Re: fundamental question - monitor profile as working RGB?
Re: fundamental question - monitor profile as working RGB?
- Subject: Re: fundamental question - monitor profile as working RGB?
- From: Chris Murphy <email@hidden>
- Date: Wed, 10 Dec 2003 19:48:16 -0700
On Dec 10, 2003, at 5:44 PM, Graeme Gill wrote:
Chris Murphy wrote:
Most of the packages are matrix and TRC based. When the TRCs are the
same for each channel, then R=G=B makes gray.
Sounds to me like it should be the other way around. If you insist
on making a profile that has the same TRC for each channel, then
the profile will be more accurate if the display is calibrated to
have perfect gray balance. But why duplicate the same TRC data
for each channel ? It's not a native concept as far as the ICC format
is concerned.
Go into Photoshop and making a custom working space. Change the red
primary so that it's yellow. Now you have a custom RGB space that's
yellow, green, blue. Yet if R=G=B, the predicted LAB value is a*=0 and
b*=0.
But that most products don't do this doesn't mean they're bad and
should be avoided.
It just strikes me as kind of awkward and fragile. It's one thing to
fiddle the analog controls on a CRT (or equivalent digital controls),
to
adjust the device gamut and native response to give best possible
control
from the frame buffer, and to match a target gamut and white point,
but it's
another to try and fake the equivalent adjustments using hardware LUTs
in the display card,
particularly if they only have 8 bit entries, and then to rely on that
adjustment for the ICC profile to be accurate.
For that matter, the number of measurements for many of the products is
on the low side to adequately sample the extreme ends of the response.
If you're willing to trade color accuracy for small profiles, and
quick profiling, then such compromises are probably fair enough. I
sort of thought
that most people in the forum would be interested in high color
accuracy though.
The independent testing of the Artisan demonstrates a simple matrix
based profile, with identical TRCs for each channel, can do the job
with outstanding quality. Small (matrix) profiles and low color
accuracy are not concomitant. The problem is that matrix profiles make
a lot more assumptions than table based profiles. If the assumptions
are true, you get good results perhaps even better than with table
based profiles for the same reasons we sometimes get better output
device profiles using fewer patches. If the assumptions are not
entirely true, it's probably not that big of a deal. If the assumptions
aren't good, then there's a problem.
Another way of doing it combines the approaches where you still have a
single measurement for the primaries, but the TRC's are defined by
points in a table instead of with a gamma function. Off hand, Optical
using the profile only option (no calibration) builds such a profile.
[Is it really the case that popular graphics cards still only have 8
bit entries in
their LUts ? Last time I was designing graphics hardware, the RAMDACs
were all
appearing with 12 bit LUT entries, and that was about 10 years ago.
Surely all
the VGA cards have 12 bits or better for the LUT data ?? Or is the
Operating
System software the problem ?)
All I've got are hearsay opinions and by mentioning them here they
become fourth hand so they should be taken with a grain of salt. But
there are a few cards with 10-bit DACs. What I'm not clear on is how
well this is supported in current OS's. I've heard that the issue has
to do with the availability of drivers to take advantage of the
10-bits. I don't know if that means with existing drivers they are
effectively 8-bit cards, or if that means the card simply won't work
unless they get a driver fully supporting those 10-bits. And I don't
know what it takes to get an OS to use such a driver. I think there are
a few variables that are making it a challenge, but I'm not privy to
them.
Chris Murphy
Color Remedies (TM)
www.colorremedies.com/realworldcolor
---------------------------------------------------------
Co-author "Real World Color Management"
Published by PeachPit Press (ISBN 0-201-77340-6)
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.