AVS Forum banner

1 - 20 of 1647 Posts

·
Registered
Joined
·
9,878 Posts
Discussion Starter · #1 ·
This thread is dedicated for posting valuable informations/recommendations for display/meter/pattern generation/software settings and explanation/guides of required steps to perform the best possible 3D LUT generation for all LG OLEDs using internal LUT capabilities or external LUT Boxes using LightSpace software.

Tips/methods/user reports or questions are always welcome.
 

·
Registered
NAD M33, Goldenear Triton One.r towers, Macpro server, Blue Jean cables, LG OLED 55
Joined
·
21 Posts
Hi all, hoping this will be a great place to share info and results.

I ran a 21^3 profile w LS using LS connect and a i1d pro. I set up for 120 nits on my LG 55 E8 because my living room is usually lit with 2 or 3 lamps etc. (wife loves to knit lol)

Anyway, using LS and Teds methods, I was able to achieve an amazing Avg dE2000 of 0.2 in Color Checker etc.

Note: I used LightSpace for calibration. CalMAN was only used for uploading LUT file and to verify LightSpace results!

I also wanted to post a brief history of how I got here.

I began originally by obtaining CalMAN Home and an i1d pro and running mobileforge on a galaxy tab s2. I could never achieve good results tho, no matter what. Then I saw results that others were getting by using LightSpace. So I purchased LightSpace HTL but still was having problems (i.e. if I got my greyscale looking good, my colors were whack or vice versa).

Then Steve Shaw of LightSpace referred me to Ted Aspiotis (displaycalibrations.com). Ted had me send him a LS profile run and discovered that my RGB separation was way off. I told him I was using a Samsung tablet and LS Connect and he suggested I try using something else as a PG as he suspected my Samsung was at fault. He was right. So, I purchased Ted's media files to use as a reference test and got good RGB sep results.

Next step, I bought a Firestick 4K and an LG service remote as Ted had suggested and results were now fine. I followed Ted's instructions on how to profile and do display characterization measurements.

Basically, I mostly use CalMAN to upload the lut file to my LG E8 when Im done but most of the calibration took place in LS CMS. I put together all the steps I used...
 

Attachments

·
Registered
NAD M33, Goldenear Triton One.r towers, Macpro server, Blue Jean cables, LG OLED 55
Joined
·
21 Posts
(My procedure I used to cal my E8)

LG OLED Calibration using LightSpace and CalMAN

Pre-Calibration

Using CalMAN:
Be sure to select the the Display Mode you wish to calibrate. Click the DDC button. Scroll to the 3D LUT Controls.
Press the Reset 3D LUT button!
(Do not ever use AutoCal as It will disable some of your tv controls!)
*see image1

Using LG service remote:
Disable auto screen dimming (ASBL)
In Start, pswrd: 0413 or 0000, 13-OLED, TPC=OFF

In LS Cal Tool Options:
These are your "goto" default settings (only for those who are using the X-Rite i1 Display Pro).

• Integration Time: 0.75
• AIO mode
• Extra Delay: 0.75
• Generic CMF (Raw XYZ in CalMAN)
• Intelligent Interpolation: 1 nit
• Stabilization: 0.75 sec with 28.28.28 custom color patch.
*see Image2

On the LG:
All TV processing off!

Set gamut wide, gamma 2.2, contrast 100, brightness 50, color 50. Use the LightSpace Calibration interface to display a 100% white patch, and use the LG 2 point High RGB sliders to fix RGB balance errors by reducing only values of RGB high sliders on tv, not all sliders, keep one slider at zero, and reduce bthe other two.

Use RGB High to fix RGB Balance, not to control luminance, you will adjust OLED light to get 100 nits (or whatever target you prefer).

Quick Profile:

Next step is to take a pre-calibration Quick Profile of Primary Only, to see that all is well. We want to verify that most of the primaries are outside of the Rec709 gamut, rgb separation is good, etc.

Pattern Gen:
Use 10% Windows with black background.

On the LG:
The gamma 2.2 setting is the TV bypass setting for the 3D LUT to work, nothing to do with what gamma you will ask from LightSpace to generate a correction, which mainly will be 2.4 gamma (the Rec709 selection)

Measure. Check results.

Display Characterization

Go to the Display Characterization window. To create a pre-roll file, select anisometric, 10 points and then click to export as color list, it will export a .cvs file. This need only be done the first time around.

While in the display characterization screen, go to Options and select this file as pre-roll, 1 sec, so LS will display these 1000 patches initially (without taking measurements) for panel to stabilize... and then it will start the profiling.

For profiling you will use 17 point cube, anisometric with drift 100.
*see Image3

Measure.

Check TV Contrast is still 100

I leave it to you to verify your measurements however you wish.

Convert Colour Space

Open the Covert Colour Space tool. In Source, select the Color Space (usually Rec709). In Destination, select your 17p measurement file. Leave Enable Drift Comp checked. Type in a name for this measurement. Select Peak Chroma. Create New. Close.

Export LG LUT FILE

Under File, select Export. Check LG (*.3dl). Enter the name of the file to create and the export to desired location.

Next, using a text editor, you will need to delete the 'space' at the end of the .3dl file header. It should be at the end of line 5. If not, you will recieve an error when trying to upload to LG. Save.

Upload .3DL File to LG OLED TV

From Calman, go back to the 3D LUT Controls menu and click the Load LUT File button. Find your .3dl file and enter.
 

Attachments

·
Registered
Joined
·
629 Posts
Hi Rob these are great results thanks for sharing them and also for you're great guide that everyone will help who is interested in.

I have some questions:
- in your first sentence you said you ran a 21^3 profile and in your guide you say to measure 17^3 so which cube size we can see in you're verification scans?
- you said to use 1.0s for the pre-roll and in your screenshot you set it up to 0.75s can you clear that out?
- youre grayscale verification shows that you measured the internal grayscale points which couldn't be selected when you manual measure the grayscale is you're scan after 1D LUT Autocal with maybe manual adjustments?

For everyone who is not using a fire TV stick as a TPG must set the correct patchscale under the meter settings tab to match the video levels the tv is expecting usually it is 16-235.

After converting the color space you must edit the cube via LS settings -> LUT manipulation -> Videoscale -> pass black.

This is only needed if you're not using a fire TV stick just to prevent some issues.
 

·
Registered
Joined
·
9,878 Posts
Discussion Starter · #7 ·
For everyone who is not using a fire TV stick as a TPG must set the correct patchscale under the meter settings tab to match the video levels the tv is expecting usually it is 16-235.
When someone will use the Raspberry Pi with PG Generator, he has to configure the device to have output PC Levels (RGB-Data/Full) and patch scale @ LightSpace options 16-235, to be able to generate accurate 16-235 RGB Triplets.

For FireStick, the default patch scale (0-255) it works, but when a user want to confirm that he is using correct settings (because there many devices where LightSpace Connect application can be installed), he can use as reference measurements taken from Ted's LightSpace CMS Calibration Disk and Quick Profile primary colors chapter; by playing the Media Files from a TV USB or from a stand alone playey with YCbCr colorspace output; ...and then compare the readings he will get from LightSpace Connect and the supported hardware device he is using, to see if there aggrement to gamma/RGB balance/RGB Separation and black/white levels also.

About PGenerator, there some guides available:

PGenerator Info: https://www.lightillusion.com/pgenerator.html

PGenerator User Guide: https://www.lightillusion.com/pgenerator_manual.html

After converting the color space you must edit the cube via LS settings -> LUT manipulation -> Videoscale -> pass black.
VideoScale Filter is required for eeColor 3D LUT Box users only, not for Lumagen Processor (or any model) users or from those who will export the LG/CM 3dl file format which will load later using CalMAN internally to their 2018 displays.
 

·
Registered
Joined
·
629 Posts
ConnecTEDDD;57446482/pgenerator_manual.html[/URL said:
VideoScale Filter is required for eeColor 3D LUT Box users only, not for Lumagen Processor (or any model) users or from those who will export the LG/CM 3dl file format which will load later using CalMAN internally to their 2018 displays.

Yes you're right my fault I didn't used the Videoscale either on for my LS LUT I uploaded.
I mixed it up with my eeColor notes *
@Rob Loney can you post you're drift chart maybe this is something that could be optimised as well
 

·
Registered
NAD M33, Goldenear Triton One.r towers, Macpro server, Blue Jean cables, LG OLED 55
Joined
·
21 Posts
I have some questions:
- in your first sentence you said you ran a 21^3 profile and in your guide you say to measure 17^3 so which cube size we can see in you're verification scans?
- you said to use 1.0s for the pre-roll and in your screenshot you set it up to 0.75s can you clear that out?
- youre grayscale verification shows that you measured the internal grayscale points which couldn't be selected when you manual measure the grayscale is you're scan after 1D LUT Autocal with maybe manual adjustments?
Yes, I did do a 21^3. It should be stated that that is optional depending on time.
Yes, I cheated and used 0.75 per pre-roll patch but the norm would be 1.0
I did not do any calibration using CalMAN. I'm wondering if anyone else has done that...either b4 or after LS LUT measurement. I suppose it is ok to do so providing you never do a DDC reset right? Otherwise, you get locked out of the WP tv controls and we dont want that. I wanted to be able to tweak but the 1D DDC values are not loaded by default. Has anyone tried this?
@BlackJoker Yes, Ted just sent me some patch sets that should help with that. I notice that my drifts always seem to start low and then rise sharply after...My guess is that is bcuz there are so many of the darker patches towards the beginning of run. I will be trying these new patch sets tonight.

First I'd like to hear whether it could be good to adjust the 1D lut either b4 or after LS.
 

·
Premium Member
Joined
·
119 Posts
Hi all, hoping this will be a great place to share info and results.

I ran a 21^3 profile w LS using LS connect and a i1d pro. I set up for 120 nits on my LG 55 E8 because my living room is usually lit with 2 or 3 lamps etc. (wife loves to knit lol)

Anyway, using LS and Teds methods, I was able to achieve an amazing Avg dE2000 of 0.2 in Color Checker etc.

Note: I used LightSpace for calibration. CalMAN was only used for uploading LUT file and to verify LightSpace results!

I also wanted to post a brief history of how I got here.

I began originally by obtaining CalMAN Home and an i1d pro and running mobileforge on a galaxy tab s2. I could never achieve good results tho, no matter what. Then I saw results that others were getting by using LightSpace. So I purchased LightSpace HTL but still was having problems (i.e. if I got my greyscale looking good, my colors were whack or vice versa).

Then Steve Shaw of LightSpace referred me to Ted Aspiotis (displaycalibrations.com). Ted had me send him a LS profile run and discovered that my RGB separation was way off. I told him I was using a Samsung tablet and LS Connect and he suggested I try using something else as a PG as he suspected my Samsung was at fault. He was right. So, I purchased Ted's media files to use as a reference test and got good RGB sep results.

Next step, I bought a Firestick 4K and an LG service remote as Ted had suggested and results were now fine. I followed Ted's instructions on how to profile and do display characterization measurements.

Basically, I mostly use CalMAN to upload the lut file to my LG E8 when Im done but most of the calibration took place in LS CMS. I put together all the steps I used...
Wow! Fantastic results. I believe these are the best results I've seen yet. Now I know what is achievable. Here is the money question: How would you describe the delta in video improvement vs out-of-the-box as the reference when viewing content?
 

·
Registered
Joined
·
9,878 Posts
Discussion Starter · #12 ·
Hi all, hoping this will be a great place to share info and results.
Hi Ron,

When you are watching the Cube Viewer, the default presentation of dots is 17-Point Cube, but 3dl file of LG is 33-Point Cube, with your right mouse click, you see other cube size options:



There ready selections for 5/9/17/21/33, but if you click you can add or reduce the cube size and the dot size also.



Select 33-Point from there to have greater visualization of your correction, for you and for the future picture you will upload. ;)

Also to the 1D LUT Viewer, with right mouse click, the menu will different setting for 1D LUT, to zoom to specific area etc.

(The pictures I used for that example came from a CalMAN 3D LUT for LG and not from your file, you can identify easily that its from CalMAN because of the yellow dots, problem explained here.)
 

·
Registered
Joined
·
248 Posts
^^^
Is there any tool available to the public that can show corrections made inside the cube?
While this is a great tool for visualization of the overall correction, IMO it does nothing for 'visualizing' what is happening on the inside. These points seen really only show the outer gamut and a general view of the corrections made. I can probably find LUT's that I have made in the past from CM that look good in this cube visualizer, but the LUT is a complete mess with clipping and banding on the inside. This visualizer is helpful, but IMO should not determine if the LUT is good or not.:)
 

·
Registered
Joined
·
2,282 Posts
When you re-profile a display with a LUT active, the cube display is a direct representation of the full volumetric accuracy of the calibration, as the Cube will be a 'perfect' cube if the calibration is accurate.
It is easy to then spot areas that are inaccurate, as the points will deviate from a 'perfect' cube display.

Looking at the actual LUT helps spot gross errors in the making of the LUT, where points anywhere in the volume of the cube are obviously out of logical place. This has often helped highlight a number of issues with alternate calibration systems, as obvious errors can be seen anywhere within the cube..

But, it will never tell you the actual accuracy of the calibration the LUT can perform, as it is showing the correct to be applied - not the result of the correction.

Both uses of the cube display can be very useful, in different ways.

Steve
 

·
Registered
Joined
·
248 Posts
Got it. So I guess what needs to happen is that we need to spend a couple more hours(depending on the meter used) to perform another 9261 point(for a true evaluation) with the LUT active. We technically should be posting these cube visualizations to get a true idea of what these displays are capable of and what LS is doing.

The screenshots of what people are posting now really does not tell the whole story as the LUT is not applied and remeasured yet.
 

·
Registered
Joined
·
9,878 Posts
Discussion Starter · #16 ·
Got it. So I guess what needs to happen is that we need to spend a couple more hours(depending on the meter used) to perform another 9261 point(for a true evaluation) with the LUT active. We technically should be posting these cube visualizations to get a true idea of what these displays are capable of and what LS is doing.
Since you have LightSpace you can use LUT Preview (and not take any measurement which will involve meter repeatability/display stability to the differences you need to locate).

From the 3 verification tools of LightSpace (1D LUT View, 3D LUT Viewer or LUT Preview), use LUT Preview and load your generated correction...then load your own images to see what is happening to grayscale for example (if you load a grayscale ramp) with the correction 3D LUT applied to that image, without taking any measurement, so this saves a lot of time from taking measurements and evaluating with real-patterns from your source.

The LUT Preview Menu opens a standard navigation window that enables the user to select any image to preview the active LUTs within LightSpace. Load any picture you like (TIFF,TGA,PNG etc.) that you will use as a reference.

Be sure that you will use the same levels, I mean if you load a 0-255 file, your correction should be 0-255 levels also (not video-scaled)

When I load my patterns which I have in PNG, I'm applying videoscale to the 3D LUT, because the images are 0-255 but black is 16 and reference white 235 to the images.



I have posted many times examples using LightSpace's LUT Preview, see there a quick link.
 

·
Registered
Joined
·
248 Posts
^^^^
Ted, I understand what you are saying bud. I have used these tools in the past. Your example shows a 'red' gradient. This is helpful to see if you have banding/artifacts in the red channel, but does not show you if 'red' is accurate at different stim/luminance levels. That needs to be re measured to verify that 'red' is accurate.


My point is that users here are posting pics of the corrected LUT via LS cube visualizer. This certainly shows a generalization, but that does not tell the whole story to tell if the LUT is accurate when the LUT is applied. It seems like the only way to tell is that you need to re measure a large amount of points after the LUT is applied to really verify. Users should really be posting those cube's.
 

·
Registered
Joined
·
1,031 Posts
Agreed. The only way to validate a LUT is to measure it in its entirety. The short way is to run a color checker but its... a short way!
 

·
Registered
Joined
·
2,282 Posts
Got it. So I guess what needs to happen is that we need to spend a couple more hours(depending on the meter used) to perform another 9261 point(for a true evaluation) with the LUT active. We technically should be posting these cube visualizations to get a true idea of what these displays are capable of and what LS is doing.

The screenshots of what people are posting now really does not tell the whole story as the LUT is not applied and remeasured yet.
The cube screen shots of the actual LUT do tell you a lot.

As most displays have what are relatively small volumetric issues (small in the sense of gross errors), the cube display will quickly point to errors in the generation of the LUT, as they stand out easily in the cube view of the LUT.

Most 3D LUT calibrations will generate a 'clean' or 'regular' looking LUT, because of the lack of gross errors.

All the errors that have been shown this way have then gone on to be proven as being calibration inaccuracies when the LUT is applied - so it is a very good way to initially asses the LUT generation.

Verification of the LUT when applied will then show the smaller errors that exist when the actual calibration has no gross errors, but is not as accurate as it could be.

Even a 10^3 LUT will give a good indication of the calibration accuracy, but the more points the better - obviously.
We always verify our client calibration with a second 21^3 pass, to prove total volume accuracy.

Steve
 

·
Premium Member
Joined
·
119 Posts
As most displays have what are relatively small volumetric issues (small in the sense of gross errors), the cube display will quickly point to errors in the generation of the LUT, as they stand out easily in the cube view of the LUT.
So, when we see errors in the cube, what we are really seeing (assuming no calibrator error) is the error in the display prior to the correction. So, how can we blame the software for creating a "bad-looking" LUT when in fact it is display (or calibrator) error?
 
1 - 20 of 1647 Posts
Top