how to measure dmax on paper

[elliot_n]

I don’t know if you are purposely trying not to understand this, or if you are really struggling with it.

The only thing I'm struggling with is your attitude. Why so angry?

[Doug Gray]

I don't find the colorimetric data difficult to either accept or understand. I accept that your 3880, and the 3800 tested on the linked website, print the same whether 1440 or 2880 is selected. But my 3880 does not. I am simply trying to figure out why this is the case, and whether separate custom profiles for 1440 and 2880 would ameliorate the problem.

I can and do make lots of profiles since I have the instruments and software, but I can't think of a reason to print different resolutions. If I'm bothering to use a good color managed process it just never occurred to me to print with different settings than the ones I made profiles with. I have not done critical printing in lower resolution modes but color accuracy was never a major consideration in those modes and, frankly I didn't bother looking at it before.

As for other printers, like the Pro1000, for the highest accuracy they need different profiles just because the highest setting yields a dark, L*=2.0 black and the next highest only gets to 3.0.  Pretty much requires two profiles though printing with Perceptual Intent or Relative with BPC, it's essentially impossible to see any difference between the two as they scale to the BP. But again, why print using a setting different than the profile was created with?

[digitaldog]

Why so angry?

That's just another of your assumptions.

[digitaldog]

I can and do make lots of profiles since I have the instruments and software, but I can't think of a reason to print different resolutions. If I'm bothering to use a good color managed process it just never occurred to me to print with different settings than the ones I made profiles with. I have not done critical printing in lower resolution modes but color accuracy was never a major consideration in those modes and, frankly I didn't bother looking at it before.

So you'll understand this. First, with a wet target at two resolutions, the average dE is less than 1. Now the next test would be to build a profile from each measurement target at the two resolutions (so, profile 1440 and profile 2880 as an example). Next we have to pass image data through both, again the TC918 (and the RI would play a role as would an actual image play a role but let's leave that for now). So there's going to be some rounding there. Next we need to actually print the TC918 from the two profiles and measure them again. So we start with a difference in just the two targets at two resolutions that when wet is an average of less than 1. What do you suppose the visual and measurement results will be going full circle from target to profiles to conversions to print and measurement again?


It might make an interesting time consuming experiment. But that the very first comparison averages less than 1 deltaE, and from a wet target at that, I don't see the point. But I encourage Elliot to imagine this and would love to see you test it but I've got to finish a really good Stephen King novel which I think is far better use of my time. 

[elliot_n]

But again, why print using a setting different than the profile was created with?

Well, because I followed the line given by Epson and Andrew, that you only need one profile per paper.

And this led me to the bad habit of getting profiles made at the best driver settings (2880, No High Speed), when in actual fact I would be mostly printing at the fastest driver settings (1440, High Speed).

It never occurred to me that switching resolutions would affect the tonality of the image. I only discovered it yesterday when making some careful side-by-side print comparisons.

[Doug Gray]

So you'll understand this. First, with a wet target at two resolutions, the average dE is less than 1. Now the next test would be to build a profile from each measurement target at the two resolutions (so, profile 1440 and profile 2880 as an example). Next we have to pass image data through both, again the TC918 (and the RI would play a role as would an actual image play a role but let's leave that for now). So there's going to be some rounding there. Next we need to actually print the TC918 from the two profiles and measure them again. So we start with a difference in just the two targets at two resolutions that when wet is an average of less than 1. What do you suppose the visual and measurement results will be going full circle from target to profiles to conversions to print and measurement again?


It might make an interesting time consuming experiment. But that the very first comparison averages less than 1 deltaE, and from a wet target at that, I don't see the point. But I encourage Elliot to imagine this and would love to see you test it but I've got to finish a really good Stephen King novel which I think is far better use of my time. 

I saw this discussion and realized I'd never bothered to compare targets colorimetrically with the higher/lower resolution on the 9800.

So I ran a test and was frankly surprised at how large the average dE's were printing the same target with 720x1440 and 1440x2880. Some time back I checked the difference between bi-directional and single directional printing. One gets slightly better alignment of fine detail but the colors didn't change much at all. IIRC, well under 1 dE. So I was somewhat surprised to see  the larger differences with the lower resolution mode. Especially as the canned 9800 profiles were supposed to be for any DPI.

[Doug Gray]

Well, because I followed the line given by Epson and Andrew, that you only need one profile per paper.

And this led me to the bad habit of getting profiles made at the best driver settings (2880, No High Speed), when in actual fact I would be mostly printing at the fastest driver settings (1440, High Speed).

It never occurred to me that switching resolutions would affect the tonality of the image. I only discovered it yesterday when making some careful side-by-side print comparisons.

Well, there you go. Printing at lower Q levels comes with a price, and on some printers, including yours and mine, that includes some color shift. I normally print at 1440x2880 dual direction, but sometimes print critical stuff single direction. But the latter is way slower. And that's why I had, in the past, checked for color accuracy. But I'd never bothered with the lower rez. But then I'd also never compared prints side by side. Viewed separately, I'd not noticed anything but I didn't print the same image that way so I wouldn't have noticed.

Andrew's printer is quite good and stable between the two rez modes and I had expected to see similar results. But my printer clearly has more variation.

[elliot_n]

2880 dual direction sounds like a good balance between speed and quality. Like you, I see no change in colour or tonality when 'high speed' is toggled on and off. (But how about 'Finest Detail'? That's another puzzle!)

I'm wondering whether my print set-up might have exaggerated the differences between 1440 and 2880? Specifically, I'm printing on a third party paper (Harman by Hahnemuhle Gloss FB), with Platen Gap set to 'Wide', and Paper Thickness set to '4'. These are the commonly recommended settings for this type of paper, but perhaps they are not optimal for getting the same tonality when switching between 1440 and 2880?

Edit to add:

No, I've repeated the test with Epson Semigloss, canned profile, standard Advanced Media Control settings (Paper Thickness 3, Platen Gap Auto) and the issue persists — 2880, on my printer (Epson 3880), prints darker than 1440. This darkening can be seen most from the mid-tones down into the shadow-tones.

[elliot_n]

Any results, Andrew?

[digitaldog]

Any results, Andrew?

Yes, as expected, even MORE of an invisible difference after try down:



Average dE:   0.84
    Max dE:   1.58
    Min dE:   0.04
 StdDev dE:   0.34


Once again, absolutely no reason to print TWO targets for TWO profiles; the differences in the resolution is invisible.

[elliot_n]

Thanks for the info.

I'll be printing two targets for two profiles as my printer behaves differently to yours.

[digitaldog]

Thanks for the info.

I'll be printing two targets for two profiles as my printer behaves differently to yours.

Knock yourself out, your paper, ink and time (and money for profiles).
What would be useful, and you can't accomplish it, would be to actually measure both targets and produce a dE report.

[elliot_n]

Knock yourself out, your paper, ink and time (and money for profiles).
What would be useful, and you can't accomplish it, would be to actually measure both targets and produce a dE report.

I'll ask my custom profile guy to do that, and I'll report back here. (Won't be for a couple weeks.)

[digitaldog]

I'll ask my custom profile guy to do that, and I'll report back here. (Won't be for a couple weeks.)

Just ask him to also email you the CGATs measurement files used to build the profile.

[elliot_n]

Just ask him to also email you the CGATs measurement files used to build the profile.

I doubt I would understand it (?)

I think I just need a simple report:

Average dE, Max dE, Min dE, StdDev dE.

[digitaldog]

I doubt I would understand it (?)

I think I just need a simple report:

Average dE, Max dE, Min dE, StdDev dE.

You would not but I would and could produce a DeltaE report from these text files.

[elliot_n]

Oh, I see. Thanks!

[Doug Gray]

Being surprised at the relatively large dE00 differences between 1440x2880 and 720x1440 printing on my Epson 9800, I ran further tests on these with bi-directional printing: 720x720, 720x1440, 1440x2880. I also ran a comparison between 1440x2880 bidirectional and unidirectional though I had previously tested the latter and found only small differences. But it was a while ago and I didn't record the specifics. The tests confirmed my recollection that they were pretty close.

Results:
Bidirectional 1440x2880 and 720x720 / 720x1440 were both different by an average of 2.2/2.1 dE00.
Bidirectional 720x720 and 720x1440 were different by an average of 1.1 dE00.
Bidirectional 1440x2880 and unidirectional 1440x2880 differed by an average of 0.51 dE00.

One curiosity is that the average dE00 between the same RGB patches printed at different locations, a test I incorporate to check printer consistency, dropped from .19 to .13 dE00 average.  OTOH, printing in that mode doubles print time and I've never seen a visual difference outside of a loupe.