> I understand the appeal of increased resolution, improved textures, overclocking to
> get rid of slowdown and flicker that occurred on the original machine, etc., it's an
> interesting new way to look at this stuff, but that should not be considered to be
> pure "emulation." Just call it what it is: modding. I'm not against this stuff per
> se, but it should never, never be considered the end goal of the emulation scene.
> Also, don't forget that home systems were meant to be viewed on low-resolution CRTs
> through analog RCA composite input (at best). Sure, the developers would have likely
> worked on RGB computer monitors, but they'd still be outputting at a lower native
> resolution with no bilinear filtering or upscaling. On a modern display, they are
> going to look like crap, and not just because graphics have advanced so much in the
> interim, but also because they have been completely stripped of the degradation in
> picture that occurred in the original delivery methods. There's something to be said
> for simulating loss of picture quality/fidelity, that's the only way to reproduce
> things like the Genesis/Mega Drive's dithered "transparency" and "extra" colors, or
> the "larger" color palette of composite mode CGA on old PC games. Sometimes you have
> to make the picture look "worse" to make it look "better."
> At least when you add a layer of CRT simulation or fake vector glow/flicker, it's
> just a "mask" on top of the base system emulation which is no less accurate than it
> would be without them.
It's not really a loss in quality that has the aesthetic effect, in my opinion. Things like composite artifacts don't lend much. The three major contributions I feel are the scanlines, the shadow mask, and the beam focus.
The beam focus gives a slightly blurred image. Even the best monitors cannot draw perfectly square pixels. The result are subtle soft edges between pixels and scanlines. This will aid in the brain's interpolation which is amplified by the scanline and shadow mask stages.
The scanline effect can be described as essentially an upsampling with zero insertion, i.e. the vertical lines are impulses of color bands. There are some caveats to this. Due to physical factors the color lines and the black lines are not the same height, but that's not really important to the overall idea, which is that zero-insertion is going to look better than stretching out the image with nearest-neighbor or bilinear filtering. With these filters, the constructed image is complete, with no room for your brain to dither things on its own. The bilinear filter will be blurry, while the nearest neighbor will be blocky. With the zero insertion, your brain fills in the black space. It can imagine what the full image 'behind' the scanlines looks like.
The shadow mask contributes to this effect as well, in both x and y dimensions, but to a lesser extent (in cases where there are no scanlines, the shadow mask has a greater effect, since dots per pixel is smaller). The shadow mask, due to its geometry, results in some distortions. This can be thought of as a lowpass filter. This is also why monochrome monitors are much sharper than color ones.