AVS Forum banner

1581 - 1591 of 1591 Posts

·
Registered
Joined
·
34 Posts
Pre-Calibrate/Profile with 16-255 and verify the post-calibration with 16-235.
For Pre-Calibration / Proifling: 2160p24 RGB-PC with patchscale 16-255, to generate 16-255.
For Post-Calibration verification :2160p24 RGB-PC with patchscale 16-255 or RGB-Video with patchscale 0-255. In both cases, Murideo with generate RGB-Video patches (16-235).
- Ted, I'm a little confused.
Five days ago, you said that it is necessary:
- Use PC-Levels ( Full Data) in SIX-G
And also 16-255 in LS / CS for both profiling and verification.

But now your says:
16-255 - when creating a profile.
And 16-235 - when checking.

I use:
SIX-G: RGB-FULL
CS Profiling: 16-255
CS Verify: 16-235

By the way, why did you recommend 2160p24 output for SIX-G?
Why is it better than 1080p24?

Which patchset have you used for profiling?
- Standard 17x3 which offers LS / CS
I found "Optimized Patch Sequences" with "improved grayscale" on Lightillusion site.
For example: cub17tub65-3.csv
But, in practice, they greatly increase the measurement time.
At the same time, they do not give any quality improvements on OLED CX

The 1000 points you are measuring doesn't contain any primary or secondary color!
Where you found that 1000p verification run?
- Actually, I don't remember where I got this "1000p VR" :)
I think once, I generated it myself, inside the Lightspace.
Or maybe I found it somewhere in this thread.
OK. If this patchset is wrong.
Then I need the right one.
I will be very grateful if you can help me with this.

Can you upload a Profile of Primary only before the profiling?
- Unfortunately, I did not save this simple measurement.
But the RGB-separation was great, visually no different from that.
In any case, the OLED 55CX is still with me, at my home.
I think for the next day.
So, if necessary, I can make any additional measurements.

But when you measured for verification 100-Point Grayscale, you found only 1 point above 1dE, and that was dangerous?
What do you mean that you have 1 point above 1dE, and that is dangerous?
- TED,
I have never used the word "Dangerous" here, anywhere :)
I just wanted to say that I see a lot of great LG OLED calibrations by LS/CS in this thread.
After studying it, I got similar, very good results, for example, on the OLED C9 series.
But on the new CX, things are not so good. Although I use the same methods and advice.
Of course, I would like to understand why.

P.S.

Here are new measurements of the same OLED55CX.

1) I made a new 10-Point Cube for testing and used it.
Now, in main mode, EXP_DARK_2.35, I can see 9 dots exceeding dE 1.0

2) I also took two Primary Only measurements on an uncalibrated Cinema mode
Just in case, one measurement was with levels (in Color Space) 16-255, the second 16-235

Here are the .bcs files
 

·
Registered
Joined
·
9,498 Posts
Discussion Starter #1,582
- Ted, I'm a little confused.
Five days ago, you said that it is necessary:
- Use PC-Levels ( Full Data) in SIX-G
And also 16-255 in LS / CS for both profiling and verification.

But now your says:
16-255 - when creating a profile.
And 16-235 - when checking.
I had a typing error in the post, sorry.

I wrote 16-255 instead of 16-235.

The original post has been corrected now.

''ConnecTEDDD said:

For Pre-Calibration / Proifling: 2160p24 RGB-PC with patchscale 16-255, to generate 16-255.
For Post-Calibration verification :2160p24 RGB-PC with patchscale 16-255 or RGB-Video with patchscale 0-255. In both cases, Murideo with generate RGB-Video patches (16-235).''


The correct is:

''For Post-Calibration verification :2160p24 RGB-PC with patchscale 16-235 or RGB-Video with patchscale 0-255. In both cases, Murideo with generate RGB-Video patches (16-235).''

I use:
SIX-G: RGB-FULL
CS Profiling: 16-255
CS Verify: 16-235
Correct, we suggest the same method from 2018; nothing has changed.


By the way, why did you recommend 2160p24 output for SIX-G?
Why is it better than 1080p24?
Since you have the capabiility to send 2160p24, why not?

We suggest 1080p24 with PGenerator as it can't output 2160p.

But if the user has DVDO AVLab or PGenerator with an HD Fury device with upscale capability (...like HD Linker or better model) then he can send 2160p24.

- Standard 17x3 which offers LS / CS
I found "Optimized Patch Sequences" with "improved grayscale" on Lightillusion site.
For example: cub17tub65-3.csv
But, in practice, they greatly increase the measurement time.
At the same time, they do not give any quality improvements on OLED CX
The optimized patch sequence with improved grayscale you see @ LI site is working only when you will profile an RGB display tech projector/TV.

LGs are very different, so you have to use the specific/optimized patchsets we provide via LG Template instructions.

No other patchset can perform better from that specific patch ordering, while the patches are the same as a regular 17-Point Cube.

So don't expect to have any better results from any different method you will try, as the team behind all these stuff, they have tested over 200 different patchsets for years until finding which one provides the best possible results for all LG models 2018-2019-2020 of any year.

The patchsets we provide for LG OLED will not work with other LG OLED panel-based TVs from other brands.


- Actually, I don't remember where I got this "1000p VR" :)
I think once, I generated it myself, inside the Lightspace.
Or maybe I found it somewhere in this thread.
OK. If this patchset is wrong.
Then I need the right one.
I will be very grateful if you can help me with this.

- TED,
I have never used the word "Dangerous" here, anywhere :)
I just wanted to say that I see a lot of great LG OLED calibrations by LS/CS in this thread.
After studying it, I got similar, very good results, for example, on the OLED C9 series.
But on the new CX, things are not so good. Although I use the same methods and advice.
Of course, I would like to understand why.
If you are using the LG instructions I send with LG Template, you have a complete idea of what to do.

If you don't have the latest LG Guide, send an email to my site, and I will send you a download link.

P.S.

Here are new measurements of the same OLED55CX.

1) I made a new 10-Point Cube for testing and used it.
Now, in main mode, EXP_DARK_2.35, I can see 9 dots exceeding dE 1.0
2) I also took two Primary Only measurements on an uncalibrated Cinema mode
Just in case, one measurement was with levels (in Color Space) 16-255, the second 16-235

Here are the .bcs files
1) For the specific procedure, as EXACTLY we suggest to instructions... when the panel doesn't have any problem, you can get a maximum 5-6 dE from 1000, with over 1dE, usually below 1.2dE

Most of the time, the results are better from these, with 1000 points, all below 1dE.

You can get these results only if you follow the instructions, not when using different patchsets for profile or verification.

When you have larger numbers of yellow, then your panel has a problem from the factory, so you have to use the method I posted to fix the factory problem.

2) Your panel can be improved before starting profiling, its clearly visible from the RGB separation you have with 16-255 measurement before calibration.

Increase Color to 51 or 52 or Contrast 84 with color 51 or 52.
 

·
Registered
Joined
·
1,089 Posts
Just to reiterate what Ted has explained above, yesterday I finally got around to my first attempt at calibrating 2020 LG OLED. It was a 55CX5LB (with a June build date and almost 600 hours recorded UTT in the Service Menu).

It was not currently updated to the latest publically available FW due to the owner not using his OLED as a gaming monitor (the latest FW fixes only the gaming features of HDMI 2.1 GPU's), and I was concerned enough about what happened to Ted's CX and DoVi playback over USB with the latest update to advise the owner not to update at this point.

Anyway, after initial setup, a Quick Profile of Primary Only showed that this was yet another CX with poor factory calibration. The RGB Separation graph was all over the place

3045771


It took some fiddling with the Contrast and Colour controls, but the best I could eventually get it was

3045772


So while not at all perfect, it was more than acceptable in my mind to begin the full display profile.

On my own C9 at home, my currently calibrated modes are done using the patch sequence that Ted describes in his post above. It is consistently great, and provides excellent results every time. For the CX calibration yesterday I used a new custom hand built patch sequence (many thanks Fabio) just to test it out really. It contains 5420 points (shown in this post), and combined with the 109 Drift Patches, the total time to display and measure 5529 patches was just 1Hr 34 Minutes thanks to the new Klein API

(The patchset I used is not required for CX calibration, the one Ted refers to is perfectly good. I used this one just to test it out. I mention it only to show the speed benefits of the new Klein API)

3045773


(Time shown is for the profile measurements only, as I performed a experimental version of Pre-Roll which took about 15 minutes)

After generating the LUT and loading it to the CX, verification over 1000 points gave results off 999 points below dE 1.00.

3045774


With the 1 single point above coming in at dE 1.0991

3045775


Avg dE over the 1000 points was 0.43, while Avg Gamma was 2.41 (targeting 2.4)

Verification of the calibration in both test patterns and content confirmed the results were as visually as impressive as those shown in the measurements.

So, the same excellent results on the CX that we have seen in the past on previous generation LG OLED's.
 

·
Registered
Joined
·
9,498 Posts
Discussion Starter #1,584
Just to reiterate what Ted has explained above, yesterday I finally got around to my first attempt at calibrating 2020 LG OLED. It was a 55CX5LB (with a June build date and almost 600 hours recorded UTT in the Service Menu).
Just for reference, the panel factory issue explained here: LG OLED's 3D LUT Profiling using LightSpace Thread

Anyway, after initial setup, a Quick Profile of Primary Only showed that this was yet another CX with poor factory calibration. The RGB Separation graph was all over the place

View attachment 3045771

It took some fiddling with the Contrast and Colour controls, but the best I could eventually get it was
I have seen many RGB separation charts from CX, and this one is the worst I have seen so far.

Which Contrast/Color values did you use to fix the factory panel calibration?
 

·
Registered
Joined
·
1,089 Posts
Which Contrast/Color values did you use to fix the factory panel calibration?
Reducing Contrast to 83 and raising Colour to 52 somewhat improved things, but I eventually settled on a setting of Contrast 81, Colour 53.

It should also be noted that the set was crushing black a little with Brightness at 50 in SDR (verified with your patterns on both USB and via PGenerator). A Brightness value of 51 resolved this.
 

·
Registered
Joined
·
34 Posts
The patchsets we provide for LG OLED will not work with other LG OLED panel-based TVs from other brands.
- This is an extremely interesting moment.
What are the reasons that the optimized patchsets are only suitable for LG OLEDs of any release year,
but do not work with TVs of other brands made on the basis of exactly the same LG OLED panels?

I mention it only to show the speed benefits of the new Klein API
- If it's not a secret, what settings for Klein K-10a did you find optimal with the new API and used in these measurements?
 

·
Registered
Joined
·
1,089 Posts
- If it's not a secret, what settings for Klein K-10a did you find optimal with the new API and used in these measurements?
It’s no secret and has been posted over in the ColourSpace thread for a few days now.

@Anger.miki did some quite extensive testing and concluded that setting a Max Integration value of 1.250, with both the Intelligent Integration and Average Low Light settings disabled produced the most consistent and reliably repeatable results for balancing speed and accuracy when profiling LG OLED. Later that same day I did my own personal testing which confirmed Miki’s conclusions were absolutely correct.

I haven’t posted my own data as I’ve had a busy weekend, but I have sent it to Miki as a courtesy. Needless to say, his settings are indeed the perfect balance.
 

·
Registered
Joined
·
9,498 Posts
Discussion Starter #1,588
- This is an extremely interesting moment.
What are the reasons that the optimized patchsets are only suitable for LG OLEDs of any release year,
but do not work with TVs of other brands made on the basis of exactly the same LG OLED panels?
As I said, the patches are still a 17-Point Cube, but with special 'shuffle' optimized for LG OLEDs.

It was tested for years different algorithms to calculate that 'shuffle' and a group of users were analyzing the results and how TV's respond.

So after thousands of different variations of these, we found the ideal one.

We can do the same for another TV brand, but it will require to find out 5-10 users available to test 100 different patches sets for months.....these users are not available....since 99% of color fidelity enthusiast has bought LG OLED since it has 3D LUT capabilities internally.

The optimized patchset for LG is to reduce the drift and 'stress' less the display during the profiling, which is a 'data collection.'

LG OLED TV to be more stable.

Using regular sorting, the panel will not be so stable during the measurements.

It will take about 10000 words to explain all that testing in more detail, but you got the idea.

You can say that we know better how LG works after studying/experimenting a lot.

Since LightSpace/ColourSpace has all the tools to evaluate fro drift plots and volumetrically measure the unlimited number of points for verification, we had all the tools to understand which patchset is better for LGs.

It's not only a patchset but also the settings we provide for the extra delay, stabilizing, pre-roll, and all these work as a combo to make it all work.

No user can even imagine how many hours a group of real knowledge people has spent to test different settings until finding out the best settings for each meter.
 

·
Registered
Joined
·
18,179 Posts
As I said, the patches are still a 17-Point Cube, but with special 'shuffle' optimized for LG OLEDs.

It was tested for years different algorithms to calculate that 'shuffle' and a group of users were analyzing the results and how TV's respond.

So after thousands of different variations of these, we found the ideal one.

We can do the same for another TV brand, but it will require to find out 5-10 users available to test 100 different patches sets for months.....these users are not available....since 99% of color fidelity enthusiast has bought LG OLED since it has 3D LUT capabilities internally.

The optimized patchset for LG is to reduce the drift and 'stress' less the display during the profiling, which is a 'data collection.'

LG OLED TV to be more stable.

Using regular sorting, the panel will not be so stable during the measurements.

It will take about 10000 words to explain all that testing in more detail, but you got the idea.

You can say that we know better how LG works after studying/experimenting a lot.

Since LightSpace/ColourSpace has all the tools to evaluate fro drift plots and volumetrically measure the unlimited number of points for verification, we had all the tools to understand which patchset is better for LGs.

It's not only a patchset but also the settings we provide for the extra delay, stabilizing, pre-roll, and all these work as a combo to make it all work.

No user can even imagine how many hours a group of real knowledge people has spent to test different settings until finding out the best settings for each meter.
Correct as a different patchset is needed for Sony OLEDs to achieve the same results.
 

·
Registered
Joined
·
111 Posts
Hi, which pattern are you looking with ''0-1'' markings, and from which player do you playback the pattern?

HDR Brightness patterns should not used for calibration/adjustments. We are using them only for evaluation.
I've just made a picture with a black background, and a rectangle in the middle with RGB level 1,1,1, and then viewed it in a pitch black room, after my eyes have adapted, i used both the internal apps, and via my HTPC (RGB limited HDMI Low) in both cases level 1 was not showing at all, i also used the black level pattern found in the AVSHD mp4 with the same outcome, so somewhere over 4-5 months the correct brightness setting changed from 50 in HDMI mode SDR to brightness 53.

I remember the last time i checked this was around 5 months ago, when the TV had around 600-700 hours used so it should have been stable, and RGB level 1 was visible at brightness 50 then.
 

·
Registered
Joined
·
34 Posts
Once again, thanks to everyone for the advices and help.

I tested RGB separation on my own LG OLED77C9
I think I finally found the reason for those "Blue Ripples" I complained about 8 months ago.
Because RGB-SEP looked like this.

After a very long trial and experiment, I managed to do this.

However, for this I had to set completely insane:
Contrast: 67
Color: 60

Of course, this led to a very significant drop in peak brightness,
which had to be much lifted by OLED Light.

The usual Color / Contrast values in the range given by Ted (80-85) did not help correct the situation.
Interestingly, if I raise the Color by at least one click above 60, everything collapses.
The RGB lines diverge very strongly immediately.

P.S.
Another one attempt.
Contrast: 82
Color: 60

P.P.S.
Sadly, even RGB-Separation corrections
didn't help solve the blue-corner problem on OLED77C9
Blue Ripples are still here.
It looks like this:



CUBE_EXP_DARK_2.35_2.JPG CUBE_EXP_DARK_2.35_3.JPG CUBE_EXP_DARK_2.35_1.JPG


I used three patchsets (Profiling, Pre-Roll, Verificaton) from the latest TED`s Guide.
Verification showed: the gray scale is in perfect order, all points below dE 1.0
CUBE_EXP_DARK_2.35_5.JPG


However, in the 1000-point check, 6 points are visible, which jumped behind dE 1.0
CUBE_EXP_DARK_2.35_4.JPG

All .bcs are here
 
1581 - 1591 of 1591 Posts
Top