Thanks for the info. Regarding filtering, if the display being profiled has non linear tracking throughout the luminance range (e.g. Yellow's non linear tracking in the capture below), should the color engine not attempt to correct the tracking? Or would this be a case of it interpreting artifacts if a correction were attempted?
How do you modify the ENV Variable for number of cores to be used?

Yellow Tracking

All the points in the image look to be valid, and will be 'corrected' when the LUT is generated.
But I would need to see the actual profile to verify as the images are difficult to assess.
You can add as a .zip file to a forum post, or email to me.
And remember, the LUT you upload will be far smaller in the number of points it contains, so the granularity of the correction will be less.
Out of interest, what probe was in use there, as that actually looks to be a potential probe error?
And taking such a high level of granularity readings is really not good, as any instability in the display/probe will cause issues in the final calibration.

And there is a Customer Download with all ENV. Variables.

Steve

Unfortunately the points were not corrected. Probe used for those measurements was a K10-A. However, a CR-100 yields the same result. None of the measurements are probe errors.

I will email you pre and post characterizations.

Thanks for the profiles.
There are some major issues in the profiles.
(Edit: See below for the correct pre-cal profiles, as I posted the post-cal here.)
The granularity of the non-linear errors is greater than the LUT size you can load into the TV.
And if it were possible to load a LUT of the required size, generating the LUT would be next to impossible due to Memory/Processing limits on just about any PC.
(Working with a large volumetric patch set with that number of points uses far, far less memory/Processing than generating a LUT)
So there is no way any calibration can correct such errors.
The Magenta is actually worse, as you can see below, as the frequency of the errors is higher.
Additionally, the TV has poor RGB Separation, so to be able to perform any sort of accurate calibration would take a volumetric profile, as is explain in the User Guides. That may actually generate a useable improvement, even with the serious errors you have.
And if you look at the RGB Separation plot you can see non-monotonic measurements, which are basically impossible unless the display/probe combination is very unstable.
And non-monotonic issues are totally un-calibratable.
So basically, that TV is un-calibratable.

Steve

Edit: What patch set did you use, and did you use Drift and Stabilisation?
The errors look like they could be heat threshold related, so actually being introduced by the profiling?
And I would like to see a CR-100 profile too, as I have previously seen such issues when they were probe issues.

Bad RGB Separation

Non-monotonic

Bad Magenta

You posted the POST cal measurements, not PRE cal. I am using a custom patchset. What sent you were one of the pre/post validation steps I use, not the actual profile. Everything is a HDR10 characterization but I have seen very similar with SDR @100nits with this model lineup.

The actual profile has roughly 72 WRGBCMY points and is approximately 10,263 measured points and is using Mapped Space to create the LUT. I am not using Drift because I have not seen where it provides a tangible improvement via my validation methodology.

It is not a heat instability issue as I have figured out how to manage that with this model. Yes there are things going on in the video pipeline of this display. But, let's take things one at a time Why didn't the ColourSpace engine correct the yellow tracking? Attached are screen grabs from the actual profile.

Measuement Points	RGB Separation	WRGBCMY
Drift

Sorry, yes, the graphs were post-cal-cal, but the pre-cal is basically the same, other than the RGB Separation which is still not good.
See below for the pre-cal graphs.
The reason the post cal has a worse RGB Separation is that is needed to try to correct the issues in the pre-cal, as defined in the User Guides, and is normal for those TVs.
But, the same non-monotonic issues remain, as does the bad magenta.

What I mean about asking about the patch set is was it Sequential or Anisometric?
And the same with the use of Stabilisation and Drift, as using those will help with heat/thermal issues.
(Drift is really no for use with LUT generation, but to show what happens during profiling - although it can be used for LUT generation if really necessary.)

But to answer your specific question, it is as stated in my previous post.
The granularity of the LUT in the TV is below the granularity of the issues the display has.

It should be noted that the issues you can see in the graphs would show as banding in images, as they are 'step changes' in the TV's response to small input signal changes, and banding is not something a 3D LUT can really correct, for the reasons defined.

So basically, the TV is unfortunately un-calibratable, as I said above.

But I would love to see a profile using a different probe, as I have only ever seen such issues when the probe has been the cause.
Out of interest, have you profiles when the TV is in a smaller gamut. The previous time I have seen such issues it was only in wide gamut.

Steve

non-monotonic

Magenta

RGB Separation

Thanks for the response. However, I think there is some misunderstanding going on. The pre cal file is sent you was the state of an LG G5 OLED using the "standard" HDR10 calibration documented in the LG calibration guides used by many, including members of this forum; grayscale only calibration, paired with the factory 3DLUT. The pre cal is not of the display with unity LUTs loaded in the LG's 3DLUT registers. 3DLUT unity files would be seen in post#8's screen grabs.

I am aware that the "zig-zag" behavior of yellow, cyan and magenta will cause visible banding. I wanted to see if the ColourSpace engine is capable of smoothing or completely correcting the "zig-zag" behavior. I guess this is not possible with the ColourSpace software- including using the manual point adjustment feature, correct?

Regarding patchset, it is not Sequential as that would not make sense to use on this display. It is OFPS.

Magenta's tracking on the luminance top is how ColourSpace's Mapped Space option adjusted the characterization for Rec2020 @ 2.2 gamma.

No, I do not have profiles of this display in anything except its native gamut and its factory 3DLUTs for HDR10 and Dolby Vision.

Yes, I am now rather confused...
Are you using a 3D LUT for calibration?
(The rest is really irrelevant.)

If so, the LUT size as used by the TV is below the granularity of the error, so will not be capable of correcting it. This is not directly associated with ColourSpace. No calibration will correct such errors.

As said, you would need a 73^3 LUT minimum to get close to that.

As for the Magenta at peak, did you use Gamut Mapping?
That is what that looks like.

Steve

Edit: I've moved the discussion to its own topic.

Hi Steve,

Yes a 3DLUT is being used in the post cal file I emailed you. Post #8 has screen grabs of a 10,263 point characterization to create the HDR10 3DLUT file on the LG OLED. The HDR LUT size in the display is 33^3 like it is for SDR. Yes the "Disable Gamut Mapping" option was left unticked for this particular session. I would need a 73^3 cube to get close to what? The results I posted or a working correction for the "zig-zag" behavior of Yellow?

Yes, to fix the banding issues the step-change causes would need at least a 73^3 LUT, if not double that, as it would likely need to have double the granularity (Nyquist) of the issue.

That is why I say the TV is unfortunately basically un-calibratable..

Steve

DeWayne - I thought I'd try top explain the issue in more detail.

Look at the attached screen grab.
This is the Cube 3D view of the profile.
The arrows show the 'step-change' issue.
To correct such a step-change the LUT would need to have points exactly either side of the step-change.
So that would mean having at least 73 points, and due the different step-change errors not being equally spaced, probably far more points.
This is actually separate to the likely need for the LUT to have Nyquist sampling, so means the LUT would probably need to be 3 or more times bigger.
Looking at this graph, the LUT would likely need to be 256^3 minimum.

Does that help?

Steve

Step-Change error

Trying to use a 3DLUT on a LG TV for HDR, not something you read about every day.

I'm curios:

• How much the white pixel is to blame for the issues?
• Does changing the mode manifest the same issues?
• Does SDR have the same issues on the same TV?

The older FSI monitors used the same WOLED screen, and they were always calibrated with 3D LUTs.
There really is no issue with the approach.
However, TVs vs. monitors tend to have more issues with heat/thermal related variations, so can be more difficult to profile for LUT based calibration, especially if the probe is slow. Getting a good patch set, with Stabilisation/Drift patches, as well as a good pre-roll is critical.
FSI monitors, as with all professional monitors, have good heat management engineered into them.

The white pixel is really no issue, at least with ColourSpace.

Steve

DaniJ
Use of 3D LUT with HDR is generally discouraged, which is why you don't find many references to it.
It was reported by Ted some time ago that the white subpixels in WOLED are always active to one degree or another, so some color dilution is constant.
3D LUT generation requires many dozens or even thousands of readings depending on the degree of correction required. This pushes the set into thermal instability, making those same readings and the corrections calculated from them not trustworthy.
Our TVs really aren't capable of performance to the HDR standard, having neither the gamut nor brightness levels required. Tonemapping is an attempt by manufacturers to compress HDR output down to a luminance range their TVs can actually produce. No LUT can change this. SDR, on the other hand, is well within the capability of modern displays and as such is much better suited to correction with 3D LUT.

This has been discussed before, and can indeed work.
https://lightillusion.com/forums/display-calibration-8/3d-lut-calibration-for-hdr-mode-of-lg-g5-1383.html

Steve