Sanity check for player led HDR on Samsung mini led tv

[blackops786187]

Hi, I’m planning to install this build on one of my boxes to get Dolby vision support for my Samsung tv which doesn’t have it and was hoping I can get a sanity check regarding the settings I require to make sure it’s accurate and correct. I did get some info from someone who id consider well read up and an expert on Dolby vision related stuff but I wanted to double check with the dev

My tv according to the edid has a leak luminance of 800 so I would choose the 807 nit value which is fine.

However I was considering the potential issue of a double tonemap scenario. Since in lldv player led Dolby mode, the output is a hdr10 signal, it will reach the Samsung with no metadata ie no maxcll fal or DML.

Samsung utilises the DML to select its static tone curves and it accepts 1000/4000/10,000 nits and ignores all other metadata. If DML is null, it seems to default to the 10,000 static tone curve

of course since the corelec box will already tonemap the image down to its appropriate value based on peak luminance and then send it to the TV without any metadata, the Samsung would then see the null DML and then decide to apply the static 10,000 tonemap unnecessarily creating a double tonemap scenario which isn’t good

the solution I thought out was to send a 1000 MDL metadata to the TV as well. The person I communicated told me that this can be accomplished using the HDR info frame override which can be generated from the excel spreadsheet

so in my case would I simply need to set MaxDL value in the spreadsheet to 1000 and input the result into the infoframe override in the VS10 settings and I should be good to go?

[sunseeker2k5]

I use static tonemapping on my Samsung S90c. Works well with the CPM build (A12) and my Ugoos Am6B+. If you install the skins that expose the detailed technical info, you can see what's passed onto the SAMSUNG and can make sure it is to you liking.

https://github.com/jamal2362/skin.arctic.zephyr.mod.cpm/releases

[blackops786187]

Is there an estuary skin that shows all this data?

Ive tried your skin and using the HDR infoframe calculator, i pasted in a 48 charachter hex core corresponding to 565 MaxDML but i cant see it mentioned at all using your Skin not sure if its being honoured.

[sunseeker2k5]

I don't use the HDR Infoframe, just the standard Dolby VSVDB block so I cannot comment on the HDR infoframe.
The point of the data exposed here is that you can see the values on a frame by frame basis and can decide if what you see on screen is in line with what you expect based on these. (DV levels metadata in the bottom right corner). CPM's own Estuary also exposes these but I don't use that skin so I am not sure to what extent.

[blackops786187]

based on my testing, samsung attempts to apply a static tonemap to a hdr10 signal if maxDML isnt present. Since we are using player Led Dolby Vision, the output is a HDR signal with no metadata attached to it as the tonemapping has been tone by the DV engine before being sent to the TV.

If maxDML isnt present, samsung applies the 10,000 static tone curve which uneccesarily dims the picture hence why im attempting to use the infoframe to send the DML

I might be wrong about this so feel free to correct me

[R3S3t9999]

Ive tried your skin and using the HDR infoframe calculator, i pasted in a 48 charachter hex core corresponding to 565 MaxDML but i cant see it mentioned at all using your Skin not sure if its being honoured.

The static HDR10 MDL and maxcll/fall displayed in the OSD is only the values from the encoded HEVC HDR10 base layer and not the actual value that is being sent to the TV. You need a device like HDfury vertex in the chain to see what's actually being sent.

If your avi infoframe is correct, I see no reason why the the Ugoos wouldn't be sending this specific value to the display.

I don't use the HDR Infoframe, just the standard Dolby VSVDB block so I cannot comment on the HDR infoframe.

You should send proper MDL/maxcll otherwise your TV will fallback to its default tone mapping curve. It would be very stupid for Samsung to fallback to 10 000nits though. It probably(most logically) fallback to 1000nits MDL.

[blackops786187]

I used some HDR10 Clipping test footage to see how the tonemapping was affected by metadata.

https://drive.google.com/drive/folders/17z94AGSQN7m0VLS1oDEkViQlr0wvpWhG

The clip which has no MaxDML looked identical to the MaxDML 10,000 nits and clipped at the exact same point.
The 1000 MaxDML video clipped around 800-900 wheras the 10,000 nit MaxDML and the Null MaxDML video clipped around 3000

DML on samsung tvs atleast seemed to be the only value that affected the choice of the static tonemapping curve

[R3S3t9999]

Well then that's a terrible/stupid implementation since 90% of the HDR movies are grade within 0-1000nits

[blackops786187]

i suspect its to handle edge cases besides HDR movies since youd always expect DML metadata to be present in videos from Movie studios

[R3S3t9999]

Well, you would expect any HDR video to have proper metadata so I don't see the logic in using 10 000nits as fallback.

Unless they updated the library, almost all the early Disneyplus web-dl is missing any HDR10 static metadata. As for bluray disc, there's one that comes to mind: Terminator 2 US UHD-BD.

[R3S3t9999]

I was curious about my C2 fallback curve so I encoded my 1000,4000 and 10 000nits patterns to a couple of different MDL/MaxCLL with the Dolby Vision version of each pattern so you should be able to figure out if the avi info-frame you're sending is working.

So on the LG C2 (800nits peak):

• the fallback curve is 4 000nits when there is no metadata.
• 10 000nits MDL= any maxCLL is ignored and the 10 000nits curve is always used which is surprising
• 4 000nits MDL= maxCLL is used if under 4000nits which I already knew.
• 1 000nits MDL= maxCLL seems to be ignored even for values under 1000nits and the 1 000 nits curve is always used.

Hisense 2022 (2000nits peak):

• The fallback curve is no tone mapping (clipping) when there is no metadata.
• MDL seems to be 100% ignored and only maxcll over 2000nits seems to have an effect. Everything else clip to 1500-2000nits

[blackops786187]

LG got quite the complicated approach. Atleast samsung makes it easy :)

Its harder to eyeball it with these dolby patterns. With luminance set to 807 on LLDV, i can resolve everything including 10,000 nits and it hard to see any distinct difference between setting the infoframe and without.