- This topic has 12 replies, 2 voices, and was last updated 5 years, 5 months ago by
.
-
Posts
-
Hi,
First, thank you for your great work ! I’m currently in charge of the color pipeline in our CG Studio in France, I’m quite new in this domain, and I need some clarifications.
From what I understood, calibration should done in two steps :
- Create video card gamma table (VCGT) to obtain the desired TRC and white point
- Create a 3D LUT to color match
It’s about the 2nd point I’m not sure, I created an icc profile with the 3D lut creator (without VCGT applied, cause I already installed it globally) and tried to use it in gimp, that seems to work, but I don’t understand why the VCGT is an icc file and the LUT is also an icc file… They look absolutely different.
Is it better to create a 3D LUT with applied VCGT (and disable the videocard one) ?
Is it harmful to calibrate with a response curve that’s not the display’s default ?
About CG imagery, I guess the good practice would be to create a spi 3D LUT and apply it with OpenColorIO as a final transform, right ? If so, which interpolation mode gives better results ?
I guess using the display’s native color gamut is the best thing to do when calibrating ?
Last, what’s the best colorimeter/sprectrophotometer for a CG/Video studio ?
Screens i use : DreamColor Z27x, Z27xG2 and a ZBook 15 with dreamcolor panel
Calibration probe : I1 DisplayPro (HP labelled)
Best regards
Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.Another thing : I try to calibrate to 6500K, the result looks more yellowish once done. Do you think it’s normal ?
The reported non calibrated gives me this with i1 display pro device :
17:54:02,711 Setting up the instrument 17:54:02,711 Product Name: i1Display3 17:54:02,711 Serial Number: I1-16.B-02.262901.12 17:54:02,711 Firmware Version: v2.28 17:54:02,711 Firmware Date: 29Jan14 17:54:02,711 dispcal: Warning - new_dispwin: Expected VideoLUT depth 11 doesn't match ↲ ↳ actual 10 17:54:02,711 Measured display update delay of 24 msec, using delay of 132 msec & 0 msec ↲ ↳ inst reaction 17:54:02,711 Quantizing to 0.094906 msec 17:54:02,711 Uncalibrated response: 17:54:02,711 Black level = 0.2429 cd/m^2 17:54:02,711 50% level = 28.78 cd/m^2 17:54:02,711 White level = 105.15 cd/m^2 17:54:02,711 Aprox. gamma = 1.87 17:54:02,711 Contrast ratio = 433:1 17:54:02,711 White chromaticity coordinates 0.3044, 0.3117 17:54:02,711 White Correlated Color Temperature = 7196K, DE 2K to locus = 2.3 17:54:02,711 White Correlated Daylight Temperature = 7206K, DE 2K to locus = 6.5 17:54:02,711 White Visual Color Temperature = 7315K, DE 2K to locus = 2.2 17:54:02,711 White Visual Daylight Temperature = 7599K, DE 2K to locus = 6.1 17:54:02,711 Effective Video LUT entry depth seems to be 10 bits 17:54:02,711 White drift was 0.000000 DE
The HP one :
17:56:48,438 Setting up the instrument 17:56:48,438 Product Name: i1Display3 17:56:48,438 Serial Number: HP-14.A-02.100551.04 17:56:48,438 Firmware Version: v1.01 17:56:48,438 Firmware Date: 12Apr11 17:56:48,438 dispcal: Warning - new_dispwin: Expected VideoLUT depth 11 doesn't match ↲ ↳ actual 10 17:56:48,438 Measured display update delay of 20 msec, using delay of 126 msec & 0 msec ↲ ↳ inst reaction 17:56:48,438 Quantizing to 0.095090 msec 17:56:48,438 Uncalibrated response: 17:56:48,438 Black level = 0.1747 cd/m^2 17:56:48,438 50% level = 27.25 cd/m^2 17:56:48,438 White level = 99.81 cd/m^2 17:56:48,438 Aprox. gamma = 1.87 17:56:48,438 Contrast ratio = 571:1 17:56:48,438 White chromaticity coordinates 0.3038, 0.3118 17:56:48,438 White Correlated Color Temperature = 7234K, DE 2K to locus = 1.7 17:56:48,438 White Correlated Daylight Temperature = 7244K, DE 2K to locus = 6.0 17:56:48,438 White Visual Color Temperature = 7320K, DE 2K to locus = 1.6 17:56:48,438 White Visual Daylight Temperature = 7601K, DE 2K to locus = 5.7 17:56:48,438 dispcal: Warning - Unable to determine effective Video LUT entry bit depth 17:56:48,438 White drift was 0.000000 DE
Does it look normal for you ? The monitor is set to 65K color temp
Just wondering if I can rely on these devices.
It’s about the 2nd point I’m not sure, I created an icc profile with the 3D lut creator (without VCGT applied, cause I already installed it globally) and tried to use it in gimp, that seems to work, but I don’t understand why the VCGT is an icc file and the LUT is also an icc file… They look absolutely different.
VCGT is to fix grey, more likely to be grey color issues rather than gamma but it can do both.
LUT3D is to transform from a source colorspace (sRGB) to display colorspace, so display recieves a set of RGB values different from RGB values in a sRGB image but that values in display represent the same color coordinates as in image
Is it better to create a 3D LUT with applied VCGT (and disable the videocard one) ?
Without if you wish to correct grey color system wide. IDNK much about GIMP LUT3D for GIMP was pretty limited in truncation causing banding in the past, so better to fix grey on GPU if your GPU has high bit depth LUTs and dithered ouputs. Then assuming a very good display that can be described with 1TRC in profile LUT3D transformations should be smoother reducieng the chances of banding. Anyway if GIMP engine has no temporal dithering some banding may appear.
Is it harmful to calibrate with a response curve that’s not the display’s default ?
Depends on your GPU capabilities, explained above.
I guess using the display’s native color gamut is the best thing to do when calibrating ?
Screens i use : DreamColor Z27x, Z27xG2 and a ZBook 15 with dreamcolor panel
Calibration probe : I1 DisplayPro (HP labelled)
Best regards
THere is an SDK to upload a lut-matrix-lut to display that performs basic colorspace simulation in that Z27x. It is not integrated in Argyll ecosystem and it seems that it will not (same for Del SDK).
If you know how to code in C++, make a simple “slow” calibration in DisplayCAL on one of the slots with HW calibration (reset to factory values first).
The set a simple target like gamma 2.2 and D65 white an a simple singlecurve+matrix profile. At the end of teh process you’ll get an ICM profile but also an intermediate product, a .cal file with grey correction.
Extrapolate that .cal correction (1table per channel) to 1024 entries instead of 256.
Now upload using SDK
prelut 3x1024x16bit linear (input=output)
matrix: 3x3x16bit identity matrix (native gamut)
postlut: extrapolated calibration for grey (3x1024x16bit values)
That will make that HW cal slot behave like it was calibrated at full native gamut, gamam 2.2, D65 white.
Now your software LUT3D for GIMP can be much simpler.
You can make a HW cal to simulate sRGB somehow accurately too (it’s not a LUT3D, just a lutmatrix), but maths are mor complex. Pre lut shoudl degamma to linear, matrix should be specific for your display to sRGB, and post lut shoud combine (multiple item by item) your extrapolated grey calibration with re-encode in display gamma. More complex… and you shoudl do the maths instead a simpel extrapolation fro ArgyllCMS data. Try native gamut first.Last, what’s the best colorimeter/sprectrophotometer for a CG/Video studio ?
IDNK Z27x backlight. Your i1displaypro + some cheap spectrophotometer (cheap = $1000 Xrite, not Jeti or CR) will do the job. Better to use CCSS corrections made at 3.3 nm and share them with this community.
You may even skip buying spectro: If it is a GB-LED, Dell U2413 correction will be fine. If it is a WLED PFS ( 99% adobeRGB + P3 flavor) use the bundled one for HP Z24x. They are 1nm and you’ll need a very expensive spectrophotometer to make it better .. and if you had such budget, you’ll have bought an Eizo CG279X and use color navigation.Also your log reports are without VCGT applied (=no GPU calibration, just what you did on OSD). Also they are not true (not “real” coordinates) if you did not correct your colorimeter with one oth the suggested CCSS bundled with DisplayCAL.
Not sure which one for laptop. If about 99% adobeRGB +98% P3 try the CCSS for the HP Z24x.Thank you so much for your long reply, that really helps. So one thing sounds strange to me, the Z27x display has a Rec709/BT1886/D65 preset, I was expecting the calibration device to detect a ~6500K white, I get ~7000K in fact. Is this something “normal” or is there an issue with the calibration device (knowing that it’s about 5-6 years old) ?
I did not know about the Z27x API, that’s great ! I know well C++ as I’m the lead programmer of our in-house lighting/rendering suit. I’ll look into that, thanks !
I get an information from HP that Z displays use KSF technology (blue LED with a different wavelength with red and green phosphors).
About GPU, I have a Nvidia Quadro K2100M, how to know the rendering bit depth ? Is it a good enough ?
Thank you so much for your long reply, that really helps. So one thing sounds strange to me, the Z27x display has a Rec709/BT1886/D65 preset, I was expecting the calibration device to detect a ~6500K white, I get ~7000K in fact. Is this something “normal” or is there an issue with the calibration device (knowing that it’s about 5-6 years old) ?
CCT, correlated color temp, is not a valid way to address white. It can be 7000K CCT and look white and at the same time other display 6500K could look green por pink (non white). Color “tint” (excluding brightness) is a 2D surface, you provided just 1 coordinate.
Also you need to use proper correction for colorimeter, otherwise measured values are meaningless
I did not know about the Z27x API, that’s great ! I know well C++ as I’m the lead programmer of our in-house lighting/rendering suit. I’ll look into that, thanks !
I get an information from HP that Z displays use KSF technology (blue LED with a different wavelength with red and green phosphors).
That points to HPZ24x WLED PFS correction bundled in DIsplayCAL.
About GPU, I have a Nvidia Quadro K2100M, how to know the rendering bit depth ? Is it a good enough ?
Tools -> Report uncalibrated display, look for results in measured video LUT entry. Do not care about actual number, just if >8 or not. It is just a test to check if limited to 8 or not.
BTW, what’s the model of the $1000 Xrite device you’re talking about ? I found the i1 display studio, but seems it has a spectral reporting of 10nm.
That is 160 euro colorimeter, more limited that yours in speed & software support.
i1Pro2. There should be models in some stores since it is discontinued because of i1Pro3. i1Pro3 is not supported by displayCAL/ArgyllCMS. i1Studio may work too.
Device reports 10nm with its vendor driver but argyll managed to get 3nm readings from it with a custom driver.
If HP Z27X is a blue led + KFS you do not need to buy an spectro (a priori), use bundled correction for HP Z24x G2.
Very interesting, does it mean that I should use the “as measured” white point, or better force 6500K, though ? Sorry, I don’t understand the CCT theory.
Looks like full custom calibration is no longer supported!
Have been years since I checked it.Very interesting, does it mean that I should use the “as measured” white point, or better force 6500K, though ? Sorry, I don’t understand the CCT theory.
No. Use 6500K AND daylight as target if you aim for D65.
I just said that “6500K” alone is meaningless. Displaycal reports show “assumed white” (blue-yellow position, by K value) and distance to actual white curves (if it is pink or green). It’s explained in wikipedia.Looks like full custom calibration is no longer supported!
Have been years since I checked it.Forget what I said. You can put manually data in files, them upload to monitor.
I do remember that in the past you can just use a C API to send prelut-matrix-postlut data directly…
Thank you so much for these precious info ! Last thing, I’m gonna generate 3d LUTs for my OpenColorIO configs, but a bit lost about ICC profiles, as far as I know, there are 2 types, one for the display card (VCGT) and another one for applications. Is the last one is installed on a specific directory and the app knows how to handle it, or do I need to load it from a file ? (I use Krita and Gimp)
Thank you so much for these precious info ! Last thing, I’m gonna generate 3d LUTs for my OpenColorIO configs, but a bit lost about ICC profiles, as far as I know, there are 2 types, one for the display card (VCGT) and another one for applications.
No. There are 2 (simplification) types of display profiles:
“idealized”: you believe taht display behaves in an ideal way, so you describe display colorspace just by ther primaries coordinates and TRC.
“detailed”: You do not believe that display is ideal so you map actual display response RGB input -> XYZ coordinates for several inputs (for example a cube with side N covering NxNxN RGB values)
Usually you put a detailed one for LUT3D input but if display once calibrated behaves very well you can use an idealized one too.
VCGT is an optional tag inside display profiles that stores grey calibration for graphic card. For ICC profiles resulting from HW calibration it stores linear ramp (input = output for each channel => DO NOT calibrate grey un graphics card)
Is the last one is installed on a specific directory and the app knows how to handle it, or do I need to load it from a file ? (I use Krita and Gimp)
Krita & GIMP should be able to find OS default profile for each display => You just need to setup color management on OS. GIMP/Krita will ask OS for it.
They may need to configured them to ask OS. Check your apps configuration.IDNK how GIMP manages LUT3D but beware not making color management twice. It may need to disable color management on GIMP (no display profile) in order to use LUT3D as correction.
Also if you manage to load cistom calibrations with HP SDK you may not need GIMP/Krita LUT3D at all. Just set as display profile in OS the colorspace target you aim for, enable colro management in GIMP and if colorspaces match (image-display) no transformation will be used (you can even turn color management OFF (only if colospaces match).
