Second Cluster Study - Knights of the Round

[0xDB]

Dennis: Why did you scale up by 700x1200 percent? That wont give you the correct ratio. You need to scale up 700x900% at some point to get the proper 4:3 ratio in the case of CPS games.

At first I tried 1200x700 which looked completely wrong, so I simply reversed width and height and went on with that without thinking about the math behind it.

@Dennis: I noticed something that looks like a structural problem with your man's left thigh. The er, bulge in the loincloth just below the belt, the straight angle, and the highlight on the leg leading all the way up to the cloth suggest to me a pole-ish shape that ends abruptly below the belt. It's more visible in the second edition. And maybe the leg could be brought back some more with darker shades? I like how you did the scales and the shell-like shoulder pad. Good stuff.

Thanks, I'll look into that as well, when I make my next edit.

Meanwhile, I've been trying to emulate a screen effect, using only layers, so it could be used to pixel with a 7x9 brush and see the effect in real-time. A picture says more than a thousand words I guess:


Ok, some words to go along with that:
There is one layer which contains the pixels themselves (at 700x900 percent scale). On top of that is one full white layer to brighten stuff up a bit, because the next layer, which emulates a black matrix shadow mask and the phosphor stripes, darkens the overall image significantly. Then there's another layer which emulates the blank lines.

After doing this, I think I even for the first time understand why there is a need for "scanlines" or blank lines anyway. It has to have something to do with the shape of the shadow mask in the screen. Thinking about it, if one would not turn the electron beam off at times to make up for the shape of the shadow mask, it would not even be possible to draw a perfect horizontal line. Even with blank lines, some of the phosphor still glows into them, more of it on bright colors because of more electrons hitting it there... hm, don't know if this is correct though, since I'm no screen engineer.

[Conzeit]

@Dennis I think shadow mask emulation is the ideal thing to do so props for that, there's just one thing we need to figure out now, how do we do this without darkening bright pixels as much as dark pixels?. I personally think that's a big part of the difference in authenticity, look at the difference between Ptoing's first and second try, there's a world of difference IMO.

BTW, What's the blending mode of the Shadow Mask layer? it's kind of complicated recreating the shadow mask without separating each channel...I dont imagine how you did it. I had thought of emulating a shadow mask by having a R layer a G layer a B layer, and then setting masks so that each channel goes to the right place in the shadow mask....if you get what I mean c.c

Ptoing, dont you think the whole scanline layer could be a gradient map adjustment layer, which goes from black to  a light grey, set only to affect luminance? that way it'll affect the white parts only lightly like we want to, and we dont have to re-create the masked scanline layer every time like you did. That still misses all the blurring and merging afterwards :p

EDIT:About the pose:
these are the two variations of the Vom tag / "from the roof", longsword stance. That's probably what they were going for ingame...too bad they made it into that stupid baseball stance.

http://www.thehaca.com/essays/StancesIntro.htm here you can see the stance and it's variations in more detail, it's the "From the roof" stance

[ptoing]

That might work, but I don't think it will give as good results.
Also, scanlines have NOTHING to do with the shadow mask or aperture grille at all. It is just that the beams get spaced out in lowres games.

A better way to do shadow mask stuff is to make a layer which has the shadowmask stuff as cyan, magenta and yellow, and the black bits as white and then set that layer to subtract and you will get proper colour values without extra darkening. You still need to ramp it up a bit usually, you can simply double the layer on top of itself (everything merged) and set it to overlay or lighten or something, play around.

[API-Beast]

I fixed a few things on my sprite which weren't as appearant on the modern pixel aspect ration (I assume that's 1:1?)

[Satsume]

i've tried too >.<

[0xDB]

@Dennis I think shadow mask emulation is the ideal thing to do so props for that, there's just one thing we need to figure out now, how do we do this without darkening bright pixels as much as dark pixels?. I personally think that's a big part of the difference in authenticity, look at the difference between Ptoing's first and second try, there's a world of difference IMO.

BTW, What's the blending mode of the Shadow Mask layer? it's kind of complicated recreating the shadow mask without separating each channel...I dont imagine how you did it. I had thought of emulating a shadow mask by having a R layer a G layer a B layer, and then setting masks so that each channel goes to the right place in the shadow mask....if you get what I mean c.c

What ptoing wrote right after your post would be a good way of recreating the shadow mask in a single layer.

By placing cyan into the position of the red phosphor stripe you could mask out everything which is not red in that position. Magenta on the green stripe to mask out everything not green and yellow on the blue stripe to mask out everything which is not blue there. White into the spaces to mask out everything in the spaces. Unfortunately, PSP8 does not have a "subtract" blend mode so I can't properly test it but in theory it should work well.

The blending mode in my previous shadow mask layer was set to multiply but that's wrong and does not mask out the individual components in any way.

EDIT:
Ok, so I downloaded the GIMP (which supports simple subtract layers) and played around with it, using the way ptoing described to emulate the black matrix shadow layer. The GIMP also allows setting individual gamma correction values per view (which I set to 1.5 in the upper left view here). Turns out it's possible to come pretty close but not quite there with just layers (comparison shot of real CRT included in screenshot). The one thing missing is the over-bright appearance of brights. See:

[ptoing]

Your RGB mask looks wonky, with the different sizes of RGB blocks. Also as far as I am concerned, proper emulation of shadowmasks and such is atm not really possible because we do not have high enough resolutions to emulate the analog effects digitally.

Did some fudging around, not feeling it.


Did these quite a while ago, not feeling that either, and again, you can see that it also is pretty high res.
Street Fighter 2
Dodonpachi

We need at least 4 times that res I think, to get a good approximation (at the same screensize. Something like a stupidly high DPI like 500 or so, where every pixel we see atm, would be about 4x4 pixels but together the same size as one is now. Then we can speak about proper emulation of this kinda stuff. If you show me something good looking before that, that really looks like a realistic CRT image, I'll be pleasantly surprised.

[Arne]

I don't know if I feel like coding this, but I think it would be accurate. Basically, each pixel lights up a number of phosphor dots. Of course, before drawing, each source pixel is separated into it's R,G,B components. The emulated phosphor dots would be images of blurry dots (R,G,B) which are drawn in light/screen mode with some overlap, with extra overbright versions for high values. The dots can be scaled down and drawn at fractional pixels so there's no need to render huge images.

Also, the electron beam is not quite aligned to the shadow mask so there'd be some pre-calcs needed for rows and columns to see which dots are covered (possibly every single pixel could have associated dots and a buncha precalc). It seems like suddenly turning down the intensity of the beam has some tapering, but this could be taken into account during precalc me thinks.



This is not something which would run in realtime of course. More of a filter.

[ptoing]

Arne, that could work to the extent of looking like a scaled down reproduction, but not like the real thing IRL. Which is what I meant when I wrote we need higher resolutions.

As far as your mockup image goes, that kinda pixel distribution is somehting which computer CRTs had, but as far as arcade monitors go the use the ||| kinda thing. At least I never seen anything else in arcades (apart from some modern ones which use LCDs as well)

Tho that is just a detail.

MAME has a pretty good HLSL filter with a boatload of sliders for things like phosphor decay and even scanline jitter. But even in that, imo, the shadow mask stuff looks ass.

[Arne]

At lower rez you're bound to get Moiré, but 320*240 emulation on a WUXGA might look OK. I was more thinking of using it for single sprites and stuff though.

Most people don't seem to use any sort of display calibration in MAME, leaving a lot of games looking quite pastel and strange. Perhaps monitor contrast/saturation was cranked up high, because the shadow mask would dull things down mixing in all that black...? My old Amiga stuff (drawn on a TV) doesn't look odd though.


Slightly on topic this time...