AVS Forum banner

241 - 260 of 2377 Posts

·
Registered
Joined
·
8,846 Posts
Highlight Handling - LLDV on NX7

I've taken some pictures showing how the NX7 handles highlights. The pictures are taken from the opening scene of the Netflix show Altered Carbon which has a MaxCLL of 10,000 nits, which of course is not typical but is a torture test for highlight handling. Note that all pictures are taken with the same camera exposure and are intentionally unexposed so that the camera itself does not clip any highlights.

The naming conventions I used are:

HDR10 - NX7 HDR10 Picture Mode, Auto Tone Mapping Off (Disabled by ATV4K for LLDV)
FAL - NX7 Frame Adapt HDR, Low
FAM - NX7 Frame Adapt HDR, Medium
FAH - NX7 Frame Adapt HDR, High

This is the first set, with the Vertex "tricking" the ATV4K into the LLDV mode using the HDFury Vertex (LG C8 LLDV Custom EDID)

I will post another set, with the ATV4K in its "normal" HDR mode.

To see the differences you may want to download all the pictures and then flip between them two at a time, for comparison.
 

Attachments

·
Registered
Joined
·
8,846 Posts
Highlight Handling - ATV4K HDR on NX7

Here's the second set, with the ATV4K sending HDR datastream. Same naming convention.
 

Attachments

·
Registered
Joined
·
8,846 Posts
Is the LLDV retaining better highlights? Is that what you're testing Dom?
That’s what I’m testing, but I’ll let the reader form his own conclusion. :)

I was also interested in comparing between Frame Adapt HDR and “static” HDR.
 

·
Registered
Joined
·
1,198 Posts
Hi Dominic.

What, if any, adjustments did you make to your baseline calibration when using the LLDV input?
 

·
Registered
Joined
·
8,846 Posts
Hi Dominic.

What, if any, adjustments did you make to your baseline calibration when using the LLDV input?
I did not make any changes. This was a direct comparison with the same calibration.

I should note that I was using the LG C8 EDID; the Sony A1 EDID makes the skin tone too red, as mentioned by @claw.
 

·
Premium Member
Joined
·
3,798 Posts
Just noticed something interesting. Enabling LLDV disables the NX7's Auto Tone Mapping (the feature that automatically sets the Picture Tone etc based on the metadata). That may be consistent with the previous comment (by claw?) that there's no metadata remaining for the display, after the player handles it.
You could try sending custom HDR10 metadata from the HDfury.

Set Maximum Luminance to either 4000 or 1000.
If you are using Auto, you might set the MaxCLL and MaxFALL to zero which should trigger Auto.
If you are using Low, Medium, or High then set MaxCLL to a value that triggers that mode.

I don't have a new JVC so I am only guessing.
 

·
Registered
Joined
·
8,846 Posts
You could try sending custom HDR10 metadata from the HDfury.

Set Maximum Luminance to either 4000 or 1000.
If you are using Auto, you might set the MaxCLL and MaxFALL to zero which should trigger Auto.
If you are using Low, Medium, or High then set MaxCLL to a value that triggers that mode.

I don't have a new JVC so I am only guessing.
That would be an interesting thing to try. I can even “reactivate” Auto setting in the HDR10 mode that way (it’s currently disabled probably due to lack of metadata).

Note that the settings in FA are Auto, High, Medium, Low; settings in HDR10 are only Auto On/Off.
 

·
Registered
Joined
·
2,027 Posts
I did not make any changes. This was a direct comparison with the same calibration.

I should note that I was using the LG C8 EDID; the Sony A1 EDID makes the skin tone too red, as mentioned by @claw.
Besides the change in Skin tone does the C8 EDID change anything else? I'm wondering if the "too dark" titles some were complaining about are still too dark to watch.
 

·
Registered
Joined
·
6,862 Posts
Discussion Starter #250 (Edited)
That’s what I’m testing, but I’ll let the reader form his own conclusion. :)

I was also interested in comparing between Frame Adapt HDR and “static” HDR.
I am seeing something interesting but aside from any observable benefit, I think it's interesting we see that level of highlight resolution from the Profile 5 layer enhancement. This confirms calibrating the display for the DV LLDV process peak white, is like calibrating for SDR, you resolve around the 1021 white clipping bar and brightest highlights are protected.

The question is, what is the best method for calibrating... reshaping the curve with a 1021 nit peak white, or using a really solid 4000 nit curve and just boosting master contrast to 1021 peak white?

Given how good either of Chad's curves look with the latter approach, Im curious if taking up any more slots for DV LLDV specific curves is really necessary or beneficial.
 

·
Premium Member
Joined
·
3,798 Posts
I did not make any changes. This was a direct comparison with the same calibration.

I should note that I was using the LG C8 EDID; the Sony A1 EDID makes the skin tone too red, as mentioned by @claw.
I watched a new Amazon series last night on my Apple TV4K. It was 4K SDR but I tried forcing Dolby Vision. With either EDID, LG or A1, there was too much red. I won't attempt to watch SDR content as LLDV again.
 
  • Like
Reactions: nathan_h

·
Registered
Joined
·
1,698 Posts
Here's the second set, with the ATV4K sending HDR datastream. Same naming convention.
Very good comparison images.. and I know it doesn't tell the whole story, but it looks to me as though you're gaining highlight resolution at the expense of overall APL and shadow detail - exactly as you can when creating custom curves.

The key here is going to be to hit a sweet spot with your curve design, but whether it will be something you want to use to watch content day to day or whether it becomes another curiosity is a different question...

It's really interesting but it looks like this is just introducing another set of variables to compensate for in gamma.
 

·
Registered
Joined
·
6,862 Posts
Discussion Starter #253
I watched a new Amazon series last night on my Apple TV4K. It was 4K SDR but I tried forcing Dolby Vision. With either EDID, LG or A1, there was too much red. I won't attempt to watch SDR content as LLDV again.
I saw similar issues with forcing 1080P BDs, as well as some other issues.
 

·
Registered
Joined
·
6,862 Posts
Discussion Starter #254
Very good comparison images.. and I know it doesn't tell the whole story, but it looks to me as though you're gaining highlight resolution at the expense of overall APL and shadow detail - exactly as you can when creating custom curves.

The key here is going to be to hit a sweet spot with your curve design, but whether it will be something you want to use to watch content day to day or whether it becomes another curiosity is different question...

It's really interesting but it looks like this is just introducing another set of variables to compensate for in gamma.
What might have moved it beyond a curiosity for me, once adjusted for APL, the presentation of specular highlights was more dramatic, dimensional. Its not something that jumps out all of the time, but when it does, the difference is apparent.

There are instances where the reproduction of the midtones -> peak white, borders on being exaggerated, such that the bright parts of the scene begin to over power the composition of the frame. Ive wondered if part of that could be resolved by a targeted curve top to bottom.
 

·
Registered
Joined
·
1,698 Posts
What might have moved it beyond a curiosity for me, once adjusted for APL, the presentation of specular highlights was more dramatic, dimensional. Its not something that jumps out all of the time, but when it does, the difference is apparent.

There are instances where the reproduction of the midtones -> peak white, borders on being exaggerated, such that the bright parts of the scene begin to over power the composition of the frame. Ive wondered if part of that could be resolved by a targeted curve top to bottom.
It's definitely going to need a targeted curve since the Sony & LG EDIDs are dictating what happens at the output and although there is merit in this since the Base Layer + RPU (Dolby metadata carrier) are processed in the player and the output is therefore "correctly" tone mapped for the respective (Sony or LG) display but I'm not so sure of the merits of then tone mapping again in a projector with completely different specs.
 
  • Like
Reactions: Bytehoven

·
Registered
Joined
·
1,235 Posts
Do you have a HDR10 calibration disc?

If yes, one option to use with your 1000nit curve, is to play with the master contrast. Dominic has noted adjusting the master contrast does more than we might want, but I have been impressed.

1st calibrate for a contrast level that resolves the highest white clipping value DV LLDV passes thru. For me on the x990, that is the 712/724 or 987/1011 nits white clipping bars. I just start to lose these bars at +12 contrast. This is with both my ChadB normal and bright custom curves. Your 1000 nit curve may result in a lower max contrast value.

Next I set up a level that resolves the white clipping value 660/516 nits bar, for me that is +25 on master contrast.

I believe this to be the range of adjustment I have in order to compensate for how HDR10 titles are handled by the DV LLDV processing.

DV LLDV appears to be remapping all white values above 724/1011 nits, to below that max value. It also appears to be doing so in a way that preserves a white diffuse/peak white relationship we have come to understand and value, in order to offer the appearance of specular highlights. If there was criticism, this is one area where the choices DV LLDV is making, might be different than what we can do on a Lumagen or with MadVR. But as I said earlier, in so far as this highlights relationship is being passed on by virtue of the DV mastering meta data, maybe it is correct.

With regard to the adjustments I am making to master contrast, this is the basis of why I am curious about tweaking a custom curve, so that the master contrast could be left at zero, while working to resolve the white clipping values of the DV LLDV rendered HDR10 test pattern. That way larger portions of the PQ curve would not be altered as they are with the master contrast unless intended.

So, try it with you 1000 nit curve. I think it will be fine, maybe better than the 4000+ nit curves we would traditionally want to use for brighter movies.

Regarding a dark movie... this is where calibrating to 660/516 nit white clipping value can come in handy. Experiment with a title like BR2049, as an example of where resolving white clipping bars below 724/1011 nit resolution, produces good results.
Yes I do have the hdr10 calibration disc..

When you say MASTER contrast do you mean with the 640's control or do i need to go into arves tools to do that.. Sorry for silly question.

when I do these changes will i still be able to use javs curves for hdr10 or they will just be good for dolbi vision

I never calibrated my projector i just installed javs curves and follow he's directions as he has them all written down.. I hope your not asking me to take out a light meter and do all that stuff with auto call. I have no clue how to do any of that.. :frown:
 

·
Registered
Joined
·
8,846 Posts
It's really interesting but it looks like this is just introducing another set of variables to compensate for in gamma.
If LLDV uses a fixed EOTF it may be worthwhile investigating. The biggest problem with the old custom curves is that one need to have multiple curves and switch between them depending on the metadata.
 

·
Registered
Joined
·
1,198 Posts
So question is how does one figure out how to shape this curve? I have no experience with this, I haven’t even loaded any custom curves on my JVC.
 

·
Registered
Joined
·
6,862 Posts
Discussion Starter #259
Yes I do have the hdr10 calibration disc..

When you say MASTER contrast do you mean with the 640's control or do i need to go into arves tools to do that.. Sorry for silly question.

when I do these changes will i still be able to use javs curves for hdr10 or they will just be good for dolbi vision

I never calibrated my projector i just installed javs curves and follow he's directions as he has them all written down.. I hope your not asking me to take out a light meter and do all that stuff with auto call. I have no clue how to do any of that.. :frown:
Yes, the Contrast adjustment on the 640. With an imported curve, there are no other options to adjust for APL accept Contrast and Brightness.

Yes to Javs curves, at least until there is feedback regarding how a Javs curve, or any curve could be fine tuned to the nuances of the DV LLDV PQ curve.

No need to learn calibrating.

What source playback devices do you have? Before you go out and spend any $$$, give it just a little more time for some others to reach some conclusions.
 

·
Registered
Joined
·
6,862 Posts
Discussion Starter #260
Very good comparison images.. and I know it doesn't tell the whole story, but it looks to me as though you're gaining highlight resolution at the expense of overall APL and shadow detail - exactly as you can when creating custom curves.

The key here is going to be to hit a sweet spot with your curve design, but whether it will be something you want to use to watch content day to day or whether it becomes another curiosity is a different question...

It's really interesting but it looks like this is just introducing another set of variables to compensate for in gamma.
Dave would ask you to keep in mind, we don't just throw any source into the HDR10 soup on our projectors, a lumagen or anything else. There is some basic calibration of the brightness and contrast, which go a long way to help the source device fit in. Now while, the brightness and contrast may not be enough to perfect the DV LLDV performance, the relative success seen so far, points the way to something far more productive by digging a little further into compensating for the PQ curve response.

In the end, is DV LLDV just a different way to skin the cat? I hear ya, but it seems like there is something else special happening and worth digging out.
 
241 - 260 of 2377 Posts
Top