AVS Forum banner

621 - 640 of 1595 Posts

·
Registered
Joined
·
34 Posts
I am also very happy for Make73!
And also extremely grateful to everyone who is trying to help me with advice.
Especially liberator72, which recently posted bcs files.
They allowed me to make many comparisons with my own results and discover a lot of strange things.

Unfortunately, my own attempts to create high-quality 3D-LUT, using Lightspace,
on OLED77C9 - not very similar to the "Christmas story" yet.

The main problem: very strange "dents" in cubes.
First of all, in the blue area.
I don’t see anything like this on Liberator72 files, his cubes are smooth, even, without the slightest sign of dents or irregularities on blue.
Whereas my measurements, whatever I try, do not turn out as smooth.

What I try to find the cause of the problem:

- I made many attempts, with cubes of both 17x3 and 21x3
- Of course, I have done many times both a complete reset of the tv, and the reset of presets and 3D-LUT via Device Control
- I used another OLED77C9. To compare LS performance on several different OLED77C9 is not a problem for me.
- I even used other sensors. Instead of "Klein K-10a + Jeti 1211" I tried a i1D3 + i1PRO2 (using the settings recommended for them).
- Finally, I built Dispcal 3D-LUT on the same OLED77C9, using 9377 "Auto-optimized" points. I did not use SIX-G, but MadTPG as the generator this time.
After all check this cube at 1000 points using Lightspace and MadTPG

All this did not help or fix the problem!

A few notes:

1) When I generate 3D-LUT from a Liberator72 file called "LG OLED65C9MLB - OOTB Full Display Profile for LUT Generation".
Lightspace reports "99% within target gamut". But when I do the same from my measurements (not 1000-points check file but the initial measurements)
17x3 or 21x3 I get only 93%.

2) I noticed that liberator72 uses a low peak brightness: 104.5 cd/m2
I usually use 140-150 cd/m2
I checked by making LS 3D-LUT at low brightness. The average dE after checking for 1000 points really turned out to be lower.
It seems to me wrong, LS should provide equally low errors both at peak brightness of 100 and 140-150 cd/m2

3) At a peak brightness of 111.25 cd/m2 with a 3D-LUT built via LS,
I managed to get:
Other: "Max 2000: 1.28 / Average 2000: 0.33

But when I see my crumpled and dented cubes, especially in the blue area,
all this is not at all pleasing.
Unfortunately, no matter how hard I try, I still can not understand the cause of these problems.
In the application, I provided a lot of bcs files with various measurements, including controls.
Also there is a 3D-LUT verification file created on the same OLED77C9, using Dispcal at 9000 points and checked by DC at 989 points.
By the way, the average dE2000 turned out to be 0.37 and the peak was 1.89

I will be very grateful for any tips and tricks that will help me find the cause of my problems.
 

Attachments

·
Registered
Joined
·
607 Posts
...The main problem: very strange "dents" in cubes.
First of all, in the blue area.
I don’t see anything like this on Liberator72 files, his cubes are smooth, even, without the slightest sign of dents or irregularities on blue.
Whereas my measurements, whatever I try, do not turn out as smooth...
Seems that there is those "blue problematic OLEDs", You have one, I have one, Leon HAD one, who knows how many are these problematic units.
I think only LG engineers could tell reason for that what cause it and if there´s anything what could solve it, maby software based solution.
If someone who have close connections to LG, could you ask this problem and if it´s not secret, then tell it also to all of us :)

And just to know, I had ~750 hours when done IN-STOP, latest 04.71.05 fw installed.

Here at Finland there´s not any chance to get it under warranty.
So if you don´t notice problem in time when you can change it or ask full refund without any reason (I had 60 days for that, it´s gone already) you don´t have any chance to get new one.

However, I send you PM.

btw, it would be interesting to see how Your unit looks with CM AutoCal, do you have license for CalMAN and could try it?

He received a 17^3 patchset based on LightSpace anisometric version which was modified to achieve better drift plot.
Is it special so unique calculated just for one unit?
Or is it more like common?
You did that? If so, thanks a LOT!

If that´s not any secret, maby someone could release .csv files(s) for these problematic units?

With special .csv I got better results, however when you take closer look you still can notice abnormal blue corner of cube.
I think it´s nothing which cause problem to real content as You can see from "3D LUT affect" picture, where is side by side original test pattern on left and right side same picture with affect of my LS created 3D LUT (Rec709, 2.4).


edit: pictures and .3dl (Rec709, 2.2) included.
 

Attachments

·
Registered
Joined
·
34 Posts
1) Wow, it seems another, very special, "panel lottery" from LG ...

2) You are in Finland, I am in Moscow, Russia.

3) Thank you so much for your help.
I tested the LS already on two copies of the OLED77C9.
Both were problematic, on both I see "blue area jams".
On both the latest firmware was installed, version 04.70.05

If I knew about the problem before, I could check on about three dozen copies of LG OLED, in different diagonals.
In the last month alone, quite a few copies of this model have passed through me.
Sometimes I simultaneously had three OLED77C9 together, in the same room :)

4) Since you are in Finland, you most likely have an OLED77C9 by Polish assembly. This is production facilities in Wroclaw, which make all LG TVs for the Europe.
Whereas in Russia we have our own LG factory, near Moscow, in the city of Ruza.
The Russian model OLED77C9, like the rest of the LG oleds, we have a local, Russian assembly.

5) It will not be a problem for me to get a new "european" OLED, Polish production. In Moscow there are many of them and they cost a cheaper than official ones.
But I must say, the reputation of the "Polish" OLED LG is bad. My personal experience has shown that, in many ways, these panels are more problematic than the “Russian” ones.
Perhaps because production volumes at the Polish factory are much larger than the Russian one. Poland produces about 7 million TV sets per year, compared to 1.5 million in Russia.

6) I have a serious suspicion that the problem of “jams in the blue area” is not a problem of specific instances,
but the OLED77C9 models at all. In the end, the OLED 77" panels, for the whole world, are produced at the same plant in South Korea.

7) Of course, I have Calman.
I am sure that on my instances of OLED77C9 it does not give a problem similar to "shark fins".
But just in case - I’ll check. I will try to do it as soon as possible.

It would be very interesting to know how many people have encountered the LG OLED problem in a blue area, similar to mine.
It is also important to understand this problem of individual instances or the entire series.
And of course it would be important to understand how to fix it.
 

·
Registered
Joined
·
607 Posts
It would be very interesting to know how many people have encountered the LG OLED problem in a blue area, similar to mine.
It is also important to understand this problem of individual instances or the entire series.
And of course it would be important to understand how to fix it.
Can´t agree more!
How about there in Russia (I say hello to neighbour :) ), is there better customer support than we have here at Finland?
Maby you could try contact them and point links to AVS if they could forward these to LG developer team?
Maby they are aware of this "problem" or maby not.
Most of ordinary users are very happy with their OLED if there is not major banding or black clip.
 

·
Registered
Joined
·
34 Posts
Yes, LG's technical support and service here is very good.
But I'm afraid that everything is decided only in South Korea, and on the way to solving such problems there is a mountain of bureaucratic problems :)
Local LG specialists are hardly able to understand such subtleties as the problems that we can deal with in the blue OLED area, and only with the help of LihgtSpace.
I'll think about how to convey this information to the central office of LG.
 

·
Registered
Joined
·
1,137 Posts
@GolorCrading

In my opinion, and this is just my opinion based solely on what I experienced with my first panel, both you and @Make73 have been struggling with a very similar issue I did with my first panel, the 65C9PLA.

At first, I too was getting those strange shapes around the blue area, among some other things.

At one point, I also tried calibrating my whole video chain, and passed my RPi PGen through my Denon AVR, but Ted discovered this was causing severe clipping in the higher measurements. My uneducated theory for this was the AVR didn't play kindly with the RGB Full output of the RPi, as theoretically when outputting YCbCr from a Blu Ray it shouldn't clip above White?? But, that's getting a little side tracked.

Anyway, it took a further week or so experimenting, each time being meticulous about setting my probes up correctly, making sure all settings were correct, and any Probe Match performed was a valid one. I tried with the RPi connected both using Bluetooth, and Wi-Fi. BT, for me, is a lot more stable. There were some sync issues using Wi-Fi from what I can tell, as I couldn't reproduce it every time. I have a very good stable network here at home, but it only takes one thing to interfere with a Wi-Fi signal to throw things off I guess.

During this time, I was in constant communication with Ted, and he got in touch with some of his contacts to see if they too were experiencing similar issues on the 2019 panels. If I recall correctly, none of his contacts at that time had, so I just continued experimenting, and listening to Ted's advice.

Eventually I just performed another Factory Reset, to make sure there was nothing left over from previous calibrations, and things began to fall into place. The results from that panel can be seen in the following post

https://www.avsforum.com/forum/139-...-using-lightspace-thread-15.html#post58844028

After I finally managed to get good results with SDR, I took a look at HDR. This is where I found my panel had a really low Peak Output on a 10% window in both HDR and DoVi compared to what the average C9 was being measured at. The maximum I could achieve, no matter what the Colour Temp, SM WB values or any other thing I could think of would get this any higher than 650 nits. When calibrated to D65, this gave me just 580 - 590 nits. This, combined with my SDR calibration woes, gave me enough to speak to my excellent retailer and get an exchange.

The results of mine you have used to compare to yours are from that replacement 65C9MLB panel. That calibration was performed brand new, straight out of the box with only 19 hours use on the panel. I tracked that calibration in 100 hour intervals to see how the panel actually drifts and changes in the first 300 hours, and then at 300 hours, reprofiled with the same repeatable results. The results from that 300 hour observation and reprofile can be seen in this post

https://www.avsforum.com/forum/139-...-using-lightspace-thread-20.html#post58985224

2 nights ago, I ran a real quick 10^3 profile for a Bright Room setting with a 150 nit peak output using probe settings provided by Anger.miki for the i1d3 with very, very good results. I know I can get better, it was just a quick one so I never took my time setting up.

I guess what I am trying to say is, some of these 2019 panels seem to have some kind of weird, funky issue. Ok, it might only be 3 reported here in the LightSpace thread so far, but that's because we are using an excellent piece of software that exposes the problem. Other software solutions do not have these tools, and as "the other" software is more widely used and has no way to show this, who knows how many more are out there. You'll see many reports in the other thread talking of banding, posterization etc, etc. And while a lot of this can be put down to the way the LUT is generated there (as show in the recent test I did in that thread), some of the more severe posterization issues could well be down to this funky issue we are seeing because we use LightSpace and have the tools to actually visualise it in the generated LUT.

Patience and perseverance will pay dividends. Set probes up meticulously. Warm the panel for at least the recommended 1 hour using SDR content. Use a Pre-Roll to stabilise the panel. I can't suggest anything for your Klein probe, but using the i1d3 then use the standard recommended settings. Just double, triple check all settings and then check again. I know it may sound a little condescending, and I apologise if it does, I really don't mean it to be. It just cannot be stressed enough, especially if you are new to LightSpace. It was how I began to learn using it and have since found I can get repeatable results each and every time.

Stick at it, you'll get there!! :)
 

·
Registered
Joined
·
607 Posts
In my opinion, and this is just my opinion based solely on what I experienced with my first panel, both you have been struggling with a very similar issue I did with my first panel, the 65C9PLA.

At first, I too was getting those strange shapes around the blue area, among some other things...

...Eventually I just performed another Factory Reset, to make sure there was nothing left over from previous calibrations, and things began to fall into place. The results from that panel can be seen in the following post

https://www.avsforum.com/forum/139-...-using-lightspace-thread-15.html#post58844028 ...
Looked file from Your 1st panel, didn´t see difference with LUT Preview, or if there was it was very faint.
But nothing like I had with CM and those "shark fins"
btw, tried again CM to get rid of those fins with different kind of setting, but no luck :confused:

Your 1st panel blue corner:


I could try at someday re-profile with one click higher brightness how that affect to blue corner.
But now I planned to do something else with my TV than calibrating it.
 

Attachments

·
Registered
Joined
·
51 Posts
Hi.

I am trying to calibrate my C9, but it is being hard to start.

- Are someone able to access the service menu using the remote of the TV? I found different sites with combinations of TV keys and remote but the C9 have one key (the TV). I am thinking about buy the remote of LG (I don't have a universal remote or infrared phone), I already have to many remotes.

- I am going to use a Lumagen Mini like TPG, output YbCbCr 422. The Lumagen doesn't let configure the bits, I read the manual and I believe the output is 10 bits (HDMI 1.4). Does someone know for sure?

- If I use YCbCr in the Lumagen, I think in patch scale Lightspace should be 16-255, in order to use 109% and avoid to apply videoscale. Is this correct? Is it better also 16-235 in Lightspace?

- I read the light objective is 123 nits in 109% with 10% window? Is it correct? Or should be a full window?

Thanks in advance.
 

·
Registered
Joined
·
1,137 Posts
Looked file from Your 1st panel, didn´t see difference with LUT Preview, or if there was it was very faint.
No that is my point, before that profile, I was repeatedly getting a very similar issue to you and GolorCrading in the blue area, but eventually, after lots of advice from Ted, continuous experimenting and just pure perseverance everything fell into place, giving me the profile you are looking at there. I honestly couldn't tell you what it was that changed, other than each time I ran a profile, I was more and more rigorous in the way I set up and checked each setting.

Unfortunately, I don't have the measurement files from that panel anymore as I deleted them once I was happy my new panel was performing as expected. My one single "nit-pick" with LightSpace is the way the file management of measurement data is handled. And I stress, it is a nit-pick and nothing more!! The Manage Colour Space section where all of that data is managed can become a little convoluted at times when making lots of profiles as it populates pretty damn quickly (e.g. when trying to troubleshoot). I could never find what I was looking for despite naming measurements something logical and memorable for the profile being performed. BUT this was down to my own misunderstanding of the manager, not the software itself. I have since realised that it is very easy to export all these unnecessary or no longer needed BCS files, store them in a clearly labelled folder on my laptop, and then delete them from within LightSpace making things a lot easier to navigate and find what you are looking for. If I ever need to refer back to them, it is a 5 second task to import them back into the software to view. Fantastic!!

As I have said many times, I am still new to LightSpace myself, in fact new to calibration as a whole, and I'm learning all the time. Had I realised the above before I'd deleted my old files, I would gladly provide them to you and you would see they look remarkably similar to yours down in the blue region of the cube. As I had never used CalMAN for SDR (before that test the other day), I honestly couldn't say if I would have had the shark fins as you did, and I'm not willing to speculate.

What I do not need to speculate about is this....Had it not been for LightSpace, the tools it provides, the ever helpful Ted, and the great bunch of guys within this thread, I never would have known about my "funky panel" in the first place and may well have ended up being stuck with a lemon. I didn't, I ended up with a fantastic TV that calibrates perfectly as expected. And that, to me, speaks volumes about the software I used, the flexibility and tools it gives you, the support that is provided for it, and the community here at AVS that surrounds it.

!!!TOP STUFF INDEED!!!

Edited to add additional screenshot

Found a leftover .dcl LUT file in my folder. Here it is in the Cube Viewer. As I said, not as bad as you were getting, nor GolorCrading, but this was a few profiles in when I was getting improvement, but you can definitely see some blue issues that aren't in the final LUT done before returning the TV
 

Attachments

·
Registered
Joined
·
2,215 Posts
Please be aware that the LUT Cube view is showing you the 'correction' that needs to be applied to the display.
So any 'dips' or 'variations' in the LUT Cube view show that there is an 'inverse' issue with the display's native colour response.

It is very possible that if a display has a 'strange' non-linear colour space error, the LUT will also be 'non-linear, but the result of the LUT will be 'correct'.
That can be tested via a verification profile, and in LightSpace making another LUT to see if the calibration is 'correct'.
(If the calibration is correct, the LUT will be a near perfect 'cube' with no variations/distortions. But, small drifts in the panel, and the 'accuracy' of the probe will mean the result will never be 'perfect'.)

However, if you send Ted your original profiles he can use ColourSpace to show the 3D Volumetric Graphs of the actual profiles, so looking at the raw profile in a similar way as the LUT Cube view, but this time looking at the actual profile errors/issues, rather than the LUT 'corrections'.

That helps isolate any underlying issues with the displays prior to attempting calibration.

Steve
 

·
Registered
Joined
·
1,137 Posts
Please be aware that the LUT Cube view is showing you the 'correction' that needs to be applied to the display.
So any 'dips' or 'variations' in the LUT Cube view show that there is an 'inverse' issue with the display's native colour response.

It is very possible that if a display has a 'strange' non-linear colour space error, the LUT will also be 'non-linear, but the result of the LUT will be 'correct'.
That can be tested via a verification profile, and in LightSpace making another LUT to see if the calibration is 'correct'.
(If the calibration is correct, the LUT will be a near perfect 'cube' with no variations/distortions. But, small drifts in the panel, and the 'accuracy' of the probe will mean the result will never be 'perfect'.)

However, if you send Ted your original profiles he can use ColourSpace to show the 3D Volumetric Graphs of the actual profiles, so looking at the raw profile in a similar way as the LUT Cube view, but this time looking at the actual profile errors/issues, rather than the LUT 'corrections'.

That helps isolate any underlying issues with the displays prior to attempting calibration.

Steve
Hi Steve.

I think the point I'm trying to make is this. Some of these 2019 OLED's seem to be severely affected around blue, and that is making them just that bit more difficult to calibrate. This is what we're obviously seeing in the Cube Viewer. Whether it is a characteristic element of the 2019 panels or not, I have absolutely no idea. It is not present at all on my replacement panel, so there is definitely some form of panel variance at play. I was still able to get fantastic results on that old TV, I just had to be a lot more rigorous in the way I did things.

I'm completely new at all of this and some of the terminology I use is very likely to be wrong at times, and I don't have the knowledge or expertise to explain myself clearly either, but using your software is proving to be a great way to learn, especially when you have such an amazing community surrounding it! (+ plus Ted of course)

I will be upgrading my license in the next day or two in preparation of the fact I have every intention of getting better probes some time in the future, and I really can't wait to get my hands on ColourSpace when it releases!!

Leon
 

·
Registered
Joined
·
607 Posts
My tests (which are valid for me) told me that 0.25 IT, 0.50 ED, 2 nits II and 0.1 SP are the best balanced settings between speed and accuracy. I verified a LUT done with those settings with a 1000p cube done with 1 sec IT, 1 sec ED, II 5 nits and 0.35 SP. Awesome results. ;)
I couldn´t resist to try BR52 setting instead of BR51 just to see how it acts to blue readings, so made profile for Expert Birght...had to do it anyway at somepoint.

I used Your settings (i1Display Pro OEM Rev.B):

Integration Time 0.25
Sync Mode AIO
Extra Delay Time 0.50
Pre-Roll 0.50 (set to 1.7s)
Intelligent Integration 2.000
Stabilisation 0.1000

Drift Comp.
Profiling 50
Verifying 0
 

Attachments

·
Registered
Joined
·
1,012 Posts
Pre-roll should always last 30 minutes, so with those settings you should set it at 1.2 sec. Anyway, if you want to test how good those settings are for your LUT, just verify the same 1.000p patchset with 1 sec integration, 1 sec extra delay and leave the other settings unaltered. It will require almost 46 minutes.
 
  • Like
Reactions: Make73

·
Registered
Joined
·
607 Posts
Already converted that to x .308 y .313 and uploaded to TV.
But I will test that for sure later.

But in future I´ll setup Pre-Roll to 1,2s.

Have you tested a lot how it affect to result even if you have watched SDR content hours before calibrating...it still requires that 30min Pre-Roll to prepare for measurements?

How you export readings to Excel and get that great chart about readings?
It´s made by You if understood correctly :)
Planned if I could make my own one.
 

·
Registered
Joined
·
1,012 Posts
That chart is a small part of a multi pages Excel spreadsheet, I created it and I am very jealous of it as I spent entire days to make it as it is now. lol

Still it needs some fine tuning (gamma calculation, for example). The latest implementation is the possibility to insert different primary and white coordinates while still getting perfect dE2000 calculation.

I was thinking about selling it, but I should find a good protection and the right price... Would you buy it for 49.99$? :D

You need to enable "Developer menu" in Excel and import the bcs, then search for all the formulas related to dE2000 (RGB to XYZ, XYZ to Lab, ecc...)
 

·
Registered
Joined
·
607 Posts
That chart is a small part of a multi pages Excel spreadsheet, I created it and I am very jealous of it as I spent entire days to make it as it is now. lol

Still it needs some fine tuning (gamma calculation, for example). The latest implementation is the possibility to insert different primary and white coordinates while still getting perfect dE2000 calculation.

I was thinking about selling it, but I should find a good protection and the right price... Would you buy it for 49.99$? :D

You need to enable "Developer menu" in Excel and import the bcs, then search for all the formulas related to dE2000 (RGB to XYZ, XYZ to Lab, ecc...)
This year budget is full (just spent 5000eur for TV and 1000eur for calibration) but let´s see at next year (six days...) ;)
Maby You should sell it to Light Illusion and get little profit for every sold LS :D
 

·
Registered
Joined
·
1,137 Posts
That chart is a small part of a multi pages Excel spreadsheet, I created it and I am very jealous of it as I spent entire days to make it as it is now. lol

Still it needs some fine tuning (gamma calculation, for example). The latest implementation is the possibility to insert different primary and white coordinates while still getting perfect dE2000 calculation.

I was thinking about selling it, but I should find a good protection and the right price... Would you buy it for 49.99$? :D

You need to enable "Developer menu" in Excel and import the bcs, then search for all the formulas related to dE2000 (RGB to XYZ, XYZ to Lab, ecc...)
It took me hours just to figure out how to make an Excel spreadsheet for myself so that I could enter meter readings to verify whether a meter correction using the Probe Match function was within tolerance levels. It gave me a headache just doing that, and I was super proud of myself when I got it right :D:D:D

You have my utmost respect for what you did!!
 

·
Registered
Joined
·
1,012 Posts
:D

In my spreadsheet I got also RGB full to limited, limited to full, RGB to YCC, YCC to RGB, RGB (0, 1) to RGB (0, 255) and (16, 235), RGB to XYZ to xyY and viceversa, XYZ to Lab and dE2000 calc, XYZ to RGB to ITP and dEITP calc (work in progress). All conversions and calcs are up to 9261 readings (21p cube).

With @ebr9999 I also created a spreadsheet that is a small calibration software: from xyY probes data, it calculates FCCM and TMC (Three matrixes calibration), it corrects and uncorrect up to 21p cube readings.

I learned so much while I was building those spreadsheets. Notions that should be daily bread for professional calibrators, but instead it is only for a few.

@ebr9999 has a very interesting spreadsheet about FCCM, but I prefer to let him talk about his creation, if he wants.
 

·
Registered
Joined
·
1,137 Posts
:D

In my spreadsheet I got also RGB full to limited, limited to full, RGB to YCC, YCC to RGB, RGB (0, 1) to RGB (0, 255) and (16, 235), RGB to XYZ to xyY and viceversa, XYZ to Lab and dE2000 calc, XYZ to RGB to ITP and dEITP calc (work in progress). All conversions and calcs are up to 9261 readings (21p cube).

With @ebr9999 I also created a spreadsheet that is a small calibration software: from XYZ probes data, it calculates FCCM and TMC (Three matrixes calibration), it corrects and uncorrect up to 21p cube readings.

I learned so much while I was building those spreadsheets. Notions that should be daily bread for professional calibrators, but instead it is only for a few.

@ebr9999 has a very interesting spreadsheet about FCCM, but I prefer to let him talk about his creation, if he wants.
Now you're just showing off ;) :)

To be honest, I enjoyed doing it because it gave me the opportunity to learn more about how to ensure I make a good correction. I also plan at some point to expand it some way, but I'm still looking into exactly what I need to do. I'm just really eager to learn as much I can about it all, it fascinates me
 

·
Registered
Joined
·
1,012 Posts
Nah, not showing off :D I’m just proud of my creations.

What have you done in your spreadsheet? x1-x2, y1-y2, ((Y1-Y2)*100)/Y1?
 
621 - 640 of 1595 Posts
Top