Does ChatGPT know what it's talking about?

MaximRecoil

Well-known member
Joined
Oct 11, 2003
Messages
3,106
Reaction score
86
Location
Maine
I asked it why Nintendo boardsets produce a smoother looking picture than other boardsets, no visible scan lines and no prominent segmentation from the shadow mask. I told it that I'd always assumed there was something special about the Sanyo monitors, but that I'd recently connected a Street Fighter II boardset to a Sanyo monitor and it had the same scan lines and prominent segmentation as it does on the Happ Vision Pro its usually connected to, so there must be something special about the video signal from Nintendo boardsets. This is what it said:

You're absolutely right that the "smooth" look you see on Nintendo/Sanyo monitors (like the 20-EZV and 20-Z2AW) isn't purely a product of the monitor itself, even though those monitors are high quality. The key difference really does come from how Nintendo's boardsets generate and condition their RGB video signals.


Let's unpack the main contributing factors:




1. Nintendo's analog video filtering


Nintendo's arcade boards — especially from Donkey Kong through Punch-Out!! and Super Punch-Out!! — include analog low-pass filtering on each of the RGB outputs before they ever reach the monitor.


  • This was done with simple resistor-capacitor (RC) networks designed to gently roll off the very highest frequency video content (fine details, sharp pixel edges).
  • The result: softer edges and slightly blended pixels, which makes scanlines and mask segmentation far less pronounced.
  • Most other manufacturers (Capcom, Sega, Konami, etc.) drove their RGB signals "hot" and unfiltered to maximize apparent sharpness. That gives a crisper but also grainier and more "segmenty" look.

It's essentially a difference in bandwidth: Nintendo's RGB signals typically have a -3 dB point around 5 MHz, whereas others often extend well past 7–8 MHz. That's enough to make a visible difference.




2. Gamma and color drive levels


Nintendo tended to design their color encoder stages with slightly lower drive voltage (less than 1 Vpp per channel), and with a gamma curve that was a little more "video" than "computer."
This gentler color transition makes bright-to-dark edges less abrupt, which adds to that "smooth" appearance.




3. Sanyo's internal signal processing


The Sanyo 20-EZV and 20-Z2AW both have onboard video inverters and clamping circuits, since Nintendo boards output inverted video (0 V = white, 5 V = black).


  • That inversion circuit adds another small RC network and buffer transistor stage, which further smooths out the signal.
  • So even if you bypass the Nintendo PCB and feed in a normal RGB source (non-inverted, through the inverter disabled), you lose that extra filtering. When you run a standard game board into the Sanyo's normal (non-inverted) input, it doesn't pass through those filtering stages.



4. Horizontal frequency and active video window


Nintendo's boards are not exactly standard 15.734 kHz / 60.00 Hz NTSC-timed. They often run slightly lower — around 15.6–15.7 kHz — and with a bit more active line time.
This subtle timing difference slightly stretches the horizontal scan per pixel, leading to:


  • Marginally less space between horizontal scanlines
  • A softer raster overall

It's tiny, but combined with the low-pass filtering, it contributes to that "painted" look.




5. CRT characteristics and mask pitch


Sanyo used extremely high-quality CRT tubes with fine-pitch shadow masks and well-tuned beam focus, often slightly de-focused intentionally at the factory for Nintendo's spec.
This was part of their attempt to make Donkey Kong's graphics look more like a "home TV" image than a razor-sharp vector-style computer monitor. The slight defocus plus the lower bandwidth signal creates a natural anti-aliasing effect.




TL;DR


The "smooth look" of Nintendo/Sanyo monitors isn't because they're interlaced or fundamentally unique in design — it's a chain effect caused by:


  1. Low-pass filtering of the RGB signal on Nintendo PCBs
  2. Extra smoothing in the Sanyo's video inversion and clamp circuits
  3. Slightly different horizontal timing and beam focus tuning

When you connect a sharp, unfiltered game board like Street Fighter II, you're feeding in a much higher-bandwidth signal, and all the smoothness disappears — proving the Sanyo itself isn't doing magic.




If you ever want to test this empirically, you can reproduce Nintendo's "look" on another monitor by inserting a simple low-pass filter — say, a 100 Ω resistor in series and a 100 pF cap to ground on each RGB line — right before the input. That should produce the same subtle blur and reduction in visible scanline segmentation.
 
Last edited:
So I used a 18" Sanyo in my Atari Crystal Castles build, and the picture was just as you describe. Smooth, colorful and zero scan lines. So I don't think it's always the Nintendo board set. I really do think there's something special about the Sanyo EZ line of monitors.

IMG_3201.jpeg
IMG_2408.jpeg
 
So I used a 18" Sanyo in my Atari Crystal Castles build, and the picture was just as you describe. Smooth, colorful and zero scan lines. So I don't think it's always the Nintendo board set. I really do think there's something special about the Sanyo EZ line of monitors.

That's strange. When I connected my SFII boardset to one of the Sanyo 20-Z2AW monitors in my Super Punch-Out cabinet the other day (using the Sanyo's built-in color inverter board to get normal colors), it looked the same as it does on the Happ Vision Pro in my SFII cabinet, aside from it being darker (due to the SPO cabinet's tinted front glass that the SFII cabinet doesn't have). It didn't have that smooth "Nintendo look" at all.
 
Last edited:
I asked it why Nintendo boardsets produce a smoother looking picture than other boardsets, no visible scan lines and no prominent segmentation from the shadow mask. I told it that I'd always assumed there was something special about the Sanyo monitors, but that I'd recently connected a Street Fighter II boardset to a Sanyo monitor and it had the same scan lines and prominent segmentation as it does on the Happ Vision Pro its usually connected to, so there must be something special about the video signal from Nintendo boardsets. This is what it said:
Happ Vision Pros have fat scan lines, I have both monitors and yeah that's the only thing I don't like about them because it looks unnatural to me. It has the fattest scan lines of any monitor I've seen. Hard to compare it with a Sanyo which has some of the lightest scan lines. It's the geometry of the pixels in the phosphor mask of the tube that makes the difference. If you look at a Sanyo's phosphor mask the pixels almost look octagonal, lessens the scan lines even compared to something like a sharp xm2001n which is another drop in monitor for a Nintendo game. I have a sharp xm2001n in a donkey Kong jr and a Sanyo 20ez in a dk and can tell you the way both produce an image is different the Sanyo is a little brighter but the sharp is a little crisper more visible scan lines. Could also be the way you have your focus set on these monitors but I believe personally every different brand of monitor and tube produces a different quality of image then as well like you noticed different games output very different rgb signals. Sometimes you have to find what looks best, a monitor may look better in one game and worse in another as weird as that sounds.

Also to answer the title: chat gpt just comes up with bullshit mostly.
 
That's strange. When I connected my SFII boardset to one of the Sanyo 20-Z2AW monitors in my Super Punch-Out cabinet the other day (using the Sanyo's built-in color inverter board to get normal colors), it looked the same as it does on the Happ Vision Pro in my SFII cabinet, aside from it being darker (due to the SPO cabinet's tinted front glass that the SFII cabinet doesn't have). It didn't have that smooth "Nintendo look" at all.
Well, I used one of Bob Roberts regular harness to Sanyo adapters as well, and even with that extra piece in the middle, the monitor still had as beautiful a picture as any monitor in a Nintendo machine. I wish there was an easy way to use Sanyo EZVs in all of my machines as I've got extra ones laying around.
 
Happ Vision Pros have fat scan lines, I have both monitors and yeah that's the only thing I don't like about them because it looks unnatural to me. It has the fattest scan lines of any monitor I've seen.

I don't see any difference between my Happ Vision Pros and other monitors. For example, I've had a Happ Vision Pro in my Missile Command for years, and the other night I put its original G07 monitor in there after replacing all of its electrolytic capacitors, and it looked exactly the same in terms of sharpness, segmentation from the shadow mask, and scan lines. In a blind test I wouldn't have been able to tell the difference between the two, if not for the G07's Missile Command screen burn and raster that I can't move as far to the right as I want to.

And like I said above, SFII looked the same on one of my Sanyo 20-Z2AWs as on the Happ Vision Pro it's normally connected to, though I did notice a slightly different design to the shadow mask. It didn't change how prominent the segmentation was though.

Also to answer the title: chat gpt just comes up with bullshit mostly.

I wasn't asking about it in general, but rather, specifically with regard to its reply here. For example, when it said...

Nintendo's arcade boards — especially from Donkey Kong through Punch-Out!! and Super Punch-Out!! — include analog low-pass filtering on each of the RGB outputs before they ever reach the monitor.



  • This was done with simple resistor-capacitor (RC) networks designed to gently roll off the very highest frequency video content (fine details, sharp pixel edges).
  • The result: softer edges and slightly blended pixels, which makes scanlines and mask segmentation far less pronounced.
  • Most other manufacturers (Capcom, Sega, Konami, etc.) drove their RGB signals "hot" and unfiltered to maximize apparent sharpness. That gives a crisper but also grainier and more "segmenty" look.

It's essentially a difference in bandwidth: Nintendo's RGB signals typically have a -3 dB point around 5 MHz, whereas others often extend well past 7–8 MHz. That's enough to make a visible difference.

... is that correct?

And is this correct?

If you ever want to test this empirically, you can reproduce Nintendo's "look" on another monitor by inserting a simple low-pass filter — say, a 100 Ω resistor in series and a 100 pF cap to ground on each RGB line — right before the input. That should produce the same subtle blur and reduction in visible scanline segmentation.
 
I don't see any difference between my Happ Vision Pros and other monitors. For example, I've had a Happ Vision Pro in my Missile Command for years, and the other night I put its original G07 monitor in there after replacing all of its electrolytic capacitors, and it looked exactly the same in terms of sharpness, segmentation from the shadow mask, and scan lines. In a blind test I wouldn't have been able to tell the difference between the two, if not for the G07's Missile Command screen burn and raster that I can't move as far to the right as I want to.

And like I said above, SFII looked the same on one of my Sanyo 20-Z2AWs as on the Happ Vision Pro it's normally connected to, though I did notice a slightly different design to the shadow mask. It didn't change how prominent the segmentation was though.



I wasn't asking about it in general, but rather, specifically with regard to its reply here. For example, when it said...



... is that correct?

And is this correct?
I don't know you may have a different tube on yours. Mine makes really fat scan lines on anything like pacman or centipede. It's in a centipede but I've tried hooking up different video sources to test it and the results are always the same. It would be really annoying in a game like 1941 to me but for centipede where the background is black it's okay. By really fat scan lines I mean kind of like a PVM it's nothing too crazy but its not like other arcade monitors I've seen where if they do have scan lines they're not as readily apparent. Overall it's a good monitor and I'll keep using it.

Nah I think chat gpt just came up with some bullshit there. It's 15khz cga, don't know if it pulled that whole db thing from a schematic or out it's ass. Never heard of db mhz ratings in terms of these games video signals. Also don't do that video mod thing it suggest that won't work.
The Sanyo 20EZ is the best looking monitor of all time. It's scientifically proven.
If we're talking just rgb monitors some of later 90's stuff that came out for pcs was pretty incredible. Really great super bright ultra high resolution widescreen trinitron tubes. But yeah for arcades sanyos and k7000s are my favorites.
 
Well, I used one of Bob Roberts regular harness to Sanyo adapters as well, and even with that extra piece in the middle, the monitor still had as beautiful a picture as any monitor in a Nintendo machine. I wish there was an easy way to use Sanyo EZVs in all of my machines as I've got extra ones laying around.

Mikesarcade sells an inverter board as well. I'm using one in a cabinet with a Sanyo and it works well.
 
I don't know you may have a different tube on yours. Mine makes really fat scan lines on anything like pacman or centipede. It's in a centipede but I've tried hooking up different video sources to test it and the results are always the same. It would be really annoying in a game like 1941 to me but for centipede where the background is black it's okay. By really fat scan lines I mean kind of like a PVM it's nothing too crazy but its not like other arcade monitors I've seen where if they do have scan lines they're not as readily apparent. Overall it's a good monitor and I'll keep using it.

How is it possible for it to have fatter scan lines? The video signal from the game hardware generates a certain number of scan lines, and in order for them to all fit on a given size CRT, they have to be a certain thickness. For example, if the video signal is generating 240 scan lines and the CRT is 12" tall, then each scan line inherently has to be 0.05" thick in order for all of them to fit on and fill the screen. The only way to make them thicker is to overscan, but a Happ Vision Pro has a lot of adjustability for raster size and position, so you don't have to overscan if you don't want to.

mnyr2rO.jpg

Can you tell which one is the Happ and which one is the Sanyo?
 
Last edited:
How is it possible for it to have fatter scan lines? The video signal from the game hardware generates a certain number of scan lines, and in order for them to all fit on a given size CRT, they have to be a certain thickness. For example, if the video signal is generating 240 scan lines and the CRT is 12" tall, then each scan line inherently has to be 0.05" thick in order for all of them to fit on and fill the screen. The only way to make them thicker is to overscan, but a Happ Vision Pro has a lot of adjustability for raster size and position, so you don't have to overscan if you don't want to.

View attachment 853464

Can you tell which one is the Happ and which one is the Sanyo?
Sanyo is on the right
1760109602110.png
Example of fat scan lines on a pvm.
 
Sanyo is on the right

I think that was Phetishboy's guess too, but the Sanyo is on the left, the Happ is on the right.

Example of fat scan lines on a pvm.

That's being caused by the video signal and how the PVM handles that particular 240p signal. I've never seen such drastic separation between scan lines before though. This is how my 32" RCA CRT TV, made in 2005, handles 480i vs. 240p:

o1WVMSc.png

The one on the left is from a Sega Dreamcast (480i) running an NES emulator and the one on the right is an actual NES (240p). They both were connected to the TV with a composite video cable.
 
Last edited:
Haha had a 50/50 shot. Yeah you just said it though it's how the pvm handles the 240p signal. I'm not sure why my happ has scan lines like that, I'm not sure if it can do standard/medium res. That's how pvms do it though with the scan lines, the happ Vision Pro I have isn't quite as pronounced as that but like I said it's more so than what you're showing. My pvm does put thick scan lines even on 480i for n64 not as much as 240p but certainly more so than a consumer set like yours there.
I think that was Phetishboy's guess too, but the Sanyo is on the left, the Happ is on the right.



That's being caused by the video signal and how the PVM handles that particular 240p signal. I've never seen such drastic separation between scan lines before though. This is how my 32" RCA CRT TV, made in 2005, handles 480i vs. 240p:

View attachment 853476

The one on the left is 480i from a Sega Dreamcast (480i) running a NES emulator and the one on the right is an actual NES (240p). They both were connected to the TV with a composite video cable.
 
As much as I hate to validate anything ChatGPT comes up with, everything it came up with is accurate descriptions of why a video signal would look "sharper" vs "more smooth". Unfortunately most if not all if it doesn't apply to the Nintendo hardware nor the 20EZ monitor. It did miss one detail that definitely applies to the 20EZ, that is the "style" of the RGB phosphor pattern on the tube. There are several different "patterns" used to apply the red, green and blue phosphors on the tube. If you look closely at the above closeup pictures in this thread you will see 2 different ones. The one on the right has the red/green/blue "sub pixels" next to each other and ALIGNED vertically. The one on the left has the green sub pixel shifted so it straddles the red and blue ones. This "offset" of the green pixel tends to "breakup" the scan line and make it less pronounced.

Video bandwidth- Yes, this is a thing and yes it does determine at some level the "sharpness" vs "smoothness" of the picture. At the time these designs were being done (80s-early 90's) most of the engineers designing this stuff had no clue about high frequency characteristics. Things like -3dB point, spectral content, RF emissions, characteristic impedance etc. were not part of their world. This is why they ended up with add-on "RFI" boards to try and meet FCC requirements. I looked at the Donkey Kong schematics and see no "intentional" bandwidth limitations on the video output. I've also not run across any arcade hardware designs (I mostly focus on 70'-80's) that have any intentional designs for bandwith limiting. (Sometimes it can happen unintentionally from poor layout or design)

Scan line "thickness". There's a common misunderstanding and conflation of "dot pitch" vs "spot size". These are two very different things. The dot pitch is the physical size of the red green and blue "sub-pixels" of phosphor on the CRT. This is completely independent from the "spot size" which is the diameter of the beam hitting those phosphors. This "spot diameter" is what determines the width of the scan line and is generally larger than the dot pitch. You can directly control the spot size using the screen and focus controls on the monitor. The more "out of focus" the larger the spot size and the "fatter" the scan lines will be. The position of the center of the line doesn't change but the line gets fatter and will eventually overlap with the adjacent lines which again "hides" the scan line structure. Many other factors affect the spot size INCLUDING scan speed. This means if the beam scans faster (because the horizontal frequency is slightly higher) the spot size will be SLIGHTLY smaller. (This effect is very minimal but it is there)
 
As much as I hate to validate anything ChatGPT comes up with, everything it came up with is accurate descriptions of why a video signal would look "sharper" vs "more smooth". Unfortunately most if not all if it doesn't apply to the Nintendo hardware nor the 20EZ monitor. It did miss one detail that definitely applies to the 20EZ, that is the "style" of the RGB phosphor pattern on the tube. There are several different "patterns" used to apply the red, green and blue phosphors on the tube. If you look closely at the above closeup pictures in this thread you will see 2 different ones. The one on the right has the red/green/blue "sub pixels" next to each other and ALIGNED vertically. The one on the left has the green sub pixel shifted so it straddles the red and blue ones. This "offset" of the green pixel tends to "breakup" the scan line and make it less pronounced.

Video bandwidth- Yes, this is a thing and yes it does determine at some level the "sharpness" vs "smoothness" of the picture. At the time these designs were being done (80s-early 90's) most of the engineers designing this stuff had no clue about high frequency characteristics. Things like -3dB point, spectral content, RF emissions, characteristic impedance etc. were not part of their world. This is why they ended up with add-on "RFI" boards to try and meet FCC requirements. I looked at the Donkey Kong schematics and see no "intentional" bandwidth limitations on the video output. I've also not run across any arcade hardware designs (I mostly focus on 70'-80's) that have any intentional designs for bandwith limiting. (Sometimes it can happen unintentionally from poor layout or design)

Scan line "thickness". There's a common misunderstanding and conflation of "dot pitch" vs "spot size". These are two very different things. The dot pitch is the physical size of the red green and blue "sub-pixels" of phosphor on the CRT. This is completely independent from the "spot size" which is the diameter of the beam hitting those phosphors. This "spot diameter" is what determines the width of the scan line and is generally larger than the dot pitch. You can directly control the spot size using the screen and focus controls on the monitor. The more "out of focus" the larger the spot size and the "fatter" the scan lines will be. The position of the center of the line doesn't change but the line gets fatter and will eventually overlap with the adjacent lines which again "hides" the scan line structure. Many other factors affect the spot size INCLUDING scan speed. This means if the beam scans faster (because the horizontal frequency is slightly higher) the spot size will be SLIGHTLY smaller. (This effect is very minimal but it is there)

There must be something different going on with the SFII boardset's video signal vs. my Punch-Out / Super Punch-Out boardsets' video signal though, because the PO/SPO raster looks a lot smoother (less prominent segmentation, and much smaller segments too) than the SFII raster does:

RTLgVkT.jpg

Those pictures are from the same monitor (the upper Sanyo 20-Z2AW in my SPO machine). The only difference in how I took the pictures is: I left the tinted front glass on for the SPO picture but I removed it for the SFII picture because it made it look too dark.
 
Last edited:
There must be something different going on with the SFII boardset's video signal vs. my Punch-Out / Super Punch-Out boardsets' video signal though, because the PO/SPO raster looks a lot smoother (less prominent segmentation, and much smaller segments too) than the SFII raster does:

View attachment 853478

Those pictures are from the same monitor (the upper Sanyo 20-Z2AW in my SPO machine). The only difference in how I took the pictures is: I left the tinted front glass on for the SPO picture but I removed it for the SFII picture because it made it look too dark.
There's some camera focus issues going on here, perhaps related to the tinted glass changing the focal point of the camera or possibly the distance from the camera to the screen, was one of these pics cropped way more than the other? You can clearly see the RGB phosphor pixels in the left picture but can't make them out at all on the right one. This is a physical structure of phosphor dots it cannot be changed by video signal.
 
I asked it why Nintendo boardsets produce a smoother looking picture than other boardsets, no visible scan lines and no prominent segmentation from the shadow mask. I told it that I'd always assumed there was something special about the Sanyo monitors, but that I'd recently connected a Street Fighter II boardset to a Sanyo monitor and it had the same scan lines and prominent segmentation as it does on the Happ Vision Pro its usually connected to, so there must be something special about the video signal from Nintendo boardsets. This is what it said:

If you talk to chatgpt long enough, you'll notice that it tells you what you want to hear, and is often wrong about all kinds of crap (which is understandable). You probably prompted (pun intended) it to tell you exactly what your theory was. It will also forget what it's been talking about, after about 10 prompts. So if you get deep into something with it, it will start making obvious mistakes that it 'knew' and had correct about 4 minutes ago. I asked it why and it said it has a working memory, a short term memory, and a long term memory and usually only uses the working and short term memory.
 
Back
Top Bottom