Print

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

[Hendrik]

.cube LUTs just being text files is rather useful.  I threw it into a spreadsheet and visualized it (the greyscale axis only, assuming I didn't screw up).



I have been testing with the Rec.1886 LUT only until now, so this is interesting.

Red is a linear response, blue is the rec.1886 cube lut, and yellow the 2.4 cube lut.
Blue/rec.1886 seems to massively boost the black level, as I'm also experiencing, while 2.4 actually reduces it slightly but is otherwise pretty linear.

One thing to note is that madVR always behaves as if the "3DLUT Gamma" in JRVR is set to Rec.1886 (eg. "no-op for SDR content"), no matter how the LUT was generated.

With that setting, and using the 2.4 .cube LUT, I get a nice dark image. Its probably a result of measuring in that particular mode. Rec.1886 is designed to boost brightness of low levels, because the black response of old SDR screens is a bit terrible. It may not be ideal to use that for better displays that have a good black point. And I believe madVR also doesn't use it, so madTPG probably presented your test pattern with a 2.2 or 2.4 gamma, without the black offset of Rec.1886.

If the gamma selection for the LUT gets too complicated to consider in all of this, I suppose I could even kill it, and just have it do what madVR does, eg. always put SDR content unmodified through it (which is equal to always setting it to Rec.1886 in 99.9% of all cases)

[mattkhan]

fwiw this is what displaycal says the lut looks like

[Hendrik]

What I would suggest to test now is this:

- Set JRVR "3DLUT Gamma" to BT.1886 and don't touch it again, regardless of how the LUT was made (this is how madVR behaves)

- Use your standard madVR 3DLUT

This should match madVR behavior and at least with the provided 2.4 LUT, I do not get a brightness increase.

[mattkhan]

I think it really *should* be quite simple for end users, just add a few lines to the wiki saying exactly what is required and how it is used, give a few pics from displaycal as examples of how to do.

One thing to note is that madVR always behaves as if the "3DLUT Gamma" in JRVR is set to Rec.1886 (eg. "no-op for SDR content"), no matter how the LUT was generated.

With that setting, and using the 2.4 .cube LUT, I get a nice dark image. Its probably a result of measuring in that particular mode. Rec.1886 is designed to boost brightness of low levels, because the black response of old SDR screens is a bit terrible. It may not be ideal to use that for better displays that have a good black point. And I believe madVR also doesn't use it, so madTPG probably presented your test pattern with a 2.2 or 2.4 gamma, without the black offset of Rec.1886.

do you mean you set 3D LUT gamma in JRVR to Rec.1886 and selected the 2.4 lut I sent? at first glance this does seem to be similar to madvr (but I'd have to measure it tomorrow to be sure)

I don't see any options in madtpg to select gamma so I don't know how it behaves exactly

I can't tell if you're saying the above is good or bad (right or wrong) btw, I *think* it's much the same topic as in https://www.avsforum.com/threads/madvr-argyllcms.1471169/page-238#post-55193072 though and that conversation doesn't actually end in anything concrete. On rereading I think it's saying that the tone curve in displaycal should be set to rec.1886, if so is that going to work for JRVR?

What I would suggest to test now is this:

- Set JRVR "3DLUT Gamma" to BT.1886 and don't touch it again, regardless of how the LUT was made (this is how madVR behaves)

- Use your standard madVR 3DLUT

This should match madVR behavior and at least with the provided 2.4 LUT, I do not get a brightness increase.

OK I'll check this in the morning. I think I want to measure again though and it would be nice to understand how the displaycal tone curve option & JRVR's configuration interact as this part remains unclear to me so I'm not sure how to handle it to get correct results without trial and error (measuring takes so long that you really need to have a repeatable process otherwise it's just extremely painful)

[Hendrik]

Ideally, after understanding how madVR handles the 3DLUT and what fhoech recommended to madshi in that same thread you linked, the 3DLUT Gamma option would just go away. Afterall madVR never had it. And we can stop thinking about it.
Without that option, the setup is essentially the same as madVR, and with support for madVR 3DLUT files you can use all the same instructions.

Assuming the result is OK after you tested again.

[mattkhan]

What I would suggest to test now is this:

- Set JRVR "3DLUT Gamma" to BT.1886 and don't touch it again, regardless of how the LUT was made (this is how madVR behaves)

- Use your standard madVR 3DLUT

This should match madVR behavior and at least with the provided 2.4 LUT, I do not get a brightness increase.

I can confirm using the madvr lut with jrvr gamma set to BT.1886 produces the same results as that same lut in madvr

note that the same lut in cube format with the same settings produces a different response, is there some difference internally to how you handle the madvr lut vs a cube lut that explains this or is this unexpected behaviour? not sure it matters but thought it worth mentioning.

[Hendrik]

There is no difference in how its applied, other then that the cube LUT is full range rather then limited, and is generally lower precision (65 vs 256 points). This may be mostly obvious in the near-black parts, because with 65 points you have no data between bands 0 and 4 (or 16 and 20 the typical chart), and its interpolated instead.

How different is the response? Drastically? A bit?

Either way, for people already setup with madVR 3DLUTs and a workflow to measure those, using those will likely be preferable in all cases. At least in SDR the behavior should match now, as I understood madshi's statements i found.

Will have to think about HDR calibration next. Might be as simple as requiring a PQ in/PQ out LUT, but I have to check if there is metadata in the madVR LUT format that actually indicates this for a HDR LUT.

[mattkhan]

ah ok, I didn't realise the madvr lut format had so many more points (though did notice the files are relatively massive). Definitely preferable to go with the madvr format in that case then.

it's an obvious difference near black for me, blacks become grey extremely quickly basically so everything looks a bit washed out. Forgot to save a pic of the gamma curve this time but it was similar to those posted yesterday.