Hello,

This is mostly a cross post with Lift Gamma Gain about my ongoing learning with the C3 LG WOLED TV. I thought that this forum might be a more appropriate location for this post.

I am using the LG as a secondary 4K monitor. My primary monitor is an FSI DM240. I do design and animation work for Cinema and my stuff always goes through a final DI. I do need to show up to DI with my colour choices intact, though.

I've worked through the LG guides, have done the SM white balance, turned off all the auto stuff, adjusted the brightness with a pattern, and set my OLED light levels to 124Nits at 109% extended white.

This is where I ended up. I does not look ideal to me: Dif EOTF is not very linear and there is separation between the colours, RGB Separation shows blue dipping outside the linear graph, rgb balance shows deviations in the low end, and - this is the big one for me, Greyscale is has a dE of around 2 in the low end.

When I look at the manual measure page with the verification pass loaded, it is giving me an average dE2000 of .63 but an EOTF dE average of 2.41

The Characterization and Verification were done with an i1D3 using the WGRB OLED (LG C3) Cor. File, Intelligent Int.@ 1.00, Integration time of .75, PGEN set to RGB Full, SDR, 1 sec Stabilization, 10 BIT extended for characterization/10bit legal for verification (using 10bit LG optimized patches),

Here is a dropbox link to my CS files for the characterization and verification pass:

https://www.dropbox.com/scl/fo/jtctf1egoqcur99ndajht/h?rlkey=cl3gd9rkqwy9jlskpx8betjdx&dl=0

I would be grateful for any input you might have.

All looks basically as expected.
The EOTF number is the EOTF average value, not a dE value.
You targeted Rec709, so EOTF 2.4 - 2.4107 is therefore a good result.
The RGB Separation is also about what you would expect for the LG TV.

Steve

It looks like you used a very bad combination of Contrast/Color values, your colors are clipped during the profiling.

Which contrast/color and brightness values have you used?

Thanks for your input Steve and Ted. The Contrast is set to 85, the colour is at 50, and the brightness is at 54. OLED pixel brightness is 31, User WB is Green+1, Blue-5. I did set the SM White Balance using their "100 IRE" pattern and I set it to 100nits for D65.

Ted, I just had a thought. I did not set the icon of the Pgen HDMI input to PC as you had mentioned in another post. In that post you stated that setting the icon to PC would ensure that the TV would process full chroma. Would this have caused the clipping errors in the characterization?

So I imagine I will need to reduce the contrast a tiny bit. I checked with the contrast chart and it looked ok, but for arugument's sake, I'll reduce it a couple units to avoid any clipping issues at the top. I don't see the clipping at all in legal range verification, just at the very top of the extended range profile.

Edit: I had asked a question about adding more greyscale information to a profile. I just read about the augment feature. I'll give that a try.

justints


BR54 is not a value that is usually correct unless your panel has a specific issue in near black that will require that adjustment.

You have to evaluate the near black in a totally black room, with all OSD OFF, your notebook screen OFF, and look at a close distance the TV.

The 16 and below bars should be invisible, and the background of the pattern should be retained in OLED black, not lift your black level.

You have to use a 109% 10 area window patch, not the IRE 100 internal patch if your TV.

Please follow the instructions of the LG Guide.

You need to enable the PC icon and repeat the whole process to prepare the TV and restore initial settings from TV menu, then CS or DC, upload reset 3D LUT, pre-calibrate 109% White.

You have clipping because you used 100% White patch (full screen internal which is wrong also)

Please follow the instructions of the LG Guide.

Please see the links inside the LG Template email for improving RGB separation in case the panel has a problem with factory calibration.

That method will improve only if the panel has a problem; it will not make any difference if you don't have that problem.

Hi, Ted.

Thanks for your feedback. I did previously work through the guide in some detail. When I calibrated the SM WB to the 109% patch from PGEN in the SM (D65/124Nits), the results were quite strange when went into User Mode: the WB in the UM was off quite far, the OLED light needed to be cranked up, the brightness was way off etc. You had mentioned in a previous post that you no longer recommend SM white balancing so I was trying to just get back to a factory-ish SM WB balance and then do everything from the UM.

Perhaps I will work through this workflow again from the guide to see if I missed anything.

I wonder if there is anything different with the 2021 models that I should know about?

The point of pre-calibration using SM or regular menu is to have 109% White pre-calibrated for 123.4 nits while you have all OSD menus OFF while you use 10% area patch.

The procedure is the same for all LG models; nothing has changed.

If you use a color temp in SM that does not match the color temp of the regular menu, then the measurements in SM WB will not match the measurements with OSD OFF.

I have worked through a number of things with the LG C3 and have arrived at what appears to be a pretty good calibration. Here are a few of my findings. I am sure they will contradict what others are doing, but this is what has worked for me. I am attaching my verification pass below:

- when setting up the SM WB, it's important to reset the UM WB, picture mode and load the unity lut. The SM seems to show the TPG through the picture mode settings.

- when going back to the UM from the SM, I found that you need to reset and upload the unity LUT again. When I do, the SM WB readings carry over to the UM.

- I found that going back and forth between the UM and SM to make sure that the OLED light level is the same in both menus makes for more consistency between the SM and UM whitebalance and white level.

- Anything from 100-109% tests very poorly for me. I have everything setup as the guides indicates and still get big jumps in the EOTF above 100%. I am not worried about it, though, as I am using this C3 in post production and am only concerned with video levels - I never use "super white".

- for whatever reason, I have had better success with 8 bit patches. I ensured consistency between the patch sets and the settings. I found that the 10bit patches gave me inconsistent results.

- When making and uploading LUTs, I had a lot of issue with consistency. For example, I would make a LUT from the same profile and end up with different results.

I found opening a new LUT tools window when generating a new LUT was helpful - even though I had "New LUT" selected. I also found that when uploading the LUT to the LG TV I had more consistent results resetting, uploading a unity lut and then uploading the calubration LUT. There is likely something I am doing incorrectly here or there is something inconsistent in the way the LG deals with uploading LUTs. I really chased my tail at this stage of the work. As an example, when creating an "Augmented Data" version of the Calibration LUT, I got more consistent results when I opened a new LUT Tools window to generate the LUT AND uploaded a Unity lut before uploading the calibration LUT to the LG.

- Normalize Data caused more issues in the 0-100% range than not using Normalize Data.

- I seem to get the most consistent results with an L20 patch.

- Using a "greyscale large" profile to create an "augmented" profile is an absolute game changer. Full stop. My calibrations went from being meh to being pretty good when using the Augment Data functionality. It's very powerful and effective.

With these notes above, I did what looks to be a pretty decent calibration which I am attaching here. It's not perfect and am sure you'll be able to point out issues, but it seems to be workable, I think. Even without doing a perceptual match to my FSI, the LG now looks very close to it. I see this clearly in highlights and mid tones. The shadows are obviously different, but not jarringly so.

The verification reports a dE 2000 average of .5064, an EOTF of 2.3971 and a pretty linear looking greyscale dE.

Attached file: verification March 23 8Bit w.zip

I have been working more on this to try and get thing running consistently so I am posting one more update.

I have a few new findings:

- based on the LI tutorial video about probe matching, I did a raw gamut probe match and a rec 709 probe match. I get better matching and readings in Rec 709 with a probe match created in Rec 709.

- Contrary to my previous post about normilization, upon further testing, I get more repeatable results and less surprises when I use normailization/normal range in the LUT creation

- Augment data is done from an extended range Large GS profile through the LUT created using normalize data. I tried adding a large GS profile in normal range to the ^17 cube, but it reports it as out of range. Not 100% sure if I did this correctly.

- I did a preroll before the verification this time.

- To my eye this verification is smoother than the previous one, but the averages are a bit worse. The method produced more consistent results, though.

- The RGB separation tracks together pretty well, but is not completely parallel to the guide line. If I turn off probe matching it tracks perfectly. I likely need to work on the probe matching further.

- I think that the normalize workflow is a good one as I can get much more repeatable results with it.

Attached file: 3_27_24 Legal Verification.zip