| Forums | Register | Polls | Search | Statistics |
 (?)  
You must be logged in to post content on this forum.
Display Calibration Light Illusion Forums / Display Calibration /  
 

ColourSpace Beta - Release Notes & User Operation Questions

 
 
Page  Page 9 of 10:  « Previous  1  2  3  4  5  6  7  8  9  10  Next »

Author webdove
ZRO
Male
#121 | Posted: 20 Mar 2021 23:29 
I happen to have photoshop

So I export a profile LUT and a Null LUT as granger rainbow images. Blend them with photoshop so the low-light pixels (unreliable) come from the null LUT and the bright pixels come from the profile. Import that blended TIF into the LUT tool and export that LUT. It does not appear that I have to do any LUT creation with that imported hybrid LUT image since it loads directly into Colourspace as a LUT.
Web

Author Steve

INF
Male
#122 | Posted: 21 Mar 2021 11:12 
The 'image' is really irrelevant.
It is just there as a visual guide during the manipulation on Photoshop (or other graphics program).

Steve
Steve Shaw
Mob Boss at Light Illusion

Author webdove
ZRO
Male
#123 | Posted: 21 Mar 2021 15:54 
In the LUT Mixing article it says the following:
"From LightSpace export ...
1) a LUT Image combined with the calibration LUT data using the 'Add LUT' function ('Edit / LUT Manipulation / Addition'),
2) or by using existing profile data to build the LUT on-top of the Null LUT Image, as explained above."

Is a third way to do this to

3) import an existing Profile calibrated LUT, add an appropriate image (e.g. Ted's near black bars) and export the resulting LUT Image?
Web

Author Steve

INF
Male
#124 | Posted: 21 Mar 2021 16:25 
With LightSpace, if you add an image into an existing LUT, the image will not be showing the affects of the LUT.
With ColourSpace, any image can be added at any time, using the Add Picture option, and the 'Apply LUT' toggle option used.

This is all explained in the User Guides.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author webdove
ZRO
Male
#125 | Posted: 21 Mar 2021 21:32 
I have a feature suggestion.

The error graphs that I have used plot errors in nits which heavily deemphasize errors at low light levels.

I would like to view a graph of perceptual error for W, R, G, and B.
This graph would have a vertical axis from -10 to +10 where 1 is a just noticeable difference and a horizontal axis from 0.0 to 1.0 which is the fraction of full scale video code (like most plots).

One way to do this would be to plot the measured error as a fraction of the Y target and use the Human Contrast Sensitivity Function vs Luminance (which varies from 1/30 at .1 nits to 1/300 at 100nits) to scale that error as a function of the target luminance.

The Contrast Sensitivity Function looks like this:
https://www.dropbox.com/s/38zmf10j6pxblpu/Contrast%20Sensitivity%20vs%20Luminance.png?dl=0
(The graph comes from http://www.umich.edu/~ners580/ners-bioe_481/lectures/charts/slides/ners-bioe_481-19-13.pdf)
Web

Author Steve

INF
Male
#126 | Posted: 22 Mar 2021 09:21 
We only use 'standards' within ColourSpace.
If the standards change, we will look to add different options to the way graphs are plotted.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author webdove
ZRO
Male
#127 | Posted: 22 Mar 2021 18:14 
I see.
What would be the easiest way to export a profile as a .csv file?
Web

Author Steve

INF
Male
#128 | Posted: 22 Mar 2021 18:28 
A 'profile' cannot be exported as a .cvs file.
The only use for .cvs files is as patch sequences.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author webdove
ZRO
Male
#129 | Posted: 24 Mar 2021 03:36 
As I worked on LUT Mixing to address the near-black defects caused by my i1Display I began to think of what LUT I should be mixing with near-black.

The factory LUTs are not null LUTs. They cannot be because the native panel gamma is not rec.709. What I was hoping to do was to download the LG C9 factory LUT for the most color accurate factory mode (which is Expert Dark) and use that to replace the defective near-black values of my i1Display profile LUTs.

I had the impression from my reading that using DeviceControl one could "backup" the 3dLUT in the C9, but I cannot find a description of how to do that and I can see no download button for the 3dLUT in the DeviceControl LG 2019 Template Rev.5.
Web

Author liberator72
ZRO
Male
#130 | Posted: 24 Mar 2021 07:55 
webdove:
The factory LUTs are not null LUTs. They cannot be because the native panel gamma is not rec.709. What I was hoping to do was to download the LG C9 factory LUT for the most color accurate factory mode (which is Expert Dark) and use that to replace the defective near-black values of my i1Display profile LUTs.

I had the impression from my reading that using DeviceControl one could "backup" the 3dLUT in the C9, but I cannot find a description of how to do that and I can see no download button for the 3dLUT in the DeviceControl LG 2019 Template Rev.5.

There is no "Factory 3D LUT" to download.

Native panel gamma is 2.2. Native panel gamut is close to DCI P3 (without the white sub pixel boost desaturating the colour volume). When the TV receives Rec709 input signaling it uses a colour gamut mapping matrix to output SDR Rec709.

When you press "Reset 3D LUT" all you are doing is preparing the TV by locking its user menu gamut selection to Wide (native) and uploading a Null LUT to replace any previously uploaded custom 3D LUT, you are not resetting a factory 3D LUT. When you press "Restore Factory Values" you are preparing the TV by telling it to bypass that colour gamut mapping matrix.

You can easily see this in action by only Resetting the 3D LUT (do not press restore factory values) and then uploading a user generated Rec709 3D LUT. You will see double mapping to Rec709, thus shrinking the overall gamut coverage.

See this below link for an example of this (with images).

https://www.avsforum.com/threads/lg-oleds-3d-lut-profiling-using-lightspace-thread.3043520/post-60015137

Never has it been claimed that you can "backup and/or restore" factory 1D and/or 3D LUTs using Device Control Templates. What you can do is store user generated 1D/3D LUTs and upload them at any time, which is a huge time saver should an issue arise where you need to factory reset your TV for some reason.

The only way to restore full factory calibration is to perform a full factory reset. No software has access to "download" the factory calibration.

What patch sequence are you using that is causing these "defective near-black values of my i1Display profile LUTs".

If you are using the sequences provided with the guides and you are targeting a peak white of 100 nits at 100% white (approx 124 nits @109%) there are only 6 points within that sequence that should measure below the X-Rite rated accuracy threshold of 0.1 nits (excluding black).



Out of these 6 points, 3 of them compare favourably and within allowable tolerance when compared to a Klein K10-A. As you go below 0.05 nits it will start to deviate further, but there are only 3 points below that range within the patch sequence. Below 0.03 nits is where the i1 Display Pro becomes unreliable.

The image above is that exact patch sequence (with target xyY) but sorted in order of target luminance based on the Rec709 Target Gamut with a Gamma of 2.4 and peak 109% of 123.3358 nits. However, when you are profiling the display it will be in native 2.2 gamma and native gamut so should actually measure brighter than the values shown above (and obviously not hitting its Rec709 target xy either).

It is part of the reason this patch sequence is recommended for LG OLED. You can delete those points if you are concerned, you can see the patch number in the sequence to the left of the image.

You should realise that with LG OLED there is a dithering algorithm at extremely low luminance output levels. It is slightly different on the 2019/2020 LG OLEDs when compared to the 2018, but it makes extremely low luminance output levels very difficult to measure even with the most expensive (what is considered "reference") meters. Instead of trying to manually manipulate a LUT, you may be better off attempting to generate a LUT to a slightly lower gamma value, I have found a value of 2.35 to work well on troublesome panels that struggle to come out of black properly.

Or, you could try Parametric Gamma

Author liberator72
ZRO
Male
#131 | Posted: 24 Mar 2021 08:54 
Oh, and it is worth mentioning also that considering you are talking about a C9 (I somehow missed this info before), there is a known issue with the 2019 LG OLEDs that seems to intermittently affect some users (not all) to varying degrees of severity.

In one of the later FW updates, an issue was introduced that seems to affect "near black". At times it will clip level 17 and affect levels anywhere from 19 - 27 (lowering their output) depending on panel. It doesn't always happen and sometimes reverts back to normal after a power cycle and will stay fine for some time before eventually happening again.

This is an LG issue, not a LUT issue. It can occur in calibrated or uncalibrated modes, and is completely intermittent. It is highly possible that you could be one of those that suffer from the issue on a more severe level than others do and no amount of LUT manipulation can fix it, because if you fix it for when the bug is active, it will be broken again when the bug is inactive.

One user has been documenting his issues with this over on AVS, even going to the extent of purchasing a new main board with his own cash to get a board with an older FW version that didn't have the issue. You can read parts of it from the following link.

https://www.avsforum.com/threads/2019-c9%E2%80%93e9-owners-thread-no-price-talk.3057634/post-60401690

Author webdove
ZRO
Male
#132 | Posted: 24 Mar 2021 18:26 
Thanks Lib.
"Native panel gamma is 2.2. Native panel gamut is close to DCI P3. TV ... uses a colour gamut mapping matrix to output SDR Rec709".
That is my understanding.

When I reset TV Expert Dark and set O29 Cn82 Br50 Cl52:
Expert Settings are Gamut Auto (unlocked), Gamma BT.1886 (unlocked). Grayscale matches rec709 in color and gamma.

When I enable calibration in Device Control:
Expert Settings become Gamut Wide (locked), Gamma 2.2 (unlocked). Grayscale now matches P3 color and 2.2 gamma.

I was mistaken in my belief that factory Expert Dark was handling gamma with a LUT. Apparently it is using a non-LUT 1D non-linear mapping enabled with the Gamma BT.1886 setting.

It appears that DeviceControl automatically sets Gamut Wide, Gamma 2.2 as soon as I enable calibration.

I have measured my Factory Expert Dark with my TE cooled astronomy camera and it is very accurate from 8bit limited codes 16 to 27.
The camera counts electrons times a scale factor and adds an offset of 100 to avoid dark clipping.
I was using 1 second exposures. I suspect the LG low-light dithering is significantly faster. I did not notice much variation on successive exposures.

Video-code Target-Y Camera-counts = Camera Y
16 0.0000 100 (100-100)/29300 = 0
17 0.0002 109 (109-100)/29300 = 0.0003
18 0.0012 135 (135-100)/29300 = 0.0012
20 0.0065 284 (284-100)/29300 = 0.0063
23 0.0249 895 (895-100)/29300 = 0.0271
25 0.0475 1430 (1430-100)/29300 = 0.0454
27 0.0737 2280 (2280-100)/29300 = 0.0744

I can clearly see when I look at Ted's dark gray flashing bars and switch from factory Expert Dark to 3d LUT corrected Expert Bright that the latter is much darker.

It appears that I need to create a LUT that maps P3 colors & gamma 2.2 (an idealized LG panel) to 709 colors & gamma 2.4 (HDTV) and then mix that LUT with my measured Expert Bright Profile LUT using the LUT Mixing method at low luminance value.

In LUT Tools if I use Destination DCI-P3 D65 - 2.2 gamma (an idealized panel) and Source rec709 I should get a LUT with the idealized mapping.

Comments?
Web

Author liberator72
ZRO
Male
#133 | Posted: 24 Mar 2021 19:28 
webdove:
It appears that DeviceControl automatically sets Gamut Wide, Gamma 2.2 as soon as I enable calibration.

Yes, it will always do this if you have previously used Device Control to reset anything in that picture mode. The only way to prevent this is to factory reset the TV to remove all traces of custom LUT resets or uploads. This is not due to how Device Control works, it is how the LG works.

webdove:
I have measured my Factory Expert Dark with my TE cooled astronomy camera and it is very accurate from 8bit limited codes 16 to 27.
The camera counts electrons times a scale factor and adds an offset of 100 to avoid dark clipping.
I was using 1 second exposures. I suspect the LG low-light dithering is significantly faster. I did not notice much variation on successive exposures.

Video-code Target-Y Camera-counts = Camera Y
16 0.0000 100 (100-100)/29300 = 0
17 0.0002 109 (109-100)/29300 = 0.0003
18 0.0012 135 (135-100)/29300 = 0.0012
20 0.0065 284 (284-100)/29300 = 0.0063
23 0.0249 895 (895-100)/29300 = 0.0271
25 0.0475 1430 (1430-100)/29300 = 0.0454
27 0.0737 2280 (2280-100)/29300 = 0.0744

I am not familiar with that type of camera or how accurately it would measure a display (especially one like WOLED). Maybe someone else can comment on it. What I can say for 100% certainty is that the patch sequences provided with the LG OLED calibration guides has been specifically selected for several reasons, one of which is that it has the absolute minimal amount patches below the read capability of the i1 Display Pro (as shown in the image I provided earlier). If you are concerned that these are somehow corrupting your profile (hint, they are not) then just delete them.

I can also tell you that I have personally compared the i1 Display Pro against the Klein K10-A, using the ColourSpace "Remote Control" function to first create a meter correction profile for both meters to a Jeti 1501 (so ColourSpace controlling all three probes simultaneously to read the exact same patch at the same time). Then, once again using "Remote Control" I have measured 1000's of points with those two profiled meters, including several custom near black sequences. The results are truly exceptional and proves there is nothing wrong with using an i1 Display Pro for calibrating an LG OLED using the patch sequences provided.

I have also measured a full display profile with the i1 Display Pro, generated the LUT and verified it with the Klein K10-A,. Again the results were far better than I expected (and I already expected them to be very good).

Maybe one day I will get around to posting the data to show this, I just don't have the time right now to compile it into a presentable post.

webdove:
I can clearly see when I look at Ted's dark gray flashing bars and switch from factory Expert Dark to 3d LUT corrected Expert Bright that the latter is much darker.

That may be the case, but it might also be that the darker version is the correct version. I know you say you are using an astronomy camera for your reference measurements, but can you be 100% certain that it is reading the light output of the TV accurately? I don't know anything about them so I have no idea and I won't even attempt to speculate. But I am measuring these displays with a Klein K10-A profiled to a 5nm Jeti 1501 Spectroradiometer, and when comparing a calibrated vs uncalibrated picture mode, I too see that the calibrated one is darker. Darker, but measurably more accurate.

But you also have to take into account that some 2019 LG OLEDs seem to have the near black bug I mentioned earlier, and if your particular set is suffering from this bug then nothing you do will correct it. You just have to wait and see if LG address and fix it.

I'm not really sure what else to say other than I don't believe your i1 Display Pro is causing defective near black values in your LUTs if you are using the sequences provided.

Author webdove
ZRO
Male
#134 | Posted: 24 Mar 2021 21:09 
I did the LUT mixing that I mentioned above and it brightened the 17-25 bars and eliminated a kink in the RGB balance. I also found that I like running the initial extended profile using the colorspace definitions from the i1Pro calibration and gamma 2.2. The resulting RGB box values for primaries and Grayscale are mostly green, they exactly fill the box and you can easily see the color errors in the primaries with tangents enabled. Running a new probe match & profile now for Expert Bright.
Web

Author webdove
ZRO
Male
#135 | Posted: 24 Mar 2021 21:20 
The astronomy camera is an SBIG ST402. This is a temperature controlled kodak CCD which I run at -10C to reduce noise. You take black images (identical in duration to the light images, but with light blocked) and bias images (extremely short duration dark images to measure pixel bias). Light images are "calibrated" using the dark and bias images to eliminate pixel to pixel variation. I then average the entire image to a single value to minimize noise. It is a scientific instrument which is designed to precisely measure the luminance of astronomical objects to look for variations over time so I expect it is comparable in accuracy to the Klein, but without the luminance conversions built in. It does have red, green and blue filters though I was not using them.
Web

Page  Page 9 of 10:  « Previous  1  2  3  4  5  6  7  8  9  10  Next » 
You must be logged in to post content on this forum.
Display Calibration Light Illusion Forums / Display Calibration /
 ColourSpace Beta - Release Notes & User Operation Questions

 

 
 
Online now: Guests - 1
Members - 0
Max. ever online: 192 [11 Jan 2023 08:39]
Guests - 192 / Members - 0