I cannot reproduce on my system, glcore + deconvergence is working fine here:
Using mesa 23.0.3 + Radeon RX 5600XT
After reading about magic glow and theories on glow and what it has to light, i tried to just sum the bloom to the content before applying the mask, and got this, pushed it alot to show better the result:
This is when it is applied after (tuned levels a bit to match the other picture):
So far, it seems that the former should be more ââ ââ ââ accurate ââ ââ ââ ?
-EDIT- Iâve to admit that it looks sexy, so I wonder if it may be worth to add (yet another) parameter to modulate between âbacksideâ and âfrontsideâ bloom.
âŚand then another question, it is not that easy, since the scanline gaps should be cleared by the backside glow at least to a certain extent, right?
4 Likes
Could the difference be down to AMD discrete vs Agpu (soc) I wonder?
Former more accurate - I think that depends on what type of tube is being emulated. My pref is pic number 2, but, as they say, itâs all down to the eye.
With you on that oneâŚtotally agreeâŚis it the case that it needs a blend of the both? Just thinking outloud.
1 Like
Further reading the thread about phosphor illumination, I think the bloom should not go before drawing the mask, it should not light the mask itself, but just the bare unexcited grid.
Dev repo just carries another parameter âBlack level of the unexcided phosphor gridâ that allows to see it.
It is enabled by default with subtle effect, but even the lowest visibility produces moiree when drawind the rgb mask at core coordinates and using curvature, because the steep mask curves.
@crtJoe (about the black level)
8 Likes
Hi there, I made 2 presets that uses 1.375x wider mask while staying at screen coords; those presets can be used with different core resolutions and the masks stay at the same size, contrary to core-sized masks.
Phosphor still width depends on input luminance tho.
If you want wider pixels, but your device has still problems, you can manually edit the setting: PIXELGRID_MUL_X . Setting it to, say, 2.0, will give you double width mask; the higher the value, the more the width and more room for phosphors to dynamicalli grow/shrink.
Those presets stay in presets-ng/Safe-to-use/
Monitor-ScreenLevel_HMaskWide-Aperturegrille.slangp
Monitor-ScreenLevel_HMaskWide-Corelevel_SlotMask.slangp
11 Likes
Hello koko3k, I donât know how to thank you. I use your version with slotmask and activate the FXAA, and the result is spectacular, exactly what I was looking for. Youâve gone beyond all expectations with your work, and I hope your final update has these presets officially available. Just one question, I wanted to add your new NTSC method, as you showed in some previous post with Sonic, as well as toggling the âNTSC: color artifact enanleâ option what should I change? Thank you so much for your work always.
2 Likes
The difference with the previous version is that CVBS bandwidth limited chroma (chromableed) can be selectively cancelled too in zones not covered by artifacts.
So you have/can enable it and then modulate how much you want it to fade away and artifacts to be blurred more.
From docs-ng.md:
**CVBS: NTSC color artifacts: **
Show NTSC artifacts mask (debug):
This will show only the part of the image that contains artifacts.
Use it to for a better visual feedback of the following parameters
Selective artifacts treshold:
Tune this to select more or less artifacts, depending on their strength.
Non-artifacts: Cancel blur (Glow):
How much the glow/blur function will skip blurring "unartifacted" areas.
Non-artifacts: Cancel Bandwidth limited chroma:
How much the Bandwidth limited chroma function will skip blurring
"unartifacted" areas.
2 Likes
Hello, here are some screens adding to your presets, as well as the FXAA, also the NTSC color alteration by performing the settings seen through some of your screens on the subject. What do you think about it ?
2 Likes
The NTSC blur effect is perfect but I would like to find a way to keep the original color. Can anyone please advise me on the optimal values to insert?
Iâm afraid i did not understood what youâre after.
If for "original colorâ ypu mean the one tjat comes out of the core, then you have to use âcancel bandwidth limited chromaâ or disable it (âcvbs limited chroma enableâ)
2 Likes
I noticed that the minimum phosphor width and mask gamma were misconfigured and wontât allow the phosphors to shrink enough, hereâs the tweaked/updated version, increased the min/max X range and tweaked X gamma:
2 Likes
Sorry, I didnât explain myself well: I meant to eliminate the color alterations that I noticed in some points, as shown in the images. As you can see, for example, Ryuâs face or the low Japan writing.
1 Like
It seems you donât like: âCVBS: NTSC color artifactsâ
2 Likes
I like everything, the final effect it gives to the image together with the FXAA is wonderful, but I would like, if possible, to obtain this effect without these alterations. Otherwise I have to look for another way, but seeing the image you posted of Sonic I hadnât seen other alterations apart from the waterfall. I apologize for my ignorance and for my difficulty in understanding certain technical discourses.
1 Like
This screen is perfect, I donât see the color alterations, and it has become my point of reference. Except the mask, the ng mask you created is absolutely perfect.
1 Like
The whole waterfall is altered in color, also:
You can modulate the treshold to under which a pixel is considered artifacted or not using âSelective artifacts tresholdâ.
Then you can choose to sharpen non artifacted pixels:
Keep in mind that if a game has not been produced for ntsc systems would probably not look good, because ntsc itself was exploited in a conscious manner, explicitely generating and, i guess, avoiding artifacts where not needed.
Try with some Megadrive game as a start.
2 Likes
Heartfelt thanks for your help. Tonight I download your new version on GitHub and do more tests. As always, your work is fantastic.
2 Likes
Does this allow you to control the intensity of the CVBS: NTSC color Artifacts
1 Like