Difference in shade of color between 8 bit versus 16 bit setting

[alan a]

Thanks in advance for any responses!  Before posting here, I searched older threads, and the difference I describe below between 8 bit versus 16 bit should not be happening -- but it is.

I recently had a problem with a print that prints darker than how it appears on the iMac screen in Lightroom.   I am printing from a Canon raw file taken with a 5D Mk III, printing on a Canon 8400.  Both the printer and the iMac have been profiled.  I'm printing on Epson Luster.

The difference is most noticeable in the blue sky, where there is a very clear difference in the shade of blue.   To rule out possible causes, I changed all of the various print settings, including changing printer profiles, using perceptual vs relative colormetric, and changed the "high" and "highest" settings. 

I finally narrowed it to the single setting that causes this result -- turning 16 bit on and off.

The print in 8 bit has a clearly lighter blue in the sky, and the print in 16 bit has a clearly darker blue in the sky.  (The 16 bit file is generally darker across the print, but the difference is most pronounced in the blue sky.) 

In my tests, I kept all settings in the print driver the same, except turning 16 bit on and off.  And got the results described above, so I don't know what other variable to attribute the difference to.

(1)  Why would turning 16 bit on and off result in such a change?  I thought the difference between 8 bit versus 16 bit was supposed to be subtle.

(2)  Could the cause be whether the printer profiles were created and printed within the Canon Photoshop plug in with 16 bit turned on or off?  Although I am pretty sure that I maxed out all the settings in the plugin when I printed the profiles, so presumably used 16 bit.

(3)  The version of the print that matches the profiled iMac screen is the 8 bit version.  This is logical, since I believe an iMac screen (as least one that is 4 or 5 years old like mine) only shows 8 bit colors.  That would mean that if there is a difference in colors between 8 bit and 16 bit, it would make sense to only print in 8 bit, so what I see on the screen will match what is printed.

Does any of this come as a surprise to the experts on the forum?  It certainly surprises me, as I thought the difference between 8 versus 16 bit was supposed to be so subtle as it would be difficult to see the difference.  That is certainly not my experience.

[digitaldog]

I recently had a problem with a print that prints darker than how it appears on the iMac screen in Lightroom.

That's a display calibration issue:

Why are my prints too dark?
A video update to a written piece on subject from 2013
In this 24 minute video, I'll cover:
Are your prints really too dark?
Display calibration and WYSIWYG
Proper print viewing conditions
Trouble shooting to get a match
Avoiding kludges that don't solve the problem

High resolution: http://digitaldog.net/files/Why_are_my_prints_too_dark.mp4
Low resolution: https://youtu.be/iS6sjZmxjY4

I finally narrowed it to the single setting that causes this result -- turning 16 bit on and off.

I find that difficult to believe. Never seen that. Shouldn't make any difference. Further, I don't think your Canon supports actually sending high bit data to the driver (the Plug-in might).
Maybe it's a bug, but what you report shouldn't be happening. Sure it's not some other setting? Got a different printer you can try? If you save through a PDF, see this issue?

[alan a]

I agree that it should not be happening.  I just made several more prints, and here are the settings in the Canon driver for the 8400 and in LR.


All test prints used --
Print quality -- highest
No color correction
Unidirectional not checked
Perceptual in LR

One print used 8 bit.  The second used 16 bit in the Canon driver, with 16 bit checked in LR.  The third used 16 bit in the Canon driver, but 16 bit was not checked in LR.

The results are the same -- a clear difference in the darkness of the blue in the sky with 8 bit lighter and 16 bit slightly darker, but that slight difference is noticeable.

I don't doubt that this should not be happening, but it is, and there are no other changes in the print driver except for 8 bit versus 16 bit.

[digitaldog]

So strange.
So if you move from 16-bit to 8-bit per color, then convert back to 16-bit, same results?
What if you use another image, something like this:
http://www.digitaldog.net/files/2014PrinterTestFileFlat.tif.zip
Again printing both ways.


So you have a specific setting in the driver for 8 vs. 16 bit? And it's the setting there, not the image data that makes a difference? If so, I think you found a driver bug.

[alan a]

I'm not "converting" to respond to your question.  I am just printing the images, and changing from 8 bit to 16 bit in the Canon 8400 driver.  Yes, it is the setting there that makes the difference.  At least in LR.

Just finished printing your test pattern, as we'll as one from Atkinson.  The difference is striking.  The 8 bit version is clearly correct in terms of facial tones, and that is especially evident in the photo of the older white woman on the right.  The 8 bit version has some pink in the skin tones, whereas the 16 bit is clearly tan. 

The other differences are even more evident.  The vertical color stripes on the left**, directly under the photos of the three women.  The stripes in the 8 bit version are gradual in in how they change color.  Not so in the 16 bit.  Of the three 16 bit stripes in the top set of vertical strikes, the stripe on the right is blue.  In the 8 bit the color transition is gradual and I would describe the color as light cyan.  In the 16 bit, it is darker blue, with a solid color in the top half, and half way down, there is an abrupt change in the blue, almost like a line.  This behavior is repeated in the next set of three stripes down, in the middle stripe.  In the 8 bit it is a gradual transition of green. In the 16 bit it again is a solid green on the top half, and abruptly changes like a line.  In the 8 bit the two dogs are white or grey as clearly intended.  In the 16 bit there is a light but clear brownish cast. 

In your test image there is a block of color, under the three women, and to the right of the vertical stripes described above.  The 8 bit prints fine.  In the 16 bit, the transition of the blue is almost like a line, and that is repeated on the green.

The Atkinson test pattern is 12 blocks of color each of which have transitions to other colors.  On three of them, for lack of a better way to describe it, it looks like the colors bleed on the edges.

I just finished printing a second time, and the results are repeated.

Anyone else with an 8400, using an iMac and OS 10.11.3, with Lightroom CC, can download Andrew's test image, and try it yourself.  It would be interesting to know if anyone else repeats these results.

**My first post said on the right, and I corrected that -- it is the stripes on the left under the photo of the woman furtherest over on the left.

[digitaldog]

Sure sounds like a driver bug.

[alan a]

So how does an average consumer like me report it?  The guys who staff the tech support line for Canon pro printers would ignore it and just say it is a flaw in my own system. 

[alan a]

(1)  It just occurred to me that based on the test prints, if anyone saw them and was asked to guess which is which -- you would say that the 8 bit print is the 16, and the 16 is the 8 bit.  Because it is the 8 bit print that clearly shows the best and gradual color transitions, and the 16 bit that is abrupt in the transitions.

Could the bug be that basic?

(2)  Am I correct that an iMac only shows an image in 8 bit in any case?  Do any high end monitors actually show 16 bit images, and if so, how many colors is that?  And even if a high end monitor could do so, and I believe that is over one billion colors, how many video cards can actually show 16 bit range of colors?

(3)  I assume the real advantage of 16 bit is less banding in solid colors, and that is a difference I did notice when comparing the lighter 8 bit and the darker 16 bit.  So it would be nice if the 16 bit setting actually worked.

But I am at a loss as to how to report this to Canon.

[Wayne Fox]

Are you printing from lightroom through the canon driver, but not through the canon plugin?

I’m not an expert on Canon, but I thought I read that canon does not support the 16 bit path through the Mac OS driver, only through the canon plugin. With an Epson printer, the visible difference between printing with 8 bit vs 16 bit is pretty much undistinguishable so I don’t think you are losing anything just staying with 8 bit.

As far as your display, iMac’s are 8 bit.  there are some 10 bit display solutions out there, no 16 bit that I’m aware of.

[Doug Gray]

I agree with Andrew. Sounds like a driver bug.

The bug issue aside, has anyone actually measured tonal smoothness improvements between 8 and 16 bit printer driver setting? There should be some, but even at 8 bits, Using 8 bits, I get a fair number of luminance reversals on a (0,0,0) to (255,255,255) at the device with a gray scale print. I've never seen a reversal on a monitor. Since I've only checked this on my Canon 9500 II, I am curious how much actual improvement going to 16 bit would provide when the ramp is relatively noisy with 8 bits. Any printer that offers 16 bit drivers should get some benefit and it would be great to see printer reviews that address this. It might be marketing hype for some but actual improvement for others.

[Wayne Fox]

Hard to believe it's a driver bug ... lots of people printing ipf printers so one would think it would have turned up long before now.

Regarding my earlier comment about being unsupported in Mac OS ... sort of dates me I guess.  this goes back to the original pif printers and before OS X 10.4.  So disregard that theory.