8-bit vs. 16-bit/channel and printers

[ggull]

This is a bit off-topic, but the relative merits of 8 vs 16 bit/channel
images used to be a subject of heated debate around here, with knowledgeable
folks on both sides.

The other day I ran across a discussion of editing (as with Photoshop) in
16-bit mode, which seemed to go with the general flow of discussions here --
it may make some difference in some cases, especially with smooth gradations
and adjusting levels.

But the following, put in as an aside, seemed far different from what I
recall --
"When you print a 16-bit image, the results are stunning."

Comments? Any idea what the author is talking about?
I thought that the available colors for printing (at least for common
consumer printers) were a tiny fraction of even a 3x8 bit image's range.
Hence, the printer can't "see" the difference between an 8 and 16 bit
version of the same image.
Could using a 16 bit image ever make a "stunning" difference? Perhaps with
poorly calibrated color management? or on very high end printers?

 

[Bart van der Wolf]

SNIP

But the following, put in as an aside, seemed far different from
what I recall --
"When you print a 16-bit image, the results are stunning."

Comments? Any idea what the author is talking about?

Without the context of that remark, it seems to be a denial that
current printers only print 8-bits/channel colors. So sending an
8-b/ch file would be all that's needed.

However, the context may have specified a more elaborate color managed
driver or RIP. That driver may perform resampling and profile
conversion, and could potentially benefit from 16-b/ch input,
especially when enlarging. It stands a better chance of avoiding
posterization under a wider range of situations.

That being said, in practice you'll be able to avoid most of the
potential pitfalls in pre-processing, and 8-b/ch will be as good as it
gets for printing.

Bart

 

[Ernst Dinkla]

SNIP

Without the context of that remark, it seems to be a denial that current
printers only print 8-bits/channel colors. So sending an 8-b/ch file
would be all that's needed.

However, the context may have specified a more elaborate color managed
driver or RIP. That driver may perform resampling and profile
conversion, and could potentially benefit from 16-b/ch input, especially
when enlarging. It stands a better chance of avoiding posterization
under a wider range of situations.

That being said, in practice you'll be able to avoid most of the
potential pitfalls in pre-processing, and 8-b/ch will be as good as it
gets for printing.

Bart

Right now only ImagePrint as a RIP claims 16 bit throughout
(till it splits at 1 or 2 bit on the color channels).
GutenPrint is 16 bit now. Canon latest models iPF5000 and 9000
have 16 bit input and 10 or 12 bit output (PS plug-in). That's it.

Ernst

--

--
Ernst Dinkla


www.pigment-print.com
( unvollendet )

 

[Bart van der Wolf]

SNIP

Right now only ImagePrint as a RIP claims 16 bit throughout (till it
splits at 1 or 2 bit on the color channels). GutenPrint is 16 bit
now. Canon latest models iPF5000 and 9000 have 16 bit input and 10
or 12 bit output (PS plug-in). That's it.

Thank you for that update.

Bart

 

[ggull]

Without the context of that remark, it seems to be a denial that current
printers only print 8-bits/channel colors. So sending an 8-b/ch file would
be all that's needed.

However, the context may have specified a more elaborate color managed
driver or RIP. That driver may perform resampling and profile conversion,
and could potentially benefit from 16-b/ch input, especially when
enlarging. It stands a better chance of avoiding posterization under a
wider range of situations.

That being said, in practice you'll be able to avoid most of the potential
pitfalls in pre-processing, and 8-b/ch will be as good as it gets for
printing.

Thanks for this info, Bart (and the update from Ernst).
As I said, it was an aside.
This was in a half-page that mentioned the (supposed) advantages of
converting to a 16 b/ch image for "extensive editing... expecially when
working with curves and levels" (it doesn't mention scanning or otherwise
starting with a 16 b/ch image). The printing comment came in the workflow:
1. open the image
2. convert to 16 b/ch
3 Edit "using the techniques discussed previously"
4. "If desired, print the image. When you print a 16-bit image, the results
are stunning."
5. save

No mention of special printer drivers. No hedging "could potentially
benefit ... under a wider range of situations" Just a bald "results are
stunning", as if the 16-bitness somehow adds magic sparkle. Seemed
overreaching to me, and from your comments, still does.

I do find it interesting, however, that some printers/drivers do take
advantage of 16 b/ch input. Useful info for the future, but certainly not
what the book was talking about.

[This was "Photoshop CS QuickSteps" by Carole Matthews, et al., by the way.]

 

[Roger S.]

I do find it interesting, however, that some printers/drivers do take

advantage of 16 b/ch input. Useful info for the future, but certainly not
what the book was talking about.

[This was "Photoshop CS QuickSteps" by Carole Matthews, et al., by the way.]

I think you've got it right.

The author isn't wrong, as erring on the side of higher bit depth is a
good thing and the results may well be stunning. Of course so may the
results from 8 bit.

 

[Ernst Dinkla]

I do find it interesting, however, that some printers/drivers do take
advantage of 16 b/ch input. Useful info for the future, but certainly not
what the book was talking about.

[This was "Photoshop CS QuickSteps" by Carole Matthews, et al., by the way.]

I think you've got it right.

The author isn't wrong, as erring on the side of higher bit depth is a
good thing and the results may well be stunning. Of course so may the
results from 8 bit.

8 bit input RIPs and drivers usually have 16 bit look up
tables internally to get better rounding off but in the end
the printer ink channels they have to drive are way lower in
total. It is only with more ink channels on newer printers +
the variable droplet size per channel + high dpi resolutions,
that the higher output could make 16 bit a sensible input
format. The Canon's with CcMmRGBYKkk (+GK) are getting closer
to that with 11 working ink channels (output drops to 10-12
bit again though). Prophoto color space would be more suited
than AdobeRGB too but that also depends on what is feeding
that workflow. It is a bit early to say what the difference
will be in practice.

http://www.redrivercatalog.com/infocenter/articles/canonprografipf5000.html
http://www.inkjetart.com/canon/wide/iPF5000.html

Ernst

--

--
Ernst Dinkla


www.pigment-print.com
( unvollendet )