DaVinci Resolve profile for MacOS with Dell U4021QW

[Nick C]

Hi,

I have been struggling a bit to find the best way to manage my color workflow with DaVinci Resolve on a MacOS with a Dell U4021QW monitor.

The system is a MacMini M1 with just the main monitor attached (Dell U4021QW), running latest ArgyllCMS 2.2.1 and latest DisplayCAL 3.8.9.3.

I’ve done a monitor calibration first for the U4021QW with Custom mode (monitor color space set to native, basically)  and got a decent result using LCD PFS Phosphor WLED IPS, 99% P3 (MacBook Pro Retina 2016) spectral correction. It’s around 94% P3 coverage and 99.7% sRGB after an extended verification @6500k.

The profile is made with BT.1886 and 100cd/m2, for general use on the Mac–I know probably sRGB/g2.2 would’ve worked for most apps, but since the work I do is mostly for YouTube, I’m interested in having it good for Rec.709 content and creation/editing.

Really happy with the results, however, I am now struggling to understand the process for the DaVinci Resolve’s Color profile/3D that I need to generate. I followed the Wiki but that leaves a bit of discussion around the specifics:

1. Since I will be using the profile already generated for the system, on the Calibration tab, the Resolve template has everything set to “As measured” except for Whitepoint, which has specific coordinates–xy for D65. Is this correct or should I also set Whitepoint to “As measured” ?
2. on the 3D LUT tab, “Create 3D LUT after profiling” is checked/ON, source colourspace is Rec709 and Tone curve is Rec.1886 g2.4 Absolute. Is this correct?
3. “Apply calibration (vcgt)” is checked/ON and also “Inverse device-to-PCS” is checked/ON. Is it OK?
4. Finally, on the verification tab, I have “Simulation profile” set to Rec709 and also “Use simulation profile as display profile” is also checked/ON. Is this OK?
5. Tone curve is set to Rec.1886 g2.4 Absolute and “Device link profile” is checked/ON. All good?

Please keep in mind, I want the 1D/single curve+matrix profile created previously to be used system-wide for all apps, and have this new 3D LUT created for Resolve only.

Since MacOS doesn’t support XYZ LUT + matrix to be set system-wide, maybe a Mac-specific section of the existing Wiki would clarify things a bit.

Thank you!

[Nick C]

Also notices that there’s a significant shift in image saturation and hue when Resolve has “Use Mac profile for viewers” option enabled. Should this be disabled when paired with the 3D LUT generated using the settings outlined above?

From what I could tell, the issue comes from the “Apply calibration (vcgt)” setting when creating the 3D LUT for Resolve. I still cannot figure out how this is interacting with the other settings in Resolve to properly get a good result. 🙁

[Vincent]

As said on other messages you should not have grey VCGT calibration applied on your system and at the same time apply that grey correction in LUT3D.
Choose one or another, do not apply it twice.. Keeping global VCGT grey calibration is useful to use other color managed apps at the same time as Resolve but its up to you

Verification:
-no additional param => check if profile matches display, typical c olor managed check.
-simulation device => check if in a color managed app that display + custom profile can show those colors (example how many and which colors my 86% AdobeRGB display cannot show in an AdobeRGB image)
-simulation device  + use it as display profile =>check if display WITHOUT ANY GPU calibration matches simulation prpifle (typical for factory or HW calibration check)
-simulation device  + use it as display profile + device link => check if display without any GPU calibration but with a LUT3D equivalent (a device link ICC) matches simulation profile (verification of a LUT3D through its ICC equivalent , a device link)

• This reply was modified 2 years, 5 months ago by Vincent.

[Nick C]

Thanks for the info, Vincent.

In this case I will load the 1D profile (single curve+matrix) and create a new profile for resolve to be used in conjunction with the system one, but without the “Apply calibration (vcgt)” checkmark.

At the moment I am happy with just the simple display calibration I’ve done that is applied system-wide for MacOS, but wanted to see if I can get better results from Resolve’s CM with the help of this 3D LUT.

[Nick C]

Ok. So, I’ve tried multiple calibrations because the results didn’t make sense to me (keep in mind that might be the issue, as I’m just wrapping my mind around how CM works in Resolve + MacOS).

After recreating the monitor calibration profile from scratch (single curve + matrix) I’ve proceeded to follow the Wiki article for Resolve, with the “Apply calibration (vcgt)” checkmark OFF.

In Resolve I’ve tried:

1. “Use Mac Color profiles for viewers” unchecked/OFF and no LUT loaded for viewers. Then I’ve exported a clip tagged with Rec.709. What I see in the Color viewer in Resolve is exactly what I get in exported clip played in MPV (since QuickTime has the whole game shift issue).
2. “Use Mac Color profiles for viewers” unchecked/OFF and the new LUT loaded for viewers. Color viewer is desaturated compared to output.
3. “Use Mac Color profiles for viewers” checked/ON and the new LUT loaded for viewers. Color viewer is hue shifted and desaturated compared to export.
4. Default Resolve for Mac, i.e. “Use Mac Color profiles for viewers” checked/ON and the no LUT loaded for viewers, hue shift in viewer compared to exported file.

Something doesn’t seem right as the only settings that seems to work fine is no LUT loaded with “Use Mac Color profiles for viewers” unchecked/OFF.

Any suggestion is appreciated.

Thanks!

• This reply was modified 2 years, 5 months ago by Nick C.

[Vincent]

You should be using MPV with color management… otherwise it’s a no go.

Anyway, if desaturation isn presnet best to check if it is actually desaturation or that you were seeing not color managed in a P3 display. Just load 100% saturation Red and green pacthes (AVS Forum Rec709 free MP4 patches) and measure them in resolve with Argyllcms free measure. Then compare CIE xy to what it should be in Rec709. Same in your MPV setup.

I mean , IDNK if your MPV is properly configured. If you doubt… measure.

• This reply was modified 2 years, 5 months ago by Vincent.

[Nick C]

MPV out of the box:

• Green in MPV
  • Result is XYZ: 27.322684 69.491697 7.241761, Yxy: 69.491697 0.262576 0.667829
• Red in MPV
  • Result is XYZ: 48.927651 22.146011 0.311034, Yxy: 22.146011 0.685408 0.310235
• Blue in MPV
  • Result is XYZ: 18.082114 7.906125 101.355643, Yxy: 7.906125 0.141994 0.062085
• White in MPV
  • Result is XYZ: 93.925887 99.396604 107.981430, Yxy: 99.396604 0.311731 0.329888

Resolve with no LUT loaded and “Use Mac Display color profiles for viewers” OFF:

• Green in Resolve
  • Result is XYZ: 27.393799 69.636840 7.261198, Yxy: 69.636840 0.262665 0.667711
• Red in Resolve
  • Result is XYZ: 48.928690 22.145383 0.311713, Yxy: 22.145383 0.685412 0.310221
• Blue in Resolve
  • Result is XYZ: 18.079196 7.911821 101.355750, Yxy: 7.911821 0.141968 0.062128
• White in Resolve
  • Result is XYZ: 93.876849 99.364521 108.167442, Yxy: 99.364521 0.311460 0.329667

3D LUT in Resolve:

• Green in Resolve
  • Result is XYZ: 35.211633 70.669495 12.260299, Yxy: 70.669495 0.298046 0.598177
• Red in Resolve
  • Result is XYZ: 40.114371 21.139436 2.580510, Yxy: 21.139436 0.628414 0.331161
• Blue in Resolve
  • Result is XYZ: 18.802173 8.266511 93.723690, Yxy: 8.266511 0.155657 0.068436
• White in Resolve
  • Result is XYZ: 93.852627 99.322696 107.794667, Yxy: 99.322696 0.311834 0.330009

Apart from the whitepoint coordinates I cannot find the RGB point coordinates xy for Rec.709. Also, keep in mind the system has an ICC profile loaded that I guess impacts both MPV and Resolve out-of-the-box. But the results are exactly the same unless I load a 3D LUT in Resolve.

Now I just need to figure out which is “correct”…or at least closer to reference on this monitor.

• This reply was modified 2 years, 5 months ago by Nick C.

[Nick C]

SO, Rec.709 illuminant-relative references are:

• White
  • 0.3127  0.3290
• Red
  • 0.6400  0.3300
• Green
  • 0.3000  0.6000
• Blue
  • 0.1500  0.0600

So, Resolve with a 3D LUT is actually correct with the output matching the Rec.709 reference.

In this case, does grading with the LUT on ensure more compatibility with most devices that support Rec.709 out there and most YouTube content?

I’m a bit baffled by this result as on my monitor now after calibration the desaturation is visible..but the probe says those are close to reference colors.

Is the ICC profile ( wp 0.3127  0.3290 + BT.1886) messing the GUI of Resolve and the output in MPV?

[Vincent]

You need to pass MPV a command line configuration to display ICC (or in home folder you may set it, I do not remember)
Read MPV docs to load display profile and make it color managed.
Red with x=0.68 is P3 red => no color managed at all, raw RGB values output to mac/secondary display with a native red close to P3 red.
Hence MPV was not properly configured.

If you still have doubts create 2 images in Photoshop, one P3, the other sRGB.  Put a big red 255 patch in each one. sRGB red is more desaturated that P3 red. x= 0.64 vs x=0.68.
Same testing with argyllcms commandline spotread or equivalent friendly tool for free measures.

(spotread needs CCSS correction too when used with an i1d3, “spotread -X path_to_p3_mac.ccss”)

LUT3D in resolve expects that display behaves exactly as described by destination colorspace, hence resulting TRC (Gamma) with VCGT loaded in your system must match whatever you put as destination profile TRC.
You can test the same way, AVS Forum, download Rec709 21 step grayscale with 5% increment in IRE samples. Put them in resolve and measure.
Anyway, if using a LUT3D without VCGT applied inside better keep system wide ICC close to native gamma (2.2/sRGB) since macOS desktop is color managed and it will be pointless loosing unique grey levels for nothing, Then on LUT3D maker choose whetever TRC you want in your video viewer (2.2/2.4… rec1886 is not useful with a limited contrast display)

• This reply was modified 2 years, 5 months ago by Vincent.

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

[Nick C]

Thanks again for all your patience, Vincent. It is much appreciated.

For MPV I know it has a .config file somewhere where I can tweak all params, so I’ll look into the one responsible for ICC loading.

For the 3D LUT creation I’ll use the Resolve template where I switch to “Current” after I’ve done the system ICC calibration profile and set it to have “Apply calibration (VCGT)” OFF when creating the LUT. This was suggested in the Wiki article comments somewhere.

Since on Mac I cannot have a 3D LUT as a system profile, I’ll have to load the VCGT from the ICC profile and hopefully that plays well with the LUT loaded in Resolve.

Since the Dell U4021QW is my only display for the Mac Mini, is it ok if I have it calibrated to use native gamut and a BT.1886 curve with D65?

Or should I follow your suggestion and calibrate to Native + gamma 2.2?

I couldn’t notice any banding in the gray tones with a BT.1886 profile, so I’m wondering if maybe if resolve is going to show me proper BT.1886 via the LUT, I might use g2.2 for all other apps which would help with normal usage (web, photos)

[Vincent]

Since the Dell U4021QW is my only display for the Mac Mini, is it ok if I have it calibrated to use native gamut and a BT.1886 curve with D65?

Or should I follow your suggestion and calibrate to Native + gamma 2.2?

It does not matter in a color managed enviroment, it is just… useless: there is no gain doing that. Why? No matter the TRC of your display, in a color managed enviroment the engine will transform from image colorspace to display colorspace (whetever it is and whatever TRC it has, as long as it is not out of gamut).
Hence, calibrate to target white and close to native gamma (or to most common image colorspace TRC or the most probable mean TRC for rall kinds of colospace you are going to use -> 2.2 is usually a good guess). Then make a LUT3D to whatever you want.

I couldn’t notice any banding in the gray tones with a BT.1886 profile, so I’m wondering if maybe if resolve is going to show me proper BT.1886 via the LUT, I might use g2.2 for all other apps which would help with normal usage (web, photos)

Safari is color managed so it won’t care abot display TRC and long  as default display ICC refects that TRC.
Color managed apps can have limited precision so with an 8bit pipeline without dither Firefox will show a mild banding on an sRGB blank to white gradient on a perfect 2.2 TRC display (with 2.2 as TRC in ICC), you cannot map 256 tones to 256 tones with different TRC unless you have more bits or dithering.

[Nick C]

OK. In this case I’ll follow your advice and recalibrate from scratch to Native gamut + g 2.2, which is the default for this Dell, as a system-wide ICC.

Then I’ll generate the LUT3D with no VCGT applied but based on the “Current” template but with TRC targeting BT.1886 for Resolve and MPV to use.

Just to be sure I don’t mess up an hour of calibrating again, in the “Video for 3D LUT for Resolve” template, I should leave all things as “As measured” for the Calibration tab and only focus on the 3D LUT tab, where I just uncheck the default “Apply calibration (VCGT)”, since it will be applied by system ICC. All else remains as default. Correct ?

Also, as a sidetone: for spectral calibration I’m currently using the Panasonic provided correction as it seems closest to what this display is actually using. The DisplayCAL database also has 2 other corrections that are lower resolution and judging by eye, the Panasonic is closest to it. Dell, DELL U4021QW (i1 Pro 2)

Are these corrections from the database more fitting than what I’m using despite the lower res?

• This reply was modified 2 years, 4 months ago by Nick C.

[Vincent]

First you need that your desired desktop calibration is active (VCGT loaded your desired ICC for desktop). Then when you set calibration as measured and button changes to “profile only” make sure you choose to embed current calibration curves (current VCGT). The result will be a detailed ICC (XYZLUT) but with the same  calibration curves as your desktop (so LUT3D without embeded VCGT will be accurate)

You can make LUT3D from main window or with standalone app in DIsplayCAL folder. So if you want later a LUT3D Rec709 2.4, you just need to feed the same XYZLUT ICC as destination and change source to 2.4.

Yes, Panasonic VVX is teh closest one. There is a tiny difference in green and blue but it’s hard to say if it is caused by device noise or if it is a little different. A well behaved i1d3 should output almost the same whitepoint with any of these two. If you see difefrent whitepoints, choose the one that looks visually whiter.

• This reply was modified 2 years, 4 months ago by Vincent.

[Nick C]

Thanks. I’m running a ICC display calibration now with Whitepoint set to “As measured” (basically using the “Default (2.2)” template of DisplayCAL and set the monitor controls to as close to D65 as I could get them. Will upload the measurements results after they’re done.

Also thinking of testing the accuracy of Office & Web template that also specifies the 6500K target for Whitepoint to see how that behaves on this monitor.

From my understanding it shouldn’t make a difference in the resulting LUT3D if I follow the steps you outlined, right?

[Nick C]

Results from the “As measured” whitepoint setting.

Attachments:

You must be logged in to view attached files.

[Author]

Posts