Digital BW, The Print

With the recent talk about number of bits/grays that should be sent to the printer drivers,
I figured I ought to try and find out and show some evidence of what's needed and can be 
seen on 
a print. Just not being able to show a difference between 8 bit and 16 bit is unsatisfying.  
So I
thought trying fewer bits to find where the difference really is might be more informative.
The prints are all made on an R2400 with Epson K3 inks and QTR software, EEM 50/50 
warm/cool profiles.
The results are pretty surprising.

Rather than using a "real" image that is unknown I've used a newly created full 16-bit, 
360ppi
21step wedge.  It's just a 16 bit grayscale gradient plus the 21 step posterized lower 
section.
A very key aspect of the exercise is the algorithm that Photoshop uses in conversions from
16-bit to 8-bit files.  The "noise" added by PS has been discussed here before but this is
not some random noise added to smooth over the transitions.  It's intelligent dithering
that can represent intermediate values very well.  This same method is used in the Color
Management Engines if you select "Dither 8-bit images" in Color Settings.  It's also what 
QTR
uses to convert 16-bit to 8-bit.

It's pretty easy to take advantage of PS's conversion for any number of bits/levels.  It can
be used for any image, here's how:
1) Set Image > Mode to 16-bit -- it's OK if the original is 8-bit
2) Run Image > Adjustment Levels, set the Output Levels to how many levels you want.  
E.g. for
   64 levels I used 101 to 164.  It's best to avoid the end points 0 and 255.
   The image will of course be very gray since you compress it all into a small range.
3) Set Image > Mode to 8-bit
4) Run Image > Adjustment Levels, to undo the compression you did above.  Now you set 
the Input Levels
   to exactly the same as you had above for the Output Levels.  This expands the range 
back to 
   the full amount.

That's it.  Check out the Histogram to see the number of levels.

I did this on the 16-bit 21 step and printed it out for 8, 7, 6, 5, 4, 3 bits to see the
extremes.    I scanned the results,  here's a 1000dpi scan of the 15 and 20 step values.
The scan is more detailed than you will see with the naked eye and even with a loupe.

http://harrington.com/bitpost.jpg

Here's what I see in reverse order.
3 bits, 8 levels of gray jumps out at you -- it's very noticeably grainy and blotchy.
4 bits, 16 levels isn't bad but once you pay attention you can see the difference.
5 bits, 32 levels is very difficult to see any difference, it takes a loupe and some
  imagination.  I doubt anyone could do a blind A/B test and do better than guessing.
6, 7, and 8 bits are all identical as far as I can tell.

My conclusion is that 8-bits going to the print driver has plenty of data and sufficient
headroom to not be an issue.  I think the most significant issue these days is the
number and densities of the inks used.  It's important how all ink densities and hues
work together to make all the gray values.  Naturally the software must give sufficient 
control to use the inks in the most intelligent way.  Probably the K7 inks are overkill
but give the most headroom for getting smooth prints.  What you give up is control of
hue on a print-by-print basis.  Most of the other inksets allow the possibility of hue
control or even full color but give up some of the headroom for smoothness.  It's a
tradeoff.

I should emphasize that none of this is meant to support using 8-bit files for editing.
For grayscale work I advocate 16-bit throughout the editing workflow.  The 8-bit dither
of Photoshop is a one-shot feature, only right before printing.  None of the editing
commands use this.

Roy

Roy,

I believe I understand most of what you are describing here. It seems
to confirm what you've said all along.

I have two quick questions though. First, do you see the potential for
sofware rip advances that WILL allow us to gain output advantage in
sending bit depth larger than 8 bit to the printer in the near future?
And, second, are any of the new large format 12 channel pigment
machines that you've investigated addressing anything larger than 8
bit capability already?

I guess what I am getting at is IF a bit depth beyond 8 bit WAS
possible in the near future, would we be seeing this capability coming
from new software, new hardware, or a combination of the two designed
in conjunction with each other.

John

John,

The current hardware (Epson x800) is already capable of printing a least 10
bits of gray (1024 levels) with IP (based on my own tests with a spectro).

Correct me if I am misquoting you Roy, but I think the point you are making
is that human vision probably struggles to perceive any difference in images
with more than 6 bits once you factor in PS dithering.

Certainly without the dithering I think I can percieve more than 8 bits
(around 9 bits) in a step wedge.
I'll do some more testing with dithered vs non-dithered images to see if
there are real world situations that the dithering doesn't handle well.

Mike




On 19/11/06, john dean <deanwork2003@...> wrote:
>
>   Roy,
>
> I believe I understand most of what you are describing here. It seems
> to confirm what you've said all along.
>
> I have two quick questions though. First, do you see the potential for
> sofware rip advances that WILL allow us to gain output advantage in
> sending bit depth larger than 8 bit to the printer in the near future?
> And, second, are any of the new large format 12 channel pigment
> machines that you've investigated addressing anything larger than 8
> bit capability already?
>
> I guess what I am getting at is IF a bit depth beyond 8 bit WAS
> possible in the near future, would we be seeing this capability coming
> from new software, new hardware, or a combination of the two designed
> in conjunction with each other.
>
> John
>
> 
>


[Non-text portions of this message have been removed]

Mike,

I'm really saying more than that.

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Michael King" 
<drmrking@...> wrote:
>
> John,
> 
> The current hardware (Epson x800) is already capable of printing a least 10
> bits of gray (1024 levels) with IP (based on my own tests with a spectro).

The point is that you can get those 1024 levels with a 6 bit file.   The issue
arises because people say 6 bits = 64 levels so you couldn't represent the 1024
levels.  This isn't the way it works.   ALL possible measurements that show
1024 levels are averaging over an area with many pixels, so measuring more
than 64 levels is easily done.

It would be great if you could take that exact setup with IP.   Take the file --
I presume it's a 16 bit stepwedge -- do the Photoshop operation I specified
making it only have 64 grays.   I maintain the file with look the same and print
the same and when you measure it you'll still get 1024 levels.

> 
> Correct me if I am misquoting you Roy, but I think the point you are making
> is that human vision probably struggles to perceive any difference in images
> with more than 6 bits once you factor in PS dithering.

Human vision will probably limit how far manufacturers will bother to advance.
But my claim about # of grays is for your high end spectro, too.

> 
> Certainly without the dithering I think I can percieve more than 8 bits
> (around 9 bits) in a step wedge.

I'm not sure what you would have in mind -- dithering is the ONLY way an inkjet 
print works.   There is no other way.  Dithering of the ink drops is needed to
give a lot of grays in the print.  The very slight dithering in PS is so minor 
compared to what happens in the driver/inkdrops/paper that the PS stuff is 
irrelevant.

Roy

> I'll do some more testing with dithered vs non-dithered images to see if
> there are real world situations that the dithering doesn't handle well.
> 
> Mike
> 
> 
> 
> 
> On 19/11/06, john dean <deanwork2003@...> wrote:
> >
> >   Roy,
> >
> > I believe I understand most of what you are describing here. It seems
> > to confirm what you've said all along.
> >
> > I have two quick questions though. First, do you see the potential for
> > sofware rip advances that WILL allow us to gain output advantage in
> > sending bit depth larger than 8 bit to the printer in the near future?
> > And, second, are any of the new large format 12 channel pigment
> > machines that you've investigated addressing anything larger than 8
> > bit capability already?
> >
> > I guess what I am getting at is IF a bit depth beyond 8 bit WAS
> > possible in the near future, would we be seeing this capability coming
> > from new software, new hardware, or a combination of the two designed
> > in conjunction with each other.
> >
> > John
> >
> > 
> >
> 
> 
> [Non-text portions of this message have been removed]
>

Roy,

If I do as you suggest, emperically I just can't see how its going to come
out at more than 6 bits of step resolution.
This seems to be no different to using a driver like ABW that only supports
8 bits of resolution and hence with ABW you only get 256 levels from the
1024 level wedge (basically every 4 steps come out at approx same value).

Note I understand that in a continuous tone gradient of 64 levels the
dithering of the printer driver will create lots of levels and I guess I
could easily find 1024 different levels. But I have a step wedge where each
step is discrete and big enough to be measured independently by the
spectro.  I am not looking for a random set of 1024 levels, I am looking for
a specific stepping of levels of approx 0.1 L* between paper white and dmax

I think we are talking about two very different measurement scenarios, or I
completely misunderstanding what is going on.

Mike




On 19/11/06, Roy Harrington <roy@...> wrote:
>
>
> Mike,
>
> I'm really saying more than that.
>
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com<DigitalBlackandWhiteThePrint%40yahoogroups.com>,
> "Michael King"
> <drmrking@...> wrote:
> >
> > John,
> >
> > The current hardware (Epson x800) is already capable of printing a least
> 10
> > bits of gray (1024 levels) with IP (based on my own tests with a
> spectro).
>
> The point is that you can get those 1024 levels with a 6 bit file. The
> issue
> arises because people say 6 bits = 64 levels so you couldn't represent the
> 1024
> levels. This isn't the way it works. ALL possible measurements that show
> 1024 levels are averaging over an area with many pixels, so measuring more
> than 64 levels is easily done.
>
> It would be great if you could take that exact setup with IP. Take the
> file --
> I presume it's a 16 bit stepwedge -- do the Photoshop operation I
> specified
> making it only have 64 grays. I maintain the file with look the same and
> print
> the same and when you measure it you'll still get 1024 levels.
>
> >
> > Correct me if I am misquoting you Roy, but I think the point you are
> making
> > is that human vision probably struggles to perceive any difference in
> images
> > with more than 6 bits once you factor in PS dithering.
>
> Human vision will probably limit how far manufacturers will bother to
> advance.
> But my claim about # of grays is for your high end spectro, too.
>
> >
> > Certainly without the dithering I think I can percieve more than 8 bits
> > (around 9 bits) in a step wedge.
>
> I'm not sure what you would have in mind -- dithering is the ONLY way an
> inkjet
> print works. There is no other way. Dithering of the ink drops is needed
> to
> give a lot of grays in the print. The very slight dithering in PS is so
> minor
> compared to what happens in the driver/inkdrops/paper that the PS stuff is
>
> irrelevant.
>
> Roy
>
> > I'll do some more testing with dithered vs non-dithered images to see if
> > there are real world situations that the dithering doesn't handle well.
> >
> > Mike
> >
> >
> >
> >
> > On 19/11/06, john dean <deanwork2003@...> wrote:
> > >
> > > Roy,
> > >
> > > I believe I understand most of what you are describing here. It seems
> > > to confirm what you've said all along.
> > >
> > > I have two quick questions though. First, do you see the potential for
> > > sofware rip advances that WILL allow us to gain output advantage in
> > > sending bit depth larger than 8 bit to the printer in the near future?
> > > And, second, are any of the new large format 12 channel pigment
> > > machines that you've investigated addressing anything larger than 8
> > > bit capability already?
> > >
> > > I guess what I am getting at is IF a bit depth beyond 8 bit WAS
> > > possible in the near future, would we be seeing this capability coming
> > > from new software, new hardware, or a combination of the two designed
> > > in conjunction with each other.
> > >
> > > John
> > >
> > >
> > >
> >
> >
> > [Non-text portions of this message have been removed]
> >
>
> 
>


[Non-text portions of this message have been removed]

Dear Roy,

A few weeks ago, a debate raged over 8bit versus
16bit imaging within the Yahoo ScanHi-End Group,
where many of the participants work with good quality
drum scanning devices daily, as part of their livelihood.

The scanning debate brought to the surface several good
points, bad points and numerous opinions, regarding
which level of information was good, better, best...

The thread is entitled ³Combed histograms from drum
scanning service.²  The thread proves to be a good read,
and the thread offers additional information that might
prove to be useful, for all of our archived files.

One day in the not so distant future, an output device
might be able to digest the abundance of data available,
and produce an image we only dream about.  Imaging
will more than likely take steps into the realm of more
and more bits, as processors become faster and more
powerful, and continually operate within 64bit or 128bit
channels.

Will we be able to discern the difference in the output
this possible higher plateau offers? Maybe not, but then
again, I will always remember the day I saw a 16X20
image produced from an 8X10 negative sitting beside
a 16X20 image produced with a 35mm negative.

Just a thought about where we might be one day...

jim k







[Non-text portions of this message have been removed]

Hi Mike,

I know it sounds non-intuitive so I'll try to explain what is going on.

Take your 1024 stepwedge.   Somewhere there are 2 adjacent steps that have the
values 400/1024 and 401/1024.   When you print with IP you will get gray patches
that differ by some amount.   Take these two patches and to the little Levels/8bit/Levels
action I outlined to get 64 levels.  The 400/1024 pixels will all go to 100/256 or 25/64.
The 401/1024 pixels will go to 94% 100/256 and 6% 104/256  (or 25/64 and 26/64).
A 402/1024 would go to 88% 100/256 and 12% 104/256.   Note that it only uses
64 unique pixel values but the average can be pretty much any intermediate value
you want.  My contention and what I think I've demonstrated is it's the average value
over the whole patch that determines the "gray" value i.e. what you see and what you
measure with the spectro.  The fact that the pixels being averaged are not identical
values has absolutely NO affect visually or measured.

Try it out.  Take your 1024 discrete steps and do Levels action.  Note this is very
different than Posterize to 64 levels -- not at all the same.  You will still have 1024
discrete steps.   You will only tell the difference in that the file is 8 bit and the histogram
has lots of holes in it.   Look at it on the screen, print it on the paper, see if there's
anyway you can tell the difference.  I haven't tried it with IP but I'm pretty confident
that you will get the same result.

Roy

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Michael King" <drmrking@...> 
wrote:
>
> Roy,
> 
> If I do as you suggest, emperically I just can't see how its going to come
> out at more than 6 bits of step resolution.
> This seems to be no different to using a driver like ABW that only supports
> 8 bits of resolution and hence with ABW you only get 256 levels from the
> 1024 level wedge (basically every 4 steps come out at approx same value).
> 
> Note I understand that in a continuous tone gradient of 64 levels the
> dithering of the printer driver will create lots of levels and I guess I
> could easily find 1024 different levels. But I have a step wedge where each
> step is discrete and big enough to be measured independently by the
> spectro.  I am not looking for a random set of 1024 levels, I am looking for
> a specific stepping of levels of approx 0.1 L* between paper white and dmax
> 
> I think we are talking about two very different measurement scenarios, or I
> completely misunderstanding what is going on.
> 
> Mike
> 
> 
> 
> 
> On 19/11/06, Roy Harrington <roy@...> wrote:
> >
> >
> > Mike,
> >
> > I'm really saying more than that.
> >
> > --- In 
DigitalBlackandWhiteThePrint@yahoogroups.com<DigitalBlackandWhiteThePrint%
40yahoogroups.com>,

> > "Michael King"
> > <drmrking@> wrote:
> > >
> > > John,
> > >
> > > The current hardware (Epson x800) is already capable of printing a least
> > 10
> > > bits of gray (1024 levels) with IP (based on my own tests with a
> > spectro).
> >
> > The point is that you can get those 1024 levels with a 6 bit file. The
> > issue
> > arises because people say 6 bits = 64 levels so you couldn't represent the
> > 1024
> > levels. This isn't the way it works. ALL possible measurements that show
> > 1024 levels are averaging over an area with many pixels, so measuring more
> > than 64 levels is easily done.
> >
> > It would be great if you could take that exact setup with IP. Take the
> > file --
> > I presume it's a 16 bit stepwedge -- do the Photoshop operation I
> > specified
> > making it only have 64 grays. I maintain the file with look the same and
> > print
> > the same and when you measure it you'll still get 1024 levels.
> >
> > >
> > > Correct me if I am misquoting you Roy, but I think the point you are
> > making
> > > is that human vision probably struggles to perceive any difference in
> > images
> > > with more than 6 bits once you factor in PS dithering.
> >
> > Human vision will probably limit how far manufacturers will bother to
> > advance.
> > But my claim about # of grays is for your high end spectro, too.
> >
> > >
> > > Certainly without the dithering I think I can percieve more than 8 bits
> > > (around 9 bits) in a step wedge.
> >
> > I'm not sure what you would have in mind -- dithering is the ONLY way an
> > inkjet
> > print works. There is no other way. Dithering of the ink drops is needed
> > to
> > give a lot of grays in the print. The very slight dithering in PS is so
> > minor
> > compared to what happens in the driver/inkdrops/paper that the PS stuff is
> >
> > irrelevant.
> >
> > Roy
> >
> > > I'll do some more testing with dithered vs non-dithered images to see if
> > > there are real world situations that the dithering doesn't handle well.
> > >
> > > Mike
> > >
> > >
> > >
> > >
> > > On 19/11/06, john dean <deanwork2003@> wrote:
> > > >
> > > > Roy,
> > > >
> > > > I believe I understand most of what you are describing here. It seems
> > > > to confirm what you've said all along.
> > > >
> > > > I have two quick questions though. First, do you see the potential for
> > > > sofware rip advances that WILL allow us to gain output advantage in
> > > > sending bit depth larger than 8 bit to the printer in the near future?
> > > > And, second, are any of the new large format 12 channel pigment
> > > > machines that you've investigated addressing anything larger than 8
> > > > bit capability already?
> > > >
> > > > I guess what I am getting at is IF a bit depth beyond 8 bit WAS
> > > > possible in the near future, would we be seeing this capability coming
> > > > from new software, new hardware, or a combination of the two designed
> > > > in conjunction with each other.
> > > >
> > > > John
> > > >
> > > >
> > > >
> > >
> > >
> > > [Non-text portions of this message have been removed]
> > >
> >
> > 
> >
> 
> 
> [Non-text portions of this message have been removed]
>

Roy,

Thanks for your patience in trying to explain this. I am off to experiment
and see what happens in practice.
It may be a few weeks before I can report back as I travelling.

Mike


On 20/11/06, Roy Harrington <roy@harrington.com> wrote:
>
>
> Hi Mike,
>
> I know it sounds non-intuitive so I'll try to explain what is going on.
>
> Take your 1024 stepwedge. Somewhere there are 2 adjacent steps that have
> the
> values 400/1024 and 401/1024. When you print with IP you will get gray
> patches
> that differ by some amount. Take these two patches and to the little
> Levels/8bit/Levels
> action I outlined to get 64 levels. The 400/1024 pixels will all go to
> 100/256 or 25/64.
> The 401/1024 pixels will go to 94% 100/256 and 6% 104/256 (or 25/64 and
> 26/64).
> A 402/1024 would go to 88% 100/256 and 12% 104/256. Note that it only uses
> 64 unique pixel values but the average can be pretty much any intermediate
> value
> you want. My contention and what I think I've demonstrated is it's the
> average value
> over the whole patch that determines the "gray" value i.e. what you see
> and what you
> measure with the spectro. The fact that the pixels being averaged are not
> identical
> values has absolutely NO affect visually or measured.
>
> Try it out. Take your 1024 discrete steps and do Levels action. Note this
> is very
> different than Posterize to 64 levels -- not at all the same. You will
> still have 1024
> discrete steps. You will only tell the difference in that the file is 8
> bit and the histogram
> has lots of holes in it. Look at it on the screen, print it on the paper,
> see if there's
> anyway you can tell the difference. I haven't tried it with IP but I'm
> pretty confident
> that you will get the same result.
>
> Roy
>
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com<DigitalBlackandWhiteThePrint%40yahoogroups.com>,
> "Michael King" <drmrking@...>
> wrote:
> >
> > Roy,
> >
> > If I do as you suggest, emperically I just can't see how its going to
> come
> > out at more than 6 bits of step resolution.
> > This seems to be no different to using a driver like ABW that only
> supports
> > 8 bits of resolution and hence with ABW you only get 256 levels from the
> > 1024 level wedge (basically every 4 steps come out at approx same
> value).
> >
> > Note I understand that in a continuous tone gradient of 64 levels the
> > dithering of the printer driver will create lots of levels and I guess I
> > could easily find 1024 different levels. But I have a step wedge where
> each
> > step is discrete and big enough to be measured independently by the
> > spectro. I am not looking for a random set of 1024 levels, I am looking
> for
> > a specific stepping of levels of approx 0.1 L* between paper white and
> dmax
> >
> > I think we are talking about two very different measurement scenarios,
> or I
> > completely misunderstanding what is going on.
> >
> > Mike
> >
> >
> >
> >
> > On 19/11/06, Roy Harrington <roy@...> wrote:
> > >
> > >
> > > Mike,
> > >
> > > I'm really saying more than that.
> > >
> > > --- In
> DigitalBlackandWhiteThePrint@yahoogroups.com<DigitalBlackandWhiteThePrint%40yahoogroups.com>
> <DigitalBlackandWhiteThePrint%
> 40yahoogroups.com>,
> > > "Michael King"
> > > <drmrking@> wrote:
> > > >
> > > > John,
> > > >
> > > > The current hardware (Epson x800) is already capable of printing a
> least
> > > 10
> > > > bits of gray (1024 levels) with IP (based on my own tests with a
> > > spectro).
> > >
> > > The point is that you can get those 1024 levels with a 6 bit file. The
> > > issue
> > > arises because people say 6 bits = 64 levels so you couldn't represent
> the
> > > 1024
> > > levels. This isn't the way it works. ALL possible measurements that
> show
> > > 1024 levels are averaging over an area with many pixels, so measuring
> more
> > > than 64 levels is easily done.
> > >
> > > It would be great if you could take that exact setup with IP. Take the
> > > file --
> > > I presume it's a 16 bit stepwedge -- do the Photoshop operation I
> > > specified
> > > making it only have 64 grays. I maintain the file with look the same
> and
> > > print
> > > the same and when you measure it you'll still get 1024 levels.
> > >
> > > >
> > > > Correct me if I am misquoting you Roy, but I think the point you are
> > > making
> > > > is that human vision probably struggles to perceive any difference
> in
> > > images
> > > > with more than 6 bits once you factor in PS dithering.
> > >
> > > Human vision will probably limit how far manufacturers will bother to
> > > advance.
> > > But my claim about # of grays is for your high end spectro, too.
> > >
> > > >
> > > > Certainly without the dithering I think I can percieve more than 8
> bits
> > > > (around 9 bits) in a step wedge.
> > >
> > > I'm not sure what you would have in mind -- dithering is the ONLY way
> an
> > > inkjet
> > > print works. There is no other way. Dithering of the ink drops is
> needed
> > > to
> > > give a lot of grays in the print. The very slight dithering in PS is
> so
> > > minor
> > > compared to what happens in the driver/inkdrops/paper that the PS
> stuff is
> > >
> > > irrelevant.
> > >
> > > Roy
> > >
> > > > I'll do some more testing with dithered vs non-dithered images to
> see if
> > > > there are real world situations that the dithering doesn't handle
> well.
> > > >
> > > > Mike
> > > >
> > > >
> > > >
> > > >
> > > > On 19/11/06, john dean <deanwork2003@> wrote:
> > > > >
> > > > > Roy,
> > > > >
> > > > > I believe I understand most of what you are describing here. It
> seems
> > > > > to confirm what you've said all along.
> > > > >
> > > > > I have two quick questions though. First, do you see the potential
> for
> > > > > sofware rip advances that WILL allow us to gain output advantage
> in
> > > > > sending bit depth larger than 8 bit to the printer in the near
> future?
> > > > > And, second, are any of the new large format 12 channel pigment
> > > > > machines that you've investigated addressing anything larger than
> 8
> > > > > bit capability already?
> > > > >
> > > > > I guess what I am getting at is IF a bit depth beyond 8 bit WAS
> > > > > possible in the near future, would we be seeing this capability
> coming
> > > > > from new software, new hardware, or a combination of the two
> designed
> > > > > in conjunction with each other.
> > > > >
> > > > > John
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > [Non-text portions of this message have been removed]
> > > >
> > >
> > >
> > >
> >
> >
> > [Non-text portions of this message have been removed]
> >
>
> 
>


[Non-text portions of this message have been removed]