Thanks you Tyler and Ted, that was a good idea!
Originally Posted by chros73
So the question is ... that do the different 19. WB point values represent the 3 tone curves?
Short answer: no. They rely on the MDL metadata value (as we saw earlier):
the "real peak nits" setting (that results in MDL and maxCLL values as well) in madvr pixelshader with "output HDR format" for the 3 different 9. WB point values:
- 1.: 669 (C8) , 669 (B8) : 100 - 760
- 2.: 696 (C8) , 692 (B8) : 761 - 2500
- 3.: 713 (C8) , 705 (B8) : 2501 - 10000
Originally Posted by chros73
5. *and* this example shows that maxCLL is ignored completely!
That's wrong: display receives maxCLL fine.
0. Test setup:
- nvidia driver version: v385.28 (passing all the HDR metadata to the display)
- madvr version: madmeasure107
- HCFR version: 220.127.116.11 (1% window, black frame insertion in every 2nd frame)
- meter: i1DisplayPro
- LG 65B8:
-- firmware: 4.10.25 (doesn't have the black crush that later firmwares introduced)
-- using: PC mode, RGB Full range
-- HDR10 Cinema preset: everything is at default setting (oled light 100, contrast 100, brightness 50), couple of processing settings are turned off, Cinema preset is uncalibrated (means factory settings)
-- DTM is off
-- peak nits using 1% windows: ~700 nits (C8 supposed to be + ~200 nits brighter)
File name conventions for the graphs: pq|rolloff-mdl-maxcll-maxfall-19point[-movieinitials]
1. maxCLL works fine, the above mentioned 3 test cases:
ro-1000-1000-512 | ro-4000-4000-512 | ro-4000-577-512
I wanted to say the MDL is not relevant when maxCLL is present and maxCLL < MDL but take a look at this when maxCLL is close to 10000:
ro-4000-10000-512 | ro-6900-10000-512 | ro-10000-10000-512
Otherwise it doesn't seem to change anything:
ro-4000-577-512 | ro-1000-577-512
ro-4000-10000-512 | ro-4000-9977-240
It seems that the TV doesn't use it at all *or* it's used by ABL algo(s).
4. Weird cases:
a) Curve is totally broken when maxCLL *and* MDL is close to 10000
, like with Spears & Munsil's 10000-10000-xxx demo video:
ro-10000-10000-512 | ro-4000-9977-240
b) Something strange is happening with maxCLL between 500 and 501
, note how bright the image gets between 501 and (let's say 670), like with Guardians of the Galaxy (2014):
ro-333-333-111 | ro-420-420-140 | ro-500-500-167 | ro-501-501-167
Above maxCLL 501 it tries to correct itself:
ro-540-540-180 | ro-580-580-193 | ro-620-620-207 | ro-690-690-230
5. Curve of some movie titles:
Guardians of the Galaxy 2014 | Godzilla King of Monsters 2019 | The Meg 2018 | Mad Max Fury Road 2015 | Arrival 2016
6. madvr with "ouput HDR format"
When you enable "ouput HDR format" in madvr then the set "real display peak nits" (e.g. 700) becomes MDL and maxCLL as well and maxFALL will be MDL / 3. So, if "real display peak nits" is 700 then: 700-700-233
ro-683-683-231 | ro-700-700-233 | ro-760-760-253 | ro-840-840-280 | ro-960-960-233
As we can see, e.g. 683-683-231 is slightly brighter than 700-700-233, so I used 700 for "real display peak nits", although only using 1% window achieved ~700 nits during continuous measurement.