How to test if a print was made at 16 bit?

[faberryman]

I repeated my tests.

I am seeing banding, admittedly subtle, when the image is printed with 16 bit unchecked in the Epson print driver*. There is no banding when 16 bit is checked.

I see this when printing from Photoshop, and also when printing from ACPU.

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

* When printing from Photoshop, I have 'Send 16-bit Data' checkbox in the Photoshop print dialog checked. I am toggling the '16 Bit Output' checkbox in the Epson print driver.

I would trust my eyes. Let Andrew make whatever decision he wants for his own work.

[elliot_n]

The banding I'm seeing is very subtle. It is significantly less than the banding visible when the image is printed after being converted in Photoshop to 8 bit (with no dither). But it does indicate that the 16 bit checkbox in the Epson print dialog is not merely cosmetic - it does do something, and it should remain checked.

[digitaldog]

The banding I'm seeing is very subtle. It is significantly less than the banding visible when the image is printed after being converted in Photoshop to 8 bit (with no dither). But it does indicate that the 16 bit checkbox in the Epson print dialog is not merely cosmetic - it does do something, and it should remain checked.

Glad you reported this as something now is going on that is VERY odd.
I took a profile target with 918 patches hoping to make a print with check box on and off, nothing more, so I could measure and produce a colorimetric report. I printed the targets three times and I keep seeing this which is rather alarming.
Before I paste the report, again, this is 918 patches of differing colors I can measure on my iSis. I printed it 'untagged' (the assumption would then be ProPhoto RGB) but also assigned sRGB to the target. The original wasn't in 16-bit so I converted it as such. That alone shouldn’t be an issue. Again, everything is identical between the two prints going off to the Epson 3880 expect the check box for 16-bit is either ON or OFF in the driver. This is the difference which is huge:



Number of Samples: 918


Delta-E Formula dE2000


Overall - (918 colors)
--------------------------------------------------
  Average dE:  11.16
      Max dE:  42.30
      Min dE:   0.21
   StdDev dE:   7.14
95th %ile dE:  15.22


Best 90% - (825 colors)
--------------------------------------------------
Average dE:   9.33
    Max dE:  20.13
    Min dE:   0.21
 StdDev dE:   4.47


Worst 10% - (93 colors)
--------------------------------------------------
Average dE:  27.40
    Max dE:  42.30
    Min dE:  20.13
 StdDev dE:   5.73


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


Doug, I believe you have a printer from Epson with this option and I know you too can print and measure as I did. I need a major reality check here. The max dE (42) and the most egregious differences are in green. But even the average is alarming and I don't know if this might be a bug with Mojave, the driver, but it's very strange.


And now that I examine the two prints I made today, using Doug's construct, I do notice the gray balance is a bit different visually. With lots of colors, it's all over the map. The targets were 'wet', right out of the printer so the dE's would be a tad lower tomorrow but nothing is going to get the two close to where they should be.
I suppose I should also try the same test but use my Gamut Test File image which is natively 16bit and print one each with (check box off and on), the difference in colors should be visible. If anyone with a 3880 can test this, it would help. I suppose it could be the driver acting up with the new OS as I can't believe the check box would make such a huge difference. I can see the differences of the color patches, there was little need to measure them but none the less, we can see the huge colorimetric differences. So anyone with this printer on Mac running Mojave can test this without the need of a Spectrophotometer. Look at greens. The lowest dE is a middle gray (0.21). So the black ball is in an area of color space (neutrals so to speak) where the color differences are the least. The driver I'm using for the Mac is 9.33, I can't find anything newer.
Odd, very very odd!

[Doug Gray]

Wow!

Those are the results I'd expect with some sort of differing color management or conversion taking place. I have no idea what's going on but that sort of thing will produce very visible difference on most any image.

What app are you using to print and how does it assume ProPhoto? Are you printing the chart in Photoshop?

In any case I can't duplicate it here - Epson's a 9800 and 8 bits only - and I've never seen anything like that on the Canon. But I'm also on Win 10 x64 which is a different beast.

Since you have an iSis, I've attached a 16 bit image file tagged in Adobe RGB formatted to be a CC and readable by iSis with the default 957 letter size. Also attached the CGATs file you can load as a patch set which makes the iSis measurement readings happy. The RGB values are also high bit Adobe RGB.

Print it in Photoshop using Absolute Colorimetric and try switching 8/16.

[Wayne Fox]

I'm curious what printer you are using. Is there an advantage to printing 8-bit?

I didn’t get through the whole thread yet, but in regards to this specific question the only thought I would have is if you are trying to print a very large print in 16 bit, I’ve seen circumstances where it has choked the printer pipeline.  I can’t explain it because I always thought the data going to the printer was streamed so all I can think of is somehow the spool file or whatever is being created becomes unmanageable.  I always print from LR on a Mac to an epson 44” wide printer (currently p9000) using the 16bit option, but on prints that are longer than about 80” and 30” wide I have had occasions where the printer appears to be printing but lays down no ink, where it just sits there and never prints and the spool file just disappears, and sometimes it starts printing a repeated pattern of a small part of the image continuously.

At this point I use 8 bit for images over about 70” long.

For some file size would also be an issue.  I do not make intermediate files and print through photoshop, I print from the original master file and let LR handle the resizing and output sharpening. But those that make files to specific output sizes before printing, 8 bit files are smaller.

[elliot_n]

Here are my latest test results. These are photographs of prints (I tried using my flatbed scanner, but I discovered that there's lots of distracting gunk on the underside of the glass).

AFAIK, I am printing with the same settings as Andrew, but, unlike him, I am seeing (subtle) banding when 16 bit is not selected in the Epson driver.

First I will post 'straight' photographs. These are not very revealing, but they show what the prints actually look like.

Next I will post 'enhanced' photographs (Blacks -100, Whites +100, in ACR). These reveal the banding. This banding can be seen by the naked eye (I wouldn't have bothered photographing the prints if I had seen no banding). The banding is subtle.

Here are the straight photographs.

(1) Epson driver, 16 bit vs 8 bit, from Photoshop
(2) Epson driver, 16 bit vs 8 bit from ACPU
(3) Converting the 16 bit file to 8 bit in Photoshop, without Dither vs with Dither

[elliot_n]

And here are the same images, enhanced in ACR to reveal banding:

[faberryman]

See my last post.

I read your post saying you can't see a difference. I am asking you whether you check the 16-bit box when you normally print or whether you leave it unchecked. You do one or the other.

[elliot_n]

I am not a colour geek, just a regular photographer who has fallen down the rabbit-hole opened up by the OP

I'm very happy for anyone to challenge my methodology.

Earlier in this thread I was confused by what I regarded as TWO 16 bit checkboxes in the print pipeline. I think I now have a better understanding of the different functions of these checkboxes. In my most recent tests (the ones I've just posted), the 'Send 16 bit Data' in the Photoshop print dialog is always checked (except for the third print, which is an 8 bit image, and hence the checkbox is greyed out). So in the first two prints, I am toggling the '16 Bit Output' checkbox in 'Print Setting' in the Epson print driver.

If you send a 16 bit file to print, and don't check 'Send 16 bit Data' in the Photoshop print dialog, then Photoshop will make an on-the-fly conversion of the 16 bit file to 8 bit. If you don't have Dither turned on in Photoshop's Color Settings, then banding will appear in your file.

[elliot_n]

Based on my tests, '16 Bit Output' in the Epson print driver should always be checked. Likewise, 'Use Dither' in Photoshop's Colour Settings.

[digitaldog]

Doug, I’m printing from Photoshop CC 2019 just released this week. I’ll try you target later thanks. Something is obviously very wrong and I need to nail down if this yet another Epson/Adobe print bug.

[Mark D Segal]

Doug, I’m printing from Photoshop CC 2019 just released this week. I’ll try you target later thanks. Something is obviously very wrong and I need to nail down if this yet another Epson/Adobe print bug.

For an experienced professional such as yourself to be getting those huge dE errors is indeed very troubling and a prominent indication that something else is amiss besides this 16-bit business. I'm staying tuned to see what you come up with, as this finding is more important than much else on this thread. If you have the time and software available there to do so without inordinate inconvenience, it may be useful to perform the identical tests using the previous and the current versions of Photoshop CC. That should nail the question about whether the update is problematic. I can't do this, because I'm still on El Capitan so I can't use the latest PS version, nor am I planning to upgrade my OS just yet.

[Doug Gray]

For an experienced professional such as yourself to be getting those huge dE errors is indeed very troubling and a prominent indication that something else is amiss besides this 16-bit business. I'm staying tuned to see what you come up with, as this finding is more important than much else on this thread. If you have the time and software available there to do so without inordinate inconvenience, it may be useful to perform the identical tests using the previous and the current versions of Photoshop CC. That should nail the question about whether the update is problematic. I can't do this, because I'm still on El Capitan so I can't use the latest PS version, nor am I planning to upgrade my OS just yet.

Couldn't agree more. On the positive side, unlike some of the past bugs, this one should get addressed pretty fast. Nothing subtle about it. Unless it only shows up with some unusual combo of hardware/software/settings.

[digitaldog]

Couldn't agree more. On the positive side, unlike some of the past bugs, this one should get addressed pretty fast. Nothing subtle about it. Unless it only shows up with some unusual combo of hardware/software/settings.

I decided to print your target in ACPU to first eliminate Photoshop as a factor. No need to measure them, I can see very clearly there's a massive difference in color with the print driver 16-bit check box on or off. So it appears to be a bug in the 3880 Epson driver. However, I do need to measure them to provide data as to which is 'correct'. Going out on a limb, I'd say OFF is wrong. I suppose I need to see if an 8-bit per color doc does the same. But I think I need to ping my boys at Epson USA and see if this is something they are aware of and care.... Could be due to Mojave, that would be my next test but if someone here has a 3880 on Mac OS, but useful for a reality check for me if they can make a print of Doug's target or for that matter anything colorful with a 16-bit doc and just toggle the 16-bit check box in print driver on and off. If they match, it points to the new OS. If not, more work ahead.

[elliot_n]

Here's a photograph of two prints of Doug's ColorChecker. ACPU. 16 bit on the left, 8 bit on the right.

Epson 3880. Epson Driver 9.33. High Sierra.

I can't see any big differences.

[Doug Gray]

I decided to print your target in ACPU to first eliminate Photoshop as a factor. No need to measure them, I can see very clearly there's a massive difference in color with the print driver 16-bit check box on or off. So it appears to be a bug in the 3880 Epson driver. However, I do need to measure them to provide data as to which is 'correct'. Going out on a limb, I'd say OFF is wrong. I suppose I need to see if an 8-bit per color doc does the same. But I think I need to ping my boys at Epson USA and see if this is something they are aware of and care.... Could be due to Mojave, that would be my next test but if someone here has a 3880 on Mac OS, but useful for a reality check for me if they can make a print of Doug's target or for that matter anything colorful with a 16-bit doc and just toggle the 16-bit check box in print driver on and off. If they match, it points to the new OS. If not, more work ahead.

Makes sense. Man, that's a serious bug!

Did you first convert in Photoshop to device space Abs before printing with ACPU? That way one will look exactly like a colorchecker and the other one won't which makes it easier to communicate, visually if necessary, the bug to Epson.

[faberryman]

I decided to print your target in ACPU to first eliminate Photoshop as a factor. No need to measure them, I can see very clearly there's a massive difference in color with the print driver 16-bit check box on or off. So it appears to be a bug in the 3880 Epson driver. However, I do need to measure them to provide data as to which is 'correct'.

I am confused. I thought a couple of days ago you said there was no difference when checking the 16-bit box or not, and that the checkbox was purely a marketing ploy perpetrated by Epson. Are you saying with the new release, the checkbox is no longer a marketing ploy, but has been implemented and implemented wrong?

[digitaldog]

Here's a photograph of two prints of Doug's ColorChecker. ACPU. 16 bit on the left, 8 bit on the right.

Epson 3880. Epson Driver 9.33. High Sierra.

I can't see any big differences.

Thanks that’s really useful information. I’ve contacted Epson and have heard back, they suspect it’s a Majove bug and are checking on if there’s a newer version. Stay tuned.

[elliot_n]

Here are flatbed scans of those two prints. The colours are the same.

(I just took Doug's tiff and printed it in ACPU - no colorspace conversion, as per his last post.)

[digitaldog]

I am confused.

Understandable.
 

I thought a couple of days ago you said there was no difference when checking the 16-bit box or not, and that the checkbox was purely a marketing ploy perpetrated by Epson. Are you saying with the new release, the checkbox is no longer a marketing ploy, but has been implemented and implemented wrong?

Do you understand what a deltaE difference is before I explain further?