30-bit Color on 24-bit Hardware

[Ron W. Silvas]

Hi.

I just wanted to get the word out there for any who may be interested
about this display fidelity extension stuff I've been working on. It's
basically about getting 24-bit hardware to perceptually display 30-bit
color.

General information is available at <http://www.silvasdigital.com/xdf>.

The "meat and potatoes" info is available in the Project Summary and
Findings Report at . . .
<http://www.silvasdigital.com/xdf/XDF Project Summary and Findings Report.pdf>

All the info is free--I don't think there's anything proprietary in
there (though I'm not any kind of legal council)--but since I *have*
spent quite a lot of time on this and I do have bills to pay, I'm only
providing the tech demo (a hi-fi image viewer for Windows) in return
for donations.

Thanks and I hope you find the info to have been worth checking out,
- Ron

 

[Rick]

Hi.

I just wanted to get the word out there for any who may be interested
about this display fidelity extension stuff I've been working on. It's
basically about getting 24-bit hardware to perceptually display 30-bit
color.

General information is available at <http://www.silvasdigital.com/xdf>.

The "meat and potatoes" info is available in the Project Summary and
Findings Report at . . .
<http://www.silvasdigital.com/xdf/XDF Project Summary and Findings Report.pdf>

All the info is free--I don't think there's anything proprietary in
there (though I'm not any kind of legal council)--but since I *have*
spent quite a lot of time on this and I do have bills to pay, I'm only
providing the tech demo (a hi-fi image viewer for Windows) in return
for donations.

Thanks and I hope you find the info to have been worth checking out,

Ron, have you written an API for extended-bit printer
drivers? Windows is limited to 24-bit output, and without
the aforementioned API your technology is of very limited
use (at least under Windows).

Rick

 

[Des Perado]

Hi.

I just wanted to get the word out there for any who may be interested
about this display fidelity extension stuff I've been working on. It's
basically about getting 24-bit hardware to perceptually display 30-bit
color.

Perceptually, would anybody see a difference?

 

[Atreju]

Perceptually, would anybody see a difference?

I know my eyes usually tune out at about 150,000 colors ;-)


---Atreju---

 

[Stephen H. Westin]

I know my eyes usually tune out at about 150,000 colors ;-)

It's not that simple. The colors from a normal display aren't
distributed evenly in the human color space, and there's a difference
between the contrast you can see between two adjacent colors and the
total number of colors in an image. So it's quite possible, depending
on the conditions, for 24 bits to be inadequate. I would expect this
more for LCD's than for CRT's, though, both because of their dynamic
range and because the nonlinearity of a CRT actually helps. I think
many LCD's use temporal dithering already, as they tend to be a bit
short of full 24-bit capability.

 

[Ron W. Silvas]

Ron, have you written an API for extended-bit printer
drivers? Windows is limited to 24-bit output, and without
the aforementioned API your technology is of very limited
use (at least under Windows).

Hi, Rick.

No, I haven't written any sort of print driver.

And, like you, I suspect that visualizing in color depths beyond what
your target output device is capable of producing might not be the best
idea, but . . .

Paper is by no means the only target medium out there these days,
right? The video game and multimedia industry produces content for
display on computer monitors and televisions, the motion picture
industry--including their animation and special effects studios--
produce content for film and 30-bit digital projection, and I believe
medical imaging usually requires greater than 24-bit color fidelity (or
greater than 8-bit grayscale, etc.).

And even if 30-bit color didn't have any practical uses–-which is not
the case–-we'd only know that by first researching it, right? I'm just
making some of that research available to those who are similarly
interested.

Thank you for asking about this,
- Ron

 

[Ron W. Silvas]

Perceptually, would anybody see a difference?

If your question is . . .
"Would anybody see a difference between the true 30-bit color (of, say,
a Matrox Parhelia) and the synthesized 30-bit color (which I'm talking
about)?"

The short answer is: No, I don't think they would–Not if the 30-bit
synthesizing is implemented correctly.

But if your question is . . .
"Would anybody see a difference between 24-bit color and 30-bit color?"

As Mononen pointed out, the short answer is: Yes they would, but only
in some images. (Thanks, Mononen. =)


More detailed answers to these questions can found in . . .
<http://www.silvasdigital.com/xdf/XDF Project Summary and Findings Report.pdf>
.. . . with the latter question dealt with in the
"24-bit Color vs. 30-bit Color" section.

Thank you for the discussion,
- Ron

 

[Ron W. Silvas]

It's not that simple. The colors from a normal display aren't
distributed evenly in the human color space...So it's quite possible,
depending on the conditions, for 24 bits to be inadequate. I would
expect this more for LCD's than for CRT's, though, both because of
their dynamic range...I think many LCD's use temporal dithering
already, as they tend to be a bit short of full 24-bit capability.

Hi, Mr. Westin.

Thank you for pointing that out--about the difference between human
vision and the color space of RGB displays. I tried to deal with that
a little in the pre-mentioned PDF's "Beyond 24-bit Color" section,
where I mention that "24-bit color allows for hue and saturation level
precision beyond human visual sensitivity but fails to naturally allow
for the specification and display of the many hundreds of
intensity/brightness levels per color that humans can visually
distinguish."

I've found that 24-bit artifacts are quite demonstratable on both CRTs
and LCDs, alike, (as can be experienced with the previously mentioned
tech demo) but I believe that, if anything, they would be more
noticeable on CRTs because the higher dynamic range of a CRT means that
an 8-bit color channel's 256 intensity/brightness levels have that much
more perceptual space to cover, right?

I believe you are correct about LCDs already using temporal dithering--
often to extend native 18-bit LCD hardware (6 bits per RGB channel) to
24-bit color. But I believe there *are* true 24-bit LCDs also, and what
I find interesting is that, apparently, some companies are starting to
use temporal dithering in *them* to extend them to 30-bit color--BUT,
unless there's some kind of good hardware reason to do this (which I
wouldn't know about), my research seems to indicate that they should
instead be using traditional *spatial* dithering for this extension
because . . . If properly implemented, spatial dithering can be used to
extend 24-bit color to 30-bit color with *none of its traditional
dithering noise visible* because of the perceptual closeness of 24-bit
color space neighbors (More about that in the PDF).

Thanks again for pointing out the color space deal, and thank you for
the discussion,
- Ron

 

[Bob Myers]

I've found that 24-bit artifacts are quite demonstratable on both CRTs
and LCDs, alike, (as can be experienced with the previously mentioned
tech demo) but I believe that, if anything, they would be more
noticeable on CRTs because the higher dynamic range of a CRT means that
an 8-bit color channel's 256 intensity/brightness levels have that much
more perceptual space to cover, right?

I've been following this thread with interest, but a couple of
comments seem appropriate at this point.

First - what "higher dynamic range" of the CRT? Neither the
color gamut or the contrast of the typical CRT display is
much different that what can be obtained with the LCD.
What I think the problem here really is, is a difference in the
response curves of the two technologies. CRTs have a very
nice (perceptually) "gamma" sort of response, whereas LCDs
have a perceptually-nastier "S-shaped" response that cannot
adequately be compensated for with only 8 bits per channel,
not if you still intend to have 8 bit/channel of perceptually linear
intensity control. The solution here is greater dynamic range
(bit depth) at the panel (i.e., at the driver level), to permit
compensation for the panel's response and thus present a
"CRT-like" 8-bit input. Fortunately, we ARE starting to see
10-bit LCD drivers entering the market.

Bob M.

 

[Rick]

I've been following this thread with interest, but a couple of
comments seem appropriate at this point.

First - what "higher dynamic range" of the CRT? Neither the
color gamut or the contrast of the typical CRT display is
much different that what can be obtained with the LCD.

Nonsense. The backlight technology used in LCDs (at least
those in the con/prosumer range) precludes the possibility that
they can match the gamut of a decent CRT. E.g. a $700
Mitsubishi 2070 easily outgamuts a $1300 Apple Cinema
Display.

What I think the problem here really is, is a difference in the
response curves of the two technologies. CRTs have a very
nice (perceptually) "gamma" sort of response, whereas LCDs
have a perceptually-nastier "S-shaped" response that cannot
adequately be compensated for with only 8 bits per channel,
not if you still intend to have 8 bit/channel of perceptually linear
intensity control. The solution here is greater dynamic range
(bit depth) at the panel (i.e., at the driver level), to permit
compensation for the panel's response and thus present a
"CRT-like" 8-bit input. Fortunately, we ARE starting to see
10-bit LCD drivers entering the market.

More nonsense. No amount of compensation will overcome
the limitations of LCD's backlight technology. One has the
choice of losing near blacks altogether, or settling for a
constant and muddy bluish or grayish glow instead of black.

Rick

 

[Ron W. Silvas]

I've been following this thread with interest, but a couple of
comments seem appropriate at this point.

First - what "higher dynamic range" of the CRT? Neither the
color gamut or the contrast of the typical CRT display is
much different that what can be obtained with the LCD.
What I think the problem here really is, is a difference in the
response curves of the two technologies. CRTs have a very
nice (perceptually) "gamma" sort of response, whereas LCDs
have a perceptually-nastier "S-shaped" response that cannot
adequately be compensated for with only 8 bits per channel,
not if you still intend to have 8 bit/channel of perceptually linear
intensity control. The solution here is greater dynamic range
(bit depth) at the panel (i.e., at the driver level), to permit
compensation for the panel's response and thus present a
"CRT-like" 8-bit input. Fortunately, we ARE starting to see
10-bit LCD drivers entering the market.

Bob M.

Hi, Mr. Myers.

It might be time for me to exit from this branch of the
discussion, as I am admittedly not very knowledgeable about the
areas being discussed (dynamic ranges, contrast ratios,
hardware behavioral differences between LCD and CRT
technologies, etc.). =)

My earlier comment . . .

.. . . was a jump to a conclusion on my part–I should have asked
Mr. Westin if he could explain to me his comment about the
different dynamic ranges before I commented on that. Totally
my mistake.

What I think the problem here really is...the response curves . . .

You may be referring to a legitimate problem that needs solving,
but I wouldn't know, so I'll leave that for you and others to
discuss, BUT . . .

The problem that this project dealt with (and successfully
overcame) was simply this . . .

Quick & Dirty:
o 24-bit color's 256 intensity levels per channel are not
enough.
(True even where gamma config'd to *exactly* correspond to
vision's intensity response.)
o 27-bit color's >500 intensity levels per channel is close,
but not enough.
o 30-bit color's >1000 intensity levels per channel *is*
finally enough.
o 30-bit color display is desirable.
o 30-bit color display on common 24-bit hardware would be a
nice bonus, and . . .

It works. On CRTs and LCDs, alike. No special hardware
requirements. No special OS support needed.

- Ron

 

[chrisv]

CRTs have a very
nice (perceptually) "gamma" sort of response, whereas LCDs
have a perceptually-nastier "S-shaped" response that cannot
adequately be compensated for with only 8 bits per channel,
not if you still intend to have 8 bit/channel of perceptually linear
intensity control. The solution here is greater dynamic range
(bit depth) at the panel (i.e., at the driver level), to permit
compensation for the panel's response and thus present a
"CRT-like" 8-bit input. Fortunately, we ARE starting to see
10-bit LCD drivers entering the market.

You feel that "more bits" is the best solution to a non-linear
response? Seems to me the panel could (and, indeed, does), compensate
for this.

 

[chrisv]

But if your question is . . .
"Would anybody see a difference between 24-bit color and 30-bit color?"

As Mononen pointed out, the short answer is: Yes they would, but only
in some images. (Thanks, Mononen. =)

Maybe, under perfect conditions.

I'd like to see these alleged "some images"...

 

[Bob Myers]

Nonsense. The backlight technology used in LCDs (at least
those in the con/prosumer range) precludes the possibility that
they can match the gamut of a decent CRT. E.g. a $700
Mitsubishi 2070 easily outgamuts a $1300 Apple Cinema
Display.

The gamut of any display is determined by its primaries.
Typical CRT primaries are generally very close to those
given in standards such as sRGB, or the "EBU" (European
Broadcast Union) phosphor set. These are as follows:

sRGB (IEC 61966-2.1):

Red - 0.640, 0.330
Green - 0.300, 0.600 (for EBU, this is 0.290, 0.600)
Blue - 0.150, 0.060

I'm not sure exactly WHICH of the "Apple Cinema"
displays you're talking about, but the latest published data
I have on the 23" claims the following:

Red - 0.640, 0.332
Green - 0.288, 0.601
Blue - 0.146, 0.065

I've never measured the Apple display specifically, but these
are pretty close to what I would expect based on tests of similar
contemporary panels.

(All of the above, by the way, in 1931 CIE xy coordinates.)

Those look very CRT-ish to me; if anything, the green is
slightly better than the sRGB/EBU standards, the red is
very slightly more saturated, and the blue might be very
slightly less. So unless you're using "gamut" to mean something
very different here, I am afraid I don't see your point.

More nonsense. No amount of compensation will overcome
the limitations of LCD's backlight technology. One has the
choice of losing near blacks altogether, or settling for a
constant and muddy bluish or grayish glow instead of black.

I don't think you understood the point here, which had to do
with the luminance response curve of the two technologies, not
the backlighting or anything related. Perhaps you could elaborate
a bit.

Bob M.

 

[Bob Myers]

You feel that "more bits" is the best solution to a non-linear
response? Seems to me the panel could (and, indeed, does), compensate
for this.

More bits AT THE DRIVERS, which does not necessarily
translate to more bits at the display interface. The problem
here is that the PANEL does not compensate for the
response - any compensation which occurs happens in the
LCD controller, which is part of the monitor's interface
circuitry (i.e., it's between the panel proper and the outside
world). The problem here is that the controller only has eight
bits per color, at the PANEL interface, to play with, and any
of that that's used in doing the response compensation there
eats into the actual range that's available at the interface to the
graphics hardware. This can (and has) resulted in some rather
obvious artifacts, which generally show up when trying to do
such things as 8-bit grayscale patterns and the like.

Bob M.

 

[Stephen H. Westin]

Nonsense. The backlight technology used in LCDs (at least
those in the con/prosumer range) precludes the possibility that
they can match the gamut of a decent CRT. E.g. a $700
Mitsubishi 2070 easily outgamuts a $1300 Apple Cinema
Display.

Really? Where did you find this out? Do you have the respective primary
chromaticities handy?

More nonsense. No amount of compensation will overcome
the limitations of LCD's backlight technology. One has the
choice of losing near blacks altogether, or settling for a
constant and muddy bluish or grayish glow instead of black.

In any reasonable lighting level, a good LCD will be at least
competitive in dynamic range with any CRT. The days of all CRT's being
better than all LCD's are past.

 

[Rick]

Really? Where did you find this out? Do you have the respective primary
chromaticities handy?


In any reasonable lighting level, a good LCD will be at least
competitive in dynamic range with any CRT. The days of all CRT's being
better than all LCD's are past.

We've been through this, Stephen, where you pull out your
highly misleading "research paper" and make highly misleading
claims based on misleading information.

The bottom line? Virtually no professional graphic shops use
LCDs. Why? Because any serious attempt at color correcting
on con/prosumer LCDs is an exercise in futility. End of story.

Rick

 

[Stephen H. Westin]

We've been through this, Stephen, where you pull out your
highly misleading "research paper"

Papers. From several sources. With actual measurements.

and make highly misleading
claims based on misleading information.

And you come back and say, "Everybody knows..." with no data (or
misleading data) to support your assertions. Is your lack of
information less "misleading" than my information?

The bottom line? Virtually no professional graphic shops use
LCDs. Why?

Because they rely on experience and conventional wisdom, rather than
measured performance of modern equipment. I am becoming increasingly
convinced of this, as I hear the arguments supporting the "LCDs are
inferior" assertion.

Because any serious attempt at color correcting
on con/prosumer LCDs is an exercise in futility. End of story.

Well, I'm not sure what you mean by "con/prosumer LCDs". To be
specific, RR Donnelly specifies the Apple 20" Cinema Display as
part of their prepress soft proofing system. Are they ignorant about
display calibration?

In short, you are totally convinced of your claims, you don't have the
data to back up your claim of inferior gamut, and you deem any
evidence to the contrary "misleading".

 

[Bob Myers]

We've been through this, Stephen, where you pull out your
highly misleading "research paper" and make highly misleading
claims based on misleading information.

The bottom line? Virtually no professional graphic shops use
LCDs. Why? Because any serious attempt at color correcting
on con/prosumer LCDs is an exercise in futility. End of story.

Well, YOU made the claim that a given CRT monitor
"outgamuts" an LCD. Stephen quite rightly asked you to
justify that claim by describing the gamut of the monitor in
question in quantitative terms. I'd also like to see that, and
am willing to bet that you can't provide numbers that support
that specific claim. Whether or not "any serious attempt at
color correcting... is an exercise in futility" is not relevant to
the claims you've made. (In other words, you may be right
in that broad statement, at least for most current monitors,
but not for the reasons you stated.)

So let's see your numbers. End of story.

Bob M.

 

[Rick]

Well, YOU made the claim that a given CRT monitor
"outgamuts" an LCD. Stephen quite rightly asked you to
justify that claim by describing the gamut of the monitor in
question in quantitative terms. I'd also like to see that, and
am willing to bet that you can't provide numbers that support
that specific claim. Whether or not "any serious attempt at
color correcting... is an exercise in futility" is not relevant to
the claims you've made. (In other words, you may be right
in that broad statement, at least for most current monitors,
but not for the reasons you stated.)

So let's see your numbers. End of story.

This debate is recurring. The last one was not long ago. The
numbers in Stephen's reference have been discussed over and
over, with the same conclusion every time. The reference's
numbers show (correctly) that CRTs have orders of magnitude
more dynamic range, especially for near blacks, then it proceeds
to claim that this isn't an advantage, because of "typical office
lighting conditions" -- which is completely irrelevant to the issue.

Those who claim LCDs have anywhere near the same gamut as
a decent CRT are generally the same people who've never even
heard of a light hood. I invite you to look up these previous
threads in Google's Usenet archive, e.g. http://tinyurl.com/623mb

Rick