Continuous tone in a digital negative

[jisner]

When I hear someone say their printing technology (color, Epson ABW, PiezoDN) makes digital negatives with more continuous tone, I know it means more complete coverage: the ink dots are closer together. How would you actually test such a claim without examining the negative under a microscope? Say you only had a UV transmission densitometer (I have a 361T). Can you think of a way of testing two negatives that would enable you to say "A has more continuous tone than B?"

 

[nmp]

OK...I'll bite...(who else)...

By my understanding, continuous tone is meant to signify distinctness between neighboring tones for a given delta of input function between them. Smaller the smallest delta the distinctness can be maintained and measured, more continuous the tone.

With that, here is one candidate method I can think of to measure tonal continousness (see if it makes sense):

Print a series of 3-step step wedges (X+d, X, X-d) each with a "d" lower by a factor (say 2: 20%, 10%, 5%, 2.5%, etc.) Then find the (dmin) where the 3 steps are no longer statistically distinguishable with desitometer (or visually.) Repeat the same with several values of X - one each for shadows, mid-tones, and highlights, for example. The printing method that gives the lowest dmin is the most continuous. Obviously, here the assumption is that each method is perfectly linear.

:Niranjan.

 

[jisner]

Print a series of 3-step step wedges (X+d, X, X-d) each with a "d" lower by a factor (say 2: 20%, 10%, 5%, 2.5%, etc.) Then find the (dmin) where the 3 steps are no longer statistically distinguishable with desitometer (or visually.) Repeat the same with several values of X - one each for shadows, mid-tones, and highlights, for example. The printing method that gives the lowest dmin is the most continuous. Obviously, here the assumption is that each method is perfectly linear.

Thank you! That's a really good definition, and it works for my problem. I made two step table negatives, one CMYK grayscale and one colorized grayscale. I plotted transmission density vs. digital tone (K) . At any tone t, the curve with the greater slope has more continuous tone. So while color gives consistently higher density over most of the tonal scale, color has areas where tone is less continuous, making it less useful (and possibly even harmful). In the plot below, the slope of the color negative curve is nearly flat around 45% and again around 65%. In these areas, there is almost no differentiation of tone. Using step tables with more steps shows that in these problem areas density becomes highly irregular and even reverses.

I'd like to understand what is causing this behavior. All I can come up with is this: the density of a color negative has two components, one physical and one optical (spectral). In the midtones of the negative, where the color is purest, optical density overwhelms physical density and creates these anomalies. Optical density tapers off toward the highlights and shadows of the negative. A grayscale negative shows only the effect of physical density.

 

[fgorga]

What wavelength are you using to make your measurements? What are you planning to use your negatives for? Density at a wavelength unrelated to the action spectra of the process you intend to use it for is meaningless.

Also, I am not sure about your ideas of 'physical' and 'spectral' density... this just does not make sense to me given what I know about the interaction of light with matter. Please explain further.

Lastly, I am a big fan of the empirical approach when it comes to the intersection between art and science. (I say this even though I am a retired chemist.) Personally, I would simply make negatives, make prints and make adjustments if the prints are not matching my vision.

 

[jisner]

What wavelength are you using to make your measurements? .... Density at a wavelength unrelated to the action spectra of the process you intend to use it for is meaningless.

I don't know the portion of the UV spectrum that the 361T measures, but I imagine it is wide enough to be useful for a large variety of industrial UV exposing applications.

Color negatives aren't analogous to color filters that you put on your lens in B&W film photography. In a color negative, color does not matter at all. The pigment (mineral) blocks UV. Different pigments have different blocking properties. As long as the pigment blocks UV, no matter what color it is, it still works. If color/wavelength mattered, we could make negatives with color dye ink. Ultrachrome black is just another pigment. It is very good at blocking UV -- maybe too good.

By "physical density" I mean simply the total amount of ink that goes on the film. In 1999, Burkholder found it impossible to print CMYK grayscale negatives on the films available at that time without the ink pooling. He discovered he could get equal density with a relatively small amount of color ink (e.g., just M and Y instead of C, M, and Y). Use less ink, but get just as good density. Maybe that's what I mean by spectral density -- density from the pigments in color ink.

Today's films can take a lot of ink, so there's no reason to avoid grayscale negatives. By 2013, Burkholder had dropped color and was making negatives with Epson ABW (PK, LK, LLK, LM, LC and maybe Y). Grayscale negatives rely on physical density and have good continuity of tone. Color negatives rely on spectral density but have continuity problems.

 

[fgorga]

You have a few misunderstanding about the physics and chemistry involved here.

The first is "density"... the amount of light blocked is dependent on both the nature of the material and the amount of the material. This is equally true for both black pigments and colored pigments. Thus your dichotomy is false.

There is fundamentally no difference between a black pigment and one that shows visible color. A black pigment simply absorbs visible light in a more-or-less wavelength independent manner. A colored pigment absorbs some wavelengths of light more than others and is thus colored.

The reason that folks found that some colored pigments were better for printing negatives is the fact that some colored pigments are better at blocking UV light better than the black pigments. Thus one could deposit less of a colored ink onto the plastic and get the more light blocked that you could with the black pigment.

One needs to remember that ink for common inkjet printers is optimized for producing photographs that are viewed by visible light sensors (i.e. the human eye) and that little if any attention is paid to how these materials perform in the UV that is important to those of us making digital negatives.

With regard to the densitometry, I am glad to hear that it is a UV instrument, but I would still be very cautious about making any conclusions based on its data. As I understand it, these devices were made for folks making plates for offset printing and that important wavelength for this application are significantly lower that the wavelengths that are important for most alt photo processes.

The exact wavelengths involved are hard to pin down, but from what I have gleaned, plate making involves wavelengths down around 360 nm. Cyanotype, the one process where I have been able to find actual data, has a very flat spectral response from about 290 nm to 400 nm and falls off steeply on either side of that. I would expect that the other iron-based processes would have sensitivities similar to cyanotype as the primary photoreduction in all of them is iron(III) to iron(II). I have never found any similar data for pure sliver-based processed such as salted-paper. (If any one know of such data, I would be glad for a citation.)

Anyway, the whole point of this discussion, is that densitometers need to be matched to the process and using a densitometer designed for process A to make conclusions about process B may lead the user astray. This is one reason why I am a fan of empirical results in this area.

 

[jisner]

The reason that folks found that some colored pigments were better for printing negatives is the fact that some colored pigments are better at blocking UV light better than the black pigments. Thus one could deposit less of a colored ink onto the plastic and get the more light blocked that you could with the black pigment.

Black pigment blocks UV better than any color. That's why it has to be used sparingly. But I'm not talking about black pigment. I'm talking about neutral ink, which is made by mixing equal amounts of C, M, and Y and possibly adding a small amount of black in the shadows. The choice is between a relatively light application of one or two color inks vs. a relatively heavy application of neutral ink made by mixing C, M, and Y. In 1999, Burkholder chose color because of the limitations of film. He did so with full knowledge that color negatives were not as smooth (not as many dots, not as continuous) as grayscale. There was simply no alternative at the time.

 

[tnp651]

I don't believe your quest has much practical value.If an inkjet printer can print a ramp in which you cannot discern banding, it is good enough for alternative processes. The tonal scale of most alt processes is narrower than that of silver prints or regular inkjets. In other words, you can print fewer perceptible differences in tone. If there's banding in the negative but it's imperceptible, it will produce an even more imperceptible banding in the print.

 

[jisner]

I don't believe your quest has much practical value. If an inkjet printer can print a ramp in which you cannot discern banding, it is good enough for alternative processes. The tonal scale of most alt processes is narrower than that of silver prints or regular inkjets. In other words, you can print fewer perceptible differences in tone. If there's banding in the negative but it's imperceptible, it will produce an even more imperceptible banding in the print.

True, true, true, and true. About my quest. I'm building a software simulation based on a mathematical model of tone reproduction. The reproduction cycle consists of three transformations: the image adjustment curve (digital tone to digital tone), the printer curve (digital tone to transmission density) and the process curve (exposure (= log 1/TD ) to Lab L*. To get somewhat realistic printer and process curves, I have been making the measurements. To make process curves, I measure L* of step wedge targets using a spectrocolorimeter. To make printer curves, I measure transmission densities of grayscale and colorized grayscale step tables printed on standard Epson P400.

I have a second P400 with piezography ink that I use to make QTR negatives for my alt process prints.

 

[nmp]

Thank you! That's a really good definition, and it works for my problem. I made two step table negatives, one CMYK grayscale and one colorized grayscale. I plotted transmission density vs. digital tone (K) . At any tone t, the curve with the greater slope has more continuous tone. So while color gives consistently higher density over most of the tonal scale, color has areas where tone is less continuous, making it less useful (and possibly even harmful). In the plot below, the slope of the color negative curve is nearly flat around 45% and again around 65%. In these areas, there is almost no differentiation of tone. Using step tables with more steps shows that in these problem areas density becomes highly irregular and even reverses.

I'd like to understand what is causing this behavior. All I can come up with is this: the density of a color negative has two components, one physical and one optical (spectral). In the midtones of the negative, where the color is purest, optical density overwhelms physical density and creates these anomalies. Optical density tapers off toward the highlights and shadows of the negative. A grayscale negative shows only the effect of physical density.

View attachment 282113

Not sure I fully understand the difference between CMYK grayscale and colorized grayscale. How are they produced...starting with a step wedge of R=G=B? or gray mode? and then how are they sent to the printer. Any icc profile used? Also in the chart is the label "color" is for colorized grayscale and "grayscale" is for the CMYK grayscale?

A little confusion on my part....

:Niranjan.

 

[jisner]

Not sure I fully understand the difference between CMYK grayscale and colorized grayscale. How are they produced...starting with a step wedge of R=G=B? or gray mode? and then how are they sent to the printer. Any icc profile used? Also in the chart is the label "color" is for colorized grayscale and "grayscale" is for the CMYK grayscale?

A little confusion on my part....

:Niranjan.

It is confusing, and maybe I've added to the confusion. I wish there was standard terminology for this.

Assume we're printing a step table, and assume the Photoshop document in Gray Gamma 2.2. Since the image is a step table, it does not need to be inverted.

If we send this image to the printer and specify Color (not B&W) the printer will make neutral tones by mixing C,M,Y, and K ink. The resulting print is what I may have called "CMYK grayscale" to distinguish it from a grayscale negative made with black ink only, which you would get if you chose B&W in the printer settings (not a good idea).

To make a colorized grayscale negative, we convert the document to Adobe RGB, then add color. The original Burkholder (1999) method was to do a Fill -> Color with the Fill operation in Color mode. The way I would do it today would be to add a Color Fill layer in Color mode. Some people use Screen mode.

If instead of a step table you were printing a real image, you would invert before colorizing.

There is another method of colorizing that is used by the "Easy Digital Negatives" (EDN) system, and I believe it explains why colorized negatives remain popular. EDN colorizes by applying a gradient map over the image (or the inverted image). The "stops" of the gradient attempt to shape the printer curve so that it's closer to linear. It's an elegant idea, but it suffers from "continuity" problems in the midtones. As tnp61 pointed out, it's not going to make much difference in a typical alt process print. But a mathematical model chokes on inversions.

I wanted to understand the history of digital negative making, and so I found a used copy of Burkholder's 1999 book on eBay. Back then, everyone was making negatives with imagesetters. But affordable inkjet printers were becoming available, and he talks about his experience using them. This is where he explains his decision to make colorized negatives, rather than CMYK grayscale negatives: colorized gives you equal density with less ink, but not as good coverage (it was Burkholder who distinguished between "physical density" and "spectral density"). By 2013 (his "Inkjet Companion" book) he had abandoned color and gone back to CMYK grayscale. A grayscale printer curve (transmission density vs. digital tone) is a thing of beauty, even on my "K1" printer.

I use your printer settings (see below) !!

 

[nmp]

It is confusing, and maybe I've added to the confusion. I wish there was standard terminology for this.

Assume we're printing a step table, and assume the Photoshop document in Gray Gamma 2.2. Since the image is a step table, it does not need to be inverted.

If we send this image to the printer and specify Color (not B&W) the printer will make neutral tones by mixing C,M,Y, and K ink. The resulting print is what I may have called "CMYK grayscale" to distinguish it from a grayscale negative made with black ink only, which you would get if you chose B&W in the printer settings (not a good idea).

To make a colorized grayscale negative, we convert the document to Adobe RGB, then add color. The original Burkholder (1999) method was to do a Fill -> Color with the Fill operation in Color mode. The way I would do it today would be to add a Color Fill layer in Color mode. Some people use Screen mode.

If instead of a step table you were printing a real image, you would invert before colorizing.

There is another method of colorizing that is used by the "Easy Digital Negatives" (EDN) system, and I believe it explains why colorized negatives remain popular. EDN colorizes by applying a gradient map over the image (or the inverted image). The "stops" of the gradient attempt to shape the printer curve so that it's closer to linear. It's an elegant idea, but it suffers from "continuity" problems in the midtones. As tnp61 pointed out, it's not going to make much difference in a typical alt process print. But a mathematical model chokes on inversions.

I wanted to understand the history of digital negative making, and so I found a used copy of Burkholder's 1999 book on eBay. Back then, everyone was making negatives with imagesetters. But affordable inkjet printers were becoming available, and he talks about his experience using them. This is where he explains his decision to make colorized negatives, rather than CMYK grayscale negatives: colorized gives you equal density with less ink, but not as good coverage (it was Burkholder who distinguished between "physical density" and "spectral density"). By 2013 (his "Inkjet Companion" book) he had abandoned color and gone back to CMYK grayscale. A grayscale printer curve (transmission density vs. digital tone) is a thing of beauty, even on my "K1" printer.

I use your printer settings (see below) !!

View attachment 282142

What happens if you eschew the use of icc profile altogether. I have started making my digital negatives without icc profile recently which anecdotally seemed to give me correction curves that are less "jumpy." The way I do that is by specifying Printer Manages Color in PS, on one hand and then specifying No Color Management the printer driver as usual. This can also be done using Adobe Color Printing Utility (ACPU.)

Another way to eliminate the ink blending all together and use pure inks would be through QTR, where a single ink can be specified for each curve. It would be interesting if those show more normal/continuous tones.

:Niranjan.

 

[jisner]

The way I do that is by specifying Printer Manages Color in PS, on one hand and then specifying No Color Management the printer driver as usual. This can also be done using Adobe Color Printing Utility (ACPU.)
:Niranjan.

The reason I don't use "Printer Manages Colors" is that I have never been sure what it means.

When you choose "Photoshop Manages Colors," Photoshop tells you to "Remember to enable the printer's color management in the print settings dialog box." That's the first problem: There is nothing called "Color Management" in the Epson P400 printer settings. There is, however, something called Mode. Let's assume Mode is Epson's equivalent of Color Management. The only choice that resembles "no color management" is Mode = Off (No Color Adjustment). Let's assume Photoshop is telling us NOT to make this choice. But we're going to ignore Photoshop and choose it anyway. What happens? Is the printer going to blow up? Probably not. Is it the same as what happens when you print with No Color Management in ACPU? I don't think so either.

When Printer Manages Color and you specify no color adjustment, you are giving the printer only two things to go on: the document profile and the Media Type. When the printer says to remember to enable color management, I think it is telling you that this is really not enough information to guarantee a good-looking print. The printer wants you to choose a Mode, perhaps by trial and error, that makes the print look good. If you choose Mode = Off (No Color Adjustment), the printer is indeed managing color, but only with the available information (document profile and Media Type). But I don't think it's the same as No Color Management in ACPU.

With no color management in ACPU, I assume Photoshop sends RGB numbers and the printer driver converts them to CMYK numbers. I wonder if the conversion is the same as what you see in the Color Picker when you type in three RGB numbers and it shows you the equivalent CMYK numbers. If so, that would be an good reason to choose ACPU. You'd know exactly how much ink was being used to make each color.

I feel more confident using Photoshop Manages Color and choosing an output ICC profile that gives the best result when printing on transparency. I have been using the output profile SC-P400 Series Photo Paper Glossy. I honestly don't know if it gives the "best" result" because I haven't tried others. Even if I did try others, I have no scientific criteria by which to evaluate the results.

 

[jisner]

I have started making my digital negatives without icc profile recently which anecdotally seemed to give me correction curves that are less "jumpy." The way I do that is by specifying Printer Manages Color in PS, on one hand and then specifying No Color Management the printer driver as usual. This can also be done using Adobe Color Printing Utility (ACPU.)
:Niranjan.

Niranjan,
I would like to analyze the effect of changes in print settings on digital negative calibration. I will use three different settings
(1) the settings that I have been using (screenshot posted above), in which Photoshop manages color (I originally got these from you)
(2) the settings which you use now in which Printer manages color
(3) your ACPU settings

Would you kindly post a screenshot of your settings for (2) and (3) so I can replicate them as closely as possible on my P400 and ACPU?

My goal is not to show which settings are best. My goal is merely to show how much difference settings make.

I think people overestimate the importance of process ("darkroom") variables and underestimate the importance of printer settings. For example, most people know that if they tweak their chemistry, they really should recalibrate their digital negatives (whether they actually do it or not). But I wonder if people have the same degree of reluctance when it comes to changing printer settings.

 

[nmp]

They are fairly straight forward to do - don't know if you need screen shots.

All 3 of them have the same exact Printer Properties - as per the screen shot above.

Method #2: In PS, select Printer Manages Color in Color Handling. RI = Relative Colorimetric (probably immaterial) and BPC is grayed out.

Method # 3: In ACPU, the only option is to go into the print driver (Print > select printer > Properties) so there are no application-side selections (except layout parameters via Print Setup)

By the way, here is the link to the work of Doug Gray over at Luminous Landscape (someone whose quantitative work I respect a lot) that shows that the methods #2 and #3 should give identical outputs.

https://forum.luminous-landscape.com/index.php?topic=124847.msg1045076#msg1045076

Finally, yes I agree, all printer parameters must be maintained for a given correction curve. I save them as a named Setting like Digineg1 etc in the printer driver so I can always go back to the exact set of parameters when printing future negatives without having to select individual ones from the various menus.


:Niranjan.

 

[jisner]

By the way, here is the link to the work of Doug Gray over at Luminous Landscape (someone whose quantitative work I respect a lot) that shows that the methods #2 and #3 should give identical outputs.
:Niranjan.

My results (see below) support Doug Gray's conclusion. Here's what I did:

I printed three 51-step tables colorized with CMYK=(0,71,71,0) on the same sheet of Pictorico, making three passes, each pass using different printer settings:

Method 1: my printer settings (screenshot posted earlier in this discussion)
Method 2: your Photoshop settings
Method 3: your ACPU settings​

I measured all three step tables with my transmission densitometer and created a printer curve for each method. I could show the printer curves, but instead I did three linearizations of a new cyanotype process, one with each method. Here are the resulting Photoshop adjustment curves (0 = black, 1 = white)

The Method 2 and 3 curves are nearly identical. Method 1 (my method, Photoshop manages color) gives a more irregular curve and introduces greater image distortion in the shadows and midtones (i.e., the corrections are more severe).

 

[nmp]

The Method 2 and 3 curves are nearly identical. Method 1 (my method, Photoshop manages color) gives a more irregular curve and introduces greater image distortion in the shadows and midtones (i.e., the corrections are more severe).

Those are great results....quick too. I take these were simulated correction curves not from actual cyano measurements. That would be real quick. Do you have an automated UV transmission measurement capability?

Also, nice to see corroboration of Doug Gray's reflective spectrophotometer-based conclusion with a second independent transmittive method. So based on this, do we conclude that the so-called null-transform method is a better way to print a digital negative?

Another question: Do you create the step-wedge in CMYK to start with, then apply the fill color layer and send the file as is or do you convert it to RGB with AdobeRGB colorspace. If so, why not start with a RGB file, given that the actual image file will be in RGB space as well. Trying to understand your sequence here: because even though printers are CMYK printers (actually cCYmM....etc) they still need RGB input. As far as I understand, if CMYK data is sent, it will still convert into RGB first at some point in the pipeline or so I have read somewhere.

:Niranjan.

 

[jisner]

Doug Gray says "The printed patch variations are exactly what one gets just printing the same image multiple times." For this experiment, I printed a single 51-step table for each method (all on the same sheet of Pictorico). I normally use a more elaborate set of step tables:

and average the results, which hopefully minimizes the variations, so I suspect if I used the more elaborate scheme, we would see the method 2 and 3 curves coincide even more closely.

I will definitely adopt the the so-called null-transform method.

Yes, these results are predicted by the model. The model does a mathematical linearization based on two curves: the process curve and the printer curve. I normally put a lot of effort into getting good curves, so I trust the results more than I would trust results coming out of the darkroom. The diagram below shows the model of tone reproduction the way we normally think of it: as flowing from the digital image to the analog print.

The software runs the model in reverse from the print to the image. The workflow is very simple. For this experiment, I simply needed to plug in a printer curve for each method. I used one of my existing process curves. You would get just as much (and possibly more) value looking at the differences in the printer curves, but most people relate better to image adjustment curves. FYI here are the printer curves:

I would really like to give you a zoom demo and get your feedback. If you send me an email, we'll set one up.

Re: the step table question. My step tables are all in Adobe RGB. They start out as grayscale (R=G=B). To colorize them, I add a color fill layer in Color mode with a chosen color. Recently I had been playing with the color Dan Burkholder used in his 1995 book, CMYK = (0,71,71,0), so that was the color I used. I could have just as easily have written RGB = (215,110,83).

 

[jisner]

Here's another test comparing the "null transform" method (printer manages color, no color adjustment) with ACPU. For this experiment, I merged measurements from a much more elaborate set of step tables (see the image in the last post), so I'm confident in the numbers.

 

[nmp]

Here's another test comparing the "null transform" method (printer manages color, no color adjustment) with ACPU. For this experiment, I merged measurements from a much more elaborate set of step tables (see the image in the last post), so I'm confident in the numbers.

View attachment 283796

They look identical - except the tiny bit at the Dmax end, which may or may not be statistically significant....interesting that these curves and the ones from before both show ACPU with slightly lower Dmax that the null transform case. No idea what to make of it. Just an observation.

:Niranjan.