software.

Phew, that was tough in a foreign language, hopefully I didn't screw this up as well.

 

[Mental Nomad]

Phew, that was tough in a foreign language, hopefully I didn't screw this up as well.

You did very nicely - good job! The extra renders also help - thanks for going through the effort.

I feel that, on my samples, i can see the brass flakes a little better, even in diffuse lighting conditions... but just a little better. Also, I may not be properly adjusting the viewing distance - I may be bringing them closer to my eyes than your images imply (because I'm viewing them without my glasses, which are for distance, and I have extremely fine close vision without the glasses.)

 

[Ray-Col]

You did very nicely - good job!

Thanks! Mucho appreciado

Regarding the flake intensity I'd take into account that we're faced with a huge dynamic range in the real world. If you're inside with, let's say 300 Lux and viewing the reflection of light coming through a windows from a diffuse sky with about 8000 Lux, our eyes adapt for it but the impression of high intensity remains and can't be reproduced with inferior 8bit images.
Nevertheless, while I'm at it - I bumped another 10% into the flake intensity (and have 15% left before hitting the limit). The flake elimination for unweighted colors is done via the same grayscale image and putting the high key at RGB 217 instead of 191 is a no-brainer. No one should complain about too much brass as long as the colors are off.

Rendered crop without DOF (otherwise the rendered part would be somewhat blurry).

I'm short-sighted too and I suppose we're in the same boat when it comes to meticulously examining detail. Although my crystalline lens gets more and more inflexible and I'm slowly loosing this lovely thing with advancing age.

 

[kmccormick100]

Some additionally comparisons/observations to new samples and pantone matches

• Red, Retro Red, Mandarin Red: I had these on the lighter, less saturated side but not by much

• Charcoal: I was a tad cooler and darker but apart from that - close

• DG Pink: We're together on this one, almost retina burning nuclear. Seems to fit sRGB barely, close to the boundary. The old samples must have been seriously degraded.

• DG Peacock is out of sRGB by a fair amount, I can't get enough saturation pumped in it for the luma level I percieve

• DG Tiger/DG Peach/DG Arc Yellow: wow, these dayglos are poppin' now, never saw that in the old samples

• Yellows: All tricky ones imho. I tend to see them lesser saturated than the pantones overall, esp. Canary.

• Lavender/Retro Lavender: Did you accidentally mixed up numbers, names or something else? Nothing fits here, hue, luma, chroma - all so much more vivid and off hue-wise compared to any of the samples in this color range. Really weird regarding that all the others were quite close.

These are the current input colors for the renderengine:

Now I'm on to fix the flake size and amount in the weighted materials and use them as visible 'dirt' in lighter colors simultaneously ...

My original lavender and retro lavender were way off, idk where those came from.

Lavender should be between 2577c and 2587c I think
Retro lavender I think 688c is actually closest

 

[ceramicmike]

But what color is the damn dress!?!?

Before or after Bill Clinton splashed the pot !!!!!

 

[Mental Nomad]

I'm short-sighted too

You mean near-sighted. (Short-sighted would mean you don't think ahead and plan for the long-term future.)

 

[kmccormick100]

Some additionally comparisons/observations to new samples and pantone matches

• Red, Retro Red, Mandarin Red: I had these on the lighter, less saturated side but not by much

• Charcoal: I was a tad cooler and darker but apart from that - close

• DG Pink: We're together on this one, almost retina burning nuclear. Seems to fit sRGB barely, close to the boundary. The old samples must have been seriously degraded.

• DG Peacock is out of sRGB by a fair amount, I can't get enough saturation pumped in it for the luma level I percieve

• DG Tiger/DG Peach/DG Arc Yellow: wow, these dayglos are poppin' now, never saw that in the old samples

• Yellows: All tricky ones imho. I tend to see them lesser saturated than the pantones overall, esp. Canary.

• Lavender/Retro Lavender: Did you accidentally mixed up numbers, names or something else? Nothing fits here, hue, luma, chroma - all so much more vivid and off hue-wise compared to any of the samples in this color range. Really weird regarding that all the others were quite close.

These are the current input colors for the renderengine:

Now I'm on to fix the flake size and amount in the weighted materials and use them as visible 'dirt' in lighter colors simultaneously ...

It's funny, I don't know if you have any Paulson samples, but their DG colors are off the charts in comparison as far as "pop" and saturation. It's really hard to convey with an accurate photo, but if the CPC samples are out of the sRGB gamut, then Paulsons aren't even in the same neighborhood. Specifically DG Pink, Arc Yellow and DG orange are unparalleled. I love CPC's colors, but they're a little underwhelming alongside Paulson :/

 

[Jeff]

And where we can't trust them (get close and stare at the black point for a while and the gray cloud completely disappears)

If I stare past this grey circle with it close to my face I see tits.

 

[Ronoh]

If I stare past this grey circle with it close to my face I see tits.

[youtube]

 

[Jeff]

75 guys are holding their phone up close to their face all over the country!