Simulating a CRT TV with C++ and OpenGL | Hacker News Reader

Simulating a CRT TV with C++ and OpenGL

nicktasios.nl

68 pointsingve51 comments

Has author actually seen a CRT TV image in real life?

Because it didn't look like that, you didn't have those clear distinct lines.

Color CRT displaying a game, no lines: https://i.imgur.com/eI2yb7Y.jpeg

Black & white CRT, you can barely make out the lines: https://d23gn3985hkc32.cloudfront.net/wp-content/uploads/202...

Some CRTs do have visible scanlines, some do not. Check out this thread https://shmups.system11.org/viewtopic.php?f=6&t=63197 . Personally I'd like the simulated CRT to have no scanlines, rather what I want is temporal ghosting (previous frame blending in the next one, used for transparency effects in games by blinking two images rapidly); and a filter to make pixels less blocky and implement colour bleed. Given the great number of CRTs out there you probably can't have a single one be the best ever, but perhaps emulating several known ones by twiddling the parameters of the shader and letting people pick would be the best.

When I look at CRT emulation, I always feel like I'm going crazy. It's been years since I last saw a CRT image but I don't remember visible scanlines at all.

I know the retro space is often specifically emulating old hardware, was thing a thing in the 80s, perhaps? Or were visible scanlines more common in America than in Europe? I know the weird colours and lowered resolution of some (emulated) games are an artifact of differences in NTSC/PAL/etc., could that explain the difference?

They seem obsessed with simulated really bad CRTs. I think you could see such scanlines if you were hooking up your Atari 2600 to a TV that was already a decade old, then sat right on top of it instead of across the living room because the cords were short and you wanted to see your games. I've seen it on some arcade games, for similar reasons. But in general, CRTs didn't have those plainly obvious lines, no, because... the engineers would have engineered them out! And did. They're plainly unsightly and it isn't that hard to remove them, after all.

I imagine part of it is the need to have visible differences strongly emphasized so it looks like the filters are doing something. A true CRT filter would be much more subtle and people might not thing anything has happened when you turn it on.

>They seem obsessed with simulated really bad CRTs. ... I imagine part of it is the need to have visible differences strongly emphasized so it looks like the filters are doing something.

The main reason for that is that games of the era were targeted at the most common hardware, using very cheap composite circuits and CRTs as medium for artistic expression. In many games, you could easily miss gradients, rainbow effects, transparency, or something else with a high-end CRT.

That seems pretty far-fetched. Do you have examples of such effects being shown wrong on CRTs without scanlines?

I don't think they're referring to scanlines specifically.

Something like the waterfalls in Sonic the Hedgehog come to mind.

https://gamedev.stackexchange.com/questions/167804/did-retro...

There's an image on this page. The waterfalls don't look very good with sharp pixels. A CRT would blur and blend, and you'd have to assume the game designers knew it would happen. It's probably not just a happy accident.

doubled112 comment is excellent.

Another example that is harder to show, is that many games used intentional flickering or other tricks to create some funky effects only possible with CRT.

For example: if you transmit the data correctly (one frame is only odd lines, next frame only even lines) and flicker sprites at certain speeds, you can achieve some fancy transparency and overlay effects.

Another trick is toy with timings to screw with resolution, color range and pixel shape.

A less reliable trick is sub-pixel anti-alias relying on the TV geometry, since each manufacturer had their own patented geometry this was more problematic (for example Sony's Trinitron was aperture grille, each pixel was composed of columns of each color. Many other manufacturers went for triangular pixels instead. With the 3 primary colors arranged as round dots in a triangular pattern).

I played spectrum games on the spectrum connected to PC . Spectrum had a graphics resolution of 256x192 px and no scan lines were visible on a 24" TV.

> They seem obsessed with simulated really bad CRTs.

I think it's quite the contrary. They seem obsessed with simulating high end CRT studio monitors like the Sony BVM line which were very sharp and had a highly pronounced negative scanline space.

The image on the left is from a BVM: https://i.imgur.com/cu4PdgF.jpg

> They seem obsessed with simulated really bad CRTs.

Good PC flat screen CRTs towards the end of the CRT era looked as good as you could ask for. By modern standards, all that was missing was 120 Hz and hidpi.

There were some who could do 120 Hz although for small resolutions. But hidpi was common. It took LCD monitors 10 years to catch up with hidpi and the colours are still problematic on the majority of LCDs.

120Hz was not that uncommon in the last years of CRTs, they could actually go higher: https://www.youtube.com/watch?v=ySQzXpYETXc

I guess it all depends on what you define as "small resolutions". My stupid ADI monitor from 97 could do 1024x768@85Hz

> 120 Hz

It always irks me that everyone seem to have completely forgotten that CRT monitors were routinely used in 75 Hz and 85 Hz (and even higher rates) and with various resolutions instead of the one and only 60 Hz in native resolution as with most LCD monitors. 60 Hz meant “I can't configure Windows” or “my video card or monitor can't handle more than that with resolution that high”.

Because it sacrificed pixel accuracy - the horizontal lines were a continuous signal and the pixels kind of blurred into each other. Looking at a CRT monitor side by side with a LCD one was quite revealing - even a last generation 2005 CRT monitor with digital controls and OSD was a blurry mess.

If you want both fast refresh rates and pixel accuracy, well, you kind of need today's technology. Not possible back then.

Really really old monitors had visible scan lines, like before 1985 or so: https://i.redd.it/o6gezzyw8dv61.jpg

It could be the opposite. Big high resolution multisync CRT monitors had to have fine dot pitch and finely focused beam. However, in lower resolutions the distance between scanlines was more noticeable, as there was limit to how wide and powerful electron beam could be. As for the lowest res CGA/EGA modes, VGA and later adapters simply had to use double scan to output anything decent on the CRT, so it was part of the standard.

So people might be trying to mimic running 1987 games on 1997 hardware and not the look of original CGA/EGA/VGA monitors. Recreating some style is an artistic choice altogether anyway, and I think it's better to discuss it this way.

I think I understood your requirements, so I tried to knock up a 2d canvas filter demo[1] which met them. Sadly I don't think I met them, but I did end up creating a crazy weird motion blur effect (that's new to me) so thank you for helping me to accidentally create it!

[1] - https://codepen.io/kaliedarik/pen/OJZzJJB ... warning: demo requests access to the device camera.

Yeah -- it seems that so many of the CRT simulations way overemphasize the scan lines. I think if they added a bit of "bloom" to each pixel, it would make them less obvious?

They over-emphasize the negative space between the scan lines without getting the details of the raster scan correct. The forward scan lines are not horizontal, they are slightly tilted, because the vertical sweep is continuous. The retrace is also slightly slanted, but less because the retrace is faster. So in the diagram on this site the solid lines should be the ones that slope, and the dashed retrace lines should be the ones that are nearly horizontal.

> The forward scan lines are not horizontal, they are slightly tilted, because the vertical sweep is continuous

You can slightly tilt the yoke to make the horizontal lines more horizontal (although then vertical lines will be less vertical, tradeoffs everywhere)

Another way to think about the horizontal scanlines is that within the lines, there is really a continuous analog signal, undergoing various (analog) filters along the way from initial detection up through final display, but between scanlines there is discrete sampling. This can be emulated by digital filters, but even if the final display doesn't show the scanlines, there should be a different point-spread-function in the horizontal versus the vertical direction.

There's also the possibility of emulating how the time resolution could be less than the refresh rate (sometimes sluggish phosphors create a kind of motion blur).

Really old tv transmissions also had an effects like a high-pass filtering (I think to show better detail within limited dynamic range)[1], and how really bright spots would be surrounded by dark halos [2]. The distinctive "look" of old TV has a lot to do not just with the CRT display, but the characteristics of the analog method of video signal generation, in a video camera tube [3] that had its own cathode raw sweeping across the source image.

[1] look at the edges of the desk at around 0:36 in https://www.newsweek.com/wmttv-black-white-color-first-time-...

[2] https://movies.stackexchange.com/questions/113358/whats-with...

[3] https://en.wikipedia.org/wiki/Video_camera_tube

I recall as a kid, getting right up close to the screen and looking at the picture on the TV and noticing the lines. Further away from the TV, then yes, the lines were not visible.

Guess it depends. I haven't had a CRT TV for many years, but I have an old multisync CRT monitor and I can clearly see the individual raster lines (close up of 320x200 progressive scan: https://i.imgur.com/Osc4ZMm.jpeg). At normal viewing distances it's not noticeable though.

Deluxe Paint, ah! fond memories!

> At normal viewing distances

Yes thats what it was all about. The pixelated game graphics were made with that and other CRT properties in mind.

Yeah, and I seem to recall reading an article about some of the good old tricks (think it was about the Aladdin game?), but unfortunately I can't seem to find it now. Doing the dithering (by hand) just right would give a good effect in practice. Things made for real higher resolution monitors looked bad on lowres CRTs though (with very distinct scanline effects) as I recall like the menubar in my screen cap, even at a distance.

> multisync CRT monitor

> 320x200

21" by chance ? Why would someone use 320x200 on a monitor when only the very old DOS games need such a mode ?

14" 1994 vintage Microvitec. Very popular monitor for Amiga systems because it supports 15KHz horizontal refresh rate modes (PAL/NTSC TV standard - for games/demos) up to 31KHz+ (VGA) for nice flicker free 640x480 60Hz display ("productivity") modes for serious stuff.

It would make sense to adjust the "line width" (focus!) of the scanlines so they blur into each other. But on a multisync monitor, this would be matched to the highest resolution it can display. Without fancy defocusing circuitry, of course if you display only 200 scanlines on a monitor capable of 5x as many, you're going to get a lot of black space between them. This bothered me personally, in those days, in my case with a 20" CRT monitor and emulation of my old C64 environment. Modern "CRT emulation" is better than the real thing, in this case.

I had a consumer CRT as a child (the family TV in the living room) whose scanlines were apparent even at 6ft away, and it was only a 19" screen.

maybe my eyes were better than most, I don't know, but I distinctly remember trying to adjust the settings so the lines would go away, thus reducing the annoyance of my "all I can see is colored lines" brother.

It seems that he didn't. And he also does not understand how a cathodic tube works (i.e. persistence).

> Color CRT displaying a game, no lines: https://i.imgur.com/eI2yb7Y.jpeg

Try loading the 4000x3000 version of that image you linked and zooming to full size. Especially on the green character's outfit, you can see the scanlines on that color CRT display.

If you put your eyeball to the screen, sure, just like you can see the pixels of an LCD screen. But it's not how you're supposed to look at it.

Along with different properties of the beam other comments mention, there was also a question of dot pitch, i. e. properties of the phosphor grid. Also, the circuits generating the signals in consumer hardware often did not have that much of the full bandwidth, or the most correct wave forms.

That is a late 90s television with a small tube. You can most definitely see the scanlines on old arcade CRTs. The shadowmask on a color CRT enhances the lines so comparison to B&W isn't meaningful.

Imho electron gun alignment topic - see https://www.javatpoint.com/computer-graphics-color-crt-monit...

If you enjoy this look, I recommend checking this out:

https://github.com/Swordfish90/cool-retro-term

This may not be perfectly realistic, but it's getting close and can be made closer. Not long ago, replacing a dead 26" colour CRT (say, in an arcade game machine) with a suitably masked 4K LCD panel would have been ludicrous. But we're pretty much there now.

I thought when incandescent bulbs pretty much went away that I would miss the bare filament in a clear bulb look. I kind of do, but not much. I have a chandelier with the fake-filament LED type bulbs in it. Despite dubious cooling of the LEDs they seem to be holding up just fine in extended use, and while it doesn't look quite traditional, it's good enough.

And emulation of older video game systems has moved into the browser!

Wow, I consider myself a pretty obsessive retro-gaming enthusiast (a custom-built arcade with 25-inch multi-scanning CRT, hacked graphics card and hand-tweaked GroovyMAME build) but damn, physically simulating a cathode ray tube is next level. Respect.

but damn, physically simulating a cathode ray tube is next level. Respect.

Definitely. While I have simulated monitor screens that were originally physical CRTs, I still simulated them as though they were digital monitors with defined pixels.

Definitely respect.

I used a CRT for the first time in a long time a few months ago (embedded in a PowerMac 7200 all-in-one), and I forgot how stunning the picture looked compared to LCD/LED screens throughout the 2000s.

Big problem though, I forgot how much they made my eyes _hurt_. I couldn't handle using it for more than 30 minutes. Increasing the refresh rate to 75hz helped a little bit, but wasn't enough to get me to stop causing a headache. I'm all for nostalgia but I think in the end it is good that we moved away from CRT technology (for many reasons).

NFSU2 on CRT looked glorious and vibrant. On LCD's it's pretty much bland and faded off.

I did a project a few weeks ago where I was trying to emulate CRT images as well. End up creating a CRT converter for still images.

https://victorribeiro.com/CRT/

I guess each one of us remember CRT a little bit differently, since my images end up with vertical stripes instead of horizontal ones.

Quite interesting. The hardest part would probably not be the simulation but actually displaying what is simulated (in real time). You would need pretty high peak brightness and an insane refresh rate.

I think modern (HDR with loads of nits of brightness, 144+ fps) gaming monitors should be able to get the effect down quite well. Nothing you'll find in a productivity oriented laptop but with the right tools you should be able to come quite close.

The peak brightness is there but not in per pixel dimmable displays (OLEDs) and you would need well above 15000 Hz to simulate a CRT.

Does anyone have experience doing this purely with FFMPEG?

yes the int10h font guy has some blog posts about it, with accompanying code on GitHub.

https://int10h.org/blog/

These seem to regularly pop up on HN but the new ones never seem to describe the ways in which they differ from the now-20-years-old xanalogtv hack.

I was hoping it would distort the image to round the corners and compress the edges.

Extra credit for added snow.

Reply on news.ycombinator.com