AVS Forum banner

861 - 880 of 1224 Posts

·
Registered
Joined
·
7,095 Posts
Interesting thoughts. Maybe some of these features can be added to CS at a later date.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #862
Excellent videos! I somehow missed the option to extract native gamut from the profile to use it as a target. It would have helped me immensely when hunting for the couple of bad readings I needed to retake (I was targeting DCI P3 instead as it is closer to native gamut of LG CX, but it is not ideal, of course).
The 3D CIE chart with tangent lines enabled, and dE filtering will help spot 'bad' readings.

'Bad' readings can appear from short integration time or short extra delay or network delay (from users with network patch generators)

However, all tools are available from the users to remove/update 'bad' readings and generally have a complete idea about their measurement data file, before generating any correction LUT.

Which pattern generator solution you used?

The method of comparing with the native gamut can showcase any active processing a display is applying.
 

·
Registered
Joined
·
149 Posts
The 3D CIE chart with tangent lines enabled, and dE filtering will help spot 'bad' readings.

'Bad' readings can appear from short integration time or short extra delay or network delay (from users with network patch generators)

However, all tools are available from the users to remove/update 'bad' readings and generally have a complete idea about their measurement data file, before generating any correction LUT.

Which pattern generator solution you used?

The method of comparing with the native gamut can showcase any active processing a display is applying.
Hi Ted. I have tested different integration times and delays (even 6s integration time and 2s delays) and could not get rid of all the incorrect readings (some readings are just totally off either in xy or Y). As I have written I blame LG 2020 iTPG for that (running it over Wi-Fi does not help, but I don't have Ethernet port for my Surface at the moment, plus there's this magenta flash with the 2020 iTPG that has a big potential of skewing the measurements too).

I am aware of the tools at the disposal, comparing against native gamut and filtering by dE values does help a lot (it would be fantastic if the points could be selected in 3D view as well), but I was thinking of some automated way of showing/filtering the biggest errors against some predictive model, Steve could even give it some fancy AI name as it seems to be all the rave now ;-)

But in all seriousness, even as it is now, ColourSpace gives a fantastic set of tools to ensure the resulting LUT is free of artifacts, so I'm not complaining, just offering some suggestions further down the road ;-)
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #864
I did have a couple (4-6) really off readings (I blame LG's iTPG for that) where the x/y values would be off by 0.2-0.4 (and I am talking bright patches here, so not really i1D3 limitations, which are clearly visible with very dark readings) and the ability to find those and delete/retake them makes all the difference between a good and a bad profile. I really think that any software missing this is useless -- you can either retake all your reading hoping it will be correct this time, increase delay/integration time even more, etc, but still might not know you have them unless you see banding/contouring/blobs in the real world material. It is time consuming and frustrating...
Hi,

Your results are excellent; if we count the fact that you used an older generation of i1Display PRO and the iTPG of LG, which adds these magenta full-field patches before generating any patch.

i1D3 (rev A unfortunately, so no AIO mode support). I used period instead of the recommended burst as it seems to 'come out' of the bad dark readings faster than in burst of frequency mode and by that I mean the xy readings hit the target at lower values than e.g. in burst mode; very low values are off in different directions in different modes, but they are 'rubbish' anyway, so hopefully ColourSpace algorithms take care of that when creating a 3D LUT. BTW, I really see no advantage of using higher integration times than 0.25, but went for 0.75 just for the peace of mind. Intelligent Int set at 2 (though again, it won't make your low-light readings magically better), again label it under peace of mind/recommended settings.
Intelligent integration is working better with REV.B i1Display PRO.

It will work with Rev. A, but due to firmware/hardware limitation of X-Rite, it will take much more time (not ColourSpace issue).

LG 2020 iTPG as my pattern generator (16-255 patches scale). On 03.10.20 firmware it causes a magenta flash every time the patch needs to be updated that I think might interfere with the readings, so I set extra delay to 0.75 (at 0.5 or even 0.25 I would get more bad readings, not a lot, but still more). I also disabled stabilization as this, again, adds an extra flash and I believe it should be set at 0.5-0.75 then (instead of the recommended 0.35), which would extend the profiling time considerably. Hopefully this will be fixed at one point by LG. Drift seems to be quite reasonable regardless (no doubt in part thanks to the optimized patch set) with drift compensation every 50 frames
I don't suggest any user use any internal generator of LG; they are always problematic.

PGenerator via Direct LAN connection is the best method for 3D LUT SDR profiling. You can use it to calibrate different TVs.

The TV firmware, which added the new iTPG to 2020 models (the version that added the magenta flashes), was tested from engineers for months, and it's released with that major issue.

It looks like LG is not bothered so much with these internal calibration functions; it makes no sense to release a broken iTPG to the public.

It can take 3-6 months for an engineer firmware for testing to different regions around the world before being available to consumers.

  • I am not touching 1D LUT to avoid banding (though my built-in one is not super smooth to begin with especially with darker values, at least it shows 17 patch at brightness=50 as it should) and only reset 3D LUT, gamma set to 2.2, all extra processing off
Even when the internal processing of the TV, responsible for the loading of 1D LUT, will not add any problem when you upload a correction 1D LUT, with WRGB OLEDs, it's not helping to perform 1D LUT.

1D LUT before doing 3D LUT, it can improve the results only if you have an RGB panel.

3D LUT will include 33-Point Grayscale inside, calculated from the measured 21-Point meter reading.

Despite this I had 5-6 very bad readings that I had to re-take (one was difficult to spot at xy values were fine, by Y was half the target, I caught this quickly by running LUT preview). I use the CIExy graph with tangent shown and filter the values by dE. 3D xyY graph is great for spotting the huge errors in brightness (too bad we cannot select a point from that view). BTW, could we have a 'Measure' button to retake a given point directly from the Point Info window?
The Point Window is under development, and it will have many changes to future releases.

You can manually measure and copy-paste the new xyY if you like.

If you like one reading to be ignored, you can add a negative symbol to the XYZ values of that point (and not delete it)

The dE calculation of the 10000 points of the 'Genious Team' can be used to spot large problems.

The selection of a point in 3D charts will be added in the future.

  • 1000-point verification done, results above
All in all, I am super happy with the results I got and enjoy my TV set much more now! :)
If you have the 1000p 'cake' dE chart from the 'Genious Team', and post it, it will be interesting to see your position between these guys who measured 1000p also:

 

·
Registered
Joined
·
149 Posts
Thanks Ted for your detailed response!

I'll address some of the points you have raised.

Intelligent integration is working better with REV.B i1Display PRO.

It will work with Rev. A, but due to firmware/hardware limitation of X-Rite, it will take much more time (not ColourSpace issue).
Good to know and that explains why I saw no big difference in various low IRE patch readings.

I don't suggest any user use any internal generator of LG; they are always problematic.
I am quite aware of that, but need to live with the limitations for now, sadly.

Even when the internal processing of the TV, responsible for the loading of 1D LUT, will not add any problem when you upload a correction 1D LUT, with WRGB OLEDs, it's not helping to perform 1D LUT.

1D LUT before doing 3D LUT, it can improve the results only if you have an RGB panel.

3D LUT will include 33-Point Grayscale inside, calculated from the measured 21-Point meter reading.
So actually my patch set contains 41pt WRGBCMY slides, so it is even higher. I have read about the rationale behind not touching 1D LUT and I am not that eager to either (as 1D unity LUT introduces black crush that Calman was unable to correct, not to mention that it gave me a jump scare when it survived the first full TV reset, so I thought for a moment my 1D LUT was screwed for good due to this and/or pixel refresh cycle that I ran at that time too; 2nd full reset restored the 'good' 1D factory LUT and patch 17 has been visible with brightness at 50 ever since). My only gripe is that lower luminance ramps display some banding. I was just wondering if 1D unity LUT or a generated one could alleviate this a bit. Again, this is only visible up close on the grayscale/color ramps, does not bother me that much (if at all) in real content. Perhaps one of those bunnies not worth chasing :)

The Point Window is under development, and it will have many changes to future releases.

You can manually measure and copy-paste the new xyY if you like.

If you like one reading to be ignored, you can add a negative symbol to the XYZ values of that point (and not delete it)
I was not aware of the negative values, great tip!
And please consider adding measure button directly to the Point Window, it will save the trip to Manual Measure for the problematic readings.

If you have the 1000p 'cake' dE chart from the 'Genious Team', and post it, it will be interesting to see your position between these guys who measured 1000p also:
Unfortunately, I do not have that nice chart, but I am sure my results are not up with the best (BTW, I only ran the verification check on the following day and took white readings a couple of times and these can really bump the dE readings up or down a bit). If someone can PM the dE Chart to me, I will happily run it on my verification profile and paste the results.

As I have said many times, all that matters to me is the real content and this looks fantastic after ColourSpace 3D LUT profiling! :)
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #866
So actually my patch set contains 41pt WRGBCMY slides, so it is even higher. I have read about the rationale behind not touching 1D LUT and I am not that eager to either (as 1D unity LUT introduces black crush that Calman was unable to correct, not to mention that it gave me a jump scare when it survived the first full TV reset, so I thought for a moment my 1D LUT was screwed for good due to this and/or pixel refresh cycle that I ran at that time too; 2nd full reset restored the 'good' 1D factory LUT and patch 17 has been visible with brightness at 50 ever since). My only gripe is that lower luminance ramps display some banding. I was just wondering if 1D unity LUT or a generated one could alleviate this a bit. Again, this is only visible up close on the grayscale/color ramps, does not bother me that much (if at all) in real content. Perhaps one of those bunnies not worth chasing
Any 1D LUT upload can degrade the picture, even a RESET 1D LUT (UNITY).

Its the reason we recommend the users to not send correction 1D or UNITY 1D LUT to LGs.

When you upload a 1D LUT to a picture mode when you reset that picture mode from the TV menu, it will only delete that 1D LUT from the TV, but it will ignore it.

If you want to delete/unload it permanently, then you have to reset the TV to factory defaults.

As I have said many times, all that matters to me is the real content and this looks fantastic after ColourSpace 3D LUT profiling!
I say the same, many times; only dE analysis is part of the success.

But when you have good dE from so many points (1000) as you measured, visualization of your generated 3D correction before sending to TV, visual evaluation with content and patterns that everything is fine, then you have for sure a successful calibration.

It will have for sure a fantastic picture.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #867
Display Profiling


An overview of some of the tools and capabilities built into ColourSpace for display profiling.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #868
ColourSpace Probes Menu


An overview of the ColourSpace Probes menu, and how different settings work with different probes.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #869
Hey ColourSpace users,

If there is any specific video guide you would like to see added, please define it here.

It recorded the 'LUT Generation' video guide now.
 

·
Registered
Joined
·
1,433 Posts
Hey ColourSpace users,

If there is any specific video guide you would like to see added, please define it here.

It recorded the 'LUT Generation' video guide now.
I saw these pop up a few days ago, and just watched them all....really insightful. Many thanks for the time and effort put into them, they are very thorough and professional, and answer a lot of my repeat questions.
I did a Lightspace Calibration a few months back, and think I can move over to the CS now (new projector bulb has sufficient hour). I will wait for you to upload the LUT Generation, then give it a go. Thanks again Ted & Steve.
 
  • Like
Reactions: ConnecTEDDD

·
Registered
Joined
·
9,575 Posts
Discussion Starter #871
ColourSpace LUT Generation Menu


An overview of the ColourSpace LUT Generation menu, and how different settings will generate different results, and assist with poor profile data.
 

·
Registered
Joined
·
7,095 Posts
ColourSpace Probes Menu


An overview of the ColourSpace Probes menu, and how different settings work with different probes.
Ted, Steve, these video's are excellent. Great information!

I'm curious how Steve measured 9,000+ points in 2 hrs when it takes me 2hrs to do a 1,000pt pre-roll and 5,000+ point (17^3) profile with my Klein K10A and Murideo 6G on my OLED.
 

·
Registered
Joined
·
18,288 Posts
Ted, Steve, these video's are excellent. Great information!

I'm curious how Steve measured 9,000+ points in 2 hrs when it takes me 2hrs to do a 1,000pt pre-roll and 5,000+ point (17^3) profile with my Klein K10A and Murideo 6G on my OLED.
He is using a broadcast display in his videos. It does not need a preroll nor additional delays like a WRGB OLED does. Also, when it comes to demo’ing features, it is not necessary to go all out and strive for absolute accuracy and precision.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #875 (Edited)
I'm curious how Steve measured 9,000+ points in 2 hrs when it takes me 2hrs to do a 1,000pt pre-roll and 5,000+ point (17^3) profile with my Klein K10A and Murideo 6G on my OLED.
Hi John,

Steve is using a professional monitor connected to PC output.

Broadcast monitors have less input lag, and the softwarel patch generator doesn't add any delay when its generating the patch.

You can use even 0 sec of extra delay when you calibrate like that way.

Pre-roll and Stabilization are useful only for Plasma/OLEDs profilings.

Even drift frames are not required for specific broadcast monitor models, but you can see that to the first profile you will take, to see if you will use drift frames to your next profile.
 

·
Registered
Joined
·
7,095 Posts
Hi John,

Steve is using a professional monitor connected to PC output.

Broadcast monitors have less input lag, and the internal generation doest have any delay in generating the patch.

You can use even 0 sec of extra delay when you calibrate like that way.

Pre-roll and Stabilization are useful only for Plasma/OLEDs profilings.

Even drift frames are not required for specific broadcast monitor models, but you can see that to the first profile you will take, to see if you will use drift frames to your next profile.
Thanks, understood and good to know.:) The videos are really good and they give a better understanding on how some of the more advanced analysis is done along with the links to the documentation.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #877
Ted, a video on the Hardware Options tab would be great. Thanks.
OK.

The next video will focus on 'Library/Manage Spaces'.
 

·
Registered
Joined
·
117 Posts
Correct me if I'm wrong but on RGB seperation the closer it is to the black line the better it is, or is it the opposite?
 

·
Registered
Joined
·
405 Posts
Just a quick question, Do you have to do a manual calibration to get the 100 ire patch the least error before doing a cube profile? I did an uncalibrated Benq ht9060 projector with no manual adjustments and I think I lost some pop in the picture (not that much though)? I am using the radiance pro to load the LUT

Also, the pattern generator for the brightness and contrast patterns do not show up using the radiance pro as the External tpg. The other patterns on the drop down patterns also do not show in the display.
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #880
Correct me if I'm wrong but on RGB seperation the closer it is to the black line the better it is, or is it the opposite?
Correct RGB Separation (or perfect RGB Separation if you prefer) would have all RGB color lines overlapping the black target line.

The definition of 'good RGB separation' is when the greyscale is an exact sum of the individual RGB values, combined with color channel independence.

Changing the value of one color channel does not affect the other two - there is no cross-talk between the color channels.

Please note this is not a simple RGB/3 calculation and can vary for each and every individual greyscale value.

ColourSpace it will require a 'Primary Only' measurement run to provide RGB Separation chart.

'Primary Only', will take 21-Point Grayscale and 20-Point of Luminance of R,G,B Primaries with 100% Saturation.

Using simple words, ColourSpace will compare the relationship of luminance per each color channel Y between Grayscale and primaries and then generate the chart.

Walter Volpatto, a veteran senior colorist now at Company 3; FotoKEM as past, (Green Book, Dunkirk, Star Wars: The Last Jedi, Interstellar, The Hateful Eight, San Andreas...and many others) was working for LightIllusion many years, he made that post:

''the idea is that the tri-stimuli necessary to present a single color/grey should be independent between each other, or in other words, if a correction is needed to tweak the greyscale for your monitor, this correction will be valid and held for the whole colorspace in a very linear fashion...

a single color A(rgb) has to be comprise of three indipendent stimuli (perfect metamerisms) as

A(rgb) = A(R, 0, 0 ) + A(0, G, 0) + A (0, 0, B)

in decoupling display, a stimuli is a blend os poor light source, poor filters that concur to form an image that is pleasing thru additions and subtractions of other color, or in other words you do not get the purity of the stimuli but a blend that approximate it.

In that case a 1D LUT is NOT a good model representation of your calibration but a more dense 3D LUT is necessary.

It is similar to the cross talk you get in the film negative/positive process...
''

For example, WRGB OLEDs, they tend to have relatively weak RGB Separation, which may or may not be associated with color channel cross-talk (before profiling).

You can have a display with a perfect RGB balance/gamma (0.1dE Grayscale average) but with significant errors in RGB separation.

When you have a perfect Grayscale, it is not enough for a WRGB panel's performance.

The point of profiling is to provide a volumetric correction without issues and without counting any dE formula during that calculation.

The ColourSpace's color engine will process all measurement data and calculate at once the whole colorspace correction.

WRGB OLED's, as they are not RGB additive displays, according to the way they work to produce the picture, one subpixel is always OFF.

Three sub-pixels can light up the same time, with various combinations.

So the whole WRGB concept works by driving the W subpixel as more as possible compared to the other sub-pixels.

W-Subpixel will always produce an unfiltered 10000K (blu-ish) white. At the same time, another two subpixels will try to balance the preserved color.

The diagram below is a quick example of the method the timing controller converts input 10-bit RGB data intro 10-bit WRGB data, for the panel.



The processing unit will zero the value of one sub-pixel, and then it will send the WRGB data to the panel pixels.

We are not doing 1D LUT profiling before the 3D LUT because it does not help the final results.

1D LUT profiling is not the same as the 1D LUT guesswork, a procedure another calibration system uses.

1D LUT profiling, calculating the whole Grayscale taking one read per different grayscale point, counting linearity per color channel, and generating at once a correction 1D LUT. Its how LightSpace works from 2009.

1D LUT guesswork is much worse, as it's trying to correct each point in real-time, based on dE formulas, trying random adjustments until a dE number is low, without counting any linearity from point to point, so for 20-Point Grayscale, it will require 100-200 meter reads or more.

We are not suggesting the users use 1D LUT with LGs if we exclude the fact that using 1D LUT, the processing of the TV will add other issues. (a detail other calibration systems still haven't understand, after many years)

When someone is using 1D LUT, it will just make a more complex the colorspace calculation, as it will add additional active processing, where the color engine of ColourSpace has to count.

When you are performing only 33-Point 3D LUT, including 33-Point of Grayscale inside, the engine will calculate the whole colorspace at once.

When you have an RGB additive display, it can help usee 1D LUT first and then perform 3D LUT, but with WRGB, it makes no sense.

When you are using 1D LUT with LG's, it will bypass the Service Menu White balance calibration values.

Since the panel natively has 2.2 gamma and 10-11K color temp, you will let the 1D LUT reduce these errors. It will require a large amount of data signal manipulation to be applied to fix the RGB balance errors.

When you pre-calibrate using SM WB, you are adjusting the panel calibration directly without using major internal TV processing, so it will not apply any manipulation of input-output data.

I will post examples for RGB Separation in a separate post.
 
861 - 880 of 1224 Posts
Top