If I may chime in, slight deconvergence makes highlights pop more imho. (It does more things to the overall color structure of the image as well, in a good way imo.)
We had a big thing about in the CRT shader show off thread, someone could find my ramblings and screenshots. (I’m about to head out for the day or I’d find them myself.)
So dont get me wrong here CRT Royale is great but a lot of what CRT Royale does seems to be largely trying to over come a shortness in brightness and low resolution displays (<= 1080p). There are a lot of kernel based blurs, blooming and noise generation that goes on that IMO doesn’t really happen in an actual CRT, or at least is that visible, or at least doesn’t happen with my PVM with a component SCART socket (it is using NTSC internally) - thats a lot of caveats.
The vast majority of the CRT image look in my experience comes from the electron beam lighting up the phosphors and trying to adapt itself at a very high frequency that its not able to do that well and so the soft image. Deconvergence absolutely gives a blurry image - I bought a PVM just before christmas and during travel up here the H STAT had got completely out with a really blurry image - just a small tweak to it gave a much much sharper image.
Modelling the displays like this has worked reasonably well so far but your point about the source signal isn’t lost on me I do get that it will be a contributing factor. Just how much for the display systems I’m simulating is the question. On a poor consumer CRT I can guess it may well play a much larger role, on a high end professional display/1990’s arcade cabinet I guess much less so. But lets see where we get to with the low hanging fruit and then tack back to noise in input signals etc.
luckily, the NTSC part is largely solved and can be run before this shader, just as is done with crt-royale.
Well it’s not that I’m not evetually going to get around to looking at the source signal, of course I will, it’s just in my opinion it’s not the largest factor and therefore highest priority in recreating a lot of my displays. Admittedly a lot of what I see are PAL and so the colours match those much more accurately (NTSC isn’t a world away though).
What an NTSC system does to the source signal in terms of noise I’m not sure but I’d imagine it varied quite a lot from display to display?
yeah, the display’s comb filter plays a large role in how grungy it looks. The older and cheaper the display, the more chroma/luma crosstalk and dot-crawl make it to the screen.
Yes for sure - this virtua fighter display is the poorest I’ve seen so far but I think its issues are in convergence rather than a poor signal BUT I may well be proved wrong on that or rather it maybe the poorer signal plays a much more significant part in this display than I thought. As you say though its dead simple just to shove a PAL/NTSC shader in front and see what effect it has.
Yeah, if it’s a JAMMA/RGB system, that would be just general deconvergence, poor focus, old capacitors, etc. When those things are new, they’re super-sharp and crisp.
Getting into simulating monitor defects is a bit of a rabbit hole, but it’s fun and really helps with the verisimilitude, since very few of us had access to ultra-high-end monitors. A lot of people are really enthusiastic about defects because that’s what they associate with CRT technology, even though it’s really just old and busted CRTs that look like that lol.
Yes absolutely - one problem with the Virtua Fighter display is that one part of the screen suffers from one issue and another part suffers from another so not only do you have to mimic each defect you also have to mimic how those defects varied across the screen. As you say a real rabbit hole.
The one thing I will say is that mimicing these defects gives a greater understanding of a CRT just as trying to fix the real thing and their very many varied problems.
lol yeah, totally.
If you’re working on horizontal deconvergence and are planning to go down the rabbit-hole, you may want to go ahead and think about radial deconvergence, as well. 
It just seems like you’re going to be over-compensating on the display side and by doing so make it where other input types won’t look right. If you’re trying to do fixed presets to exactly match a specific complete hardware setup I guess that doesn’t matter, but in that case you’re going to need a much fancier camera… and then probably do a big tour of the US and Japan to get the pictures you’d need for that level of reproduction.
But also it would be nice to have something adjustable, and where people could try unrealistic setups, even stupid things like “what would PSX with RF look like on an arcade monitor?”
Well I’ve done vertical (VSTAT) convergence already. The horizontal one (HSTAT) I’m going to do is dead simple. All the other ones like corner bow, horizontal tilt, y bow, h. amp, y cross, bottom pin, top pin, pin phase, pin amp etc etc I’m going to save for a (very) rainy day. 
I’m reasonably happy with this level of accuracy:
I dont feel there’s a massive need to go hyper realistic/exact atm just a shift up in realism (and a lot cheaper/simpler shaders) over what we’ve had due to better displays.
The colours are slghtly off but I do plan to come back to this.
Yes totally agreed - all my shaders so far have had slight deconvergence in the vertical direction just I havent added it in the horizontal direction yet. Its absolutely intrinsic to the look of a CRT.
PAL arcades shouldn’t be a thing because these are RGB monitors running at the refresh rates and resolutions needed, including non-standard signals (e.g. 57 Hz). CPS 2 games are supposed to run at 59.63 Hz.
Yes so this is virtua fighter and a number of other 3D Sega games (Crazy Taxi, Star Wars, Virtua Striker, Virtua Tennis) I looked at weren’t quite matching up to 1/60 and seemed to capture better at 1/50. That’s absolutely not to say they aren’t all NTSC systems and that they aren’t running at some odd frequency near 60fps but it does add some amount of doubt in my mind.
EDIT: Possibly what you are saying there is that PAL/NTSC has nothing to do with these arcade systems?
It’s probably a stretch to say there is no connection, because most older 15 khz arcade monitors will loosely adhere to the NTSC standard (e.g. most games running at 224 active lines etc), but afaik arcade monitors are similar to PC monitors i.e. your location doesn’t matter, there is no PAL/NTSC color decoding and 50/60 Hz difference dependening on your location.
So going back to the conversation above about arcades and NTSC signals and what not, it may or may not be that relevant to simulating the Virtua Fighter screen depending on what the arcade cabinet did. I’m guessing they picked and chose the parts of the system that was cheapest/gave best results etc and ditched the rest because it wasn’t needed.
General about CRTs condition, I recently openend up two displays (never did that before) for tweaking the FOCUS and Screen knobs after adjusting Brightness. These seemingly trivial adjustments were making huuuge differences.
Yeah, there’s no region to worry about with arcade stuff. It’s all region-free, and while most are ~60 Hz, they vary wildly. Virtua Fighter, for example, is “medium” res, which means: 25 kHz [email protected] Hz (super-weird)
Some other notable weird ones are R-Type and other IREM games at 55 Hz, Mortal Kombat and NBA Jam at 53 Hz and Bad Dudes, Robocop and other Data East games at 57.4 Hz.
Righty ho my first attempt at horizontal convergence and its a bit too discrete - when you actually change this value right now it pops over at the triad level.
I need something that continuously changes to get some more smoothness in the shapes.
I also think I may need to let the horizontal axis effect more than one scanline away.
However given all that I think this is a marked improvement over the first attempt (maybe in true Microsoft style at attempt three I get it right(ish))
Both LCD photos are of the same frame just at different ISO’s to show the effect double the brightness has. Not sure what the dark patch below the number is (reflection of my black phone possibly?)
CRT Photo: OnePlus 8 Pro Camera: Pro Mode, ISO 200, WB 5000K, Aperture Speed 1/60, Auto Focus, 48MPixel JPEG.
LCD Photo1: OnePlus 8 Pro Camera: Pro Mode, ISO 400, WB 5000K, Aperture Speed 1/60, Auto Focus, 48MPixel JPEG.
LCD Photo2: OnePlus 8 Pro Camera: Pro Mode, ISO 200, WB 5000K, Aperture Speed 1/60, Auto Focus, 48MPixel JPEG.
