madVR - high quality video renderer (GPU assisted)

[Niyawa]

So, I'm trying to test Overlay and everything but when I enable it in madVR and try to play anything I get this:

http://i.imgur.com/DNZeEiO.png

Anyone know why? I have Intel GPU with Windows 8.1 if that matters. It used to work a few months ago and I didn't change any settings so not sure what it could be.

[THX-UltraII]

If I don't remember it wrong, the CCC allows us to select any HDMI port as the default audio output. So it might be possible to make an audio-only link to AV receiver via another HDMI port...

But there is no HDMI input on my old-fashioned AV receiver (Yamaha RX-V1500). So I cannot make a test to verify that with my HD7970. :p

Sounds interesting. But there is one problem: My HD7950 (Asus Royal King) has only one HDMI output (and 2 DVI outputs). Feeding the VW1000 with the DVI output (and my AVR with HDMI) won t work because the DVI outputs do not support 4096x3112 output, only the HDMI output does.

But maybe one of the 2 DVI outputs support high.res. audio! This would mean I can connect one of the DVI outputs to my AVR and the HDMI output directly to my 4K projector. The specs of my card say that the card has one Dual-Link DVI-I output and one Single-Link DVI-D output. You have any idea if these support high res. audio?

[michkrol]

Anyone know why? I have Intel GPU with Windows 8.1 if that matters. It used to work a few months ago and I didn't change any settings so not sure what it could be.

I'm on Windows 8.1 and it works here on Intel HD 4000 (Ivy Bridge). I'm using drivers version 15.33.5.64.3316.
What drivers are you using and what's the exact model you're using? You might have issues with Sandy Bridge CPU/iGPU, since they don't seem to have Win8.1 drivers available.
Also try Windows Update to download newest updates and drivers. It's usually not recommended to download drivers from WU, but who knows, it might work.

But maybe one of the 2 DVI outputs support high.res. audio!

DVI doesn't support audio output at all, so you'd have to check it then connect it with HDMI either directly or not.

[Razoola]

I have a 3D related question with madVR.

I know that currently madVR does not support3D (and I think that's also right for MPC-HC). That said I can play a 1080p SBS using MPC-HC and madVR and my plasma will detect the 3D image. If not I can still force it to 3D. Am I right in thinking that I'm loosing half the detail due to madVR resizing the 2160 SBS hortz pixel source to 1080 (due to panel res set to 1080p) and then the plasma re expanding half that back for each L/R 1080 frame for each eye?

[huhn]

DVI doesn't support audio output at all, so you'd have to check it then connect it with HDMI either directly or not.

you can run full hdmi over an dvi slot just use an adapter nothing new about that. there where a lot of nvidia cards without an hdmi connector they still can output audio over dvi.

so no need for an need avr just make 2 connections.

gpu -> hdmi -> avr -> beamer
gpu -> dvi-> hdmi -> beamer

even if audio is not working in this case it is still working
your beamer got 2 hdmi inputs or ?

or you can use dvi direct dvi-d can handle easily 2560x1440x120

3840x2160x30 is a joke for dvi

[Q-the-STORM]

Am I right in thinking that I'm loosing half the detail due to madVR resizing the 2160 SBS hortz pixel source to 1080 (due to panel res set to 1080p) and then the plasma re expanding half that back for each L/R 1080 frame for each eye?

yes... but that should apply to any other player as well... unless they can output the full resolution and your TV accepts it... but where do you get 2160p SBS content?... maybe you mean OU...

[Niyawa]

I'm on Windows 8.1 and it works here on Intel HD 4000 (Ivy Bridge). I'm using drivers version 15.33.5.64.3316.
What drivers are you using and what's the exact model you're using? You might have issues with Sandy Bridge CPU/iGPU, since they don't seem to have Win8.1 drivers available.
Also try Windows Update to download newest updates and drivers. It's usually not recommended to download drivers from WU, but who knows, it might work.
My laptop is somewhat old. I have GM45 Express Chipset with 4500MHD. The drivers are modded to work with Windows 8 but nothing else. I'll try using overlay with default WDDM drivers and see if that fixes it.

Edit: No difference, unfortunately.

[Razoola]

yes... but that should apply to any other player as well... unless they can output the full resolution and your TV accepts it... but where do you get 2160p SBS content?... maybe you mean OU...

Ahh yes I'm sorry. I though 1080p SBS rips kept 1080p for each eye. I see they only have 960 hortz pixels for each.

[pie1394]

Sounds interesting. But there is one problem: My HD7950 (Asus Royal King) has only one HDMI output (and 2 DVI outputs). Feeding the VW1000 with the DVI output (and my AVR with HDMI) won t work because the DVI outputs do not support 4096x3112 output, only the HDMI output does.

But maybe one of the 2 DVI outputs support high.res. audio! This would mean I can connect one of the DVI outputs to my AVR and the HDMI output directly to my 4K projector. The specs of my card say that the card has one Dual-Link DVI-I output and one Single-Link DVI-D output. You have any idea if these support high res. audio?

Isn't the following DVI -> HDMI adapter provided with your card? :p

http://www.overclock3d.net/gfx/articles/2012/11/19113235847s.JPG http://www.overclock3d.net/gfx/articles/2012/11/19113236941s.JPG

My card was made by Lantic. Yet the CCC shows 6 audio devices from this card in total. It means the proprietary DVI connectors are actually wired with SPDIF signal pins if it runs at single-link mode --- exactly as HDMI. Of course the correct adapter is required because standard DVI connector does not have SPDIF pin definition.

[THX-UltraII]

Thanks for your reply pie1394. So just buy an adapter like that, connect that to one of the two display port outputs of my card and run it to my AVR? I really never thought that a display port output can output HD audio?!

I found this on the site of my card as optional adapter:

http://www.club-3d.com/index.php/products/reader.en/product/33.html

[THX-UltraII]

huhn, don t you mean:
gpu -> hdmi -> avr
gpu -> dvi-> hdmi -> beamer

instead of what you are saying:
gpu -> hdmi -> avr -> beamer
gpu -> dvi-> hdmi -> beamer

This won t work anyway. Check the specs of my card here:
http://www.club-3d.com/index.php/products/reader.en/product/radeon-hd-7950-royalking.html

You can see here that:
Maximum Resolution analog: 2048x1536 (via Dual-Link DVI to VGA adapter)
Maximum Resolution Single-Link DVI: 1920x1200
Maximum Resolution Dual-Link DVI: 2560x1600
Maximum Resolution HDMI 1.4a: 4096x3112
Maximum Resolution Mini DP 1.2: 4096x2160
Maximum Outputs Simultaneously: 4

So DVI => HDMI => beamer will have a max resolution of 2560x1600 while I need 4096x2160

[huhn]

this should work anyway cause u use the dvi slot but it behavours like a hdmi slot.

i'm pretty sure all slots are display port slot read about eyefinity what its called.

the card just detects what type of connect it is.

so just try it out.

an passive adapter works for sure in this case.

[e-t172]

My card was made by Lantic. Yet the CCC shows 6 audio devices from this card in total. It means the proprietary DVI connectors are actually wired with SPDIF signal pins if it runs at single-link mode --- exactly as HDMI. Of course the correct adapter is required because standard DVI connector does not have SPDIF pin definition.

SPDIF and HDMI Audio are two completely different things. For example you can't output LPCM 5.1 over SPDIF, but you can over HDMI. Some cards support HDMI Audio over a DVI output with an adapter (sometimes you need to use the right adapter that has a shunt to "signal" the card it's actually HDMI), some don't.

[THX-UltraII]

this should work anyway cause u use the dvi slot but it behavours like a hdmi slot.

i'm pretty sure all slots are display port slot read about eyefinity what its called.

the card just detects what type of connect it is.

so just try it out.

an passive adapter works for sure in this case.I just connected one of the DVI outputs to my AVR with this cable:
http://www.beamershop24.nl/hdmi-kabel/oehlbach-real-matrix-hdmi-und-dvi-kabel-12-meter/

I get no sound........ @pie1394None of the 6 HDMI audio devices 'lit up'.....

[THX-UltraII]

SPDIF and HDMI Audio are two completely different things. For example you can't output LPCM 5.1 over SPDIF, but you can over HDMI. Some cards support HDMI Audio over a DVI output with an adapter (sometimes you need to use the right adapter that has a shunt to "signal" the card it's actually HDMI), some don't..
But do I just order a load of DVI=>HDMI adapters and try and see if it works? No adapter came with the card.

[huhn]

try the onboard for audio and amd for picture.

both normal hdmi connections.

i have audio option for every connection in my nvdia control center and amd should works the same way... and on top of this i have 4 hdmi audio output devices in my sound's settings.

[djfred93]

THX-UltraII, you need a hdmi splitter or a mini DP to hdmi active adapter.

Splitter :

gpu -> hdmi -> splitter -> avr
splitter -> Sony

Mini DP to hdmi active adapter :

gpu -> hdmi -> avr
gpu -> mini DP -> active adapter -> hdmi -> Sony

[ryrynz]

Between maxDif 1.9 and 2.1 there is a slight difference in debanding quality, I find 1.9 is just a little too low sometimes. MaxDif at 2.0 could be a better compromise, I don't think there is a difference between 2.0 and 2.1. But it must be confirmed by other testers.

I'd ideally like for it to be set at 1.9, and yeah changing just about anything down affects debanding effectiveness but you always get more detail back as a result.
It's a slider you just have to find the most acceptable position for and I think the compromise here is worth it. I tried lower settings also and a touch less on the maxangle but the difference was too great to be worth considering.

If only the areas of "real" banded content was affected more by these settings it would be great, but what I see is enabling deband changes much of the picture quite subtly (almost 50% of the image gets changed in some way by enabling debanding at the low preset) so it's no wonder things look softer. Changing maxdif down a bit brings back a touch of those steeper gradients, so yeah the debanded area will look ever so slightly less blended but remember the rest of the image gets back what would have been lost as well with smoothing as well. I wouldn't be mentioning it if I didn't think it was worth it :) I like debanding as much as the next guy and have been using 3kdb for quite some time.

Just wondering, is with the content you feel with maxdif being a touch too low at 1.9, should this content not be better served by the medium preset?
For anime I'd like to keep low on by default maybe even with 10 bit content noticeable banding doesn't happen that often (unless you have low quality content), I do however also like some of the minor blending that occurs in
very minor banded areas, just not too much that it takes away the pop from the picture. Also content is just going to improve in quality as time goes on, more bitrate, HEVC etc so I'd like to be on the side of detail personally and let medium preset take care of those harder to deal with bands.

I was thinking perhaps some years down the track the presets may need to be looked at again as banding becomes less of an issue.

[Razoola]

Have you checked that your card even has onboard audio? I know my first card I though I could get audio from DVI but the card never even supported audio over HDMI. Maybe you have already done the following but I suggest going into device manager and confirming there are audio drivers installed for your GFX card. You should also make sure that audio output device is set correctly in reclock if you use it (or your media player) to force sound over the gfx card audio.

[baii]

Another silly question.

Using a relative weaker Gpu (HD 6650m, actually have no problem on a 6750m, only difference is ddr3 vs ddr5 vram) for smooth motion on 23fps on 60hz screen, and smooth motion almost double the rendering time.

Is there options to make smooth motion less "expensive'?

[Niyawa]

Is there options to make smooth motion less "expensive'?
Having "use a separate device for presentation" enabled halves my rendering time with smooth motion. But that's about all that made any significant difference.

[Werewolfy]

Just wondering, is with the content you feel with maxdif being a touch too low at 1.9, should this content not be better served by the medium preset?
For anime I'd like to keep low on by default maybe even with 10 bit content noticeable banding doesn't happen that often (unless you have low quality content), I do however also like some of the minor blending that occurs in
very minor banded areas, just not too much that it takes away the pop from the picture. Also content is just going to improve in quality as time goes on, more bitrate, HEVC etc so I'd like to be on the side of detail personally and let medium preset take care of those harder to deal with bands.


The content I feel maxDif at 1.9 is just a bit too low is an anime ;) The medium preset will blur some details so I try to deband as much as possible with the low preset while preserving details and I don't see a difference there between low or high value for maxDif even if I'm very picky.

Like I said, other testers or madshi need to confirm that. But do you see the difference on every content? Or is there some contents where the difference is far more pronounced? If that's the case, a sample video would help.

[pie1394]

SPDIF and HDMI Audio are two completely different things. For example you can't output LPCM 5.1 over SPDIF, but you can over HDMI. Some cards support HDMI Audio over a DVI output with an adapter (sometimes you need to use the right adapter that has a shunt to "signal" the card it's actually HDMI), some don't.

Urm.. You are right... There is something wrong with my memory. @_@

The HDMI audio data frame is also over TMDS signal pins -- in different packet format. So does DP's audio.

I just connected one of the DVI outputs to my AVR with this cable:
http://www.beamershop24.nl/hdmi-kabel/oehlbach-real-matrix-hdmi-und-dvi-kabel-12-meter/

I get no sound........ @pie1394None of the 6 HDMI audio devices 'lit up'.....

The regular DVI-HDMI cable does not map all HDMI signal pins or signal impedence to the card's DVI connector. So it is always detected as the DVI device. It needs the card vendor-designed adapter to use the DVI connector port as standard HDMI output --- just like the cable picture I attached previously provided with Asus Royal King card.

[Mangix]

The regular DVI-HDMI cable does not map all HDMI signal pins or signal impedence to the card's DVI connector. So it is always detected as the DVI device. It needs the card vendor-designed adapter to use the DVI connector port as standard HDMI output --- just like the cable picture I attached previously provided with Asus Royal King card.

The weird thing is that on my nvidia card, a generic DVI-HDMI adapter provides sound. I am not sure what the exact situation is for AMD cards but I do also have an ATi adapter which works as well. The difference between both adapters is that the ATi one has a DVI-I interface while the generic is just regular single-link DVI-D.

[huhn]

The weird thing is that on my nvidia card, a generic DVI-HDMI adapter provides sound. I am not sure what the exact situation is for AMD cards but I do also have an ATi adapter which works as well. The difference between both adapters is that the ATi one has a DVI-I interface while the generic is just regular single-link DVI-D.

i used the nvidia adapter with an singel link dvi-i cabel and it works too...

yeah i put the adapter in the tv ...

[ryrynz]

But do you see the difference on every content?

Yeah I see it on all content. I don't expect many would pick up the difference so I'd rather spare some pixels the change. It may be that a calibrated screen helps out somewhat here.

[pie1394]

The weird thing is that on my nvidia card, a generic DVI-HDMI adapter provides sound. I am not sure what the exact situation is for AMD cards but I do also have an ATi adapter which works as well. The difference between both adapters is that the ATi one has a DVI-I interface while the generic is just regular single-link DVI-D.

Now I have understood why the special adapter is required for some ATi cards. :p

As the source information says, such vendor-made adapter contains a specific I2C EEPROM wired on the I2C bus (for DDC function). So I guess it is how the GPU driver can decide if it should make such video port's function with HDMI capability or just regular DVI one.

But I don't know if all ATi card manufacturers are focibly to use such feature to design the DVI connector's functions on their cards...

[DragonQ]

Some nVidia cards also had audio output over DVI using a special adapter. Quite clever really but I think modern cards just have HDMI outputs instead.

[LigH]

My only problem with HDMI is that it is not fully compatible with DisplayPort.

There are DP-HDMI adapter cables ... but: Someone told me that you may connect DP out ports with HDMI in ports, but not vice versa. I did not know before; a DP monitor connected to a HDMI graphic card reports "no signal".

[jmonier]

This LONG discussion about audio over HDMI has nothing to do with madVR. Maybe it could be moved to a new thread?

[Guest]

Indeed it is off topic for this thread. Further off topic posts may incur strikes, so please stay on topic.

[THX-UltraII]

Yesterday I finally installed my HD7950 GPU. I put the card on 3840x2160@23Hz and put on some 1080p Blu-Ray content. My VW1000 now reports that it is receiving a 3840x2160 signal. Pretty impressive to see this for the first time :)

I settled on Jinc3+AR for chroma upscaling, Jinc3+AR for luma upscaling and Spline3 for luma downscaling. Going any higher than this causes stutter. When I was on 1080p HTPC=>1080p on projector I could easily do Jinc8+AR in madVR. So it is clear how heavy it is to upscale 1920x1080 to 3840x2160. Of course this makes sense to me because when doing 1080p=>1080p only chroma upscaling does something and if I am correct, with 1080p=>2160p, luma upscaling also comes into play.

But I just found another setting that I m interesting in how this works exactly. This involves the calibration options in madVR. Please do not blame me if I did not read all 1000 pages where my answer might be hidden somewhere!
I own a Sony VPL-VW1000ES projector and I ISF calibrated this display. My calibration reports are pretty impressive by the way. If someone is interested in the reports send me a PM so I can send them to you (this is way off-topic for this thread :D). So what do in my case, the first option (disable calibration controls for this display) or the second one (this display is already calibrated). What I find most interesting is WHY i should choose the first or second one and WHAT it actually does.

[sebdelsol]

@madshi
I've developped on windows 7 a player for my HTPC using madVR (great stuff BTW).

Everything's fine, yet in fullscreen (either windowed or FSE) the madVR's child IvideoWindow is stuck behind its openGL parent UI window.
It only happens when the UI window is double buffered. It seems to be a issue with WGL' swapbuffers().

I've tried a lot of stuff: put the player and the UI in the same thread, pump waiting graph messages, give the focus to the IvideoWindow,
send every relevant windows messages I could think of to the IvideoWindow (directly or with NotifyOwnerMessage)... to no avail.
I've even tried to destroy the openGL context, it works but stops the UI.

The only reliable hack I've found is to prevent madVR to go fullscreen at start by changing the parent UI window's size by one pixel, build the graph, start playing,
then switch the UI window back to its original fullscreen size when i'm sure swapbuffers() has stopped being called.
It works... but it's nasty, causes some flickerings and delays somewhat the beginning of the video.

I'm not even sure it's a madVR bug, since EVR seems to behave the same and it only happens on some video cards (ATI actually).
I still ask, in case it would ring a bell :)

[bacondither]

Is it just me or is angleboost about 5x heavier on the gpu than localcontrast?
Yes angleboost seems to require much more then just the "normal" algorithm. I ran three tests of the rendering times and averaged and normalised them so the no debanding category had an index of 100.
No debanding 100%
Debanding, angleboost off 114%
Debanding, angleboost on 153%


Ouch! 53% longer rendering times compared to 14% longer with angleboost off.
And with my material that have low to moderelaty banding i fail to see any improvment with angleboost on. Only material with heavy banding seems to get any improvement.

Madshi, please keep the custom debanding in the stable version! :)

[madshi]

could you tell me what the difference in PQ will be when using DXVA (HW) decoding and when NOT using any HW decoding (software decoding?).
None. I prefer software decoding myself.

Anyone know why? I have Intel GPU with Windows 8.1 if that matters. It used to work a few months ago and I didn't change any settings so not sure what it could be.
Don't know, try reinstalling the GPU drivers. I haven't installed 8.1 yet, but with 8.0 overlay works just fine.

Am I right in thinking that I'm loosing half the detail due to madVR resizing the 2160 SBS hortz pixel source to 1080 (due to panel res set to 1080p) and then the plasma re expanding half that back for each L/R 1080 frame for each eye?
Well, not every source has full 1080 image detail, so maybe it's not always half the detail you're losing. But with better sources you're definitely losing a bit of resolution/detail, yes.

If only the areas of "real" banded content was affected more by these settings it would be great, but what I see is enabling deband changes much of the picture quite subtly (almost 50% of the image gets changed in some way by enabling debanding at the low preset) so it's no wonder things look softer.
What you say sounds logical, but it's only partially so. The way the algorithm works is that every pixel in big gradient areas is modified. The less banding a source contains, the more pixels are actually likely to be replaced. So you should not aim to limit the number of pixels changed, you should aim at adjusting the tresholds so that the *right* pixels are changed. Which no statistics will tell you. The number of changed pixels has no real meaning. It's not like if you reduce the number of modified pixels to less than 20% of the image then suddenly all detail is preserved, and if the number is more than 20% then everything gets blurred over. In some scenes 90% of the pixels could be replaced and the detail might still be preserved. In other scenes only 10% of the pixels could be replaced and there could still be some detail loss.

Is there options to make smooth motion less "expensive'?
There is a "trade quality for performance" option for that. And generally, fullscreen exclusive mode is recommended for smooth motion. Other than that there's not much I can do.

Yesterday I finally installed my HD7950 GPU. I put the card on 3840x2160@23Hz and put on some 1080p Blu-Ray content. My VW1000 now reports that it is receiving a 3840x2160 signal. Pretty impressive to see this for the first time :)

I settled on Jinc3+AR for chroma upscaling, Jinc3+AR for luma upscaling and Spline3 for luma downscaling. Going any higher than this causes stutter. When I was on 1080p HTPC=>1080p on projector I could easily do Jinc8+AR in madVR. So it is clear how heavy it is to upscale 1920x1080 to 3840x2160. Of course this makes sense to me because when doing 1080p=>1080p only chroma upscaling does something and if I am correct, with 1080p=>2160p, luma upscaling also comes into play.
FWIW, I've never recommended Jinc8. Jinc3 is all you ever need.

So how does Jinc3 AR compare to Sony's upscaling algorithm? Have you compared yet?

what do in my case, the first option (disable calibration controls for this display) or the second one (this display is already calibrated). What I find most interesting is WHY i should choose the first or second one and WHAT it actually does.
If you disable the calibration controls, basically madVR will disable some algorithms. If you tell madVR how your display is calibrated, it can auto-correct different source gamuts to match your display gamut. E.g. DVDs usually have BT.601 or PAL/EBU primaries, but your display is probably calibrated to BT.709 primaries. If you disable calibration controls, madVR will playback DVDs with their native primaries and you will probably get ever so slightly incorrect colors as a result. If you tell madVR how your projector is calibrated, madVR will convert DVD content to the primaries your projector is calibrated to, so you will get 100% correct colors for DVDs, too.

I've developped on windows 7 a player for my HTPC using madVR (great stuff BTW).

Everything's fine, yet in fullscreen (either windowed or FSE) the madVR's child IvideoWindow is stuck behind its openGL parent UI window.
It only happens when the UI window is double buffered. It seems to be a issue with WGL' swapbuffers().

I've tried a lot of stuff: put the player and the UI in the same thread, pump waiting graph messages, give the focus to the IvideoWindow,
send every relevant windows messages I could think of to the IvideoWindow (directly or with NotifyOwnerMessage)... to no avail.
I've even tried to destroy the openGL context, it works but stops the UI.

The only reliable hack I've found is to prevent madVR to go fullscreen at start by changing the parent UI window's size by one pixel, build the graph, start playing,
then switch the UI window back to its original fullscreen size when i'm sure swapbuffers() has stopped being called.
It works... but it's nasty, causes some flickerings and delays somewhat the beginning of the video.

I'm not even sure it's a madVR bug, since EVR seems to behave the same and it only happens on some video cards (ATI actually).
I still ask, in case it would ring a bell :)
Good to know that EVR shows the same problem. That saves me some worry. My suggestion would be to temporarily disable madVR's fullscreen exclusive mode by using madVR's custom interfaces (see developers\interfaces\mvrInterfaces.h). And then to re-enable it once your windows are all fully setup etc. This way you don't have to do weird things to the window size, which might reduce/remove the flickering. Other than that I don't know what to say...

Yes angleboost seems to require much more then just the "normal" algorithm. I ran three tests of the rendering times and averaged and normalised them so the no debanding category had an index of 100.
No debanding 100%
Debanding, angleboost off 114%
Debanding, angleboost on 153%


Ouch! 53% longer rendering times compared to 14% longer with angleboost off.
And with my material that have low to moderelaty banding i fail to see any improvment with angleboost on. Only material with heavy banding seems to get any improvement.
Maybe I can switch the angle calculation to 8bit instead of 16bit. I'll check if that reduces the GPU load. In any case, I was planning to offer a "trade quality for performance" setting for angleBoost. But that also means we'll need different low/mid/high presets with and without angleBoost. <sigh>

[6233638]

I settled on Jinc3+AR for chroma upscaling, Jinc3+AR for luma upscaling and Spline3 for luma downscaling. Going any higher than this causes stutter. When I was on 1080p HTPC=>1080p on projector I could easily do Jinc8+AR in madVR. So it is clear how heavy it is to upscale 1920x1080 to 3840x2160. Of course this makes sense to me because when doing 1080p=>1080p only chroma upscaling does something and if I am correct, with 1080p=>2160p, luma upscaling also comes into play.Jinc 3 is better than Jinc 8 anyway, and you can probably drop Chroma to Bicubic 75 AR rather than Jinc 3 AR to reduce the load.

I own a Sony VPL-VW1000ES projector and I ISF calibrated this display. My calibration reports are pretty impressive by the way. If someone is interested in the reports send me a PM so I can send them to you (this is way off-topic for this thread :D). So what do in my case, the first option (disable calibration controls for this display) or the second one (this display is already calibrated). What I find most interesting is WHY i should choose the first or second one and WHAT it actually does.Typically displays are calibrated to BT.709. When you tell madVR this, it will then be able to perform any necessary colorspace conversions so that content which is not BT.709 encoded looks correct. (e.g. SD content)

[huhn]

What you say sounds logical, but it's only partially so. The way the algorithm works is that every pixel in big gradient areas is modified. The less banding a source contains, the more pixels are actually likely to be replaced. So you should not aim to limit the number of pixels changed, you should aim at adjusting the tresholds so that the *right* pixels are changed. Which no statistics will tell you. The number of changed pixels has no real meaning. It's not like if you reduce the number of modified pixels to less than 20% of the image then suddenly all detail is preserved, and if the number is more than 20% then everything gets blurred over. In some scenes 90% of the pixels could be replaced and the detail might still be preserved. In other scenes only 10% of the pixels could be replaced and there could still be some detail loss.

isen't madvr converting both bt 601 and 709 to srgb ?

i can't see a difference between a dvd version and a bd if both are right flags.

and is a tv in pc mode still "waiting" for bt 709 or srgb? and what about pc monitors?

[e-t172]

isen't madvr converting both bt 601 and 709 to srgb ?

Nope. It is converting to what you're telling it to in the options.

i can't see a difference between a dvd version and a bd if both are right flags.

That means it's working.

and is a tv in pc mode still "waiting" for bt 709 or srgb? and what about pc monitors?

BT.709 and sRGB are not directly comparable because BT.709 does not specify gamma, only primaries (gamut). That being said, BT.709 has exactly the same primaries as sRGB. We can compare BT.709+BT.1886 vs. sRGB, in which case they are very similar since the BT.1886 curve is very similar to the sRGB curve on typical LCD screens. So it doesn't really make a big difference what the TV expects.

[huhn]

Nope. It is converting to what you're telling it to in the options.

but it is on "disable calibration for this display"

[turbojet]

Has anyone found a video where gradient 4.0 helps over 2.0?

It really destroys details and that's about it here. Seems others are sticking with the really high gradient values though. If that's the case in the final high setting I really hope there's a way to customize it. Suggested low and mid values show a lot of banding to me, using them is not an option.

[Ver Greeneyes]

BT.709 and sRGB are not directly comparable because BT.709 does not specify gamma, only primaries (gamut).
Rec. 709 uses the following transfer function:
https://upload.wikimedia.org/math/e/c/3/ec35257e2949acc4d559fe436e1187a2.png
So it uses a gamma curve with power 1/0.45 = 2.22 and has an effective gamma of 1.95.

[e-t172]

No. BT.709 only provides the opto-electronic transfer function (camera gamma), which we don't care about. What we need is the electro-optical transfert function (display gamma), and to get it you need to look at BT.1886.

[ryrynz]

The less banding a source contains, the more pixels are actually likely to be replaced. So you should not aim to limit the number of pixels changed, you should aim at adjusting the thresholds so that the *right* pixels are changed.


I'm not aiming for a number of pixels changed. I can inspect the differences and see which pixels are excluded and if they are the right pixels to be adjusting. The first and foremost question is, "what does it look like?" then "what was changed?"

Maxdif is one of the less heavy handed settings, with everyone pretty much agreeing upon all the others it's fairly hard to find something that produces similar results with more focused debanding but it looks like dropping maxdif down and substituting with higher avgdif seems the way to go..


The number of changed pixels has no real meaning. It's not like if you reduce the number of modified pixels to less than 20% of the image then suddenly all detail is preserved, and if the number is more than 20% then everything gets blurred over.


I was just mentioning that to show the difference between settings, It was only meant as a comparison.


In some scenes 90% of the pixels could be replaced and the detail might still be preserved. In other scenes only 10% of the pixels could be replaced and there could still be some detail loss.


Yup, you could have a incredibly banded sunset for example where you need to have 50% of the pixels changed in some way for it to actually have effective debanding. In these particular frames though I felt the amount of change was unnecessary as minor as it was. I want to reduce this as much as possible whilst still maintaining deband quality.

Madshi, I'm wondering what are the chances are of having debanding affect files within a particular folder rather than just using filenames? It would be pretty cool to maybe have it say disabled by default but files within a subfolder say cartoons for example having a particular deband preset loaded automatically, this was always one of my wishes for ffdshow automatic preset loading.

[pie1394]

What you say sounds logical, but it's only partially so. The way the algorithm works is that every pixel in big gradient areas is modified. The less banding a source contains, the more pixels are actually likely to be replaced. So you should not aim to limit the number of pixels changed, you should aim at adjusting the tresholds so that the *right* pixels are changed. Which no statistics will tell you. The number of changed pixels has no real meaning. It's not like if you reduce the number of modified pixels to less than 20% of the image then suddenly all detail is preserved, and if the number is more than 20% then everything gets blurred over. In some scenes 90% of the pixels could be replaced and the detail might still be preserved. In other scenes only 10% of the pixels could be replaced and there could still be some detail loss.


Noise reduction first, and then the resolution-enhancement again? :p

I guess it might need the method of comparing with material pattern library to distinguish what level of processing should be applied. As I have heard, it is part of the super resolution engine processing in Sony's X-Reality PRO (XCA7 for 2K, XCA8 for 4K) and Toshiba's.

X-Reality_Pro (http://www.sony-asia.com/microsite/bravia/assets/X-Reality_Pro/)

But it means more GPU computing power is required to do the comparison if it also contains the temporal frame's comparison. Maybe extra 300% ~ 500% computing / memory bandwidth loading? :p

[THX-UltraII]

If you disable the calibration controls, basically madVR will disable some algorithms. If you tell madVR how your display is calibrated, it can auto-correct different source gamuts to match your display gamut. E.g. DVDs usually have BT.601 or PAL/EBU primaries, but your display is probably calibrated to BT.709 primaries. If you disable calibration controls, madVR will playback DVDs with their native primaries and you will probably get ever so slightly incorrect colors as a result. If you tell madVR how your projector is calibrated, madVR will convert DVD content to the primaries your projector is calibrated to, so you will get 100% correct colors for DVDs, too

I ISF calibrated my display @BT.709. There also appears an option where I can choose between 'PURE POWER CURVE' or
'BT.709/601 CURVE'. For both I can choose 1.80-2.60 with 0.5 steps. Which curve does what? And do I set the gamma value on the average gamma result of my calibration?

Here is the calibration I did for my Sony VPL-VW1000ES projector:
http://s15.postimg.org/wgaa2iwav/D65_6_0_3_0_0_0.jpg (http://postimg.org/image/wgaa2iwav/)

http://s14.postimg.org/8mvku176l/color_50.jpg (http://postimg.org/image/8mvku176l/)

[THX-UltraII]

So how does Jinc3 AR compare to Sony's upscaling algorithm? Have you compared yet?Did not have time yet for extensive testing but my first findings are that the HTPC (madVR) scaling outperforms Reality Creation on the VW1000.

Could you explain a bit more on the 'scale in linear light' checkbox?

[THX-UltraII]

And Madshi, when you have some spare time left, could you check this thread on the AVS community where I started a discussion about BTB and WTW clipping:

http://www.avsforum.com/t/1174468/the-official-chromapure-thread/4800_100#post_23876711

Read from post #4822 to #4830.

What you think about it? It is suggested there to put madVR to 16-235 so BTB and WTW does not get clipped......

[James Freeman]

Good News.

JanWillem32 from this board helped (actually he written all of it) me write a customizable 3 section gamma control.

http://forum.doom9.org/showthread.php?p=1650230#post1650230

Works like charm for a highly custom BT.1886 curve.

No more over bright BT.1886 shadows !!!
You can change from what point YOU like, to how much Brighter YOU like. :D

I use only the lower part to brighten the shadows (emulate BT.1886 to my taste).
Looks better and a lot faster (no waiting for calibration).

No more Crushed Blacks, OR , Raised Blacks.
Set it exactly where you like it (matches your viewing environment and Monitor).
You can change the black level below a certain point without touching anything above it.

You can also raise the MadVR gamma to 2.6 (for a deeper picture) and NOT LOSE ANY SHADOW DETAIL.
This is just perfect !!!

Again, thanks to JanWillem32.

[6233638]

Has anyone found a video where gradient 4.0 helps over 2.0?Yes, I have a number of videos which show improvements from gradient values above 2.

EDIT: Here's an example:

http://www.abload.de/thumb/nodebandetsv2.png (http://www.abload.de/img/nodebandetsv2.png) http://www.abload.de/thumb/grad20j9sll.png (http://www.abload.de/img/grad20j9sll.png) http://www.abload.de/thumb/grad59wcsei.png (http://www.abload.de/img/grad59wcsei.png)

It really destroys details and that's about it here. Seems others are sticking with the really high gradient values though. If that's the case in the final high setting I really hope there's a way to customize it. Suggested low and mid values show a lot of banding to me, using them is not an option.Well a "high" setting would be for when debanding matters more than detail preservation.

I ISF calibrated my display @BT.709. There also appears an option where I can choose between 'PURE POWER CURVE' or
'BT.709/601 CURVE'. For both I can choose 1.80-2.60 with 0.5 steps. Which curve does what? And do I set the gamma value on the average gamma result of my calibration?You seem to be calibrated to 2.40 gamma. It shouldn't affect anything unless you also enable gamma processing though, which you should not need when it's already calibrated externally.

Could you explain a bit more on the 'scale in linear light' checkbox?The short version is: don't enable it for upscaling.

What you think about it? It is suggested there to put madVR to 16-235 so BTB and WTW does not get clipped......Below black and above white information should not be visible.

[Werewolfy]

@madshi

I was thinking about it from a long time and I really think that you should keep the custom mode in the final build. There will always be some users that prefer to decrease someting or increase someting else. Personnaly I would set Gradient at 1.3 for the low preset for example.

There's no disadvantage to anyone to keep the custom mode, you can make it optional in the menus for example. If you don't tick the option, you can't modify the presets so a beginner won't be "bothered" by it and an advanced user won't have to live with a preset he doesn't really like. In that case, everyone could make his own set of presets. It will also prevent some very long discussion to find the "perfect" preset if everyone know that he can modify it.