This is mad-scientist territory,
I am not mad! Don't call me mad!

Maybe a bit angry, or possibly a trifle peeved. Or maybe irked. Yes, irked, that's it...

I'm an irked scientist!

but what about mapping the image to a gradient that goes between the weakest UV blocker and the strongest UV blocker? In theory, that would give you the most ink in all the parts of your negative.
That will work, if your printing process is in the control of a RIP. For example, to set this up with QTR, print their standard cal chart in color, print that chart onto the paper of interest (PT, PD, cyanotype, etc) and measure the densities of the print. Then subtract all the densities from DMAX, because the QTR calibration process won't work with print density that decreases as ink density increases.

I am very encouraged with the look of the negatives using the imageprint RIP. They looks smoother under a 10x loupe than the Epson driver negatives. I am going to do some printing tests this weekend to see if this turns out to be the case. If this works, it means that one adjustment curve can be used for any printer that Imageprint supports, since they profile all of their printers to behave the same way. I'll report in.
They don't profile anything to "behave the same way" in the UV, or in the transparency mode. That's the big drawback with making digital negatives with colored inks to begin with, you never know when Epson is going to make a little change in the UV absorbtion of the cyan ink, or in the visible blue blocking ability of the yellow ink. (those are both real examples from about 18 months ago).

Best thing in the world for digital negatives is still lots of dilutions of neutral carbon black.

The point that I was trying to make was that it may be better (smoother) to use a different combination of inks for the low densities than that used for the high densities.
You're quite right. One approach I used a couple of years ago was to print a thousand square chart, all possible combinations of red, green, and blue at 10 values.

0, 28, 57, 85, 113, 142, 170, 198, 227, 255

This was printed on transparency material at 9x7 with 1/4 inch squares, contact printed, scanned on a 4800 dpi flatbed, and dumped into a matlab program that spend the next few days chewing on a 1.4 gigapixel image computing the mean, variance, and standard deviation (use to judge the "smoothness" of each square) and then give me a list of the smoothest colors for 15 target density neighborhoods.

After that, computing the curves was a snap.

And, the next time Epson changes the UV blocking characteristics of the green without changing the other colors, you can flush the whole thing down the toilet.

While I haven't seen the output from IJC/OPM rip, I have had a chance to compare the output between Imageprint RIP, Quadtone RIP and the Epson driver on the 7800. They all have slightly different dither patterns and the dots just look different when you stick on loupe on the paper. I'm guessing that IJC/OPM will have yet another slight variation in the way it makes the printer lay down its inkload.
Quadtone RIP and BowHaus Inkjet Control / OpenPrintMaker should be very similar, as both use versions of the Gimp Print engine. I don't know what version IJC/OPM uses, but QTR uses version 4.3.5, a development "fork" of the 4.2 platform that was a precursor to 5.0.

You're quite right. One approach I used a couple of years ago was to print a thousand square chart, all possible combinations of red, green, and blue at 10 values.

0, 28, 57, 85, 113, 142, 170, 198, 227, 255

This was printed on transparency material at 9x7 with 1/4 inch squares, contact printed, scanned on a 4800 dpi flatbed, and dumped into a matlab program that spend the next few days chewing on a 1.4 gigapixel image computing the mean, variance, and standard deviation (use to judge the "smoothness" of each square) and then give me a list of the smoothest colors for 15 target density neighborhoods.

After that, computing the curves was a snap.

And, the next time Epson changes the UV blocking characteristics of the green without changing the other colors, you can flush the whole thing down the toilet.

An excellent and bold analysis attempt to be sure! You're way ahead of the curve...to pardon a pun. I have a pile of questions.

What did you learn from printing out those 1000s of squares? Was there an advantage in doing this? Did you end up with using a single hue as fill? How did you apply the colour to the negative after matlab processing revealed the smoothest colours? Was the curve even necessary? Did any of the study involve analysing the final print outs, did you use a UV densitometer? Was the smoothness in visible light often different than that under UV light? Can you post some excerpts or some abstracts? Thx.
~m

An excellent and bold analysis attempt to be sure!
Glad you enjoyed it.

You're way ahead of the curve...to pardon a pun. I have a pile of questions.

What did you learn from printing out those 1000s of squares?
I learned that the different visible light and UV absorbtion of the different colors of ink makes negatives act very strangely. You may find one color that prints the smoothest light grays, but is much grainier than another color in the dark grays.

Was there an advantage in doing this?
Yes. I was able to find the color that produced the smoothest print at each desired density.

Did you end up with using a single hue as fill?
No, I had a complex "zigzag" of hues through density space. Light colors were black, as density increased it went green, then cyan, then red (if memory serves)

How did you apply the colour to the negative after matlab processing revealed the smoothest colours?
I output three comma separated value lists of curves, then import those into PhotoShop through a utility that converts .csv to .acv files (I think it's called curvalicious).

Was the curve even necessary?
Very much so. Three curves, red, green, and blue, instead of just one.

Did any of the study involve analysing the final print outs, did you use a UV densitometer?
Yes, and no. The equivalent density was computed by measuring print density where the print also contained a Stoufer wedge, and interpolating density values to the curve for the Stoufer.

Was the smoothness in visible light often different than that under UV light?
Yes. The colors required for a negative varied dramatically depending on whether printing a UV or visible light process.

Can you post some excerpts or some abstracts?
Not at this time. It's a long abandoned project.

Thx.
~m
You're welcome

Hi Wiz,

I've found your posts very informative and helpful. Thanks!

I'm curious about your use of many different dilutions. What is the advantage of your method over using, say, three dilutions, but all the cartridges - so on a 2200 maybe two dilute cartridges, four medium and one full-strength?

I'm working on the assumption that the more nozzles I can get firing to contribute to a particular tone, the smoother that tone will be. Is this a good assumption? It means that, in order not to have too much ink laid down, each ink channel is working at fairly low volume. Is this better than one or two channels at higher volume to produce a given density?

I've put together a spreadsheet routine for QTR that roughly models the total blocking and total ink volume, based on measurements of QTR calibration strips for each ink (only in blue, with a calibrated scanner, I don't have a UV gizmo). This seemed to work pretty well for the UC inks once I figured out the correct curve shape to give good tone separations in the print. It can be tweaked with a few trial prints of a step wedge. The final result is still grainy, as I didn't do the research you described above on least grainy colors/densities. This was an interim stage - I'm moving on to a more monchrome set-up and an 1800 next, so would love to have your input from you or others on these questions.

Oh, and one more - what's your experience of the different dither options on the blocking and grain?

I'm printing in Ziatype right now, but I may want to do some silver stuff later.

Best,

Ben