DOSbox-CRT | Hacker News Reader

DOSbox-CRT

mattiasgustavsson.itch.io

154 pointsakavel81 comments

You can also use DOSBox as a LibRetro core [https://docs.libretro.com/library/dosbox/] - this enables to use any of the filters the LibRetro provides on the frontend, including CRT.

This pattern should be the future of emulation development, where possible.

RetroArch is a pretty terrible UI, though - it takes everything bad about that Sony UI and it's not particularly performant, either.

Lots of stability and performance issues. Had difficulty getting a 3ghz Core 2 Duo to emulate SNES effectively with it... might've chosen the wrong engine out of the thousand alternatives, though.

> Had difficulty getting a 3ghz Core 2 Duo to emulate SNES effectively with it

That's very odd considering retropie runs snes perfectly.

Yes and no. RetroPie can run SNES games using comparatively inaccurate emulators (SNES 9x, ZSNES; mostly sound bugs), but can’t run more accurate emulators like Higan acceptably.

True but that sort of is the issue with things like Higan - proper CPU architecture emulation is REALLY slow in software.

RetroPie (or libretro) does not have a ZSNES core; ZSNES, apart from being wildly inaccurate, is largely written in x86 assembly and so there is no ARM version.

Additionally, the more modern snes9x cores have essentially the same sound emulation as higan (blargg's SNES_SPC library). I believe higan has done some modification of that code since importing it but regardless, the sound emulation in snes9x is very, very accurate.

Higan is indeed unparalleled for those who truly care about extreme accuracy, but snes9x is 99% there for the vast majority of users who just want to play games.

The op said effectively, not perfectly.

That's really bad... I beat Chrono Trigger on a 486

wtf ? I ran SNES and Sega Genesis emualtors on a 486DX4 a long time ago, and was running perfect!

No, certainly not perfect.

A great article from the author of higan (whose core OP probably used): https://arstechnica.com/gaming/2011/08/accuracy-takes-power-...

Genesis was pretty good on 486 with Genecyst but certainly not SNES.

Yes, you probably used a bsnes or higan core. bsnes is all about accuracy and that IS performance-hungry. Use a snes9x or something else if you want. See https://docs.libretro.com/meta/see_also/#snes

Frankly i don't get why it even exists.

Best i can tell it simply slaps a UI on top of already existing, free standing, emulators.

It's odd to me that someone would want this filter that much on DOSBox. I understand why it's generally included in game emulators - they were generally played on displays that looked somewhat like this. But DOSBox is for DOS games, and to my knowledge, most weren't played on displays that looked like this. At least, I don't think so, unless I'm completely misremembering my childhood, since almost all of the images shown are for games I actually played on a DOS computer using a SVGA display when young.

The 1980s with its CGA, EGA and early VGA displays must be from before your time then? I definitely played games looking like that. I prefer not to use such filters, but some people might be nostalgic for it. One interesting application might also be for games that make use of the overscan border, which is not normally visible in DOSBox (and people have been trying to work around this), causing problems in games that need it such as Crystal Caves.

I'm referring less to the curve, and more to the scanlines[3]. The bleed in CRTs means that's not what it actually looked like when displayed on era accurate hardware, so it's always confused me why people would opt for a filter that claims to be "authentic" that does this, especially when emulating the visual experience enough to show screen rounding.

For example, here's a picture of an actual DOS ANSI output, from a CRT monitor[1], and here's the equivalent display from this filter[2]. In what way is emulating the scanlines making this more authentic?

Edit: clarified a bit

1: https://steemit-production-imageproxy-thumbnail.s3.amazonaws...

2: https://img.itch.zone/aW1hZ2UvMjU1NzcwLzEyMjU1NTQucG5n/origi...

3: The correct terminology is escaping me. The interlacing? Whatever is responsible for the skipping of lines in the output.

It's tricky to make a comparison of 'authenticity': for one thing, some displays had more visible scanlines and for another, it's the sort of detail that's easily lost in photos because you're taking a picture of a small, very-high contrast area and the artifacts often bleed together.

Well, it can be tricky once you start getting accurate, but I'm not sure any display ever looked like these screenshots, or even very close to it. There was always some visual bleed to eyes as well, not just pictures.

I'm pretty sure even LCD panels just repeated lines and didn't leave them blank, so some portion of people actually used an LCD to play these originally might have close to perfect accuracy if they repeated lines.

The only times I can particularly remember scanlines being really bad and noticeable were on the old CGA displays in the 80's. And those didn't look like this representation by a long shot, there was a lot of bleed in those as well, the lines were just big enough the bleed didn't obfustate it quite as much. I can't comment on EGA/VGA, we went straight from a CGA monitor to a SVGA one in the late 80's or early 90's, I don't recall exactly when. It's possible there was a display type in there that I didn't use much which actually makes this representation more accurate.

I think the effect tends to be exaggerated in most of these implementations my points are just 1. visible scanlines were totally a thing. 2. it's really hard to take a representative photo of a CRT.

Also, human memory is a funny thing, and these sorts of things are easy to forget from your remembered experience, too. So to some it is exaggerated because they only remember it at its best and others it is spot-on because they only remember it at its worst. Because we notice problems more than things working, there may be more people in that second camp than in the first.

I've had CGA, EGA VGA (MCGA :) ) at that time. I clearly don't remember so much scanlines. There may have been some present, but they were not that visible (IIRC). So I tend to confirm what you say.

Also bear in mind that for CGA the difference between NTSC composite output and RGB output matters quite a lot.

I remember some monochrome monitors having visible scanlines like this. I think that the shadow mask on color monitors caused this to be blurred out.

It's not like everybody bought the same monitor back then. The hardware was all over the place quality wise. A lot of the hardware back then was really bad unless you were willing to pay literally thousands of dollars for a system too. You could definitely get smearing from too-long VGA cables or shitty controller boards or just hardware getting old and falling out of spec.

So the guy that had the honest to god IBM PS/2 that cost $10k thinks all of these artifacts are anachronisms, while the guy with the bargain basement hand-me-down Packard Bell thinks it looks perfect.

And someone who kept a radio next to their CRT with a big honking speaker might remember the scan line deviation being much bigger than someone who didn't.

https://www.youtube.com/watch?v=3AfFaGwFNIg

Seems to be one guy that takes CRT emulation seriously...

Wow, that's absolutely fantastic.

My first PC (an IBM XT clone circa 1985) came with a monitor which looked very much like the second image. Later PCs I owned with higher-resolution monitors looked like the first screenshot.

Most SVGA CRT monitors I've used in 90's and early 00's actually displayed visibly discete scanlines in any video mode with less than 480 vertical lines, this includes classic VGA 80x25 text mode (actually 720x400, and in this mode the resulting effect is in fact somewhat pleasing to look at) and mode 13h (320x200 in 256 colors with double scan) which was popular for 90's games, the effect was even more pronounced with EGA's 640x350 (which is not double scan mode).

Edit: also this effect was easily visible on Windows 9x startup/shutdown screen, which IIRC uses 320x400 unchained mode.

Edit2: somewhat relevant to this is that many late 00's LCD monitors with only VGA input are not able to reliably keep synchronisation lock in double scan modes and there is significant subset of current monitors that will simply reject any 400-line mode (and there is significant subset of modern graphics cards that output classical VGA text mode as 640x480 or 720x480)

Visible scanlines, or to be more precise, visible gaps between scanlines is usually a sign of a misadjusted monitor or one not running at the correct resolution. I don't exactly remember the name of the control (focus? bleed?) but there is one which adjusts the diameter of the beam, and is supposed to be set such that each scanline just touches its neighbours; you don't want them to bleed into each other, nor be so far apart as to approximate the IBM logo.

I'm extremely confident that picture 1 is of an LCD. The edges align with the bezel too perfectly, and the screen area is flat, not convex.

I played Crystal Caves when I was (very) young and wasn't actively aware of the overscan border being relevant—what was it used for?

Overscan border was set to red in each level, and would switch to green when you collected all the crystals.

It would be great to get a filter that replicated the days when I'd play Sim City on the IBM PC Convertible's 640x200 monochrome un-lit LCD screen.

Oh, wait. No it wouldn't. That was terrible!

Yet, somehow I'm nostalgic for it.

Back on my first computer a 486, I had a flat Trinitron monitor. So at least in my case, the curvature would actually make the experience less authentic if I felt like reliving the Doom and Duke3d gaming of my youth.

My CRTs up into the late 90s had some amount of curve/distortion, it definitely got better in the late 90s, but by then I was on Windows 95+. When doing mostly games in DOS, I had a pretty curvy 14 inch CRT.

> But DOSBox is for DOS games, and to my knowledge, most weren't played on displays that looked like this.

DOS games most certainly were played on CRT monitors, it wasn't until, IIRC, 2000 that LCDs really became mainstream for desktop use.

> since almost all of the images shown are for games I actually played on a DOS computer using a SVGA display when young.

Being an SVGA display isn't relevant -- there were plenty of CRTs running that resolution and beyond. Perhaps you're thinking of something else when you say "displays that looked like this"? Otherwise I'm not really sure how to interpret your comment, because if you were playing DOS games at the time of their release, you were almost assuredly playing on a CRT -- unless you're talking about a 90s era laptop; those were LCD.

I'm referring less to the rounding and more to the way it's representing scanlines/interleaving of output, which while "signal accurate" isn't really "experience accurate", and if you're going as far as to emulate the rounding of the monitor, you really don't care what the signal was, you care what the person saw, and because of CRT bleed, that's far worse than some other display options (even repeating lines looks more accurate in some circumstances, I think).

Oh. Well I agree, CRT emulation does a pretty poor job of actually being representative of the CRT experience -- the scanlines are just so ridiculously exaggerated.

They are appropriate for a late 80's EGA monitor. It looks a bit weird with VGA.

Not really, check this image of an original IBM CGA: https://i.imgur.com/M3gKlfA.jpg

The scanlines aren't visible. The most important "mistake" those filters do is that they apply a uniform shade for each scanline, but in all CRTs i have around here, the scanlines do kinda merge together when the colors are bright.

This looks like CGA card connected to composite monitor. Can you find any examples of those exaggerated scanlines on EGA monitors?

http://www.vcfed.org/forum/showthread.php?26680-EGA-monitor-... Third picture shows how it actually looked like on ~1985 hardware. Barely noticeable. By the time VGA hit the streets it was gone.

Having had one of those - appropriate sure, but accurate? No. They're super exaggerated.

The vast majority of CRT monitors have a curved glass screen. Flat monitors were still pretty common, but they came at a premium.

I played on CGA and EGA monitors in the late 80's and early 90's. This filter does feel a little exaggerated to me but maybe I simply forget how bad they were.

Those scanlines were bread and butter of low budget 15KHz systems displaying non interlaced signal. That means all home computers/consoles using TV, and CGA/EGA in low res mode displaying 200 lines on cheaper monitors. VGA cards scan double when emulating CGA/EGA modes and always display at least 350 lines, no visible scanlines.

You are absolutely right, example images are pure nonsense, because they represent late nineties VGA era of DOS gaming. UFO, Warcraft, Rayman, Dungeon Hack, not a single one of those supports CGA/EGA.

You're right, with a caveat. While it's true that 320x200 VGA DOS resolutions didn't have scanlines, arcade games and home computers did, and generally looked much better. A MS-DOS arcade port for example would look very blocky, while the Amiga counterpart would be fine in comparison. Scanlines are integral to that correct old-school look IMO, so it's very cool to have the option of replaying old classics like the Lucasgames stuff with that vibe.

Indeed, scanlines are definitely more of a "TV" thing - e.g. the old 8-bit / 16-bit home machines and games consoles. PC monitors were always a lot sharper. Then again, my earliest PC experiences were in the 486 era when SVGA was already a thing.

I don't really recall the memory of scanlines. It's not my memory of how CRT's looked. The reason I think LCD's are jaggier and feel overpixelated today with old games is because I think CRT's were part much smaller (a 24" CRT was an absolute monster in size and weight, many sat with 15" ones!) and part much less sharp. You didn't exactly have a physical pixel per... pixel.

But yeah, it's hard to simulate a smaller size... And if it's still larger, simulated softness will just look awkward, at least as far as I have seen with such filters. Like rounded but still big pixellated blocks. It can even look worse than without filters...

My solution? Sit a little farther away from the monitor than usual or play windowed. Unbeatable compared to filters.

The audio production world which goes to great lengths to reproduce audio equipment, I would love to see shaders faithfully replicate classic monitors.

Seconded.

I think it would be interesting to mix the two ideas and have a (Dolby Atmos/Windows Spatial) spatial sound map model for some of these games that includes the hum and fuzzing noises of the CRT scanner and the terrible locations of old Harmon-Kardon PC speakers right beside that CRT and a tiny bit of random electromagnetic interference. Really go for that 90s PC gaming experience.

And a special key binding to make the screen briefly distort with an audible "THUNK! bzzzz...."

"Hold on, the monitor is acting up again, let me open up its menu three levels deep to force it to degauss."

I used to like pointing the back of the monitor at another monitor in the lab, and hitting degauss - just to watch the other monitor get influenced by it

And a turbo button for my PC

We probably need 8k monitors for that to make sense.

A 4K one should be enough for example to emulate the tiny grill wires visible in the old Trinitron screens. Those lines were very hard to see and a lot thinner compared to a then single pixel thick line.

The Iiyama Vision Master Pro 454 should be one of those. And the Sony FW900 24" wide-screen trinitron. I had both those screens.

The Sony FW900 is without doubt the best CRT monitor ever built (IMHO). It was a ridiculous monitor, both in size, sharpness, colour accuracy and most of all weight.

I remember buying a used one for about £300 (I recall the retail price was close to £1000) and my housemate and I carrying the thing up to my room, almost breaking our backs. It was easily 40kg. But it was an outstanding monitor. The best CRT screen I've ever used by a long shot.

FW900 started at a cute $1999.99 in 2001

Dunno, I really hated Trinitron screens due to both horizontal strips as well as seeing individual pixel triplets and gaps between them (I guess my eyes are super sensitive to parallel/perpendicular lines; can't look happily at non-retina LCDs either). I found Samsung's Invar variants in their SyncMaster (757DFX and higher going up to 2048x1536) superior and much more natural (likely my eyes aren't that sensitive to diagonal non-perpendicular directions).

I didn't mind those lines across the screen. I could definitely see them but didn't find it bothered me. I couldn't go back to a non-retina LCD screen now though. And I suspect going back to my old FW900 might be something I wouldn't enjoy these days. But I remember at the time being blown away from it after having a variety of average to slightly above average CRT screens.

Actually, one of my earliest memories of using CRT monitors was my first PC (a Packard Bell 286) with the monitor having a dot pitch of about 0.3, or possibly even higher than that. The pixels were like bricks.

Eventually upgraded to a monitor with a dot pitch of 0.26 and I remember it felt like going from a standard LCD screen to a retina screen. It was quite the revolution for me at the time.

Me and some friends acquired some cheap Eizo 21" CRTs back in the day. It was an awesome display, but weighted >30kg.

The amount of heat it dissipated was huge, tho.

Not the same, but you might enjoy searching in Shadertoy [0] for tags such as TV or CRT.

[0] https://www.shadertoy.com/view/4dlyWX

I was just thinking, something like this could be pretty cool in VR; the curvature of the screen could be 3D instead of a 2D warp. Having modern higher dpi monitors inside VR is not great because of the resolution of the headsets, but this low resolution would work I think.