[argyllcms] Re: Color errors on < 100% stimulus

  • From: Kristian Jörg <krjg@xxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 04 Apr 2012 08:43:35 +0200

The 75% Rec 701 CMS calibration is available in Chromapure and is used whenever you are calibrating a TV which has non-linear color response. If calibrated at 100% intensity the color errors could be larger and larger as you go down the saturation scale. If that is the case calibrating at 75% saturation may even out the errors and achieve an overall better compromise. Chromapure has a diagnostic tool that lets you plot out the accuracy at 25, 50, 75 and 100% saturation and intensity levels to see how the color errors occurs in a CIE diagram. I have only tried saturation diagnostics yet since that is the only patterns I have on disc. For the intensity patterns I have to use the built-in pattern generator, and I need a 5 m hdmi cable for that. Will try that one day. The sad thing is that there is no CMS controls to fix errors in other points on the scale. There is only the one setting for all and it is here I feel Pana is letting us down with a subpar CMS implementation. Not that I need more controls. I need better linearity in the CMS decoding. One thing I found was that if I do not lower color intensity to a normal level as the first step in color calibration but instead use the default setting - saturation linearity is almost perfect! I will then need to decrease lightness a lot for all primaries and secondaries. What it does to intensity linearity though is unknown as I havent't done that diagnose yet. The result is somewhat pleasing and somewhat dissapointing depending on video content. I haven't found a method to calibrate color on this TV yet that gives a better color response than the THX setting. I may end up just fixing the THX white point in SM and be done with it. ISFccc seems impossible to get a good result out of.

For other maybe reading this I speak of a Panasonic VT30.

For the other question. It so happens that I am a Java + HTML programmer... :) What exactly did you have in mind. I did not get it from your explanation. DispcalGUI is writen in Python btw...


János, Tóth F. skrev 2012-04-03 19:53:
Yes, HCFR supports multi-point primary/secundary color gradient
readings. I forgot about that.
I guess that's more likely for benchmark/validation purposes than
something you should base your calibration on. (But it could be a good
solution if it works for you.)

By the way, I am trying to find somebody who is intereated in HDTV
calibration and good in Java+HTML programming.
With some small changes, the profile validation feature of DispcalGUI
could serve as a great tool for that.
It's already usable for TVs with HSL color space and 2p grayscale
controls and you can also guess RGB color settings from the HSL errors
but it's hard to guess RGB settings for the 10/20p grayscale controls
from HSL errors (or is it...? May be it's only hard for me...). So a
few indicators (RGB balance graphs) could make the process easier,
more convenient and faster.

2012/4/2, Kristian Jörg<krjg@xxxxxxx>:
János, Tóth F. skrev 2012-03-26 00:56:
I still don't understand the so-called „color decoding issue”.
My Pana G30E has relatively good and significantly better color
accuracy than my Samsung D550 had.
According to avsforum threads there is an issue with at least the
American versions of the GT and VT models. The issues are visible with a
video sequence with a woman in purle dress that was decoded as blue - on
a otherwise perfectly calibrated set.
There has also been reports on lightness nonlinearities over different
intensities/saturations - this is supposedly fixed by named patch.

The problems I have myself is that if I calibrate my set to perfection
with the tools and methods as usual, I end up with a set that measures
perfectly with dE<  1.0 over the scale - but visibly has color errors
with normal content! I just now calibrated my PC monitor (at last) and
it shows no color casts whatever. Perfect calibration/profiling with
Argyll. So it is not my meter that is malfunctioning... For the moment I
have reverted to the stock THX mode which is fairly accurate. I'll try
again to calibrate some day.

How do you know that the 2012 models will not be improved with regard to
CMS? They have hardly come out yet...

Chromapure has options for calibrationg at 75% saturated patterns. These
can be found on the AVSHD 709 disc under the HCFR header and the
"Saturation" patterns. Use a 100% stimulus white pattern and then the
75% saturated.

Actually, I haven't found any instructions on how to use the 75%
intenity color patterns or when they are appropriate?

2012 won't be very different except the VT50.

75% saturated pattern?
I never heard about such thing. On the other hand, 75% stimulus is
very common for HDTV calibration.

2012/3/21, Kristian Jörg<krjg@xxxxxxx>:
Graeme Gill skrev 2012-02-13 14:41:
Kristian Jörg wrote:
Well, the reason I was a bit unimpressed is because this is supposed to
be one of the very best TV
sets to buy, and it includes full ISF controls for managing grayscale
10% stimulus steps. Color
however is only at 100% stimulus. So I can nail in the grayscale and
gamma to mostly perfect result
(according to the patterns used, i.e full windows cause lightness to
a bit). Color however
seems a bit more difficult. Now it is possible I have used the CP
software wrongly at the 75% color
intensity. It has a 75% Rec. 709 setting that I assumed should be used
for the 75% W+ 75% RGB
measurements. But maybe I should measure all those at 100% Rec. 709
setting since i assume the Rec
709 standard is linear? When using 75% patterns all that happens is
luminocity falls a bit,
but it is all realite to white which is measured as a baseline... ?
I'm afraid I have to smile a bit at all that. Rec 709 is an idealised
additive display.
Idealised displays have a perfect black, something no real display can
So (as usual) you are faced with the issue of how do you deal with the
non-perfect black,
given that the transfer curve assumes that is is perfect. Non-LCD based
displays are not
so bad in this regard though. Then there are the viewing condition
assumptions inherent
in Rec 709.

The channel combination targets aren't too hard to compute - add the XYZ
of the individual channel
target or measurement values, since Rec 709 is additive.

Graeme Gill.
Just a short follow up to this little thread that I started. It was all
due to a misunderstanding of what patterns to use for the 75% Rec 709
calibration settings in Chromapure. It should be 75% saturated, not 75%
intensity. With correct patterns my results where correct. However if
anybody tries this, there is a confirmed bug with it with at least
version 2.0.10 (and probably 2.0.11) in CP, so wait for the next release
before trying this.

And there is also strong indication of color decoding issues in the 2011
models of Panasonic plasmas. Something that should be fixed in the 2012
models. Bummer. There will not be a fix to the 2011 models, or even any
more firmaware updates at al,l according to those who "know".


Other related posts: