AVS Forum banner

181 - 200 of 1596 Posts

·
Registered
Joined
·
600 Posts
I tested the LG template from device control in combination with the latest Lightspace release and could achieve stunning results. At this stage I can say that I really have no need to upgrade to the C9 this year because this is a huge step in picture quality for SDR.

So how does it work?

Setup:
- LG with 4.10.31 firmware (needs engineering mode in Service Menu)
- ASBL deactivated
- DVDO TPG @ YCbCr 4:2:2 - 12 bit
- Lightspace 10.0.0.2922
- CalMAN Ultimate 5.9.2.59 for verification and cross measurements
- Device Control with LG template
- id3 profiled in Hi-Res Mode with i1Pro2
- 21^3 LUT

LG TV:
- All processing off
- Contrast 100 for measurements
- Contrast 85 for verification
- OLED Light 25 ~ 115 nits

First of all I updated to the upcoming firmware release of the LG C8 4.10.31 because they fixed the near black problems and changed the dithering also. I measured the difference and it's also easily to see with the eye.

Firmware 4.10.15



Firmware 4.10.31



After that I reseted 1D and 3D LUT via Device Control which locks some of the internal controls of the LG. Then I measured a Grey only quick profile and created a 1D LUT which I uploaded to the LG via Device Control. Unfortunately I had problems to create a new LUT with Aug data so I skipped that part.

Then I created a 21^3 cube with a custom patchset. I also used the pre roll function to set the display in the right power consumption conditions which helps to reduce the drift later.

It took me little more then six hours with my setup but it is worth it



The drift is good and stable just at the end the temperature drops just a little bit















Ive never had a better pq before and the combination of the new firmware with the Lightspace LUT is beyond everything I achieved before with CalMAN. I hope they fix they 3D LUT problems in the future. Anyone who really wants to get out the best in terms of pq needs to try this. BTW I didn't used Aug data after 3D LUT upload because it messed up the gamma maybe someone else can check that either. My next try will be a 17^3 with LS standard patch sequence and I will not reset 1D LUT I just do a 2p adjustment and will only reset 3D LUT via Device control because that only locks the gamut setting. But not now because for the moment I just want to watch movies without thinking about my calibration and relax.

Thanks to Ted, Steve and all the others that maked this possible :)
 

·
Registered
Joined
·
49 Posts
Setup:
- LG with 4.10.31 firmware (needs engineering mode in Service Menu)
- ASBL deactivated
- DVDO TPG @ YCbCr 4:2:2 - 12 bit
- Lightspace 10.0.0.2922
- CalMAN Ultimate 5.9.2.59 for verification and cross measurements
- Device Control with LG template
- id3 profiled in Hi-Res Mode with i1Pro2
- 21^3 LUT

LG TV:
- All processing off
- Contrast 100
- OLED Light 25 ~ 115 nits
Thanks for posting your results.

You can actually download the 4.10.31 firmware directly from the LG website now. Also, can you explain why you used Contrast 100? I'm getting mixed info on whether I should use 85 or 100 when using the LS workflow.

I plan to do a full reset tomorrow and generate new LUTs. Is the new Devicecontrol finally available or do you have early access to it?
 

·
Registered
Joined
·
600 Posts
I used contrast 100 to have more headroom for the LUT. When you do the verification later you set back to 85.

When you use CalMAN for example you can not use contrast 100 because every time when you hit AutoCal it will set contrast back to 85 it's hard coded because the internal LUT needs 85 to bypass.

With LS you don't have this limitation. I had early access to the template but I'm sure it will be released soon.
 
  • Like
Reactions: clifak

·
Registered
Joined
·
9,575 Posts
Discussion Starter #184
Thanks for posting your results

Also, can you explain why you used Contrast 100? I'm getting mixed info on whether I should use 85 or 100 when using the LS workflow.
I used contrast 100 to have more headroom for the LUT.

When you use CalMAN for example you can not use contrast 100 because every time when you hit AutoCal it will set contrast back to 85 it's hard coded because the internal LUT needs 85 to bypass.

With LS you don't have this limitation.
About the question of the Contrast setting of 100 Fabio used, which look like an unusual setting for CalMAN users, as CalMAN recommendation were saying the Contrast should be set to 85 and brightness to 50 for LUT’s to work as designed. (bypass settings)

We found (with help of Maciej Koper from HDTVPolska.com also) that the bypass setting of Contrast 85 works when you are using 16-255 patch generation.

When you are using 16-235 patches, you have to set contrast to 100 when you will measure, and after the LUT generation and everything loaded internally to the TV, then you can restore it back to 85 when you will take post-calibration verification.

But you need to set the OLED Light in a setting where you will have about 20% more nits (for example to 120 nits) then you will have Contrast @ 100, because later when you will reduce Contrast to 85, it will reduce your luminance output also, so you will have 100 nits @ verification.

Measure for post-calibration (with above method), it will provide this slightly better results from keeping Contrast @ 85 when you are using 16-235 patch generation.

As Brightness/Contrast are positioned before 3D/1D LUT's, you don't want any change of these controls to be performed to your input signal while you will taking measurements for LUT's generation.

So for these controls to work as bypass, when you are using 16-255 patches, the Contrast 85 is the bypass while when you will use 16-235 patches the Contrast 100 is the bypass.

Image processing chain seems to be: Input Signal -> Contrast/Brightness -> 3D LUT -> 1D LUT (gamma preset, white balance, CMS) -> OLED Light -> Display's Panel.

With this processing chain, you must set contrast this way to make incoming signal levels match LUT level during profiling.

But after profiling you can change Contrast, it will not mess the LUT calibration because it processes the signal before LUT is applied.

The same procedure has to performed with all displays that have internal LUT full range, its not only for LG.

The general rule is that everything before LUT's (from input signal side) must be bypassed before profiling, and can be adjusted after LUT calibration. Everything placed after LUT can be adjusted before profiling and can’t be changed after calibration. OLED LIGHT can’t be touched after 3D or 1D LUT, and 1D LUT must be done first (if used) with 3D LUT bypassed (UNITY).

The input signal must match the dynamic range of a display to make the LUT align to signal levels correctly.

When profiling with CR:85 / BR:50 the signal must be delivered in 16-255 range.

When profiling with CR:100 / BR:50 the signal must be delivered in 16-235 range.
 

·
Registered
Joined
·
13,714 Posts
I tested the LG template from device control in combination with the latest Lightspace release and could achieve stunning results. At this stage I can say that I really have no need to upgrade to the C9 this year because this is a huge step in picture quality for SDR.

So how does it work?

Setup:
- LG with 4.10.31 firmware (needs engineering mode in Service Menu)
- ASBL deactivated
- DVDO TPG @ YCbCr 4:2:2 - 12 bit
- Lightspace 10.0.0.2922
- CalMAN Ultimate 5.9.2.59 for verification and cross measurements
- Device Control with LG template
- id3 profiled in Hi-Res Mode with i1Pro2
- 21^3 LUT

LG TV:
- All processing off
- Contrast 100 for measurements
- Contrast 85 for verification
- OLED Light 25 ~ 115 nits

First of all I updated to the upcoming firmware release of the LG C8 4.10.31 because they fixed the near black problems and changed the dithering also. I measured the difference and it's also easily to see with the eye.

Firmware 4.10.15



Firmware 4.10.31



After that I reseted 1D and 3D LUT via Device Control which locks some of the internal controls of the LG. Then I measured a Grey only quick profile and created a 1D LUT which I uploaded to the LG via Device Control. Unfortunately I had problems to create a new LUT with Aug data so I skipped that part.

Then I created a [/b]21^3 cube with a custom patch-set[/b]. I also used the pre roll function to set the display in the right power consumption conditions which helps to reduce the drift later.

It took me little more then [/b]six hours[/b] with my setup but it is worth it



The drift is good and stable just at the end the temperature drops just a little bit















Ive never had a better pq before and the combination of the new firmware with the Lightspace LUT is beyond everything I achieved before with CalMAN. I hope they fix they 3D LUT problems in the future. Anyone who really wants to get out the best in terms of pq needs to try this. BTW I didn't used Aug data after 3D LUT upload because it messed up the gamma maybe someone else can check that either. My next try will be a 17^3 with LS standard patch sequence and I will not reset 1D LUT I just do a 2p adjustment and will only reset 3D LUT via Device control because that only locks the gamut setting. But not now because for the moment I just want to watch movies without thinking about my calibration and relax.

Thanks to Ted, Steve and all the others that maked this possible :)
Thanks for this summary. So a 21^3 cube (9261 pts) took ~6 hours and your next try will be a 17^3 cube with 4913 pts should take about 3 hours right? You are using the smaller profiling LUT to save time and to check whether you see any difference in the outcome, right?
 

·
Registered
Joined
·
168 Posts
Wow, so the brightness and contrast controls are not affecting the 1D LUT? That's a strange choice made by LG.

According to the link from hometheaterreview, the CMS controls on the LG are working in linear space, and require a gamma-space -> linear-space and back conversion. This means the CMS controls can't affect the 1D LUT. This processing must happen somewhere else.

How about when calibrating for the full range 0-255 8 bit RGB signals. What should be the contrast value in this case? Any ideas?
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #187
Thanks for this summary. So a 21^3 cube (9261 pts) took ~6 hours and your next try will be a 17^3 cube with 4913 pts should take about 3 hours right? You are using the smaller profiling LUT to save time and to check whether you see any difference in the outcome, right?
It takes such time because except the delay before each patch read you have to use the black or gray frame insertion (stabilization) before each patch is displayed. (+ Pre-Roll add some time).
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #188
Wow, so the brightness and contrast controls are not affecting the 1D LUT? That's a strange choice made by LG.
This is not what I posted, read again, if you do that specific procedure, restoring back to 85 the contrast after the end of calibration, it will harm your results, (becasue 85 is the bypass setting)

For 0-255, its working with LightSpace only (CalMAN don't have that option). Contrast should be at 85.
 

·
Registered
Joined
·
168 Posts
As Brightness/Contrast are positioned before 3D/1D LUT's, you don't want any change of these controls to be performed to your input signal while you will taking measurements for LUT's generation.
Ok, so what have I misunderstood in this statement? It says "Brightness/Contrast are positioned before 3D/1D LUT's". Did you mean something else?
 

·
Registered
Joined
·
13,714 Posts
Hi Miki,
One way would be to use Excel, or similar, to merge the selected data into a third file.
All .bcs files are just XLM, so easily human readable.

An alternative, is again to use the LUT Image, and in Photoshop lay one image over the other, and then key out the different range from one LUT to the other.

Steve
Steve, I have a question that may require this same merging of profile data in Excel.

I understand that it's recommended to run 21^3 or 17^3 profile LUTS in the interests of time (no way to read a 33^3 LUT with 36,000 points ;)).

But I'm interested to know if there is any easy way to take readpoints at full 33^3 density near-black (granularity of 3% IRE) at least for the first 5 points near black (0-15%) and then merge those finer-granularity readpoints into a less-dense cube, say a 17^3 cube reading at every 5.9% IRE beyond 15% IRE.

Another way of thinking about what i'd like to do is to take a full 33^3 profile cube reading only one point for every 4 after the first 5 sets of near-black measurements (which I suppose could total to as many as 5445 points) meaning 5445 points at 3% IRE near black (up to 15% IRE) and then a lower-density cube for the remaining 85% containing 14^3 points measured at 6% (2744 points).

This could be through merging of 2 different density and different-sized profile LUTs using Excel or it could be through a custom patchset, but I'm interested in ways I can use a higher-density of readpoints at lower IRE levels near-black (15% and lower) combined with a lower density of readpoints beyond 15% IRE.

Any thoughts as to how I can take maximal advantage of LGs internal 3D LUT, especially in the near-black region (which supoorts 3% IRE granularity being 33^3) appreciated.
 

·
Registered
Joined
·
2,234 Posts
No, that is not possible.
And regardless, unless you have a very accurate probe, that can read low-light patches very, very accurately, the results would be worse, as all you would have is a lot more bad data points.

Steve
 

·
Registered
Joined
·
13,714 Posts
No, that is not possible.
And regardless, unless you have a very accurate probe, that can read low-light patches very, very accurately, the results would be worse, as all you would have is a lot more bad data points.

Steve
OK, thanks. So a profile LUT essentially needs to be uniform-density.

Here is an RGB graph of near-black measurements taken off of my 65C6 with an i1DisplayPro with HCFR (below):

Are you saying these measurements are meaningless? I achieved this level of linearity and balance followng a great deal of effort screwing around with low-level controls on the C6 and when I was done, the measurements came out like this consistently (after temperature stabilization) and the image also looked better to my eyes.

Perhaps the 2019 WOLEDs have improved near-black linearity so much that this level of effort is no longer needed, but until that's confirmed, my experience is that the greatest bang-for-the-buck in terms of WOLED calibration comes from focusing attention on near black

LG provides 3D LUT control points with 3% IRE resolution and I'm interested in any thoughts you have as to how to best to capitalize on that control granularity in the
 

Attachments

·
Registered
Joined
·
13,714 Posts
I used contrast 100 to have more headroom for the LUT. When you do the verification later you set back to 85.

When you use CalMAN for example you can not use contrast 100 because every time when you hit AutoCal it will set contrast back to 85 it's hard coded because the internal LUT needs 85 to bypass.

With LS you don't have this limitation. I had early access to the template but I'm sure it will be released soon.
BlackJoker, seems as though you are truly on the forefront here as far as being an owner/enthusiast (no a pro) using LS to calibrate LG internal LUTs.

I'm interested in the near-black accuracy of your 3D LUT calibration and wonder if you could measure 1% - 5% Near Black following your next calibration.

I measured the attached using HCFR. Hopefully Calman supports a similar measurement (1%-5% Near Black displayed in RGB).
 

Attachments

·
Registered
Joined
·
2,234 Posts
OK, thanks. So a profile LUT essentially needs to be uniform-density.

Here is an RGB graph of near-black measurements taken off of my 65C6 with an i1DisplayPro with HCFR (below):

Are you saying these measurements are meaningless? I achieved this level of linearity and balance followng a great deal of effort screwing around with low-level controls on the C6 and when I was done, the measurements came out like this consistently (after temperature stabilization) and the image also looked better to my eyes.

Perhaps the 2019 WOLEDs have improved near-black linearity so much that this level of effort is no longer needed, but until that's confirmed, my experience is that the greatest bang-for-the-buck in terms of WOLED calibration comes from focusing attention on near black

LG provides 3D LUT control points with 3% IRE resolution and I'm interested in any thoughts you have as to how to best to capitalize on that control granularity in the
 

·
Registered
Joined
·
13,714 Posts
Thanks for the clarification, Steve. More questions below:

No, the profile doesn't need to be a uniform density.
But the LUT generated will be.
I understand that the 33^3 3D LUT genetated for download into TV will be uniform density - is that what you are referring to?

If the 'profile' does not need to be uniform density, do you have a pointer to where I can understand how to generate a non-uniform profile?

That will kill processing times when you have a lot of points in a given area of the profile.
Not cettain if you are referring to the 'processing time' assiciated with using the 3D LUT when processing video input or the processing time associated with genetating a 3D calibration LUT from a non-uniform/dense characterization profile...

Your graph is just the grey scale, so has maximum RGB levels for the given patch.
The same patches that are, say, mainly blue will be a lot lower in brightness, so will not measure anywhere near as accurately.
Gotcha, so while an i1DisplayPo may be able to provide a useful measurement at 3%, 3%, 3%, it will be unable to measure 3%, 0%, 0% (for example). Since in the shadow-detail region I am interested in, it is primarily luminance I care about (rather than chrominance), I'm still interested in the best accuracy possible at 3% grey, even if measurements of 3% blue or 3% red or 3% green are meaningless. Is there a way to achieve that? (follow-on questions below).

For enhanced grey scale you can use the 'Augment' process with LightSpace.
That will enable the use of more grey scale point measurements,

Steve
Yes, I've understood how the Agment process allows a 3D LUT to be refined. So it sounds like that is the process I want to use. Will Augment support, for example, generating a 33^3 LUT from a 17^3 profile amd the augmemting with a 5^3 LUT captured only over the darkest 15% IRE?

I'll poke around more on the LS website to try to learn more, but any pointers appreciated ;).

P.S. The i1DisplayPro colorimeter has three photosensors, each filtered to correspind to one primary: https://displaycalibrations.com/x-rite_i1_measurement_solutions_info.html

Doesn't that mean reading 3% Blue in the presence of no Red or Green light will be exactly the same as far as Blue sensitivity to reading 3% Blue as part of 3% Grey (with 3% Red and 3% Green light as well)? :confused:
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #196
Ok, so what have I misunderstood in this statement? It says "Brightness/Contrast are positioned before 3D/1D LUT's". Did you mean something else?
I was talking for the patch generation, when you display patches what is happening.
 

·
Registered
Joined
·
168 Posts
I still don't understand. In your post, it also says this:

Image processing chain seems to be: Input Signal -> Contrast/Brightness -> 3D LUT -> 1D LUT (gamma preset, white balance, CMS) -> OLED Light -> Display's Panel.
My take on your words is that the brightness and contrast controls cause some sort of transform to be performed before the LUTs. And that in order to be able to calibrate the LUTs, one has to configure that transform so that the input values which reach the 3D LUT are full range, since the LUTs are full range.

In other words, the brightness/contrast controls are not used to compute the 1D LUT values, they are configuring a transform in front of both LUTs.

I don't have any inside knowledge about the color pipeline internals in these TVs, this could very well be how it is.

Am I misunderstanding what you were saying?
 

·
Registered
Joined
·
600 Posts
@fafrd of course I can measure 6 Points near black but be aware because I'm using an id3 which is limited in this area. What I can say for sure is that it's definitely worth to measure a 21^3 Cube with the 2018 sets. My 17^3 has some more problems with the edge points of the color gamut especially with red and cyan.

With my eecolor I didn't had this issue so the internal LUT is definitely suffering of some kind of signal problems. Also the Aug data didn't heped to improve the results it makes it even worse so for now I can say to skip that I hope for feedback from other users as well later on.

I still need to do some more tests I think there's still some potential for tiny improvements. Like Ted already said you have to measure with contrast 100 to bypass the signal and switch later on back to 85 for verification but then the gamma is way off. So I will try a cube also with contrast at 85 from the beginning this is how I used it in the past with eecolor also.

Next week I get a review sample of the 2019 models and I will do my best to upload a Lightspace LUT into it as well Im pretty sure it is possible too.

Stay tuned.
 

·
Registered
Joined
·
13,714 Posts
@fafrd of course I can measure 6 Points near black but be aware because I'm using an id3 which is limited in this area.
Thanks. Those near-black measurements I posted were measured with an iD3 (i1DisplayPro).

What I can say for sure is that it's definitely worth to measure a 21^3 Cube with the 2018 sets. My 17^3 has some more problems with the edge points of the color gamut especially with red and cyan.

With my eecolor I didn't had this issue so the internal LUT is definitely suffering of some kind of signal problems. Also the Aug data didn't heped to improve the results it makes it even worse so for now I can say to skip that I hope for feedback from other users as well later on.

I still need to do some more tests I think there's still some potential for tiny improvements. Like Ted already said you have to measure with contrast 100 to bypass the signal and switch later on back to 85 for verification but then the gamma is way off. So I will try a cube also with contrast at 85 from the beginning this is how I used it in the past with eecolor also.
Good luck!

Next week I get a review sample of the 2019 models and I will do my best to upload a Lightspace LUT into it as well Im pretty sure it is possible too.

Stay tuned.
Double good luck (and I suspect you're probably right - word of the 'hack' didn't get out until long after the 2019 WOLED design was frozen...
 

·
Registered
Joined
·
9,575 Posts
Discussion Starter #200
I still don't understand.
Contrast @ 85 and Brightness @ 50, you should have all the time when you will use internal LUT capabilities.

What I posted was a Contrast change during patch generation procedure.
 
181 - 200 of 1596 Posts
Top