madVR - high quality video renderer (GPU assisted)

[James Freeman]

OK, I have tested the BT.1886 standard against my regular 2.2 power curve calibration.

I don't like it.

Yes, the image has more discernible shadows.
But it destroys the depth of the image I am used to, and looks unnatural.
It also lowers my already low IPS panel contrast.

The default settings in Argyll (Gamma: Relative, Black Offset: 100%) give me the best results.
Because it works best with what a monitor can give natively, and not force it to conform to a standard.

Quote from the manual:

The default way of handling this (equivalent to 100% black output offset) is to allow for this at the output of the ideal response curve, by offsetting and scaling the output values.
This defined a curve that will match the responses that many other systems provide and may be a better match to the natural response of the display, but will give a less visually even response from black.

I don't care for "less visually even response from black"...
I care for deep contrasty popping image, which I don't get when calibrating to the BT.1886 standard.
Yes, 1886 gives a more even grey scale and close to CRT black response but at the expense of image depth.

I can see ALL the black steps either way (No black crush what so ever).
I don't think you can set any setting that resulting in a pure power curve below 10% that crushes the blacks, argyll always has some kind of offset to the blacks.
Only its selectable between Input & Output Offset (0% to 100%).



I think that if a display has less than 1000:1 contrast ratio, BT.1886 can do more harm than good.

A "Power Curve + BLC" gives me a nicer deeper picture.
At least with my setup.



Any thoughts?

[6233638]

Originally Posted by James Freeman

OK, I have tested the BT.1886 standard against my regular 2.2 power curve calibration.

A power curve will only have shadow detail down to around 5% black on a display with 1000:1 contrast.

Originally Posted by James Freeman

Yes, the image has more discernible shadows.
But it destroys the depth of the image I am used to, and looks unnatural.
It also lowers my already low IPS panel contrast.

The point is that you're no longer crushing or clipping shadow details when using BT.1886

Originally Posted by James Freeman

I don't care for "less visually even response from black"...
I care for deep contrasty popping image, which I don't get when calibrating to the BT.1886 standard.
Yes, 1886 gives a more even grey scale and close to CRT black response but at the expense of image depth.

Then you need a higher contrast display.

Originally Posted by James Freeman

I can see ALL the black steps either way (No black crush what so ever).

Black crush does not necessarily mean that all steps are indistinguishable, but that they are compressed and more difficult to distinguish in high contrast situations.
The old methods of black level compensation did not do a good job with this, and also had a tendency to reduce midtone contrast.

See this post for comparisons between 2.2, BT.1886 and 2.2 with black level compensation for a 2000:1 contrast display: http://www.avsforum.com/t/1409045/ho...#post_22253013
You can see that 2.2 with black level compensation does not do well to distinguish gray levels near black. A 1000:1 display would be even worse.

[James Freeman]

6233638,

Yes, the blacks are more compressed with "Power law + BLC" but they are still visibly all there.
I don't think there are higher contrast displays in IPS type panels (above 1000:1) at least not in my budget.

Based on viewing few dark blu ray movies:
I think (in my opinion only) the movies are mastered with "Power law + BLC".
I can still see all the shadow details even in high contrast scenes (yes, even below 5%).
The blacks are compressed to a smaller range and that makes them visibly "smoother" and NOT even.
If the movie mastered that way, it's just fine.

I think BT.1886 is too aggressive with the black compensation (as stated before).

Do you think the movies are mastered on a CRT?
Or a BT.1886 calibrated LCD?

I think there is a problem with Argyll's calculations.
When calibrating with argyll to BT.1886 specs, the lowest black on my monitor (0.23 to 200 cm/m2 = 870:1 ) turns to 0.29 and as a result lowers the contrast ratio.
I don't know why but it should use the lowest black no matter what.

This partially made me look negatively up on BT.1886.

[NewmanHD]

Hey guys! Currently I use Niyawa description (MPC-HC, madVR, LAV, ReClock, xyVS). Thinking about buying a VGA card next to a G840/i3-2120 (Asus B75 MoBo/8GB RAM). The best picture settings will prove adequate for a HD7730? Later prove to be enough? I do not play games and I'm sorry for my English. Thanks.

[6233638]

Originally Posted by James Freeman

I think (in my opinion only) the movies are mastered with "Power law + BLC".

Films are typically mastered on high contrast displays which do not have a need for black level compensation. LCD monitors have not been widely adopted.

Originally Posted by James Freeman

I can still see all the shadow details even in high contrast scenes (yes, even below 5%).
The blacks are compressed to a smaller range and that makes them visibly "smoother" and NOT even.
If the movie mastered that way, it's just fine.

I think BT.1886 is too aggressive with the black compensation (as stated before).

Do you think the movies are mastered on a CRT?
Or a BT.1886 calibrated LCD?

BT.1886 is the new standard for mastering. To calibrate to anything else is wrong.
If you want a higher contrast image, you have to either compromise on shadow detail and go with a less accurate image, or replace it with a higher contrast display.

Originally Posted by James Freeman

I think there is a problem with Argyll's calculations.
When calibrating with argyll to BT.1886 specs, the lowest black on my monitor (0.23 to 200 cm/m2 = 870:1 ) turns to 0.29 and as a result lowers the contrast ratio.
I don't know why but it should use the lowest black no matter what.

That doesn't sound right, but I have never had good success with ArgyllCMS. I'm hoping that there will be CalMAN support eventually.

[leeperry]

Originally Posted by James Freeman

Do you think the movies are mastered on a CRT?

Well, SONY Pictures promised that CRT monitors in their mastering houses would be replaced by OLED through the end of the year, but yes indeed CRT still rules there and here's a real world picture of a french TV show production control room in 2013, it's one of those.

That's the very reason why some/most/many HD movies look greenish and overly saturated in the REC.709 gamut as some mastering facilities would convert from SMPTE-C/EBU to REC-709 via a high precision 3DLUT but IRL some wouldn't. Once everyone goes OLED, hopefully the SMPTE-C orangey red will be history and the hero of Disney's Cars will finally be as deep red in the anime as it is on the poster

Originally Posted by James Freeman

I think there is a problem with Argyll's calculations.

Using Argyll's mail list is a bit of a pain, but Graeme Gill does read this thread FWIW and he's the only one that can help/fix the issue you're encoutering.

Originally Posted by NewmanHD

Thinking about buying a VGA card next to a G840/i3-2120 (Asus B75 MoBo/8GB RAM). The best picture settings will prove adequate for a HD7730? Later prove to be enough?

madshi made clear that he might add more demanding algorithms as he sees fit, and for instance no performance figure has yet been announced regarding the forthcoming debanding filter.....so it wouldn't be too wise to buy a monster board in order to hopefully anticipate on what mVR might require later.

If you want Jinc AR scaling@1080p for both luma & chroma upscaling, possibly SmoothMotion on top of it and being able to process 1440x1080i@59.94, a GTX650Ti will barely have the muscles......not sure about AMD boards as I believe madshi said that their texture processing was (sometimes) faster, but I guess you would be looking at a HD7850+ in order to remain on the safe side for a (possibly little) while and their new GPU generation is about to be released so resellers will have to get rid of their 7850/7870's in a timely manner

[madshi]

Originally Posted by James Freeman

Is there any way to use ICC Color Profiles (like MPC-HC does natively) without converting to 3D LUT?

No, you need to create a 3dlut for madVR to do any calibration.

Originally Posted by StephaneM

I'm not sure I know how to put the drain on madVR ?

This is what I'm doing for now :

[...]

The "videoFrame" is a WPF Control which hosts the WindowsFormsHost, this host does have a Window, this is this window handle that is exposed through videoFrame.VideoHwnd and forwarded to madVR (with put_Owner)

This window is created on the WPF UI thread (I have no other choices), while madVR is created on a dedicated thread (STA or MTA, with message pumping)

Well, looks alright to me. I'm not sure why you have the problems you have. I can only imagine that the WPF UI thread doesn't react to the messages as quickly as it should, but I don't really know...

Originally Posted by 6233638

Looks promising! I assume there will be a new tag so it can be selectively enabled/disabled?

Of course.

Originally Posted by e-t172

Ideally madshi should remove BT.709 from the gamma options (because it is confusing and doesn't actually make sense) and add BT.1886 support

Wouldn't I need to know the display's black level in order to do that? I don't know that, unless you provide madVR with measurements. And if you do that, you can just as well provide a full 3dlut which does the BT.1886 stuff internally, can't you? I'm just not sure if it makes any sense for madVR to offer BT.1886.

Originally Posted by ryrynz

Will it have values that can be adjusted?
Will the deband be done before or after resizing? And have you compared quality of deband before and after resizing?
What's the general performance impact?
Will there be a an HQ mode at all?

There will be a "low" and a "high" setting. Debanding is done before resizing, but (currently) after chroma upsampling. On my HD4000 the "low" setting costs about 7ms per 1080p frame. The "high" setting costs about 14ms. The "high" setting removes more banding, and has a slightly more complicated algorithm to avoid too much detail loss. The "low" setting is comparable to f3kdb's default settings (in high bitdepth).

I haven't compared the quality before/after resizing, and I don't think it's necessary. If the source is banding free before the resizing, then it will still be banding free afterwards (except if you use DXVA scaling with a GPU which only supports 8bit). IMHO the right place to do debanding is before scaling.

Originally Posted by baii

madvr report 23 and 24 as well, just curious, is it possible that the monitor scaler scale the 23/24 hz into 60hz?

Yes, that is possible.

Originally Posted by n2k3

Since early this month I've been having blue screens (3 total now) and I think it's related to madVR.

Related to maybe. But there's no way a user mode software like madVR should be able to produce a blue screen. So the blame is most probably the GPU drivers. I'd suggest updating the drivers, maybe the BIOS.

Originally Posted by NewmanHD

Hey guys! Currently I use Niyawa description (MPC-HC, madVR, LAV, ReClock, xyVS). Thinking about buying a VGA card next to a G840/i3-2120 (Asus B75 MoBo/8GB RAM). The best picture settings will prove adequate for a HD7730? Later prove to be enough? I do not play games and I'm sorry for my English. Thanks.

I don't know if the 7730 is fast enough for highest settings at all framerates. You may want to check renethx' posts in the AVSForum. He's done some checks which GPU can do which framerates with which algorithms. The 7730 is probably not fast enough to do the highest settings for 60fps, but I don't know for sure...

[StephaneM]

Hi,

Originally Posted by madshi

Well, looks alright to me. I'm not sure why you have the problems you have. I can only imagine that the WPF UI thread doesn't react to the messages as quickly as it should, but I don't really know...

I've found a workaround for this : when adding madVR to the graph, it has to be done on the main UI thread.

At this point I thought I would put back the whole graph on the UI thread, but while it works, it does requires many tricks to build the graph while having WPF animation (and keep them smooth). It's much easier to have the graph in a dedicated thread.

So you can have madVR added to the graph in the UI thread and have the graph in another thread (preferably with a message pump). Then the only thing you have to do is to place some madVR calls on the main UI thread (well I had to do it only for SetWindowPosition, so I guess that anything that will somehow change properties on the madVR window requires to be called on the UI thread)

[NewmanHD]

Originally Posted by madshi

I don't know if the 7730 is fast enough for highest settings at all framerates. You may want to check renethx' posts in the AVSForum. He's done some checks which GPU can do which framerates with which algorithms. The 7730 is probably not fast enough to do the highest settings for 60fps, but I don't know for sure...

Thanks for your reply. Could you give me a link about the post? I can not find.

[James Freeman]

I want to quote a few short posts from AVSforum: How power law gamma calibration can lead to crushed blacks Thread.

Yes, at times. The trick is to avoid "A/B' comparisons with a pwr-law setup.

I don't know what "pros" are actually doing, but at this point they really *should* be doing one setting with PowerLaw w/BLC and another with BT1886 (as opposed to "Day" and "Night" modes) ...
Subject to the customers' wishes, and the displays' capabilities, of course.

Straight power law (with no black level compensation) should be a non-starter in the flat panel age.

This is exactly what I've wanted to read.

As I like the Power Law + BLC better because I get a more Contrasty image without "Visually-Non-Satisfactory-Blacks".
Note that I did not use the term "Black Crush" because they are NOT (visually for me), they just live in a smaller space, yet all of them are perfectly visible.

Well... BT.1886 is a relatively new standard that tells "everyone" to conform to it, yet, not without a compromise.


Cheers.

[6233638]

Originally Posted by James Freeman

Note that I did not use the term "Black Crush" because they are NOT (visually for me), they just live in a smaller space, yet all of them are perfectly visible.

That's still black crush. BT.1886 relies on a lot of research that has been done for just-noticeable difference in a number of scenarios.

No-one says you have to calibrate your display to follow the standards. Set it up however you like, if that's what you prefer. It does sound like something is going wrong with ArgyllCMS if it's raising your display's black level.
But further discussion on this probably belongs on AVS Forum rather than the madVR topic.

The real solution to your problem is to move away from a display with only 870:1 contrast, which is very low.

[fairchild]

Some people don't have the option to calibrate to BT.1886 I would need to use a 3dlut which would rely on my spectrometer which drifts with heat over time, and the process of creating the 3dlut would probably give me inaccurate results anyway.

Hence, i can only calibrate my display using it's controls to 2.22 gamma since my set also lacks a 10-pt gamma adjustment. There is NO standard for gamma adjustments anyway, different authoring houses calibrate with different displays and different values.

[6233638]

Originally Posted by fairchild

Some people don't have the option to calibrate to BT.1886 I would need to use a 3dlut which would rely on my spectrometer which drifts with heat over time, and the process of creating the 3dlut would probably give me inaccurate results anyway.

Hence, i can only calibrate my display using it's controls to 2.22 gamma since my set also lacks a 10-pt gamma adjustment. There is NO standard for gamma adjustments anyway, different authoring houses calibrate with different displays and different values.

You may not be successful using ArgyllCMS with your Spectro, but you would be able to implement a custom gamma curve using the 3DLUT interface.

BT.1886 is the standard for gamma now. There was no standard.

[leeperry]

Originally Posted by chuuey

Well I think my Vsync issues on XP are a problem related to the Nvidia drivers. Even with latest WHQL drivers, the refresh rate reported by madVR is all over the place, sometimes falling to 57Hz on a 60Hz screen and sometimes jumping into 62 range, and frame drops occur everywhere.

Originally Posted by petran79

Problem lies in how the CPU and GPU handle threads in Windows XP.

If you guys are still around, my contact at nvidia just confirmed that these new beta drivers might fix the refresh rate detection problem in mVR on XP with the 660: GeForce 331.40 beta drivers

Too bad I sold my Asus 660 at no loss as I was under the impression that it would take forever to see a fix released and was starting looking at HD7850's.

[fairchild]

Originally Posted by 6233638

You may not be successful using ArgyllCMS with your Spectro, but you would be able to implement a custom gamma curve using the 3DLUT interface.

BT.1886 is the standard for gamma now. There was no standard.

Can you point me in the right direction as to how to actually create a 3dlut to use with MadVR to have a BT.1886 gamma curve? I'm leaving this in this post as I think it would be helpful to MadVR users who are in the same predicament as myself. (if madshi requests this be moved to PM's no problem doing that.)

BTW I have read posts from calibrator's who are still calibrating displays to 2.22 even with the ability to calibrate to BT.1886 Either because the picture overall comes out too dark or just preference. (referring to both pro and DIY calibrators on the AVS forums.)

[Asmodian]

I started with this post from AVSforum but changed some options for BT.1886 output. This uses Argyllcms to do the calibrations. Rec709.icm can be found in the Argyll directory, the rest of the files are generated.

Open madTPG.exe from the MadVR install folder
Enable "use fullscreen"
Enable "disable VideoLuts"
Enable "disable 3dlut"

dispcal.exe  -v  -dmadvr  -Yp  -qu  -m  -G2.4 -f0  -w 0.3127,0.3290  "dispcal"
targen.exe  -v  -d3  -G  -s40  -g128  -f2000  "MadVR"
dispread.exe  -v  -dmadvr  -Yp  -K dispcal.cal  "MadVR"
colprof.exe  -v  -qh  -bl  -aX  "MadVR"
collink.exe  -v  -3m  -qh  -et  -Et  -Ib:2.2  -G  -a dispcal.cal  -iaw  Rec709.icm  MadVR.icm  3DLUT.icm

For very dark viewing conditions I like this change to the collink line:

collink.exe  -v  -3m  -qh  -et  -Et  -Ib:2.4  -G  -a dispcal.cal  -iaw  Rec709.icm  MadVR.icm  3DLUT.icm

I also add the -ye option to both dispcal and dispread. This uses the meter correction file for my i1 Display 3 Pro and a white LED backlit LCD monitor but you need to install the correction files for Argyll to use them. You can do dispcal -help to see what options there are available for -y on your system.

Any suggestion or improvements very welcome.

edit: I am currently happy with the results of the calibration on my 1000:1 contrast ratio, ~180 cd/m2 IPS monitor but nothing is ever perfect.

[andybkma]

Greets, can someone please confirm this madVR Full Screen Exclusive (FSE) behavior for me please? I need to find out if this is being caused by mVR or something on my system....

Win7 64-bit, with FSE On (using new path, Present Several Frames in Advance checked), HDMI or VGA out to an external monitor or projector, if I put the computer into Sleep (Standby) and then come out of Sleep, the first time (and only the first time) I go into FSE mode I seem to lose graphics output to the external monitor (the monitor goes into sleep mode) for about 3-4 seconds before finally going into FSE. If I use the old FSE path (Present Several Frames in Advance unchecked) then I have this problem when coming out of FSE mode instead of going into. Very odd. This happens only with Sleep, if I reboot the computer I don't have this problem.

I have tried checking & unchecking numerous mVR settings; I have tried using different media players Zoom Player and PotPlayer with both Nvidia graphics and Intel HD4000; and have tried various settings under Device > Display Mode (on/off). All did nothing, I still have the same problem so this is leading me to believe this is something possibly with mVR itself and not my laptop. Just odd that it happens only the very first time I flick from windowed to fullscreen FSE mode after coming out of Sleep. If FSE mode is unchecked then all is good so I know it's something with FSE.

Anybody else experience same? Cheers...

[oddball]

Bug report.

http://pdl.warnerbros.com/wbmovies/gravity/trailer2/GRAVITY_TRAILER_5-2k.mov

Plays as red and shades of black. If I switch to EVR it plays fine.

[sneaker_ger]

It works fine for me. It's ProRes 10 Bit 4:2:2 - do you have any intermediate filters between LAV and madVR? Do you have "P210" output ticked in LAV Video? ("v210" output seems broken for me, not sure if LAV or madVR fault)

[6233638]

Originally Posted by oddball

Bug report.
http://pdl.warnerbros.com/wbmovies/g...AILER_5-2k.mov
Plays as red and shades of black. If I switch to EVR it plays fine.

Please report bugs to the bug tracker: http://bugs.madshi.net

But Madshi is aware of the problem: http://www.avsforum.com/t/1490573/gr...#post_23762599

Originally Posted by sneaker_ger

It works fine for me. It's ProRes 10 Bit 4:2:2 - do you have any intermediate filters between LAV and madVR? Do you have "P210" output ticked in LAV Video? ("v210" output seems broken for me, not sure if LAV or madVR fault)

I'm guessing you don't have any display calibration options enabled.