8 Bit or 16 bit , sRGB or Pro Photo

Started Sep 11, 2019 | Discussions
Ken60 Senior Member • Posts: 2,294
8 Bit or 16 bit , sRGB or Pro Photo
1

In recent forum exchange I have found much comment re this subject of bit depth and gamut.

I mentioned my practice of using 16 bit data and Canon XPS driver, thus allowing output direct to print and instant manipulation of files in their photoshop format .

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned. Good news , it seems in February of 2007 (12 years ago ) Andrew Rodney wrote a piece on this subject. Back when Photoshop was pre CC , and we were shooting the great Canon 5D, he alluded to very similar concepts ........ I guess things have evolved quite a bit since then. for some !

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

DotCom Editor Veteran Member • Posts: 7,610
ProPhoto for me, whever possible
6

Ken60 wrote:

In recent forum exchange I have found much comment re this subject of bit depth and gamut.

I mentioned my practice of using 16 bit data and Canon XPS driver, thus allowing output direct to print and instant manipulation of files in their photoshop format .

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned. Good news , it seems in February of 2007 (12 years ago ) Andrew Rodney wrote a piece on this subject. Back when Photoshop was pre CC , and we were shooting the great Canon 5D, he alluded to very similar concepts ........ I guess things have evolved quite a bit since then. for some !

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

Modern inks can make colors beyond the limits of sRGB. Thus, I nearly always work in ProPhoto RGB so I can start with as broad a range of data as possible. Why cripple a printer by using a narrower color space that doesn't allow it to reach its full capability?

If I need to squeeze the color space to sRGB in order to make a JPG for website posting or emailing to someone, so be it. (Think of a funnel analogy, wide input at the top and narrow output at the bottom.) But, I consider creating a JPG a final output step only, never an intermediary post-processing step.

My rule is simple: Never, ever throw data away. Disk storage is so inexpensive as to be almost free.

My opinion on this is not universal.

 DotCom Editor's gear list:DotCom Editor's gear list
Canon EOS 5D Mark II Canon EOS 5D Mark IV Epson Stylus Pro 3880 +3 more
OP Ken60 Senior Member • Posts: 2,294
Re: ProPhoto for me, whever possible

Interestingly DotCom Editor, Andrew mentions that he feels the Epson K3 inkset from that era could exceed aRGB 1988 ! 12 years later , and many claims by every printer manufacturer of improvement ............ Again I was interested to see the comment

" if you utilize a wider gamut working space, you might want to use more bits"

It seems the thought is that as the field becomes larger (GAMUT), the spaces between the players ( data/colours ) becomes larger , and more can be accommodated to better describe the image.

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

NAwlins Contrarian Veteran Member • Posts: 5,283
Working versus final for printing
1

In recent forum exchange I have found much comment re this subject of bit depth and gamut.

I mentioned my practice of using 16 bit data and Canon XPS driver, thus allowing output direct to print and instant manipulation of files in their photoshop format .

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned. Good news , it seems in February of 2007 (12 years ago ) Andrew Rodney wrote a piece on this subject. Back when Photoshop was pre CC , and we were shooting the great Canon 5D, he alluded to very similar concepts ........ I guess things have evolved quite a bit since then. for some !

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

IMO we really, really need to separate out:

(1) (a) color space (e.g., sRGB, Adobe RGB, ProPhoto RGB, etc.) from (b) bit depth (e.g., 8 bits per channel, 16 bits per channel, etc.); and

(2) (a) intermediate working files (e.g., raw, PSD, etc.--the original and intermediate files we process in the digital darkroom) from (b) final files used or sent for printing.

Although sRGB is the most common color working space, arguably the most appropriate for web display, and required by many printing services, anybody with a lick of sense knows that some monitors and many printers can print colors outside of sRGB, and therefore use of a wider working color space is better when circumstances allow you to do so.

Although I think the consensus of expert opinion is that 8 bit per channel files provide all of the depth that any of us can see in prints, I don't think anyone with a lick of sense thinks that it isn't better to use higher bit depth files for original capture and all intermediate processing, only truncating (or rounding) down to 8 bits for printing.

Just for starters, I suggest you go read:

https://forum.luminous-landscape.com/index.php?topic=123579.0

where the very same Andrew Rodney you mentioned weighs in (as "digitaldog").

 NAwlins Contrarian's gear list:NAwlins Contrarian's gear list
Nikon Coolpix S30 Canon PowerShot S120 Sony Alpha DSLR-A580 Tamron SP AF 90mm F/2.8 Di Macro Tamron SP 70-300mm F/4-5.6 Di USD +5 more
technoid Senior Member • Posts: 2,276
Re: 8 Bit or 16 bit , sRGB or Pro Photo

Ken60 wrote:

In recent forum exchange I have found much comment re this subject of bit depth and gamut.

I mentioned my practice of using 16 bit data and Canon XPS driver, thus allowing output direct to print and instant manipulation of files in their photoshop format .

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned. Good news , it seems in February of 2007 (12 years ago ) Andrew Rodney wrote a piece on this subject. Back when Photoshop was pre CC , and we were shooting the great Canon 5D, he alluded to very similar concepts ........ I guess things have evolved quite a bit since then. for some !

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

Canon's Pro1000 XPS driver does not print using 16 bits in Windows 10 on my system.  And Canon's Print Studio Pro doesn't print 16 bits either even though there is an option that supposedly does that.

It is extremely hard to test whether a printer is actually using a 8 bit drive or 16 bit driver since actual photo images have enough noise in them (it only takes a bit at the LSB) that you can't see a difference. You have to use noise free, synthetic images.

I've tested this by printing RGB 254, 254.25, 254.50, 254.75, and 255 RGB neutrals. Then scanned it (you can't see a difference! )  then increased the contrast 20x or so and guess what? it's just printing 8 bits.

But printing this from Photoshop and enabling 16bit to 8 bit dither in the color settings and then you see visible differences after contrast bumping.

OP Ken60 Senior Member • Posts: 2,294
Re: Working versus final for printing
1

NAwlins Contrarian wrote:

In recent forum exchange I have found much comment re this subject of bit depth and gamut.

I mentioned my practice of using 16 bit data and Canon XPS driver, thus allowing output direct to print and instant manipulation of files in their photoshop format .

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned. Good news , it seems in February of 2007 (12 years ago ) Andrew Rodney wrote a piece on this subject. Back when Photoshop was pre CC , and we were shooting the great Canon 5D, he alluded to very similar concepts ........ I guess things have evolved quite a bit since then. for some !

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

IMO we really, really need to separate out:

The wheat from the chaff

(1) (a) color space (e.g., sRGB, Adobe RGB, ProPhoto RGB, etc.) from (b) bit depth (e.g., 8 bits per channel, 16 bits per channel, etc.); and

You don't need to , read the PDF, its all there for you .

(2) (a) intermediate working files (e.g., raw, PSD, etc.--the original and intermediate files we process in the digital darkroom) from (b) final files used or sent for printing.

working files are best kept in high bit depth to avoid inducing 8 bit banding , it also talks about that in the PDF

Although sRGB is the most common color working space, arguably the most appropriate for web display, and required by many printing services, anybody with a lick of sense knows that some monitors and many printers can print colors outside of sRGB, and therefore use of a wider working color space is better when circumstances allow you to do so.

sRGB is not the most common working space, it might have been many years ago , but the science and technology has moved on. Monitors are now more able than the big glass back  cathode ray things. My graphics card is now able to display  way more than 24 bit, and modern monitors are able to go way beyond Adobe RGB 1988. The mobile internet is now going 5G and bandwidths are designed to eat video !

Although I think the consensus of expert opinion is that 8 bit per channel files provide all of the depth that any of us can see in prints, I don't think anyone with a lick of sense thinks that it isn't better to use higher bit depth files for original capture and all intermediate processing, only truncating (or rounding) down to 8 bits for printing.

So if all it takes is a lick of sense , then why all the above ?

Just for starters, I suggest you go read:

No I am off to bed , I am done with historic knowledge and folk grabbing onto what was.

As for LL, is it still open after MR's passing.

https://forum.luminous-landscape.com/index.php?topic=123579.0

where the very same Andrew Rodney you mentioned weighs in (as "digitaldog").

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

OP Ken60 Senior Member • Posts: 2,294
Re: 8 Bit or 16 bit , sRGB or Pro Photo
1

There is note that Photoshop does not do 16 bit , they say it is 15 bit +1

What is important is that it goes beyond 8 bit . Our cameras produce 12 - 14 bit , so why degrade this data.   I am sure the Canon XPS driver is rendering greater than 8 bit , possibly more like the 12 that their cameras produce.

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

bmoag Senior Member • Posts: 2,650
Re: 8 Bit or 16 bit , sRGB or Pro Photo

I am all in on capturing raw data in the biggest color space your camera allows and processing it in a big color box. There are learned screeds about how PS actually handles or doesn't all 16 bits. I have no idea what other programs do that claim to process in a big color box because I use PS.

None of that matters although I do not understand why Adobe chops the numbers the way they do.

It should be obvious that you could easily test things for yourself to see if there is a reward or a penalty for printing in 16 bits, something one only sees in the OSX drivers. As well as comparing prints made directly from the prophoto file to a properly converted sRGB file--the prints can be on the same piece of paper.

Presuming you have an sRGB monitor you will never see any more colors than the sRGB gamut while crunching data in Prophoto/16 bits. So if you indeed have a printer whose ink gamut can exceed sRGB you will never match the color you see on your monitor to the print if the print has a wider gamut than the monitor and the printer is getting more than an 8 bit data stream--the print can have colors you never saw on the monitor. I am not sure why that would be desirable.

It would seem that using a full RGB monitor, as well as having excellent color vision which most people do not, should be de rigeur for WYSIWYG printing if your printer can indeed print colors outside the sRGB gamut.

My point is that there is a valid reason for capturing and processing in a big color space but properly converting to sRGB for printing so that the print can best match what you spent your time and energy on based on what you thought you saw on your finely calibrated monitor. The whole reason for capturing and processing in that big color box is to minimize color banding when the image is used in a smaller color space, and it works! I am not the first to point this out.

 bmoag's gear list:bmoag's gear list
Nikon D70 Nikon D3300 Panasonic Lumix DMC-GX8 Nikon Z6
technoid Senior Member • Posts: 2,276
Re: 8 Bit or 16 bit , sRGB or Pro Photo
1

Ken60 wrote:

There is note that Photoshop does not do 16 bit , they say it is 15 bit +1

That has zero to do with whether Canon's Pro 1000 can print anything higher than 8 bits.

I am sure the Canon XPS driver is rendering greater than 8 bit , possibly more like the 12 that their cameras produce.

Really? I'm curious why you are "sure?"  I thought they did too since they claimed to do so. Then I actually tested it and discovered it only printed to the driver in 8 bits. I found that rather annoying.

I described how I tested it? How did you test it?

BTW, RGB numbers like 254.25 (referred to 8 bit 0:255) require only 10 bits of accuracy so the 15+1 limit of Photoshop does not apply.

technoid Senior Member • Posts: 2,276
Canon's Print studio pro and the xps driver fails 16 bits

Well, here's pretty strong evidence that the Canon Pro1000 doesn't print beyond 8 bits.

There's two attachments of 16 bit scans of prints with a high key gradient but the scans are saved as jpegs.

The 16 bit tiff image has a gradient from RGB (245,245,245) to (255,255,255) .

One was printed using Canon's Studio Pro ostensibly at 16 bits. The other printed directly with the driver but with the ColorSettings 16->8 bit dither selected.

One attached image is just the scan converted as is to jpeg. The other is the scan with high key contrast enhanced to bring out banding. This was done by using "Curves"
 and setting the input (which is scaled to 0) to 220. This also magnifies the tint since the paper white is slightly bluish.

It's pretty obvious that the Canon Studio Pro does not even print 9 bits, let alone 16 and the dither that Photoshop injects when printing through the driver works quite well.

NAwlins Contrarian Veteran Member • Posts: 5,283
Then I'll be more direct with you.
1

I was trying to suggest gently that you appear to be confused about some fundamentals, and ought to review some things. That having evidently gotten nowhere, I'll be more direct. Above you wrote:

It was my feeling that modern inks should be able to exceed the 8 bit data from sRGB files that have been the traditional print feed.

This reflects a clear mixing up of (1) gamut, i.e., what range of colors you (a) can describe in a file, which is a function of the working space, like sRGB or Adobe RGB, and/or (b) print onto a page, which is a function of those inks, the paper, and the printer / its driver, with (2) how finely you divide up that gamut with mathematical representations, which is a function of bit depth (and, to the extent it differs, gamma, see, e.g., https://en.wikipedia.org/wiki/Gamma_correction).

Whether you use an 8-bit JPEG or a 16-bit TIFF to send your photo to the printer does not change the printable gamut. Whether your file is encoded in sRGB or ProPhoto RGB does change the printable gamut, at least where the photo has colors outside of sRGB but within the printer's gamut. (Of course, your printing software and/or printer driver, if color-managed, has to convert your ProPhoto RGB or whatever to what the printer + paper combination can actually print--another wrinkle / complication.)

Also, there is no real connection between "8 bit data [and] sRGB files". You can have a 16-bit TIFF encoded in sRGB or a JPEG (8 bits) encoded in ProPhoto RGB. And the latter is almost always fine for printing, you just don't want to use it for editing.

Then you wrote:

This comment and claim seemed to be rejected by some as heresy, and my excommunication was warned.

Being a bit dramatic?

 NAwlins Contrarian's gear list:NAwlins Contrarian's gear list
Nikon Coolpix S30 Canon PowerShot S120 Sony Alpha DSLR-A580 Tamron SP AF 90mm F/2.8 Di Macro Tamron SP 70-300mm F/4-5.6 Di USD +5 more
technoid Senior Member • Posts: 2,276
Re: Then I'll be more direct with you.

NAwlins Contrarian wrote:

Also, there is no real connection between "8 bit data [and] sRGB files". You can have a 16-bit TIFF encoded in sRGB or a JPEG (8 bits) encoded in ProPhoto RGB. And the latter is almost always fine for printing, you just don't want to use it for editing.

Yep.  Almost always. I recall only one jpeg image saved in high quality in ProPhoto RGB showing evidence of banding that was a photograph. It was in a darkish brown area IIRC and one had to look pretty closely to see it. The image had an exceptionally low level of noise which tends to naturally mask banding.  OTOH, I've never seen an actual photo in 8 bit tiff RGB ProPhoto that showed any banding though it's easy to make synthetic 8 bit tiff images that show slight banding. That natural shot noise in photos covers many sins.

And, of course, editing is a different story and should be done in high bits.

Your full statement is, I think, dead on in all respects.

OP Ken60 Senior Member • Posts: 2,294
Re: Then I'll be more direct with you.

Nawlins it seems you did not read the attached PDF, if you had you would see the Gamma described as the potential area, and the bit depth as the definition level. I again say this in a post here ......... all clearly to those that read.

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

gscotten
gscotten Senior Member • Posts: 1,972
Re: 8 Bit or 16 bit , sRGB or Pro Photo
1

bmoag wrote:

It should be obvious that you could easily test things for yourself to see if there is a reward or a penalty for printing in 16 bits, something one only sees in the OSX drivers.

Windows users can also use Canon's included 16 bit XPS driver, so this is not just of interest to OSX users. The XPS driver is not installed by default, but included.

I have tested both 8 and 16 bit printing, although perhaps not as rigorously as others in this thread.  Since I can detect no difference either in speed or image quality, I see no harm in using the 16 bit driver as my default driver just in case there is some benefit under some conditions.

I am curious as to why Canon would include both an 8 bit driver and a 16 bit driver if the printers don't actually use 16 bits. If it is just for marketing, why not just call the standard driver 16 bit?

-- hide signature --

George

 gscotten's gear list:gscotten's gear list
Canon EOS RP Canon RF 24-240mm F4-6.3
OP Ken60 Senior Member • Posts: 2,294
Re: Then I'll be more direct with you.

https://www.youtube.com/watch?v=vLlr7wpAZKs

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

NAwlins Contrarian Veteran Member • Posts: 5,283
Re: Then I'll be more direct with you.
1

it seems you did not read the attached PDF, if you had you would see the Gamma described as the potential area

No the gamut is the potential area (of colors); the gamma is a mathematical encoding that takes typically linear raw sensor data and redistributes it so the brightest part doesn't get so much more bandwidth than it needs and the darkest part gets enough bandwidth so that it still has reasonable tonal smoothness. Read the Wikipedia article on gamma I linked above.

and the bit depth as the definition level.

I don't know what "the definition level" level is. That's not a term I've seen used in this context. Basically, you just need to use enough bit depth to avoid banding. For printing (as for the old TV show) 8 is enough (well, 8 per each RGB channel). The reason you don't use 8 for editing is that if you adjust the light-dark or color relationships much, then you'd risk creating banding.

I again say this in a post here ......... all clearly to those that read.

Andrew Rodney's digitaldog website has a lot of good stuff, and I've read a lot of it. I think you appear to be misunderstanding what you've read.

 NAwlins Contrarian's gear list:NAwlins Contrarian's gear list
Nikon Coolpix S30 Canon PowerShot S120 Sony Alpha DSLR-A580 Tamron SP AF 90mm F/2.8 Di Macro Tamron SP 70-300mm F/4-5.6 Di USD +5 more
OP Ken60 Senior Member • Posts: 2,294
Re: Then I'll be more direct with you.

Nawlins you seem too interested in arguing semantics, a common fault on DPR ... did you download the test file and follow the print process as suggested after watching the video.  ?   Far more productive than endless posting about which term is used !

Try it with your Canon printer and 16 bit driver.

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

OP Ken60 Senior Member • Posts: 2,294
16 bit printing in Profoto

For anyone that watched the video linked, and that tried printing the test file linked...

Here are two images shot of the printout  with my Samsung phone.... wish you could see the differences first hand !

One is printed with the Canon XPS 16 bit  printer driver, from a Pro Photo test chart on Canson Platine using perceptual, black point comp, and a custom profile that I had made for that box of paper and my printer.

The second is from an 8 bit sRGB of the same test chart using Canon Pro 1000 driver ( 8 bit)  and with the same settings.

So which one of the two is this ?

-- hide signature --

Gear ... what I need to get the job done , after all you don't see mechanics listing their brand of spanner as a qualification .

NAwlins Contrarian Veteran Member • Posts: 5,283
The difference is ...
2

Here are two images shot of the printout with my Samsung phone.... wish you could see the differences first hand !

One is printed with the Canon XPS 16 bit printer driver, from a Pro Photo test chart on Canson Platine using perceptual, black point comp, and a custom profile that I had made for that box of paper and my printer.

The second is from an 8 bit sRGB of the same test chart using Canon Pro 1000 driver ( 8 bit) and with the same settings.

The visible difference there is caused by ProPhoto RGB versus sRGB. Even the with Bill's Balls test image that you used,* I'd bet that $100 that to our eyes the differences between 16-bit and 8-bit files will not show, and the differences between a TIFF and a best-quality JPEG will not show. So I'm back to where I started (at https://www.dpreview.com/forums/post/63089378),

[W]e really, really need to separate out ....

Probably also worth noting, Bill's Balls is not a photo, but a wholly synthetic 'torture test' image.

Last but not least: in a color-managed workflow (necessary to reproduce color accurately), you don't print in either ProPhoto RGB or sRGB; instead you (or your software automatically) convert from the working space (ProPhoto RGB or sRGB or whatever) to the ICC profile of the printer + paper. How that conversion is done can have a big effect on the resulting colors, especially with an image like Bill's Balls, which has a lot of color at or beyond the extremes of normal gamuts. With something like that, whether the conversion is perceptual or relative colorimetric or whatever, and whether there is black-point compensation, can make a huge difference in the printed result. You did not tell us about some of those details of your test, and those details matter a lot.

*To be precise, the page you printed is actually Andrew Rodney's Gamut Test File, the bottom of which incorporates the smaller, 14-ball version of Bill Atkinson's balls, the full version of which has 28 balls. Your photos show the left 5 of the 7 balls on each row.

 NAwlins Contrarian's gear list:NAwlins Contrarian's gear list
Nikon Coolpix S30 Canon PowerShot S120 Sony Alpha DSLR-A580 Tamron SP AF 90mm F/2.8 Di Macro Tamron SP 70-300mm F/4-5.6 Di USD +5 more
ToshDog Forum Member • Posts: 57
Re: 16 bit printing in Profoto
3

Ken60 wrote:

For anyone that watched the video linked, and that tried printing the test file linked...

Here are two images shot of the printout with my Samsung phone.... wish you could see the differences first hand !

One is printed with the Canon XPS 16 bit printer driver, from a Pro Photo test chart on Canson Platine using perceptual, black point comp, and a custom profile that I had made for that box of paper and my printer.

The second is from an 8 bit sRGB of the same test chart using Canon Pro 1000 driver ( 8 bit) and with the same settings.

So which one of the two is this ?

The differences you see in the output are solely due to the color gamut, not the bit depth. What you see is to be expected due to the color gamut used prior to conversion to the printer color space.

Color Gamut defines the range of 'colors'.

Bit depth defines the numbers of color values. And some color values are not actual colors, some differing color values are the same color. But that's another story for now.

ProPhoto RGB has a much wider range of colors it can contain than sRGB. No matter the bit depth.

The bit depth simply defines the numbers and how they are divided.

Think of Color gamut as a pie that's 12 inches wide and weights 4 pounds.

Think of the bit depth as how many slices you cut into that pie for serving to your guests.

You can cut the pie such that each slices is one pound (1/4 of that pie). That could be 8-bits per color. Now suppose you decide to cut the pie so you end up with 8 slices of pie. That's "16-bits" in our analogy. The pie (Color Gamut) is still 12 inches wide and weights 4 pounds.

Think of the 12 inch 4 pound pie as sRGB, think of a 14 inch pie that weighs 6 pounds as Adobe RGB (1998) if you will. You haven't sliced up the pie, the bit depth isn't a factor yet.

The two are independent of each other.

Keyboard shortcuts:
FForum MMy threads