Using the current Version 2 preset for the 27…40? Aperture Grille, RWGB, couldn’t find a Mask Accurate Mode setting.
300 TVL:
600 TVL:
I can never get Megatron to look good:
My HDR Guest Advanced preset with Mask Strength set to 0.90:
2 Likes
Just to say chaps (@PU786) the save slot thing has nothing to do with me (at least its very unlikely). You need to raise it on github as an issue so that someone closer to that system can fix it.
2 Likes
Have a play around with the colour spaces see if it makes any difference to you (under Your Display Settings in the shader parameters)
This picture was taken on Super, and was the best I could get the Megatron parameters set to 600 TVL. I’ve read that you don’t really support WOLED’s, and the Z95B is certainly that, so there’s probably something about the panel preventing the full use of Megatron’s capabilities.
And that’s okay, this is why we have so many different shader options to choose from. Your work is still of great value and the ways you’ve improved HDR in retroarch shine with the other HDR presets I’ve made.
1 Like
So Super isnt a Colour Space option on the Sony Megatron: its either r709,sRGB,P3,Adobe or r2020. If the main menu is affecting the colour then… Ah I know whats happening youve got Guest Advanced using Super via the menu and because the Sony Megatron ignores that setting and uses its own colour spaces its using what us equivalent to Accurate which is much more desaturated. So as I say go into the Sony Megatron’s parameter settings and change Colour Space to r2020. That should make things the same as Guest.
Thank you very much.
Slow down a bit there guys. @MajorPainTheCactus, @PU786 is not reporting an issue with colour spaces or anything like that.
He’s reporting something to me, just as @Azurfel did not too long ago because I wanted to refresh and document my knowledge on the capabilities and limitations of the WOLED RWBG subpixel layout.
@PU786, ever since I learned that the new tandem RGB OLED TVs used a different, updated BWRG subpixel layout, I have wanted someone to verify and assist in getting proper subpixel shader support for them.
Due to this, I always recommended the G4 as the best OLED TV for CRT Shader emulation and the G5 with this huge caveat of it possibly needing an update to CRT-Shader Support to look as well as it could.
So now that we actually have a willing participant in you, @PU786, can you confirm how the same close up shots look with the other available Display’s Subpixel Layout, i.e. RGB and BGR? It’s no surprise that RWBG/WOLED doesn’t look or work as intended on your TV as that was designed for previous WOLED TVs, which use a different RWBG layout to your Z95B and the LG G5, G6 and C6H.
Once we see those results, we can know if an existing subpixel layout works fine, or if we need to add a new one with swapped/rearranged subpixel order in order to accommodate the newer model tandem RGB displays.
What Mask, Mask Size, Mask Zoom and Mask Layout are you using for the CRT-Guest-Advanced pics? Also, can you increase the Mask Strength to 1 for those please?
1 Like
This is as close as I could get.
Still a ways off but, slightly better? I was noticing it says RWBG instead of RGWB…I’ve always kinda presumed that was the issue.
The pictures aren’t honestly capturing the night and day difference between brightness. One is with my calibrated settings (Guest Advanced) the other is with my panel cranked to the max brightness, which, while beyond overkill and washing out most things like movies/etc. still leaves Megatron washed out and dim.
Guest Advanced is set to Mask 11 with default zoom/mask layouts. This is with Mask Strength to 1, which doesn’t make much of a difference…
This is BGR Layout, which just narrowly lowers some specular highlights in the orange bricks:
Legit my camera is hiding differences/adding differences that aren’t here. Like the difference in Sonic’s blue from the Mask Strength change, and not quite capturing the difference in the orange blocks. Not super important stuff.
1 Like
I’m not really interested in the full screen shots.
Take a look at the post I shared that @Azurfel made for an example of the ideal format. You can use the older Sony Megatron v1 Presets if you wish, since they definitely have a Colour Accurate/Mask Accurate shader parameter but v2 is supposed to be always Mask Accurate so it should be fine as well.
Please don’t confuse what I’m trying to do here with anything colourspace related. We just want to focus on the subpixel layouts with the aim of getting possible support for the new layout in the future.
That Mask 11 if I recall correctly is a B&W mask so it’s subpixel agnostic. There are no R,G or B, Phosphors to be found there.
Update: here is an updated CRT-Guest-Advanced Mask and Layout list:
If you want to post relevant comparisons using CRT-Guest-Adanced, you have to use the R-G-B subpixel Masks, which are Mask 6 (R-G-B), Mask 10 (R-G-B-X) and Mask 12 (RR-GG-BB-X).
For camera settings, you can use Pro/Manual Mode.
ISO low enough to not overexpose the subpixels at close range, so 100 - 250 or so.
Shutter Speed matching your content refresh rate or half of it i.e. 1/60 or 30 or 1/50 1/25.
Focus Auto or Manual
WB 5000K, Auto or whatever you think looks closest to how it looks in person.
Seems like I should have asked this question in the “OLED Subpixels, How do they work?” thread.
2 Likes
After reviewing your screenshots, everything looks as expected. 1080p 300TVL uses the R-G-B-X subpixel layout, which would probably look great at 1440p and 4K.
1080 600TVL uses Magenta-Green.
4K 800TVL uses R-G-B/B-G-R and that is great for 1080p.
1 Like
Yes this is probably expected: have you noticed that black frame insertion does a similar thing?
Its along similar lines as in to get more similar to a CRT you have to turn off most pixels or in this case sub pixels. You mentioned when using Guest you’re using a soft mask - as in this will be switching on sub pixels that wouldnt be switched on if you are to accurately simulate the phosphor triad using the OLEDs sub pixels i.e an RGBX mask.
Take pure white, here’s its sub pixel pattern for a RGBX mask: (1.0, 0.0, 0.0),(0.0,1.0,0.0),(0.0,0.0,1.0),(0.0,0.0,0.0). Out of 12 sub pixels only 3 are on: a 75% reduction in brightness, then add in the vertical scanline simulation i e dim towards the outside and bright in the middle and its even more:say 90% reduction in brightness. So softening the mask has a huge impact on brightness at the expense of simulating a CRT less accurately: those off sub pixels in the above mask are on. This gives a ‘blurry’ look to scanlines.
This is entirely up to you what you prefer but it is to be be expected. Guest Advanced will do the same with strong pure masks no doubt.
1 Like
Just as a side note these macro screenshots arent really that useful: instead try taking pictures of Sonic zoomed right in so we can see whats happening at the sub pixel level. That way I can see what is happening in Guest and what is happening in Megatron.
1 Like
@PU786 Ah OK something is broken! I just tested the Megatron on Sonic and r2020 colour space is broken as in we’re not getting an equivalent colour space to Super so you’re probably still comparing Accurate (r709) to Super (r2020). Ill get this fixed ASAP so that we can get a fair comparison.
1 Like
Sadly this just affected the _config versions of the Megatron but its fixed now.
2 Likes
Since newer WOLED monitors are getting RWBG, and apparently the G5/Z95B Tandem panels have BWRG subpixel layouts, are those going to become options in the parameters?
Apparently my panel is BWRG.
2 Likes
@PU786 It’s hard to tell from those photos, but it looks like you might be seeing the same color issues as i have been observing in more recent versions of Megatron.
Please try this version based on Megatron 2.6.2: hdr262-AzMods
I have included presets for all three supported subpixel layouts (RGB, RWBG/RBG, and BGR), but they are otherwise identical. Try all three.
A new “BWRG/BRG” setting may need to be added to allow Megatron to work properly with the BWRG WOLED subpixel layout.
2 Likes
As a user, I usually never make any kind of request unless I feel like it’s something that will be appreciated by other users and also worth the effort from the developer, but I think this fits:
@MajorPainTheCactus could you create a crt-sony-megatron-v2-default-sdr-NTSC.slangp preset? This would add NTSC to the (currently) only Sony Megatron v2 SRD preset, and serve as a walkaround with NTSC for Linux users while HDR does not work correctly on Linux.
I’m aware that the Sony Megatron v2 detects whether HDR is enabled or not and works accordingly, but since the v2 SDR preset exists with a few parameters tweaked, I think it would be good to have it with NTSC as well.
But I totally understand if it’s not worth the effort or you’re not interested. Thanks again for the work on these shaders!
Yes so although the sub pixel layout is important for spacing/ordering of the phosphor triad it wont effect colours/brightness. You can see this by rotating through your sub pixel layout. What it will do is show incorrect fringing to objects.
I took some time to look into your colour issues tonight but I couldn’t really see any issues apart from the _config.slang version not having the correct colour space order. In r2020 I get the same values as you would without the shader for the peak point of the scanlines. In order to get the right colours on my QD-OLED 1600nits display (using Super/r2020 colour space as you were using) I need to put the displays brightness up to max and set Megatron’s (_config version) peak and paper white luminance to 1600 and 800 respectively i.e half peak luminance.
2 Likes
So the first thing to point out is that @PU786 is using Super colour space whereas I believe you’re interested in Accurate? This changes things quite radically as there is no r2020 to r 709 conversion going on and so you’d see SDR pixel patterns but with far more saturated colours.
I’d be very interested if you can see your colour shifting in this r2020/Super colour space.
I did investigate this extensively and I havent been able to find any colour shifting from a shader output point of view (i.e raw values output by the shader as seen in RenderDoc). What goes into scanline generation comes out the other side in the peak of the scanline.
By what goes in I mean when scanlines turned off in the built in shaders (which use the same scanline generation code as Megatron). The peak point of the scanlines have identical values to that original image. BUT I will say the non scanline image (that all the custom SDR shaders use) is much more desaturated because of HDR brightness so maybe its just a gamma issue - but then you said you were happy with the image when scanlines are turned off in the internal shader.
Another thing to try for me is scRGB - Windows doesn’t appear to rotate into r709 and so the colours aren’t accurate in ‘Accurate’ mode - you can see this in greens as you rotate through the colours spaces - the greens dont really change much when in fact the oppsite should be true - they should change the most.
One thing I will say is that I do get quite different behaviour across devices and screens and as I haven’t got a W-OLED screen maybe this is something specific to that type of screen Im just not able to see.
1 Like
I have not noticed any desaturation in non-scanline HDR with Luminance set to 100, in comparison to my C1’s calibrated 100 nits Rec709 SDR, or in comparison to my iPhone 12 (with brightness matched to my C1 in SDR), or in comparison to the color seen with RetroArch running in SDR, converted to HDR by Windows 10 (gamma doesn’t match the Windows 10 conversion, but that is to be expected, since the Windows SDR to HDR conversion uses piecewise sRGB gamma rather than power.)
I think it is possible this is where our disconnect lies. It sounds like you are comparing non-scanline and scanline using the same high Luminance value required for scanline? I am comparing non-scanline at 100 Luminance to scanline at ~670 Luminance.
100 nits was the closest thing to a standard brightness for CRTs in the pre-HD era, and it is still the standard brightness for SDR today, thus that is my target equivalent baseline for both brightness and color.
Non-scanline Colour Boost: Super is still a good color match for Megatron with Colour System set to r2020 (tho Megatron seems to require lower Luminance values for r2020 compared to the narrower gamuts, which can make it look oversaturated in comparison to non-scanline when using the higher Luminance values needed for Rec709/Rec601/NTSC-J/EBU/etc).
scRGB makes scanline colors match non-scanline colors for Accurate/Rec709. I have known this straight along, and i suspect this is because it has that bug that effectively undoes the mask changes from 2.6.5.
That, combined with Super and r2020 matching is part of why i have so strongly suspected the mask changes were the issue.
I think this is because the Colour System and Phosphor settings don’t work as intended (and they never have, which is why my testing forks hijacked the old Colour Boost setting as an alternate route to add additional gamuts).
Please take a look at hdr262-AzMods and cycle through the “Color Gamut” options. Greens change just as much as they do for non-scanline, at least on my C1.
And the difference between non-scanline Accurate/Rec709 and Super/Rec2020 looks about the same as using my C1’s gamut override in SDR to swap between Rec709 and Rec2020 (tho i can’t side by side that obviously. I should verify with my colorimeter at some point come to think on it.)
1 Like