Would LOVE Verification on settings for Resolve 3D LUT

[Kyler Boudreau]

Hey, thanks in advance! Been doing a lot of reading here and other places. Will do my best to be concise.

SETUP
ASUS ProArt PA247CV fed off a BMD UltraStudio Monitor 3G (to bypass GPU) / Display Pro HL (Calibrite)

GOAL
To create a 3D LUT for Resolve for monitoring (during grading) Rec.709 Gamma 2.4 film

WHAT I DID
Updated to latest ArgyllCMS (3.0.1)
Tools > Detect Display Devices and Instruments
Tools > Correction > Import Colorimeter… Check i1 Display Pro and hit AUTO

SETTINGS SELECT
Video 3D LUT for Resolve (D65, Rec.709)

DISPLAY & INSTRUMENT
Display = Resolve
Instrument = i1 Display Pro, ColorMunki Display
Mode = LCD (generic)
Override Minimum Display Update Delay = 1000ms
Output Levels = Full Range (matching what’s set for Resolve/Monitor)
Correction = Auto (None)

*Note: I do see a CCSS file in the database for the PA247CV monitor, but it’s for the iPro 2 colorimeter or something like that — not the new model I have. However the monitor is an exact match. Better to use?

CALIBRATION TAB
Interactive Display Adjustment = Checked
Observer = CIE 1931 2
Whitepoint = Color Temperature / 6504 / Reference=Daylight
White Level = Custom / 100
Black Level = As Measured
Tone Curve = As Measured

*Is that right? To set the target whitepoint or leave it “as measured” or do chromaticity and take a sample reading here vs later? This is something I just couldn’t sort out.

PROFILING TAB
Profile Type = XYZ LUT + matrix
Black point compensation = Unchecked
Profile Quality = High
Testchart = Auto Automized

3D LUT Settings
Check “Create 3D LUT after profiling”
Source colorspace = Rec709 ITU-R BT.709
Tone curve = Rec. 1886
Gamma = 2.4 / Absolute (Relative better?)
Rendering Intent = Absolute colorimetric / white point scaling
Input and output encoding = full range (since Resolve handles?)
3D Out Resolution = 33x33x33

Again, THANK YOU to anyone willing to check this out. I’ve done my best to try and dial this in accurately.

I’m grading a film show on the BMPCC4K. BRAW files and grading in DWG but delivering Rec.709 Gamma 2.4 (and obviously monitoring as well).

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

[Vincent]

PA247CV

White LED

*Is that right? To set the target whitepoint or leave it “as measured” or do chromaticity and take a sample reading here vs later? This is something I just couldn’t sort out.

Use D65 coords to get desired white in monitor using RGB gains. Beware destroying contrast. IDNK if that asus uses default value = 50 in 0-100 or 100 in 0-100. For 100 as default keep one channel at 100. For 50 as default do not push high above 50 without checking channel clipping (test visually with a non color managed chart like lagom’s white level opened in MS paint).

Tone curve = Rec. 1886
Gamma = 2.4 / Absolute (Relative better?)

It’s unlikely that you want rec1886 win a low ocntrast IPS display. Maybe better to aim to 2.4 relative balck outut 100%

[Kyler Boudreau]

Thanks Vincent! Much appreciated.

1) Is it better to just always use White LED (assuming that’s the makeup of the display) vs trying to locate a CCSS file that matches? I’m a little unclear as to how this is playing into the whole thing. For example, I also have a BenQ PD2700U and it has several CCSS file matches. But it’s a White LED display too. Is it more accurate to just use the White LED option? And if so, why the CCSS files?

*I’m not questioning you…just wanna understand how this is working.

2) Cool, just did the custom white point thing by hitting the RGB icon and allowing it to scan. It adjusted the values. Can you shed light on what this is doing? Well, here’s the question: If I set this to 6500K I’m assuming it would read the display, have me balance RGB/Brightness manually and then make any other required adjustments in the LUT. So when I do a custom read like this and use chromaticity coordinates…how’s that playing in? Hopefully I’m asking that right.

And btw…this display is a 50/100 contrast. And yeah, I saw things go south faster yesterday messing with that. Just leaving it at 50.

3)  This is GREAT info on the contrast for my display. So if I had a 2000:1 contrast display I could do 1886, but for this 1000:1 display I should do: Tone Curve Custom / Gamma 2.4 / Relative with black output offset pushed to 100% – Did I get that right?

THANK YOU for your help.

Also, do you have any input on the size of the LUT? For monitoring HD / Rec.709 is a 17 point LUT fine, or do I need 33? And what’s the purpose of 65? Thx.

• This reply was modified 2 years, 9 months ago by Kyler Boudreau.
• This reply was modified 2 years, 9 months ago by Kyler Boudreau.

[Vincent]

Thanks Vincent! Much appreciated.

1) Is it better to just always use White LED (assuming that’s the makeup of the display) vs trying to locate a CCSS file that matches? I’m a little unclear as to how this is playing into the whole thing.

sRGB only is very likely to be White LED

For example, I also have a BenQ PD2700U and it has several CCSS file matches. But it’s a White LED display too. Is it more accurate to just use the White LED option? And if so, why the CCSS files?

Those sRGB only PDs from Benq are exception, they are PFS phosphor but limited to sRGB.

*I’m not questioning you…just wanna understand how this is working.

sRGB only is very likely to be White LED, blue led + yellow phosphor bc sRGB it’s all they can do, so most sRGB only modern led displays will be White LED.

Chooing a wrong correction for i1d3 colorimeter may result in an error while measuring whitepoint. Mangitude of error depends on stuff that is too complex to explain in a friday night.

2) Cool, just did the custom white point thing by hitting the RGB icon and allowing it to scan. It adjusted the values. Can you shed light on what this is doing? Well, here’s the question: If I set this to 6500K I’m assuming it would read the display, have me balance RGB/Brightness manually and then make any other required adjustments in the LUT. So when I do a custom read like this and use chromaticity coordinates…how’s that playing in? Hopefully I’m asking that right.

IDNK what you are asking with “So when I do a custom read like this and use chromaticity coordinates…how’s that playing in?”.

And btw…this display is a 50/100 contrast. And yeah, I saw things go south faster yesterday messing with that. Just leaving it at 50.

As a general rule do not modify contrastsettings on a digital TV. You can modify brightnesss at will.

3)  This is GREAT info on the contrast for my display. So if I had a 2000:1 contrast display I could do 1886, but for this 1000:1 display I should do: Tone Curve Custom / Gamma 2.4 / Relative with black output offset pushed to 100% – Did I get that right?

Google UIT-R BT.1886, you’ll se the spec and why actual “gamma” value in rec1886 depends on black level.

THANK YOU for your help.

Also, do you have any input on the size of the LUT? For monitoring HD / Rec.709 is a 17 point LUT fine, or do I need 33? And what’s the purpose of 65? Thx.

Those nodes are interpolated values. Choose default. They will be computed from profile measuements…. and they run to the 3rd power. 10 measures per channel ramp = 10x10x10 =1000. 17 measurements per channel ramp = 5000. (assuming evenly distributed patches).
Use auto optimized on a 1st run.

[Kyler Boudreau]

Thanks Vincent, much appreciated.

The question I didn’t ask clearly regarding chromaticity coordinates: As someone new to this I thought if I set it to Custom / 6504K DisplayCAL (during scan) would look at the monitor, see where it’s off and work the math into the 3D LUT to get me to a Rec.709 Gamma 2.4 accuracy.

So scanning now and going by the discovered white points of the monitor doesn’t make sense to me. Which means the above process I’m assuming must not be correct?

[Vincent]

DisplayCAL corrects greyscale, that’s all. Greys will have the same color as white and a desired target gamma.Same for Xrite or calibrite software.  Nothing more.
Greyscale correction is loaded into GPU (for typical GPUs, not decklinks). Then, with correction applied (grescale calibration) app measures display behavior and creates a profile. For decklinks & Resolve it’s the same but there is no grayscale correction loaded into GPU.
Photoshop relies on ICC info, or macOS desktop.

Resolve and other apps may use a LUT3D instead because of some speed constraints may be better suited for that task. Too long…
LUT3D creation relies on that profile with display behavior. It transforms from source colorspace (video content colospace) to destination (display) colorspace.

On a mac you can order Resolve to use Apple’s color management (ICC on regular GPU) or a LUT3D.
If you use a decklink you’ll use a custom LUT3D (typical scenario).

So:
-set your display to D65, desired brightness.
-read faq for Resolve and create a profile, with profile a LUT3D will be created. Typically  LUT3D will be from rec709 g2.4 to display custom profile.
-read faq and place LUT3D on Resolve folder, open Resolve, open your project, configure it to use LUT3D
-if you want to be sure open a 100% saturation RGBCMY Rec709 patches + 10 IRE greyscale samples (mp4 files) or run a verification report.

[Kyler Boudreau]

Vince, thanks for explaining.

I’m using an UltraStudio Monitor 3G to bypass the GPU entirely. Only reason I have Calibrite installed is for the other thread you commented on. And I’m also about to use a BMD Micro Converter 12G to load the LUT file.

So here’s my 3 questions if you’d be so kind:

1) I just let DisplayCAL do a LONG scan for a LUT. Loaded that into Resolve for the monitor and it darkened the picture a little. Is it normal for a long scan vs a shorter LUT scan to do that?

2) If DisplayCAL only corrects grayscale, then why does it show so many color swatches during the profiling? So are you saying that regardless of what it sees, the 3D LUT file that is created has no RGB data?

3) Is my understanding right on this: On the Calibration Tab if I set my Whitepoint to Chromaticity coordinates and do a scan, then DisplayCAL is taking that data and putting information in the LUT to get my display to a proper white point? Because if that were the case, what if I chose custom there and entered 6504? What would it attempt there, because that’s messing with RGB.

Thanks for any clarifying. I like to understand what’s happening, and don’t yet.

[Vincent]

Read again

DisplayCAL corrects greyscale, that’s all. Greys will have the same color as white and a desired target gamma.Same for Xrite or calibrite software. Nothing more.
Greyscale correction is loaded into GPU (for typical GPUs, not decklinks). Then, with correction applied (grescale calibration) app measures display behavior and creates a profile.

Color patches are shown afrer grey calibration, to measure display behavior and create a ICC profile storing that behavior.
LUT3D creator uses that display behavior description, like a 3D mesh (RGB <-> CIE XYZ) to compute a color transformation to another ICC profile ( behavior RGB <-> CIE XYZ). So LUT3D is just a map RGB to RGB to preserve CIE XYZ values but with different RGB numbers.

If this is too complex take a course or webinar on color management.

[Kyler Boudreau]

Not too complex, I understand color management. What I’m new to is doing my own calibration.

The second part you added this time was the gold!

So colorimeter/DisiplayCAL measures grayscale, then sends a bunch of color patches. And based on the monitor’s behavior it creates a 3D LUT to convert what the monitor can do to my desired Rec.709 Gamma 2.4.

Good? Thanks man!

[Vincent]

Not too complex, I understand color management. What I’m new to is doing my own calibration.

The second part you added this time was the gold!

So colorimeter/DisiplayCAL measures grayscale,

Then ArgyllCMS computes grey calibration (if needed, if you are using a common GPU), then ArgyllCMS loads grey calibration

then sends a bunch of color patches.

And measuring those patches you get a map “raw RGB” <-> CIE XYZ that describes display behavior

And based on the monitor’s behavior it creates a 3D LUT to convert what the monitor can do to my desired Rec.709 Gamma 2.4.

Good? Thanks man!

Yes

• This reply was modified 2 years, 9 months ago by Vincent.

[Author]

Posts