how to measure dmax on paper

[elliot_n]

This issue was dealt with in this blog on the 3800. 

http://www.outbackphoto.com/printinginsights/pi045/essay.html#20070206

Reassuring reading. However it doesn't tally with my experience. I've recently been running tests with my Epson 3880, comparing the rendering of fine detail with the settings '1440/high speed on' vs '2880/high speed off/fine detail on'. Even with a loupe, it is hard to see any difference in the presentation of fine detail (though if pressed, I'd say that the latter settings give a sightly superior result). However only a casual glance at the two prints is necessary to reveal a difference in colour. One print is warmer and more saturated than the other. So on my setup it seems that separate custom profiles will be necessary if I'm going to switch between different print driver settings (1440/2880, High Speed On/Off, Finest Detail On/Off). — I'm surprised by this, and I'm still exploring...

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

Further exploring...

I've now done some more careful tests, running nozzle checks after each print to make sure nothing is amiss. The source image is a sharp Nikon D800 file, downsized to 5"x7.5" at 720ppi. The prints were examined with a loupe, and also scanned on a flatbed for closer examination. In total, four prints were made, with the following driver settings:

1. 1440 High Speed
2. 1440
3. 2800
4. 2880 Finest Detail

Observations:

1. There is no difference between prints 1 and 2. Using 'High Speed' has no negative effect on print quality.

2. From a distance, prints 3 and 4 look the same as each other, and they look different from prints 1 and 2. The difference is that they have slightly less density in the midtones. In Photoshop terms, they can be corrected to match prints 1 and 2 by adjusting the middle Levels slider from 1.00 to 0.96. Therefore on my printer I would say that switching from 1440 to 2880 has a subtle but significant effect on the tonality of the image, and to compensate for this separate profiles should be made.

3. Viewed close-up, with a loupe and (especially) with a high res flatbed scan, there is a significant difference in the rendering of fine detail between prints 3 and 4. Checking 'Finest Detail', contrary to the Epson manual, renders photographic detail more finely.

[datro]

There used to be something called "Companding Calculator" on that website - not any more, apparently for whatever reason. 

I used to use the Companding Calculator for conversions between L* and Density before it was removed from Bruce Lindbloom's website.  My understanding is that he had to remove it when modern browsers stopped supporting running JAVA apps in the browser for security reasons (note here that I'm talking about JAVA, NOT Javascript which is something entirely different).  This was at least several years back now.  In any case, as Mark pointed out in his earlier post, i1Profiler provides the density information from patch measurements and that is what I use now.

More recently I had a need to actually do the calculation myself in some code I was writing for a "Linearity Checker" application for use in linearizing my QTR/Piezography quad files.  After quite a bit of googling and technical background reading, I came to understand how to do it.  Lightness (L*) and Visual Density are slightly different animals and the conversion requires adjusting for those differences and human visual response characteristics.  The key is first calculating reflectance (R); once you have that then density is simply  -log10(R).

So, in case anyone has a need to do the calculation themselves (in code or just by hand), I've attached my short Python function that does the conversion.  I've validated it with i1Profiler data and similar tools and it tracks accurately.  Hope this might be useful.

Edit: Added missing minus sign in front of "log10(R)".

 

[nirpat89]

I used to use the Companding Calculator for conversions between L* and Density before it was removed from Bruce Lindbloom's website.  My understanding is that he had to remove it when modern browsers stopped supporting running JAVA apps in the browser for security reasons (note here that I'm talking about JAVA, NOT Javascript which is something entirely different).  This was at least several years back now.  In any case, as Mark pointed out in his earlier post, i1Profiler provides the density information from patch measurements and that is what I use now.

More recently I had a need to actually do the calculation myself in some code I was writing for a "Linearity Checker" application for use in linearizing my QTR/Piezography quad files.  After quite a bit of googling and technical background reading, I came to understand how to do it.  Lightness (L*) and Visual Density are slightly different animals and the conversion requires adjusting for those differences and human visual response characteristics.  The key is first calculating reflectance (R); once you have that then density is simply log10(R).

So, in case anyone has a need to do the calculation themselves (in code or just by hand), I've attached my short Python function that does the conversion.  I've validated it with i1Profiler data and similar tools and it tracks accurately.  Hope this might be useful.

 

Never used Python so you made me look it up....learn something new everyday....

Comparing your code with the Excel formula I posted earlier, I can see the first part is identical.  The second part (where the condition is not met) was not immediately apparent to me as equivalent so I plugged your formula in Excel and compared with earlier results.  They turned out to be identical (to at least 2 decimal points.) 

By the way, the condition can be simplified as L* > 8.  So there is one equation for L* above 8 and another for equal or below. 

Is the Y here the same as that of the CIE XYZ fame? 


:Niranjan.

[datro]

Is the Y here the same as that of the CIE XYZ fame? 

Yep.  In the code, "Yn" represents the max luminance tristimulus value and "Y" is the calculated value based on the L* input.

[NeilPrintArt]

2. From a distance, prints 3 and 4 look the same as each other, and they look different from prints 1 and 2. The difference is that they have slightly less density in the midtones. In Photoshop terms, they can be corrected to match prints 1 and 2 by adjusting the middle Levels slider from 1.00 to 0.96. Therefore on my printer I would say that switching from 1440 to 2880 has a subtle but significant effect on the tonality of the image, and to compensate for this separate profiles should be made.

Interesting!! Can I ask how you made the profile? I am wondering if the number of patches used in the test chart would have an effect? Maybe if one used a large (2000+ in i1Profiler) number of patches for the target then that would eliminate the differences you have observed?   

[elliot_n]

Interesting!! Can I ask how you made the profile? I am wondering if the number of patches used in the test chart would have an effect? Maybe if one used a large (2000+ in i1Profiler) number of patches for the target then that would eliminate the differences you have observed?   

My test prints were made on Harman by Hahnemuhle Gloss FB, using a custom profile I had made several years ago. The profile chart was a standard one of about 1,000 patches, and was printed at 1440, High Speed. I will shortly send off for new profiles for this paper, and including charts for both (1) 1440, High Speed and (2) 2880, Finest Detail. I'm interested to see if separate profiles can eliminate the (subtle) differences in tonality I am seeing with the different print settings.

[nirpat89]

Interesting!! Can I ask how you made the profile? I am wondering if the number of patches used in the test chart would have an effect? Maybe if one used a large (2000+ in i1Profiler) number of patches for the target then that would eliminate the differences you have observed?

In theory, whether or not there is a discrepancy between different modes shouldn't be affected by how the profile was made, or how accurate it was.  I would think this would manifest itself regardless of what profile was used...no?

[nirpat89]

My test prints were made on Harman by Hahnemuhle Gloss FB, using a custom profile I had made several years ago. The profile chart was a standard one of about 1,000 patches, and was printed at 1440, High Speed. I will shortly send off for new profiles for this paper, and including charts for both (1) 1440, High Speed and (2) 2880, Finest Detail. I'm interested to see if separate profiles can eliminate the (subtle) differences in tonality I am seeing with the different print settings.

Can you see a difference if you print a simple step-wedge with different conditions side by side on the same paper, I wonder.

[elliot_n]

Can you see a difference if you print a simple step-wedge with different conditions side by side on the same paper, I wonder.

I can only assume so, if I can see the difference on two separate prints.

I'm not a colour geek, and not particularly technically minded, but I have a keen eye for subtle differences in print quality, and for me the change in tonality when switching from 1440 to 2880 is enough for me to reject the 2880 print. Getting two separate custom profiles would fix the problem, but I wonder whether I'll ever actually want to print with settings other than 1440 High Speed.

[digitaldog]

I can only assume so, if I can see the difference on two separate prints.

I'm not a colour geek, and not particularly technically minded, but I have a keen eye for subtle differences in print quality, and for me the change in tonality when switching from 1440 to 2880 is enough for me to reject the 2880 print. Getting two separate custom profiles would fix the problem, but I wonder whether I'll ever actually want to print with settings other than 1440 High Speed.

There's no need to make any colorimetric assumptions; you have the tools to know exactly the differences in the targets used to create a profile.
1. Print out a target at 1440 and let it dry. Do the same for 2880.
2. Measure one (say 1440) two times in a row. Create a deltaE report ideally an average and with worst dE. I'd use dE2000. This report provides the noise inherent in all measuring devices. It should be well below 1.0 dE.
3. Measure the other target and now compare the dE between 1440 and 2880, keeping track of the dE noise you should subtract from the two.
Now you know exactly the differences between the two resolutions. And you know the dE noise inherent in the measuring process.

[elliot_n]

There's no need to make any colorimetric assumptions; you have the tools to know exactly the differences in the targets used to create a profile.

I don't have the tools. (I send out for profiles.)

[digitaldog]

I don't have the tools. (I send out for profiles.)

So you're not making printer profiles? In which case, stick to the advice from those that do 😀

[elliot_n]

So you're not making printer profiles? In which case, stick to the advice from those that do 😀

Your advice - 'one profile for all' - doesn't work for me. If I use the same profile for 1440 and 2880 prints, the prints looks different.

[digitaldog]

Your advice - 'one profile for all' - doesn't work for me. If I use the same profile for 1440 and 2880 prints, the prints looks different.

What you see and what is measured can differ! I'll not comment on your visual accuracy but could on actual measurements for profile creation!

"I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind".
-Lord Kelvin

[elliot_n]

What you see and what is measured can differ! I'll not comment on your visual accuracy but could on actual measurements for profile creation!

I don't have a spectrophotometer but I do have a flatbed scanner and this confirmed my visual assessment. The 2880 print is lighter in the midtones than the 1440 print.

[digitaldog]

I don't have a spectrophotometer but I do have a flatbed scanner and this confirmed my visual assessment. The 2880 print is lighter in the midtones than the 1440 print.

Sorry no, the scanner isn’t appropriate and IS NOT measuring.
“There are two ways to be fooled. One is to believe what isn't true; the other is to refuse to believe what is true.” -Søren Kierkega

[elliot_n]

Sorry no, the scanner isn’t appropriate and IS NOT measuring.

Why is it not appropriate? I can lock the scanner off, so it scans in a fixed state. Sure, a spectrophotometer would be handy. But if my eyes tell me the two prints are different, and my scanner supports this observation, that is sufficient for me - the 1440 and 2880 prints are different.

[digitaldog]

Why is it not appropriate? I can lock the scanner off, so it scans in a fixed state. Sure, a spectrophotometer would be handy. But if my eyes tell me the two prints are different, and my scanner supports this observation, that is sufficient for me - the 1440 and 2880 prints are different.

For the same reasons you don't use a scanner instead of a Spectro to build printer profiles. But it's moot: you don't have the tools but do have the instructions to correctly evaluate the differences that those of us that do have done!

[elliot_n]

For the same reasons you don't use a scanner instead of a Spectro to build printer profiles. But it's moot: you don't have the tools but do have the instructions to correctly evaluate the differences that those of us that do have done!

When I next send out for profiles I will ask the guy who makes them to give me a report of the difference between the 1440 and 2880 charts. I will be surprised if they are the same.

Just to clarify - you would expect them to be identical?

[elliot_n]

In the meantime, here are scans of two prints, one printed at 1440dpi, the other at 2880dpi.

The 1440dpi print is darker than the 2880dpi print.