Fade to grey – The pigment balance problem

Open any book or web page on color printing and it’ll say at some point that all pigments or dyes used in the process should mix to a neutral grey or black – at least in theory. It’s an issue I’ve been ignoring throughout my color carbon adventures so far. Well, not exactly ignoring, but I didn’t spend sufficient attention to it, certainly not in writing. Allow me to make up for this, at least in part. In this blog, I’ll explore the issue of pigment balance and try and work out a way to determine, at least with rather coarse resolution, a usable pigment balance for color carbon tissues.

“Make it up as you go along” is something that really applies to my color carbon journey. Mostly it’s because I have a mind of my own, and while I’m aware of the theory on many things and the fact that there are established procedures, materials and equipment to do certain things, I try to do it my way. Usually because that allows me to (1) fully understand what I’m doing, as I’m apparently an experiential learner and (2) instead of shopping my wallet empty, I generally try to make things work with the simplest materials or preferably the stuff I have on hand first. If I then end up buying something because I can’t get around it, at least I’ll understand pretty well which requirements it needs to meet.

With the pigment mixing issue, there’s not all that much in terms of practical guidelines I could go by. I did ask a question on Photrio about it, but it was somewhat hidden in a rather opaque posting of mine, and online searches also didn’t yield much information. At least not for color carbon printing. It’s somewhat different for gum bichromate. For instance, Katharine Thayer shares some thoughts on the matter and I’m sure there are many other sources as well. Sadly, Grier remains pretty much silent on the matter in his otherwise fairly comprehensive gum printing manual. The problem, in part, is also that while with gum you could adjust as you make the print, for instance by applying an additional compensation layer of some color to make up for a color cast. This is pretty much what Thayer appears to suggest. With carbon transfer, this is of course impractical, as it would necessitate creating a dedicated tissue to fix a color balancing issue and everything that goes with it. It would be far better, and indeed the only viable option, to get the balance between the C, M and Y tissues right the first time.

What happens if you don’t get it right? Well, look at pretty much any example I’ve posted so far along my color carbon journey. This recent color checker is a good one:

Problematic color balance – note how every color is its own hue…but they’re all part of the yellow family.

Yellow is dominant in this one, with just about any hue leaning distinctly towards it. Even black is a murky yellowish-brown. It’s not so visible in this scan, but the real print in daylight shows it very clearly. Magenta also seems to win from cyan, although not as strongly as yellow. In short, it’s a mess. Which is no surprise, because for the tissues I used on this print, I did the color balancing in a bit of a haphazard way. I did do it alright, more or less along the lines of what I’m going to explain below, but I rushed through it and probably cut a couple too many corners here and there.

So how to get this balanced better? Well, perhaps not many practical guidelines are published online about this because it’s so straightforward. At least, what I tried so far really isn’t rocket science – the only thing I don’t know yet, is how well it works. But here’s my line of thinking: firstly, it would make sense to have pigment concentrations in the C, M and Y tissues so that if they are all exposed to give let’s say 50% density and you overlay these three prints, they would mix into a neutral grey. In other words: getting the actual pigment ratios between the tissues right.

The second part, which I expect to be doing in finetuning, is based on curves: apply an adjustment curve to two of the colors (depending on what kind of color casts the tissues produce by themselves without correction) so that any minor balance issues are negated. Note that these are not the linearization curves I wrote about before, but an additional (more linear, I anticipate) curve that specifically adjusts the overall density of a color layer so that it matches the other layers. Think of the kind of ‘boost’ and ‘chop’ curves I discussed in my article on crossover – but now applied purposefully and not as a defect that happens to exist in an image.

To be frank, I haven’t tackled the second part yet. I thought I was getting to that a few days ago, but I was set back a couple of steps because I ran into, err…something. Because my linearization effort was well underway, I had already designed a test to figure out what kind of color matching curves I would need on the tissues I was working with. It looks like this:

Color mixing chart designed for determining color correction/matching curves

This is one of those points where the graphic arts industry has already solved my problems eons ago; I’ve been making test charts like the one above based on some reasoning and gut feeling, but they evidently already exist somewhere. They’re likely also better than mine, but hey, I actually quite like a puzzle like this once in a while.

Anyway, the chart above is really the composite of three color layers (in negative) that create the pattern above if you overlay them on top of each other. Here’s what they look like, individually:

Individual layers of the color mixer chart above; cyan, magenta and yellow from left to right.

What I did here was really create a 3-dimensional chart. Since prints are 2-dimensional, I had to figure out a way to add the third dimension, so I made the cyan and magenta layers the rows and columns respectively of a table, and then repeated the table with different densities of yellow. When you overlay the layers above, you’ll notice that each little square in the resulting tables makes a combination of 10% through 90% density of cyan, magenta and yellow. Cyan and magenta come in 10% spacings, and yellow in 20% spacings. The latter was a compromise; I do these tests on 4×5″ tissues and repeating the same table 9 times for the yellow densities instead of 5 times like I did now would have resulted in the individual color patches being too small to evaluate reliably.

You might have realized by now that in this color chart, there should be exactly one patch somewhere that mixes neatly into a neutral grey. In the digital mockup, it’s (of course!) in the spot that represents 50% density of cyan, magenta and yellow. Don’t believe me? Try it out for yourself! Take the color image above and measure the 50/50/50 patch that you’ll find near the center in the top-right table. I know, your eyes trick you into believing it’s some undefinable shade of blueish green or so, but it really is a perfectly neutral grey: the a and b components of a Lab measurement really do show zero for both.

So the theory is I print this with my fledgling color carbon printing skills, determine which patch is a neutral grey in the print, and then create compensation curves on this basis. Well, that’s easy enough (famous last words)! Here you go:

First color carbon print of the color checker chart

Well, that looks pretty much as expected if you look at the first image example in this article. Rather yellow, a bit heavy on magenta and overall not quite balanced well. Check out in particular the greyscale bar on the right side of the image, which should ideally mix into neutral grey. It’s kind of orange, in reality, and so are all the black borders. Not to worry, that’s exactly what this chart is for, right? It’s just a matter of finding the grey patch and create adjustment curves in GIMP.

Right. Finding the grey patch. Turns out, there isn’t any. Yeah, there are a couple of patches that come close and that should be good enough. The problem is, the patch that comes the closest to neutral grey (with a and b out of Lab of -0.9 and 1.4 respectively) is at 50% cyan (so far, so good), 20% magenta (ermmm…) and 10% yellow (ah…) Yes, you’ll find it nearly all the way to the left of the top left square. In other words: the color balance is so skewed that any correction curve would have to be pretty extreme, especially on the yellow channel. After all, it would take only 20% of the yellow density I have now and 40% of the magenta density, but all of the cyan density to give a neutral color rendering. Huh, 20% and 40%? Yes, because 10% yellow gives the density that 50% should be giving, so that’s only 20% of what I have now. And for magenta, it’s the 20% patch that gives what should be 50%, so that’s 2/5 or 40%.

So reduce yellow density by 80% and magenta by 60% and we should be close. You know, that sounds like throwing perfectly good pigment out of the window for starters. Moreover, I could apply a minus 80% curve on the yellow channel in theory, no problem. In practice, however, this will compress the entire tonal scale of the yellow layer into 20% of the density range I can make on the digital negative. Apart from this being rather costly in terms of inkjet ink (every digital negative would end up pretty much all a very dark grey!), it would also create tonal res0lution problems. Put differently, tonal transitions on the yellow and to a lesser extent the magenta images would be unnecessarily poor.

This is what color-corrected tissues would look like for the unbalanced carbon tissue; cyan, magenta and yellow from left to right. Note how magenta and yellow are pretty much all black & dark grey negatives. Costly on inkjet ink and giving poor tonal resolution.

In short – this is just not good enough. My haphazard way of establishing a pigment balance in these tissues was apparently a bit too haphazard, indeed.

Back to square one: how to get that pigment balance right in the first place? Well, this is the fun part, because it means I can get away from the computer screen for a little bit. This is low-tech at its best, really! First, I took all the relevant pigments (which I use in the form of watercolors and some gouache paints) and created a 1% solution of each. So I measured out 250mg of each watercolor/gouache and mixed it with water to make a 25ml volume. And boy, does that make for a nice display:

A 1% solution of all the paints I’ve been using and considering so far for color carbon.

In case you’re curious, this is what they are, from left to right:

  • Cyan:
    • Winsor & Newton watercolor ‘Winsor Blue Green Shade’, pigment PB15:3
    • Winsor & Newton gouache ‘Primary Blue’, pigment PB15
    • Winsor & Newton watercolor ‘Cobalt Turquoise Light’, pigment PG50
  • Magenta:
    • Daniel Smith watercolor ‘Cobalt Violet’, pigment PV49
    • Winsor & Newton watercolor ‘Quinacridone Violet’, pigment PV55
    • Winsor & Newton watercolor ‘Quinacridone Magenta’, pigment PR122
    • Royal Talens gouache ‘Permanent Rose Magenta’, pigments PR122 & PV19
  • Yellow:
    • Winsor & Newton watercolor ‘Winsor Yellow’, pigment PY154
    • Van Gogh watercolor ‘Transparent Yellow Medium’, pigment PY128
    • Winsor & Newton watercolor ‘Transparent Yellow’, pigment PY150

Wouldn’t it be nice to see how these actually look like on paper? Well, we can do this, you know. How about a little swatch of each of these colors (using the 1% solution I made) on the same paper:

My apologies for the different order and the small mistake, but you can track down which paint is which based on the pigment used. The mistake is the PB15 marked ‘Talens’ which really is the Winsor & Newton ‘Primary Blue’ gouache, and the PB15:3 is the W&N ‘Winsor Blue GS’ watercolor.

If you recall my blog on picking pigments for color carbon, you’ll find that I was particularly interested in the pigment set PG15:3 for cyan, PR122 for magenta and PY128 for yellow. Well, for now, I ended up picking PY150 for yellow because there is something to that W&N ‘Transparent Yellow’ watercolor that’s just so appealing – it has insanely high tinting strength and somehow just mixes very nicely. I’m not sure if I’ll stick with it and I might actually move to PY154 instead, but that’s a story for another day. I did go with PG15:3 for cyan in the form of W&N ‘Winsor Blue GS’ and PR122 for magenta from the W&N ‘Quinacridone Magenta’ paint. Previously I used the Talens concoction of PR122 + PV19, but I don’t like the unnecessary (for carbon transfer) addition of opaque fillers in gouache paint and in the end, the Talens gouache doesn’t have as a high a tinting strength as the watercolors do. I marked my current paint selection in bold in the list above.

On a side note: in that earlier blog I cited above, I also expressed interest in an alternative pigment set using PG50 for cyan and PV49 for magenta. It’s now easy to see why I’m not in a hurry to pursue that line of thought with practical experiments. Firstly, both pigments appear to be either rare or expensive, or both, since the paints containing them aren’t always very easy to find, and they cost a pretty penny. Secondly, looking at the paint swatches, observe how weak they are compared to the other ones. They’re all 1% solutions of the paint in water – I don’t really know the actual pigment concentration of course, but at least when considering the entire paint vehicle, the tinting strength of the paints containing the PG50 and PV49 pigments turns out to be very low. It would take a lot of paint to make usable tissues with these. For now, that option is out of the window. Coincidentally, both of these paints appeared to be loaded heavily with opaque filler (usually titanium dioxide) and I suspect the actual pigment concentration is very low to begin with.

Back to our pigment set of PG15:3, PR122 and PY150. The swatches of these paints look quite promising indeed: they’re nice and bright. The solutions also look quite nice; especially the PY150 solution is quite transparent and heavily tinted, and there is very little to no sedimentation even when letting it sit still for a couple of days. This is different for the PG15:3 and PR122 solutions; these remain heavily tinted, but larger pigment particles start dropping to the bottom already within an hour. For tissue making, this means frequent stirring is necessary – something to keep in mind, as it’s usually not very necessary when using e.g. India ink for black tissue.

Sedimentation in the 1% solution of W&N ‘Quinacridone Violet’ PR122 watercolor. Sedimentation like this usually points towards unequal pigment particle size distribution. The larger, heavier pigment particles are the first to drop to the bottom. Very small particles remain dispersed due to the continuous motion of the water molecules themselves.

Now that we’ve got fancy colors, let’s see if we can fancy things up a bit more. After all, we’ve got these primaries so that we can mix them and create all kinds of hues – and we can actually give that a little try using some pipettes and a brush. For instance, here are the primaries and secondaries of the PB15:3, PR122 and PY150 set:

A very rudimentary color wheel made with the PB15:3, PR122 and PY150 pigment set

I think this immediately shows the strengths and weaknesses of the pigment set: yes, it’s possible to make very nice looking intermediate / mixed hues using these pigments. Isn’t it a pretty picture, after all? I quite like it! But if you’re critical, you would also note the lack of saturation in the red, green and violet patches. These represent the mixes of the individual pigments, and mixing comes at the cost of reduced chroma. There is, however, also one more good thing I’d like to point out: the grey patch in the center is not made with black paint, but an impromptu mix if the three pigments. Look how neutral it actually seems! If I sample it in GIMP, I get a and b values (from Lab) of 0 and 6.7 – but paper white also measures -1.4 and 4.6. So it doesn’t only look pretty neural, it actually is quite close to it. This is good news, because it proves that a decently neutral grey is at least feasible by mixing these pigments.

Shall we do one more? Not that it adds much more to the argument, but at least it makes for a pretty picture. Here’s another version of the color gradient, but with a little more variation. I made this by mixing drops from a Pasteur pipette; the number of drops are indicated in the picture.

Droplet mixes of PB15:3, PR122 and PY150. Left column: PB15:3 : PY150 droplet ratio. Center column: PB15:3 : PR122 droplet ratio. Right column: PY150 : PR122 droplet ratio.

Well, actually this does add something to the argument, but you have to know what to look for. In the left hand column, notice how only the top patch is blue (it’s pure PB15:3 after all), but yellow becomes dominant pretty much immediately. Even at a 2:1 ratio of blue:yellow, a convincing green already arises. You’d expect this to happen more in the middle of the mixing range if the pigments were to have the same tinting strength. In other words: the tinting strength of this particular PY150 solution seems stronger than that of this PB15:3 solution. In the right-hand column, this is even more pronounced – note that I already accounted for the much higher tinting strength of PY150, but even so, magenta gets pretty much pushed aside in the mix as soon as a little yellow is added to it. In the center column, we can also see that PB15:3 wins from magenta – note again that I used twice as much magenta and even so, the blue becomes dominant rather quickly.

Now, this is all a bit offhanded guesswork and squinting through eyelids, but at least it introduces two important things: (1) it’s possible to do mixing experiments with these 1% paint solutions and (2) there are real differences in tinting strength. Both are kind of obvious; the latter already was quite apparent from the mishap this post began with (i.e., the rather dominant yellow in the test print), and the former, well, it doesn’t take Einstein to figure that one out.

Droplet mixing primaries to attempt to create a neutral grey.

So how about trying to figure out a neutral grey by mixing droplets of 1% paint solutions? Well, why not indeed! It’s a bit of work, but the results are insightful at least. Here’s an example, again using droplet ratios, in the order C:M:Y (PB15:3 : PR122 : PY150):

Mixing grey using 1% paint dilutions

The 1% dilution turns out to work quite well for this because the resulting tone is fairly easy to evaluate visually. Keep in mind the human eye isn’t as good at accurately determining color in very low densities, let alone very high densities. In other words, it’s fairly easy to determine neutral grey, but determining neutral white or neutral black is much more challenging.

In the swatch chart above I basically did some ring around tests around ratios that seemed to work well. In the top two rows, I started with a 2:2:2 ratio, observed it was too green, so started to add magenta until I overshot the target. I then took the promising 2:6:2 ratio and used that to vary the cyan content, by mixing ratios of 2:12:4 (I doubled the magenta and yellow count to gain some resolution) through 7:12:4 in the third and fourth rows. I did the same thing for the yellow, resulting in rows five and six. The final single patch is an attempt to simplify a more or less neutral mixing ratio, but frankly, that one didn’t turn out too well. Note how especially patches 6:12:4 and 7:12:4 look very promising – they’re quite neutral grey ones. Here, the mixing ratio is close to ideal.

Now, back to the Einstein remark. Mixing droplets isn’t exactly a groundbreaking invention. And yet, it does have its complexities. I noted this when I tried to use the droplet ratio as a basis for a true volumetric ratio. In other words: could I take e.g. a 3:6:2 droplet ratio and get the same hue by mixing 3ml of cyan, 6ml of magenta and 2ml of yellow? You’d expect so, right? Well, turns out it doesn’t work this way. Maybe it has to do with the inaccuracy of plastic Pasteur pipettes, or perhaps the actual pigment or differences in the paint vehicle cause differences in the surface tension and/or viscosity of the solution, which will affect droplet size, and thus make for different numbers of droplets per milliliter. I did a quick test and sure enough, I counted different numbers of drops per milliliter for each solution. The differences were small, but real, and sufficiently significant to create problems. So important observation: droplet ratios are a nice starting point, but ultimately not reliable.

Well, not to worry, I did some more tests that are conceptually similar to the one above, but using milliliter measurements instead of droplets. Using this approach, I arrived at a ratio of 4:10:3 for the paints containing PB15:3, PR122 and PY150. I proceeded to mix some glops using this ratio as a starting point, and departing from the concentration of the yellow paint in my previous tissue version. But let me explain briefly.

With the previous tissues, I noted an overall lack of saturation. In the previous blog post, the one concluding the linearization game, I observed that the overall pigment density was on the low side, especially compared to the earlier tissues I made. While this does make for nicely subdued hues, it should also be possible to mix very saturated hues. If I also want CMY (instead of CMYK) tissues to work, and I’m actually starting to lean in that direction, I’ll need sufficient pigment density to create a true black as well. In other words – I don’t want too much pigment, because there’s only so much that’s necessary, but I also don’t want too little pigment as it creates saturation problems. I also observed that yellow was too dominant in the previous tissue version, but that I needed a lot more cyan.

Now, my previous color tissues actually used the Talens gouache PR122/PV19 mixture for magenta and the W&N gouache PB15 ‘Primary Blue’ for the cyan, and I don’t use these in the current tissue version. So I couldn’t really use the earlier magenta and cyan concentrations as a starting point. Luckily, yellow stayed the same (which actually was no coincidence as you might understand now) so I had a benchmark to depart from. I applied some guesswork (which I have not yet been able to verify for correctness) and estimated that the previous yellow concentration of 0.8% was probably on the high side and that I could get away with around 0.5%. Using that as a starting point, I determined the other concentrations, because I now knew the correct ratio to mix a neutral grey. So that’s how I made the glops for the current C, M and Y tissues.

Color carbon glops using PB15:3 for cyan, PR122 for magenta and PY150 for yellow. Also a black glop using 0.5% Royal Talens India ink, but this pigment concentration is just a wild guess as of now.

But there was one more bump. To verify the color ratio between the glops, I took half a milliliter of each while they were outgassing, mixed this together really well and applied it to paper. And immediately it was apparent that the result was distinctly green. Not a pure black or grey, but a very greenish hue. Darn.

So I did some additional impromptu experimentation, adding some more magenta glop to the small quantity of three-color glop mix, and this way worked out that I needed to add around 20% more magenta pigment to get closer to neutral. I quickly measured some more magenta paint, dispersed in water and added it to the magenta glop. I then verified the ratio by taking a drop of each glop (assuming the viscosity of glops wouldn’t be all too different given the significant effect of the gelatin in there), mixing those drops and applying to paper. Much better!

Left: green ‘black’ obtained by mixing the glops initially; right: result after correcting the magenta glop by adding 20% more pigment. Note the different paper base, which also is a bit more greenish in the left example. Quick and very dirty indeed!

The scan of the corrected black swatch reads a and b values from Lab of around 5.3 and 4. This isn’t perfect by any means, but I hope that it will be close enough to correct the final bit with correction curves as my initial plan dictated. So another lesson learned: 1% watercolor dilutions in water don’t necessarily mix the same way as the gelatin-based glop does.

Maybe this has to do with the color of the gelatin itself; I’m using a technical grade gelatin that’s rather yellowish, and maybe that color influence really cannot be ignored – although I’d expect a shift to yellow and not to green as a result of this. Maybe it has something to do with refractive indices. I just don’t know, but it seems like a good idea to do the ultimate mixing tests with an actual gelatin solution that approximates a real glop.

For now, I can only wait for the tissues to dry so I can proceed with my testing. The consequence of the pigment changes and the changes in pigment concentration mean that I will most likely have to do the linearization again. I have good hopes (a sense of realism isn’t my strongest trait) that this will be a quicker process now that I’ve figured out how to approach that. Once that’s done, I can give the color mixer chart another go and see if I can get the process to print a grey patch on there somewhere – and see if I can actually locate it.

Leave a Reply

Your email address will not be published. Required fields are marked *