Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[jmone]

I don't think I've seen any native DCI-P3 content, just DCI-P3 in 2020. 

Unelss there is an "easy" fix to 2390 highlights, I'd also agree that (for now) moving on to add the other features like Overlays, Passthrough, Scaling Options etc makes sense.  Tonemapping has consumed madVR for years and the endless combinations of content, HW permutations makes it hard to see when it would ever end.

[jmone]

Tested on the Epson TW9300 from a NUC Win11 OS HDR on in a dark room.  Same Hable settings as previously.  No apparent issues.  Looked good.

[jmone]

Tested on the Sony OLED from a 1660ti Win11 OS HDR on in a light (but not bright) room.  Same Hable settings as previously.  No apparent issues.  Looked good.

[Hendrik]

I have been looking into DCI-P3 values, and I noticed there are two options, so for anyone with a device that might care, are we talking about DCI-P3 (Cinema) or DCI-P3-D65/Display-P3 (which is DCI-P3 with a D65 white point)?
I would assume the D65 one would be more appropriate for PC usage, and its apparently also what madVR uses for its DCI-P3 mode.

[jmone]

While, I've only ever seen the description of the encoding being DCI-P3 in general not the DCI-P3-D60 and DCI-P3-D65 variants, which I think have white point changes for different display technologies.  I see that wikipedia lists the differences as:
- P3-DCI (Theater): DCI-P3 uses a simple 2.6 gamma curve, a white luminance of 48 cd/m2, and a whitepoint with a correlated daylight temperature of ~6300K, though it is not on the Planckian locus and is slightly greener, the result of optimizing for best light output with xenon lamp projectors. Blue is the same as in BT.709.
- P3-D65 (Display): It uses the DCI-P3 primaries but with a D65 white point which is much more common among computer-display colorspaces  It is also used for some of Netflix deliverables, including HDR and without BT.2020 container.
- P3-D60 (ACES Cinema): No idea

Not much help.... but I'd guess that as we don't use xenon lamp projectors, that PS-D65 would be the go.

[jmone]

...if you do implement it, I'll be able to feed it to the JVC and see what it reports.

[jmone]

...also found this description

What is Display P3? Display P3 is a combination of the DCI-P3 colour gamut with the D65 white point together with the sRGB gamma curve. It originated from the DCI-P3 colour gamut’s implementation in digital cinema projectors, as this standard offers more vibrant greens and reds than the traditional sRGB colour gamut. The white point of the original DCI-P3 is tinted green, and the gamma curve is 2.6. These parameters made it suitable for theatre viewing, but not for closer viewing, such as on monitors. Hence, Apple proposed changing the white point to D65 and the gamma curve to the sRGB curve, and named the new set of attributes “Display P3.”

Looks like I'll need to re-render my UHD footage to P3-D65 (I'd been using the older / original P3-DCI and I don't think my "content" is destined to ever be viewed in a Theatre!!!)... which thankfully is easy.

[mattkhan]

@ mattkhan - I don't think I have a DCI-P3 option on my JVC 7500.  I can set it to 2020 (then ST.2084 for Gamma) which worked very well in my last round of testing.  I'll dig further tonight when it is dark.

it's an option on the N series, I believe you needed to use autocal with a custom colour profile on the older range (e.g. as per the 1st post in https://www.avsforum.com/threads/jvc-calibration-software-v6-for-2015-models-x9000-x7000-x5000-rs400-rs500-rs600.2246658/) though I think it might be the cinema2 mode that is built in? I imagine the answer lies in one of those giant AVS threads somewhere

I have been looking into DCI-P3 values, and I noticed there are two options, so for anyone with a device that might care, are we talking about DCI-P3 (Cinema) or DCI-P3-D65/Display-P3 (which is DCI-P3 with a D65 white point)?
I would assume the D65 one would be more appropriate for PC usage, and its apparently also what madVR uses for its DCI-P3 mode.

yes it's the D65 one

[Hendrik]

...if you do implement it, I'll be able to feed it to the JVC and see what it reports.

It won't "report" anything, just receive different colors. There is no metadata attached, which is why its important to set these properly based on external information.

[jmone]

While, I've only ever seen the description of the encoding being DCI-P3 in general not the DCI-P3-D60 and DCI-P3-D65 variants....

I'm, wrong.  I checked what MediaInfo reports on a bunch of the latest UHD Blurays and they are all reported as using
Matrix coefficients                      : BT.2020 non-constant
Mastering display color primaries        : Display P3
Mastering display luminance              : min: 0.0050 cd/m2, max: 1000 cd/m2

The stuff I'd been encoding was:
Matrix coefficients                      : BT.2020 non-constant
Mastering display color primaries        : DCI P3
Mastering display luminance              : min: 0.0001 cd/m2, max: 1000 cd/m2

So Display P3-D65 it is! 

[jmone]

Tested on the JVC 7500 from a 1660Ti Win11 OS HDR on in a dark room.  Same Hable settings as previously.  No apparent issues.  Looked good.

[jmone]

it's an option on the N series, I believe you needed to use autocal with a custom colour profile on the older range (e.g. as per the 1st post in https://www.avsforum.com/threads/jvc-calibration-software-v6-for-2015-models-x9000-x7000-x5000-rs400-rs500-rs600.2246658/) though I think it might be the cinema2 mode that is built in? I imagine the answer lies in one of those giant AVS threads somewhere

Thanks - I've looked at the Autocal thread and even bought the Colourmeter for it..... but have always used a 3rd party calibration person so never done it on the PJ myself (but that was now some time ago, so it will have drifted).

I now have a DCI P3 and DCI P3 D65 (Display) version of one of my clips, so can test if I can even see a difference (when/if Hendrik adds it) and report back.  I certainly can not see any difference using the latest Hable settings even with side by side screen shots.

[jmone]

Initial Testing comparing DCI P3-D65 vs 2020 vs 709 (all using the stock new Hable Settings + Auto for Gamma on my High Nit screens in Windows 10 with HDR OS on).

WOW - the P3-DCI call looks A+ , the comparison in order is:
- Off line Davinci Resolve Conversion to 709
- PS-D65 : Looks almost identical to DR, a bit more saturated but looks good
- 2020 : Less saturated with a olive tint to the skin tone, does not look great
- 709 : Much more saturated but has a nice "pop" for those that like the look

Need to run over some "real" content and maybe find some colour charts but a very worthy addition!

[Hendrik]

If you want tonemapping with a bit less saturation because you feel thats more faithful, you can tweak the desaturation parameters. Strength is the overall desaturation strength, and exponent controls how much it depends on the peak brightness of the image. The original of 0.9/0.2 would give you stronger desaturation with less dependence on the image - overall a bit less saturation. Don't use the colorspace to tweak for taste!

[jmone]

yeah - I started looking at colour chips, and quickly found there is all sorts of cross dependences and impacts between all the settings (both in JRVR, OS, HW)..... but I think there are plenty of options to tweak settings to dial in what someone wants (or likes) and no glaring issues.

[jmone]

...and I quite like the sRGB target gamma with 2020 and DCI P3-D65

[tij]

I think P3 has a bit more pop (not too much like 709) ... i use it in MadVR for animated stuff ... it might not be faithful, but a bit of pop on animated stuff is nice (dont want to play with desat as it affects details in very bright scenes)

Overall ... i really think Hable is a winner - especially in preserving details in bright areas like skies.

Will do a bit more testing with darker scene ... and then some fire and explosion

... will try to test it on SD TV later on.

[jmone]

Yeah - I don't mind a bit of pop either, and it's a good point about aiming for a pleasant vs a 100% faithful image.  For me it is:
- Skin Tones have to look "right" as we all know how they "should" look
- Not blowing out highlights to a wad of white
- Not crushing shadows to a lump of black

Apart from that I've stopped trying to look at colour chips, as I don't really care if the shade of red on a bike helmet is faithful to what it looked like on set as I will never know. 

Looking forward to seeing how HDR Passthrough goes.

[tij]

100% faithful is impossible with HDR unless your screen can do 100% BT2020 What we are doing is a compromise.

I agree with skin tones ... that part, i think, have to do with not shifting hues.

Blowing highlights ... that part, i think, has to do with desat ... my understanding is, you need to desat colors (add other colors without shifting hue) to boost brightness

Honestly ... hable is great atm ... MadVR still produces sharper details in higlights (i think due to options to tweak fire and highlight recovery) ... but hable is not limited to Windows, which is a big plus

Who knows ... one day Hable might get to where madvr is ... and it will be multiplatform

Question to Hendrik ... measuring peak luminance ... is that get used to adjust tonemapping to reduce compression?

[Hendrik]

Question to Hendrik ... measuring peak luminance ... does that get used to adjust tonemaping to reduce compression?

Yes, Peak Detect overrides the static metadata to limit compression and avoid brightness loss as good as it can.
Without it, it only has the static metadata to work with, and maybe the HDR10+ dynamic metadata if I hook that up, but that is rather uncommon to be found in files.

[lello]

Tested on the Epson TW9300 from a NUC Win11 OS HDR on in a dark room.  Same Hable settings as previously.  No apparent issues.  Looked good.

Why did you enable Windows HDR? I knew that it should always be left disabled and in fact, if I activate it, the colors are completely off; or maybe I got it wrong?

Or maybe with Win11 things have changed?

[Hendrik]

With Windows 11 you can leave it enabled constantly if you want. I'm not sure yet if I want to, on the other hand Netflix can't toggle it and it doesn't do HDR if its not enabled beforehand already, so that might be a nice solution to get it working as well.

[mattkhan]

Yes, Peak Detect overrides the static metadata to limit compression and avoid brightness loss as good as it can.

Is it frame by frame or has some scene detection algorithm?

[Hendrik]

Is it frame by frame or has some scene detection algorithm?

It does scene change detection as well as uses a IIR low-pass roll off to avoid too sudden changes and smooth the response.

[tij]

Peak Detection works for sure

Blade Runner 2049 ... around 00:05:05

[tij]

It does scene change detection as well as uses a IIR low-pass roll off to avoid too sudden changes and smooth the response.

Can it be called dynamic then? ... Or there are other criteria that needs to be fulfilled to be called dynamic?

HDR10+ metadata you mention can be interesting ... though not many titles have it ... i remember reading something about DV metadata parser - though not sure how useful those DV metadata can be if they are proprietary

[Hendrik]

Peak Detection makes the tonemapping rather dynamic indeed. It adapts to the actual scene brightness instead of sticking to what the metadata dictates.

[jmone]

A bit of a ramble.... but AFAIK, there are two parts to the compromises that may need to be made to watch "HDR" content like that found on UHD BD:
- HDR: ability to show an expanded luminance range (both with lower black level details well under 0.1nits and currently as high as 10,000 nits) than SDR (traditionally only encodes from 0.1 to 100nits)
- Colour Space: ability to cover a wider colour space (say 2020, DCI-P3) than 709.  Note: you can have an SDR display with DCI-P3 coverage and accept a 2020 signal (aka 2020 SDR which is a "thing")

As we don't have domestic screens that can do 10,000 nits or cover a 2020 colour space we have to compromise on both aspects, and hence the work Hendrik is doing.

My thinking was that as a lot of UHD content has been mastered to DCI-P3 (in a 2020 wrapper) and max 1000nits;
- Use Passthrough with screens like my Phillips (DCI-P3 Coverage: 97.6%, DisplayHDR 1000 and UHDA certified),
- Tonemapping for the PJ and
- ?? for the OLED. 

So to check what content I actually have, I ran SOT over my 250 odd UHD BDs and populated MC with the values for,
Color range: 100% were "Limited"
Color primaries: 3 were BT.709 and the rest were BT.2020
Transfer characteristics: 3 were BT.709 (same discs as above) and the rest were PQ
Matrix coefficients: 3 were BT.709 (same discs as above) and the rest were BT.2020 non-constant
Mastering display color primaries: 13 were BT.2020 and the rest were Display P3
Mastering display luminance: 103 had a max of 4000, 2 had a max of 1100, the rest were 1000

If I'm reading these tags correctly, it looks like 40+ % are mastered for 4,000 nits and 5% with a full BT2020 colourspace.... so tonemapping will still be needed (either JRVR or the displays in built tone mapper) and will be for sometime regardless of the display.

Anyway.... as I've now got this metadata into MC it could even potentially setup Zones based the Luminance Values / Color Primaries to switch between Tonemapping and Passthrough depending on the content. 

To answer an earlier question is I'm trying out leaving OS HDR ON to prevent the display change that happens and let Windows Tone Map UP the Desktop and SDR content to HDR to the Display that is running in HDR.  So far it seems to work pretty well and everything just looks "correct" (I guess it is easer to map up SDR to HDR levels than the other way around)..... but this is all a new concept.

[mojave]

- HDR: ability to show an expanded brightness range (both lower in the blacks and currently as high as 10,000 nits) than SDR (traditionally only 100nits)

The black level is determined by the display technology and both SDR and HDR can output to the lowest black level of any display. HDR is therefore not "lower in the blacks."

[tij]

Now that 28.0.80 has been set loose ... would be interesting to hear some feedback on HDR tone mapping ... no need to comment that NGU of MadVR is sharper than JRVR ... we know that

If something is amiss ... movie title and time stamp of troubled scene as well as your JRVR settings would be most helpful

[jmone]

The black level is determined by the display technology and both SDR and HDR can output to the lowest black level of any display. HDR is therefore not "lower in the blacks."

I think SDR is typically encoded for 0.1 to 100nits while HDR can go much lower than 0.1 (eg if you look at the luminance encoding values in my pic above you can see min levels much lower than 0.1 and about 2/3rds of my UHD HDR discs are 0.005nits and the other 1/3rd go lower again).  So more details and less banding in dark scenes if your display can resolve these additional dark gradients.  SDR will map all content from 0-0.1 as Black, where HDR has many more shades.

 

[Hendrik]

I think the important distinction here is that while absolute black is certainly going to be equal in HDR and SDR (off is off), HDR can resolve more detail in low-light then SDR could, due to the different gamma curve the HDR PQ encoding offers.
Of course better black detail can only really be used on OLED screens, but HDR PQ is designed in such a way that the screen can tonemap to its physical capabilities, both for black and peak levels. SDR is more of an absolute signal.

[jmone]

The question I have from my earlier rambling : Say if you have a display capable of DCI-P3 Coverage and HDR1000 (more and more common) what would you use?  Such displays met the required Colour Space coverage for most content (all but the few 2020 discs) and for half(ish) of the Luminance used (1000 nit vs 4000nit discs).

I've tried setting JRVR to PCI-P3 Display and 1,000 Nits but it is clearly still tonemapping HDR Content mastered to DCI-P3 with 1000nit max as HDR Passthrough on madVR has a much better picture.

Just thinking ahead on the "Just Works" HDR concept, could it be possible for the JRVR logic to only tonemap if needed, eg when the source material exceeds the display capability (else just passthrough)?



Hendrik enjoy your break!

[jmone]

Thanks - updated my earlier statement to "HDR: ability to show an expanded luminance range (both with lower black level details well under 0.1nits and currently as high as 10,000 nits) than SDR (traditionally only encodes from 0.1 to 100nits)"

[Hendrik]

Just thinking ahead on the "Just Works" HDR concept, could it be possible for the JRVR logic to only tonemap if needed, eg when the source material exceeds the display capability (else just passthrough)?

Maybe, but rather would just let passthrough happen in all cases then if thats a mode your display supports.

Tonemapping right now always targets SDR, which is very different to a HDR signal, even if you increase the target nits. Outputting a HDR signal after tonemapping to some degree is quite a different problem, and not something I'm currently interested in. Rather have people pass-through HDR then, or entirely tonemap to SDR.

In other words, you can't "pass through" in SDR, because HDR PQ always needs a certain amount of processing to make it SDR. By setting the gamut to BT.2020 or DCI-P3, and increasing the Target Nits, you can already very much limit the amount of compression it has to do.

So in the name of "just working", I would actually recommend either always pass-through or always SDR tonemapping, trying to do both based on some parameters gets .. complicated.

[jmone]

Just a thought as we wait for Passthrough.

FYI - for those wanting to see what your display can do, chances are it will be listed here:
- Brightness: https://www.rtings.com/tv/tests/picture-quality/peak-brightness : No surprise that modern LCD Flat Screen typically do above 1,000 nits in HDR Mode, and even OLED in the latest generations are creeping closing to 1,000 nits (HDR Real Screen Brightness)
- Colour Space: https://www.rtings.com/tv/tests/picture-quality/wide-color-gamut-rec-709-dci-p3-rec-2020 : Just about everything has pretty good DCI-P3 coverage but still way off on 2020.

[jmone]

Maybe....
So in the name of "just working", I would actually recommend either always pass-through or always SDR tonemapping, trying to do both based on some parameters gets .. complicated.

Got it - thanks!

[mojave]

SDR (traditionally only encodes from 0.1 to 100nits)"

But that isn't true either.  Black isn't encoded at 0.1 nits on SDR. If it was, then the maximum contrast ratio of SDR would only be 1000:1. Again, black is determined by the display. An OLED can reproduce SDR black below .001 nits. My JVC RS500, 620, and NX7 could get close to that, too, with the dynamic iris. I can measure down to .0007 nits with my Colorimetry Research CR-100.

The following provides good information:
 https://www.lightspace.lightillusion.com/uhdtv.html

[jmone]

...but I don't think in SDR you can specify any steps between 0 and 0.1.  In HDR you can.

Edit, Sorry I'm not trying to suggest that black is 0.1 (it is obviously 0 in both HDR and SDR), but that brightness starts at 0.1nits on SDR material and much lower on HDR.... but I'm certainly no expert, just trying to get my head around it.  From what I've read, HDR can resolve far more dark gradients than SDR (eg it's not all just how "bright" it can get, but also how many shades it can specify and at the dark end small nit changes are actually pretty large % changes).

[JimH]

OK, Hendrik's on vacation now for a week, so don't find any problems with JRVR.

[jmone]

Ahhh, we are way off topic and in the weeds anyway overthinking this (or at least I am)!

It will be good to see others test and report what they think of JRVR on their setups and with various media.  As tij suggested feedback on what the clips are and timestamps of anything that does not look right (or does look right) would be great.

[lello]

I'm trying the new V80 on my Tw9400 projector and I have to say that I am very satisfied with the image produced with 4K HDR movies.

I left almost everything by default except the gamut set to DCI-P3-D65 BT-709 and the Peak Nits target set to 180. About the latter, should not be set with the nits that reaches my projector and that is 65? If I do it, I burn the high lights.

A problem to report. I can't open an old 480p DVD that I copied to the hard disk: depends on the fact that I saved it as ISO? Or is the problem 480p definition?

[lello]

As for me, it would be enough for me to add the possibility to see what is happening (the ctr+j of madvr) and the possibility to move the OSD from the bottom to the active area of the film.

[tij]

I'm trying the new V80 on my Tw9400 projector and I have to say that I am very satisfied with the image produced with 4K HDR movies.

I left almost everything by default except the gamut set to DCI-P3-D65 and the Peak Nits target set to 180. About the latter, should not be set with the nits that reaches my projector and that is 65? If I do it, I burn the high lights.

A problem to report. I can't open an old 480p DVD that I copied to the hard disk: depends on the fact that I saved it as ISO? Or is the problem 480p definition?

I don't think nits setting are actual nits of source (madVR allows max to be 100 nits) ... i think JRVR principle for nits is similar to MadVR ... lower values makes image appear brighter (more details in shadow) ... while potentially blowing out highlights

Higher value deals better with bright higlights ... but can crush blacks for lower nits display.

I dont think i have problem with HD/DVD files ... but mine are in mkv ... i have not tested JRVR with disc menus (mostly because i don't rip them that way)

[tij]

As for me, it would be enough for me to add the possibility to see what is happening (the ctr+j of madvr) and the possibility to move the OSD from the bottom to the active area of the film.

Hendrik plans to do those after his holiday

[tij]

A problem to report. I can't open an old 480p DVD that I copied to the hard disk: depends on the fact that I saved it as ISO? Or is the problem 480p definition?

I don't have DVD menus to test ... but i have some DV UHD in folder format that awaits transfer to mkv (now that mkv  supports DV)

I cannot play its disc menu with JRVR too (Casino Royale 2006) ... it just gives me black screen - both Hable and BT ... MadVR can play the menus

Something for Hendrik to look at when he is back

PS. Movie plays find from folder structure using JRVR  .. it just cannot play disc menu ... anybody else can confirm disc menu playback?

[Hendrik]

DVD menus should work as long as you don't use Hardware Decoding (incompatible with D3D11 decoding right now, which would be fixed at some point in the future). Blu-ray menus are tied to the overlay functionality thats coming soon.

[tij]

These screenshot is making me scratch my head (movie is Tenet ... and timestamp is on one of screen shots) ... seems that MadVR 0.92.17 is really showing its age in terms of tone mapping

Will try beta 113 to see what improvements madshi did (but even that is few years old ... and latest beta are time limited)

PS. JRVR are default settings at 480nits with peak detection ... MadVR matches Hable if set at around 200nits

PSS. Not sure how this scene suppose to look ... but prefer Hable at 480nits and MarVR at 200nits (unless Tenet creators really wanted to hide details of main characters)

[froschling]

Hi everyone,

Much, in fact most, of this thread is beyond me but I can report that using the new video rendering setting has fixed the problem I've had with incorrect colours being displayed when playing video.

This is a very good thing and I hope it will continue.

Many thanks to those developers who worked on the new renderer.

John

[thorsten]

Hi,

I have the same "problems" of understanding everything 

But, I can report: on my system (Ryzen5 2600X, Win10/64, lame AMD GPU, MC v80/32), 4K HDR feeding my JVC X7900 with SDR, the convertion runs really good and flawless.
Only during standby or osd infos (like volume), the picture stutters and gets for a fraction of a second green or blue or only partly, it differs from one to the next.

Normal 1080p24 movies also works.

But I'm very happy that it works! Thank you Hendrik! Up and running only 3 weeks after the first demo.

Greetings,

Thorsten