Low-res screenshots (Página 2 de 4)

I had actually nearly forgotten about these doubled screenshots. When you write that they have the usual issues, do you mean the general issues of any doubled screenshot, or the specific issues of the examples I provided you a while ago? Because the above examples shouldn't have the same PNG-related colorimetric problems. In fact, doubled and quadrupled screenshots now feature nearly identical colors. So if you still don't like these doubled screenshots, the only reason I can see is the vertical interpolation... Here is a doubled screenshot made without vertical interpolation, if you want to compare. Of course, the beam spot doesn't appear round anymore and is not symetrically depicted:
http://img197.imageshack.us/img197/9963 … terpol.png

As for the quadrupled screenshots, the scanlines may be too dark. But it is very easy to adjust (opacity value in step 5). Do you think the screenshots should be a bit more blurry?

I think non-scaled screenshots should only be corrected by increasing the contrast (possibly using the curves tool for more faithfulness). Slightly upscaled screenshots would not look very sharp, except if you don't use any vertical interpolation, but then you would get ugly artifacts because of the few vertically-doubled pixels. Any screenshot should of course be adjusted to a 4:3 aspect ratio with horizontal interpolation. I guess you already know very well about this, but I will be glad to try to help you if I can. So just let me know when you need such screenshots.
No need to talk about them here and now, but here are already three examples: non-scaled, slightly upscaled using vertical interpolation, and slightly upscaled without interpolation:
http://img195.imageshack.us/i/1xtfiv.png/
http://img191.imageshack.us/i/12xtfivinterpol.png/
http://img534.imageshack.us/i/12xtfivjaggy.png/

OK, here we go. I think I've gotten the hang of the basics:

http://www.soundshock.se/dump/SCR_MD5.png

What confuses me though is games that run at custom resolutions (like SFC Prince of Persia or certain cinematics, which run at 512*448 - should I resize that to 597 x 448?):

http://www.soundshock.se/dump/POP_SF3.png

Another thing is PC88 games (and several other computer games I'm sure) that run at different weird resolutions, and with different aspect ratios it seems:

http://www.soundshock.se/dump/WER1.png

Some tips and comments would be very much appreciateed.

Maybe it's a stupid question, but how come? I don't understand the theory behind this. Is it because image editing method's distortion is too obvious with such small resolutions?

Hmm, scratch that. I forgot I had doubled the resolution already in the emulator settings.

So what's the procedure when creating these larger CRT simulations? Just to follow Ronan's instructions?
Should I start with the game's native resolution, no double, filter or scanline options?

Shit this is confusing. One thing I don't understand is why emulators aren't set to emulating the 4:3 aspect ratio by default? I mean the difference is huge!

It is.

Ah, that's why I got weird results with that intro sequence in that SFC System Sacom car game? The emulator removes the borders for us?

The emulators have default scanlines you mean, there must be some way to disable them though? I haven't the deciphered the options menu in M88 yet.

One more thing: when resizing these images to fit my website's design, is there something to keep in mind? All images will ofcourse link to the full-size version, but I'm sure there's a golden rule to make them look acceptable when scaled down?

Sorry for all the confusion...

Oh, right, I used linear. Bilinear basically means linear in two dimensions.

Strange. The PNG for the scanline effect has an alpha layer (i.e. transparency), but it seems that you don't use it... Can you check this? If you don't manage to do it, I will create two PNGs without alpha layer. That should work fine.

Could you send the original screenshot?

Well, it is just that your screenshot looked strange, so I wondered what could have gone wrong. Since the PNG I linked has some transparency, I thought that maybe your software didn't take this into account. I checked, and that seems to be the case, for whatever reason...

Anyway, I have just made two new PNGs, without any transparency, that should do the job. Use the same procedure as exposed above, but copy the first PNG below in the layer in Overlay mode, and the second in the layer in Multiply mode:
http://img687.imageshack.us/img687/6259 … aoverl.png
http://img208.imageshack.us/img208/3949 … amulti.png

You should get something like this:
http://img217.imageshack.us/img217/9649/test2a.png

I made an attempt with Metal Slug, but it doesn't look quite the same.

http://fc27ya.blu.livefilestore.com/y1p … RQYuoYEB5i

I'm guessing that either the scaling is different (chose Bilinear for this), or the overlay doesn't seem to get applied in the same way.

Hello low-res lovers !

First, I'd like to congratulate Recap about this excellent website.
Few years ago, the very first thing that took my attention was the care for the screenshots, a huge difference between every other sites with the usual jaggy or blurry pictures. Then I read the reviews and others stuff, and definitly I knew I was in the good place.

I've been working on the subject for long time, looking for the perfect scanline ( right after "the perfect beat" :-P ). I've been working on many different CRT, taking tones of pictures, understanding the complex way to properly set-up those screens. The goal was to be sure of what is possible and/or desirable to recreate with a raw output from an emulator. I'm still working on a huge tutorial about the set-up of CRT, especially the difficult part of convergence.

Many people in the emulation scene wants to recreate the feel of the CRT, but they have in mind that the picture tube gives "genuine" dirty visual, with lots of defects. They focus on the bad points (not on the right behavior of the beam under the best conditions). I can't stand this, because I'm an ardent supporter CRT precisely because this technology gives you the best picture at any resolution, if you feed it with a good signal. For example, take a look at those beautiful pictures of KOF XII sprites at their native resolution (640x360) on a Trinitron PC monitor:

http://raster.effect.free.fr/31khz/KOF_ … 66x768.jpg

http://raster.effect.free.fr/31khz/KOF_ … 360p_s.jpg

http://raster.effect.free.fr/31khz/KOF_ … 60p_2s.jpg

(others: http://raster.effect.free.fr/31khz/KOF_sprites_360p/ )

If only you could see it in front of the monitor, it looks georgeous. Very "15kHz-like", very "video game graphics". The deep black with the perfect colors of the tube give you an unmatched result. I can't have enough of the scanline effect. ^_^
And on this subject, I must say that Ronan has achieved the best result to recreate those beloved scanlines. His method is even better than mine, because it shows the behaviour of the beam at any greyscale. My method is more precise and "tweakable", but if I want to handle all greyscales, it becomes very laborious. His method is really simple compare to mine. So, nice job, Ronan. "You're doing great !" :-)

By now, I'm working with the wonderfull "soft15kHz" software that allows you to drive your graphic card at 15 kHz, with every possible resolution in this frequency, progressive and interlaced. Fuckin'great, I'm like a little boy in a big toy store. :-P
Very helpfull for working on non square pixel (like 448x224 or 512x240).

Before I log out, just for the pleasure, a full shot of a 15 kHz picture (Trinitron, low curvature):

http://raster.effect.free.fr/tv/photos_ … -Leo_s.jpg

See ya'

Most multi-sync/31 kHz arcade monitor have the same large pitch as usual TVs driven at 15kHz. That means you can't obtain such precise scanline (as on the picture shown before) at 31 kHz on these.
The electron gun is conceived for bright and large beam spot, and you have limited number of phosphore triads. You need a 21 or 22" PC monitor with thin electron gun and precise grid (but small screen size...) or an insanely expensive 32" CRT Broadcast monitor that can really display 720p.  You can also have good results with CRT projectors (but not as good as direct view CRT).

I don't own an Xbox 360, but I'm pretty sure that the downscaling to 480p is filtred. Anyway, even if it's not as precise as the picture posted above, it will alway be more pleasant to the eye than rough cubic pixels at different resolution of the background, and of course blurry filtred sprites.

This picture [>] is based on a nice .png screenshot of KOF XII (720p), reduced to 360p with nearest neighbor option. Perfect pixel for the sprites and the background, displayed in heavenly conditions on a performant CRT. Each time I look at this, and then go to see the game on 768p or 1080p flat screen (because there are very few true 720p flat pannels out there...), I'm so angry. I just want to take a hammer and smash the screen. :P

Considering that KOF XIII only runs with filtred sprites upscaled at 150%, there's definitly no chance to try any tricks to display a nice picture on the best CRT displays.

Just for the tears:

http://raster.effect.free.fr/tv/photos_ … s_comp.jpg

> Part of a 720p KOF XIII screenshot on the same Trinitron monitor, driven at 31 kHz. On the left, a native res' sprite of Iori, and on the right, the sprite with 150% filtred upscaling.
:-[

Even if you take the signal and converted it from 720p to 852x480 (a 66% downscaling in order to recover the original size of the sprites and have the same definition between background and sprites), it's impossible to obtain the same precision and gorgeous visuals as the KOF XII 360p picture.

Men, I hate fixed res' pannels, and I hate the "everything scaled and filtred" shit.

What's the point to have easily adressable pixels (the only advantage compare to CRT) if most time you get blurry picture and artifacts ? Flat pannels are just good for internet browsing and word processing. For everything else, you need true multi-sync monitor. And CRT is the only technology that can gives you this.

Yeah, I suspected you would already know the software ^^, or have tried an arcade VGA or something else.
I knew there were solutions to drive a graphic card at 15 kHz, but for many years I didn't want to hear anything from PC stuff and emulators. I was fed up with, and went to original hardware: nice CRTs, real consoles and arcade PCBs.

The reason I finaly try soft15khz few day ago was the creation of the best patterns for convergence and geometry adjustement.

You can check this:

(others: http://raster.effect.free.fr/15khz/CRT_set-up/ )

Very useful, and much more efficient than any stuff you can find in PCB service menu, or any regular convergence pattern available.

Without modesty, I can say that I'm the expert of CRT photographing. :P
With my advices, you will achieve such quality  as this :

http://raster.effect.free.fr/tv/photos_ … G-SvsC.jpg
(it's reduced at 50%, the original picture is much precise, but very big, about 5 MB)

Near perfection CRT set-up and proper way to capture it with your camera.

This is the best you can obtain from a curvated shadow mask picture tube at 15 kHz. The beam spot is precise at every greyscale, even on bright red.
And about the matter of the red beam:

The red beam is alway driven stronger than the green and blue (because red phosphore is less efficient and needs more current), but with a proper set-up of every parameters, you can obtain precision at all levels.
So, the usual blur and merging that occur with bright red portions are not a characteristic of the CRT picture, it's just a question of the quality of the adjustment. Hence, we see that carefull observation is not enough, you need to know the way to set-up a CRT to determine what to do for recreating a CRT picture in digital environement.

Most TVs and arcade monitors are not set-up to their maximum possibility (especially for convergence), because you need time to do it, and because 15 kHz screen are primarily conceived for bright picture (for use in bright environment) and interlaced signal, with a necessary melting of lines. Precision is not the main goal for 15 kHz interlaced picture tube. But we, pixels lovers and users of non-interlaced signals, we want it, because it's possible !!! :P

Lots of people think that each field of the interlaced signal ("half picture", at half the resolution) is supposed to be displayed with a large space between the lines in order to fit the lines of the next field. That leads to think that "scanlines" are plain black lines with the same size of the line displayed. Hey no ! First, scanlines are the lines that are displayed (that are... scanned), and the black space between them is not regular, not as large. But for convenient sake, people continue to call "scanlines" the black space they see in 15 Khz picture, and they continue to view it as regular lines. It's difficult for them to understand why the beam vary in size. But this is nothing less than the fundamental part of the CRT technology, and the most important thing to achieve a good CRT simulation. But when you see some atrocity like this:

http://www.bogost.com/games/a_televisio … ator.shtml

Then, you know that people are very far from understanding the point...

I've been working on a way to easily recreate the variation of the beam:

(example of a bright pixel)

(example of a dark pixel)

Even if in reality, the spot would have different sizes, for digital environment, I choose to use the same size and create the illusion with shades of colors.
A pixel that would have been a 4x4 square in a classic 400% nearest neighbor scaling becomes a 6x4 sample (the core is 4x4, and you have two extra columns depending if the beam size increases of decreases).

The goal is to "interpret" the picture in only one pass (to have minimum use of resources and minimum lag), directly taking each original pixel and "construct" the final result according to a table that store small sets of values ( 1A to 4F in the examples above) for every variations of greyscale. There would be only one comparison with each previous pixel per line, to determine if the beam increases or decreases. Well, it's a simulated scanning (from top to bottom and left to right) to obtain the data, but the result is made of a full frame (no layers, no transparencies), constructed for fixed resolution.
It may not be very clear, because it's not so easy to explain it in french even for me... ^^
But I think this is the best way. The difficult part is to create the table, you need a perfect screenshot to start from. For this, I will use my method, because it's very tweakable and precise. I need to adust every layer according to the observation of the behavior of the real beam in the best condition : the proper set-up of the screen. For this last point, I'm ready now, I know. :-P

I need to explain the trick and convince somebody to write a bunch of code lines, somebody who wants to recreate a perfect CRT picture, not somebody who think that CRT screens produce genuine dirty pictures with lot of defects that you need to focus on. In fact, that's the tricky part ! ^_^

So, wait and see...

Moving this here from the development subforum since I guess it will be of interest for non-contributors too:

Old-computer screenshots is indeed quite a tricky subject. Most of them from 1983 onwards--PC88 included--supported both, 15 kHz (low res) and 31 kHz (hi res) modes. For PC88 and PC98, almost always you'd find either, 640 x 200 or 640 x 400 (they were actually 640 x 240 and 640 x 480 once the monitors' benchmarks were set as I explained earlier, but let's just leave that apart for now). PC88's hi-res mode was pretty limited color-wise, so we must assume most games used 640 x 200. So an actual frame-buffer screen from your game should be like this:

By that (which I guess it's the first PC88 screen ever at its original resolution [laughs]), you can understand why the emulator--conceived, sadly, only for stardard PC video cards--automatically line-doubles the image (1st screen in your post) -- the aspect ratio through a standard PC is way too distorted. The game, you know, despite the resolution, had a 4 : 3 aspect ratio (not really 4 : 3 if we count the excluded borders, but close enough). It could be possible--I never had a PC88--that the game was line-doubled on the original frame-buffer for a 'fake' hi-res display so that the emulator is actually mimicking that behaviour, but given that there also existed 15 kHz-only PC88 monitors, I don't find that too plausible. That's also the reason for a universal scanlines simulation feature on the emulator (your 2nd screen) -- it's the most natural way to replicate the original low-res display on hi-res video hardware (therefore rendering the line-double method quite useless), though we've already explained that the final result here--no matter the type of monitor you're using--lacks the beauty of actual 'scanlines' for several reasons (hence the point I've been claiming for years of never using a standard PC for proper emulaton, if you're really into that).

Now PC98 is a bit different. It was especially designed for hi-res (most of its monitors indeed couldn't do 15 kHz) and we must assume that most games were rendered at 640 x 400. This was a problem for simultaneous PC88/PC98 releases or PC88 - PC98 ports -- you now needed to double the original lines for a hi-res display, and that's what many devs did:

PC98 Xak

PC98 Fray

(They didn't need any excuse though -- many original PC98 games were also designed at half the display resolution, saving RAM demands and well, effort.)

Fortunately others decided against this awful method and invented the scanlines simulation -- instead of drawing the same line twice, even lines would be black lines. That PC98 Tritorn II screenshot you mention, to answer your question, doesn't have scanlines added by the emulator; the game was rendered exactly like that. You can check these other examples for more color:

PC98 Lessern Mern

PC98 Vain Dream II

Notice the scanlines are applied only on the action area or the background -- there's no way an emulator could do that. Of course they went even further and invented different patterns which, applied on the line-doubled picture, would alleviate the issues of design - display discrepancies:

PC98 Digan no Maseki

So that's why you won't find (my guess) a scanlines simulation feature on any PC98 emulator.

This pretty much covers, not just your questions, but the whole core of what was going to be part 2 of the Scanlines -- Ab Initio article, so it's now done for anyone's interested. Thank Macaw. The only remaining thing would be illustrating with actual photographs the wonders of 15-kHz RGB CRTs to let everybody see what they've missed, and that will come with the new Site's content itself. So yeah.

Edit: Grammar bits.

It's quite unique, at least. Needs some corrections once I get the time to check the Japanese interweb. For instance, I've learned that the 640 x 400 mode of these is not 31 kHz, but 24 kHz. That technically is not 'high resolution', but 'extended resolution', as some put it. But pretty much it's all like I pictured.

I disagree. The more the resolution, the more the details needed. Simple tiles would get complex unless you go with 'empty-style' backgrounds and 'flat' colors. Scanline effects just helped to avoid this style, which generally has little to do with the 8-bit style they were trying to evolve. And let's not forget about RAM and even CPU limitations -- in this sense, it indeed did provide 'extra performance' on games.

I hardly believe that they used square pixels-based video hardware in the early years to design game graphics. They most likely used hardware capable of the same video modes as the target platforms -- remember that work stations and PCs weren't always like they are today and that there was a era where the pixels' aspect ratio was indeed a quite changeable thing thanks to CRT technology.