Closed BMF54123 closed 2 years ago
Do you have a link to this test rom?
This should be fixed in the latest unstable build.
Well, there's some color now, but it is really faint compared with the real Famicom capture above. Tested both the unstable build and the latest build from 9/10/21:
A side-by-side comparison of range $x0 makes it much more noticeable (MiSTer on left, Famicom on right):
this is my version: NES.zip
it has slightly different tints on some enphasis, but at least it gives different tints in all groups. Currently core operates in rgb pixel level for tint while it should: 1) use original phase coded color. 2) use YIQ color space.
Each emphasis bit uses 1/3 of whole phase range and thus it doesn't affect cleanly R, G or B. So enphasis has "dirty" tint. That's why real NES has dirty green in "011" tint mode while my version (using RGB color space on pixels level) produces logically Yellow tint. Probably, instead of making tint over-complicated, we can hardcode tint substraction table. it won't be 100% correct but will be hard to distinguish.
p.s.: correct term would be de-emphasis. So each bit is de-emphasing by some tint (by 1/3 of phase range) but after phase and color space conversions it turned to RGB emphasis which is technically wrong point of view.
this build has some adjustments: NES_adj.zip
That is definitely a huge improvement! However, it appears that the entire image is now shifted right by one pixel (e.g. the right-most pixel on the screen is missing, and there is an extra black pixel at the left edge). Cropping and resolution settings don't seem to affect it.
That's easy to correct. I will proceed with new emphasis.
is this resolved? There's only a handful of games that use emphasis at all, mostly for just dimming the screen.
https://github.com/MiSTer-devel/NES_MiSTer/commit/abddbc94824d46cdde0d7dbea90bd39046d03b14 - Sorgelig committed this last year, so it should be fixed. Testing to be sure:
- Kitrinx Palette on makes it a little too subtle on the right to show that it's working.
- Kitrinx High Saturation Palette shows it.
A little better. Not sure how to replicate @BMF54123's results completely since it seems to be palette dependent, but the emphasis bits are showing.
This is an interesting result: - the PC-10 Better Palette shows all black for the right column. However if we go to the original PC-10 Palette: - The right column does show up. I think this indicates that the PC-10 Better Palette should probably be adjusted somewhat?
this is not a bug
333,014,006,326,403,503,510,420,320,120,031,040,022,000,000,000
555,036,027,407,507,704,700,630,430,140,040,053,044,000,000,000
777,357,447,637,707,737,740,750,660,360,070,276,077,000,000,000
777,567,657,757,747,755,764,772,773,572,473,276,467,000,000,000
You'll notice the 0x0D column does not have the two lowest values. See here for more information: https://www.nesdev.org/wiki/PPU_palettes
In the MiSTer NES core, enabling two color emphasis bits simultaneously causes all three channels to be darkened. As a result, you cannot get yellow, magenta, or cyan tints.
MiSTer core:
AV Famicom (stock composite output):
This behavior occurs regardless of which palette is selected in the menu.