Calman and other observers

[Guillaume]

Hi ! I use the ColorHCFR software to calibrate manually and I want to use the CIE 2012/2015 10 deg observer for manual calibration. But this observerver isn’t aligned to the CIE 1931 graph. I obtain very consistant white results between displays. My smartphone set to D65 via whitescreen app is the same as my QD-OLED D65 calibrated monitor with my I1Pro2. The Red Green and Blue readings are wrongly located compared to Displaycal. The I1Pro2 and I1 Display Plus are both using 2012 10 deg observer and the correction matrix has been generated by HCFR using the 2012 10 deg observer. Is there a matrix to go from 2012 10 deg to CIE 1931 ? And the main question is : does Calman make the correction automatically ? By using the methodology I mentionned Displaycal makes a good looking Gamut.

[Guillaume]

Nevermind ! I managed to fix this problem.

I made a tailored matrix from DisplayCal to HCFR :

1.192023 -0.161968 -0.018385
0.110225 0.955541 -0.055514
-0.049848 0.078630 0.970592

It alignes the cie 2012-10degs to the standard CIE 1931 graph. I tested the smartphone and my Agon Pro AG27UZD and the matrix is accurate enough

[DaniJ]

A different CMF will give you different XYZ/xy values. So it will be like comparing apples to oranges.

The Red Green and Blue readings are wrongly located compared to Displaycal

What do you mean? Did you use the same observer in HCFR and Displaycal and got considerably different values?

[Guillaume]

Yes I got different RGB xy values from HCFR and Displaycal. I think that if you do a correction matrix with a spectrophotometer for a colorimeter and you set both measurements with the same CMF, Displaycal is able to compute the gamut as if it was under a 1931 2 deg CMF. HCFR don’t seem to be able to do this trick because it’s based on manual calibration or something else. There are others CMF that are tagged as “(Aligned)”. The Stiles and Burch 1955 is an RGB based CMF so it has to be recomputed to make it CIE 1931 compatible, there is also a CIE 1964 10 deg aligned. And these one look more like the 1931 2deg, especially in the blue part.

Here are the pictures : https://www.mediafire.com/folder/k4eu4agcwmvdc/2015_10_deg_CMF_Displaycal_vs_HCFR

My correction matrix takes the DisplayCal as ref by editing manually red green and blue XYZ values in HCFR and measure the one with uncorrected 2015 10 deg values. In the end I obtained what wanted and it works with the sample I measured. I gonna try with a CCFL backlit LCD to see if it makes the same white and the same RGB.

[DaniJ]

I see argyllcms provides 1964_10c selects a version of the CIE 1964 10 degree observer that has been adjusted using a 3×3 matrix to better agree with the 1931 2 degree observer.

Might be worth asking on their mailing list how that 3×3 matrix was computed and if it makes sense to compute one for CIE 2012/2015.

I would be very curios how such a correction helps when measuring displays with vastly different spectra.

• This reply was modified 6 months, 1 week ago by DaniJ.

[Guillaume]

I did my testings with an old CCFL monitor and I targeted the D65 white point with its CIE 1931 2 deg and the result I had was very close to the QD-OLED D65 with xy 0.3138 0.3313. The result had a tiny blue tint but as far as I know the CIE 2015 10 deg values reduces mainly the blue component by 1 or 2 percent for the D65 white point. I checked red green and blue visualy with HCFR built in measured and reference comparator. And guess what ? The measured patches hues matched the CCFL monitor. So the QD-OLED CIE 2015 10 deg calibration lead to correct colors.

My matrix works well but I need to do more testings with the CCFL TV from 2009/2010 we’re currently using. I tried today but there was some technical problems. I’ll also do a try with an old CRT TV I have in my bedroom.

I would be very curios how such a correction helps when measuring displays with vastly different spectra.

Me too ! That could be a better way to calibrate manually. Due to the fact Displaycal/Argyllcms compute this well, it needs less to be taken into account. The matrix must be done via HCFR and the used observers need to be specified in the CCMX.

CCMX

DESCRIPTOR “i1 DisplayPro, ColorMunki Display & AOC Agon Pro AG276UZD 2012 10”
INSTRUMENT “X-Rite i1 DisplayPro, ColorMunki Display”
MANUFACTURER_ID “AOC”
MANUFACTURER “AOC”
OBSERVER “2012_10”
REFERENCE_OBSERVER “2012-10”
DISPLAY “KA241Y”
TECHNOLOGY “QD-OLED”
DISPLAY_TYPE_BASE_ID “1”
DISPLAY_TYPE_REFRESH “NO”
REFERENCE “I1 Pro2”
REFERENCE_HASH “md5:e4ca4b2d17ec8b36c99af27f66d8828c”
TARGET_HASH “md5:ed0e2de550871a76a74fa7edabf840fd”
FIT_METHOD “ΔE*00”
FIT_AVG_DE00 “0.242857”
FIT_MAX_DE00 “0.600000”
FIT_AVG_DE94 “0.285714”
FIT_MAX_DE94 “0.600000”
ORIGINATOR “HCFR THC”
CREATED “2025”
COLOR_REP “XYZ”

NUMBER_OF_FIELDS 3
BEGIN_DATA_FORMAT
XYZ_X XYZ_Y XYZ_Z
END_DATA_FORMAT

NUMBER_OF_SETS 3
BEGIN_DATA
1.036748 -0.022047 -0.005005
0.048415 0.945583 0.006995
-0.038860 0.038309 1.015411
END_DATA

Calibrite Display Pro HL on Amazon  
Disclosure: As an Amazon Associate I earn from qualifying purchases.

[Guillaume]

Due to the fact Displaycal/Argyllcms compute this well, it needs less to be taken into account. The matrix must be done via HCFR and the used observers need to be specified in the CCMX.

In fact right now I disagree with this statement I did. It would be nice to select the observer on colorimeter correction measurement tab too. So the CCMX will be tailored with the same observer and we wouldn’t need to make a correction with HCFR and copy past the matrix into the CCMX.

[Author]

Posts