SNES/Famicon PCB Revisions and RGB Video

[Einzelherz]

If I didn't have to work for a living I'd gladly collect all of the chipset versions of every console, recap them, and spend my time doing video quality analysis. It seems to me that a lot of the common knowledge regarding these is anecdotal and wrought with unaccounted for variable (age, small batch samples, etc.).

[blizzz]

Yea, I wonder if a recap or something else would improve the quality of my APU console. But since I either play on my 1CHIP SFC or modded PAL console I don't use it. So it's not worth it for me to recap the APU one.

[Blair]

well, I ended up getting a one chip. I didn't expect the output to be so bright and detailed. I'm not sure if it's really better looking than my UN25. but it's definitely different.

I'll have to take some comparison screenshots when I get some time later, I'm pretty happy with it though.

[FBX]

I pulled a screenshot taken from a video capture I did of my 1CHIP-03 SNES scaled to 4x on my Framemeister. Keep in mind the original image is even cleaner because this is taken from the converted video capture:

http://www.firebrandx.com/downloads/scr ... srgb4x.png

[Einzelherz]

well, I ended up getting a one chip. I didn't expect the output to be so bright and detailed. I'm not sure if it's really better looking than my UN25. but it's definitely different.

I'll have to take some comparison screenshots when I get some time later, I'm pretty happy with it though.

Can you open them both to tell us what boards are in each?

[blizzz]

FBX, is there any post-processing in the screenshot? The white is pure 255-255-255, which is a lot higher than what I get from my consoles on the Framemeister. Could you post a screenshot of the RGB color ramp screen of the 240p test suite? A screenshot of the pure white screen in the 240p test suite would also be interesting.

[FBX]

FBX, is there any post-processing in the screenshot? The white is pure 255-255-255, which is a lot higher than what I get from my consoles on the Framemeister. Could you post a screenshot of the RGB color ramp screen of the 240p test suite? A screenshot of the pure white screen in the 240p test suite would also be interesting.

No post processing has been done. However, it's coming from the Framemeister, and so there's two different conversions going on before it gets to the final screenshot:

1. Framemeister and my profile settings:

Brightness = 25
Gamma = 19
Black = 0
A/D = 131

2. Avermeida Game Capture HD II at optimal quality video conversion.

I'm in the middle of uploading a video for a friend, but later on today I will do a screen cap of the color ramp from the 240p test suit.

[blizzz]

My settings are:
Brightness = 25
Gamma = 19
Black = 2
A/D = 130
And of course YCbCr output.

The pure white from my 1CHIP comes out at 230 (out of 255). Here's how it looks: rgb_color_bars.png (emulator screenshot)

I'm really interested in your color bars. Does your console have an RGB Amp Bypass?

[FBX]

My settings are:
Brightness = 25
Gamma = 19
Black = 2
A/D = 130
And of course YCbCr output.

The pure white from my 1CHIP comes out at 230 (out of 255). Here's how it looks: rgb_color_bars.png (emulator screenshot)

I'm really interested in your color bars. Does your console have an RGB Amp Bypass?

It's only got Csync restored, which I did myself. It's an original design SNES console with the 1CHIP-03 revision.

But yeah, I will get to the color ramp soon. The video I'm currently uploading is 1GB in size, so it's taking a while.

[RGB32E]

I pulled a screenshot taken from a video capture I did of my 1CHIP-03 SNES scaled to 4x on my Framemeister.

Stating that's it's a 4x scale is misleading, as you have columns of pixels that alternate in a pattern between 4 and 5 pixels in width (e.g. 4/5/5/4/5/4/5/5/5/4). I'd imagine you're going for a 4:3 aspect ratio, but it gives you uneven horizontal scaling. I'm sure you're already aware of that...

What lead you to use a gamma value of 19?

[RGB32E]

In preparation of testing the OSSC I tried to get the capture of my Framemeister right. One step was to compare the brightness levels of a 1CHIP vs a 3-chip console, since recently people have been saying that 1CHIP consoles are too bright. At least for my consoles that's not true, I get the exact same brightness from both consoles. But it reminded me again of how abysmal my APU-01 console's video output is.

Just have a look at this: 1CHIP-02 vs APU-01 (cables, PSU, settings and processing are the same)

The 1CHIP-0* revs do appear brighter/more intense than the earlier revs. The 1CHIP mini in particular generally appears even brighter than the 1CHIP-0* consoles and it was only the "resistors to ground" or ">75 ohm output resistors" approaches that lower output levels that are closer to the 1CHIP-0* models. From what I've seen, most of the "too bright" 1CHIP output stemmed from people modding mini 1CHIPs that did not account for the higher voltage output. That said, comparing a SHVC-CPU-01 to a 1CHIP-0* should give a noticeable picture level difference.

[FBX]

I pulled a screenshot taken from a video capture I did of my 1CHIP-03 SNES scaled to 4x on my Framemeister.

Stating that's it's a 4x scale is misleading, as you have columns of pixels that alternate in a pattern between 4 and 5 pixels in width (e.g. 4/5/5/4/5/4/5/5/5/4). I'd imagine you're going for a 4:3 aspect ratio, but it gives you uneven horizontal scaling. I'm sure you're already aware of that...

What lead you to use a gamma value of 19?

Sorry I meant 4x scale vertical, but 4:3 correction on the horizontal. The 19 Gamma was the result of trying to find the least amount of solid color noise for most of the games I play (Mega Man X, Zelda, etc). Of course it doesn't really get rid of the noise, only just pushes it other shades of color.

[blizzz]

That said, comparing a SHVC-CPU-01 to a 1CHIP-0* should give a noticeable picture level difference.

I wish I had a SHVC-CPU-01 instead of that APU-01 console. At least fagin's CPU-01 has amazing video quality:

[FBX]

I got to the bottom of the 255 whites issue on my screenshot: VLC's screenshot function was the culprit. Instead of using the built-in feature, I did a print-srcn capture and the color ramp showed the top white value to be 242. When I used the built-in screencap function, the top white jumped to 255, and several of the top end colors were crushed. Gee, thanks VLC. Thanks a lot for screwing my screencaps up!

BTW, forcing YCbCr output causes my capture device to significantly dull the colors. The top white was only 225 when I recorded footage from that mode.

[blizzz]

I take all my screenshots with MPC-HC (the Kawaii Codec Pack version). It's the only player I found that gets the colors right for any video I throw at it. VLC likes to change colors and do other stupid stuff.

YCbCr is always limited range on the Mini. So you have to expand the 16-235 color space to 0-255 on PC. MPC-HC does this automatically. On my devices RGB has too intense greens, but red and blue are slightly too dark. Just what you would expect from a YCbCr conversion with Rec.601 instead of 709. My guess is that the Mini uses YCbCr internally and uses the wrong color matrix to convert YCbCr to RGB for RGB output.

The RGB color bar screen should make it easy to compare the colors in YCbCr and RGB.

[FBX]

Cool, I'll download MPC_HC and give it a try.

[FBX]

Ok so I tried out the Kawaii pac and here's my results:

The RGB output video had to be switched to 16-235 color range in order for it to look correct. Top color value:

Red = 220
Green = 255
Blue = 220

The YCbCr output video had to be switched to 0-255 color range in order for it to look correct. Top color value:

Red = 245
Green = 220
Blue = 245

In both cases, a few of the top-end colors for each primary were crushed, whereas when shown live on my Sony display, each color is distinctive. I'm guessing this is due to the capture device having to convert/compress the color range, and this results in the less optimal looking colors.

Edit: Testing out direct feed to my display, it turns out the top-end colors were getting crushed there as well. In order to just make out a change in shade from the top green to the next bar over, I had to set A/D to 119. No other adjusting of brightness/Gamma/black level would work, just A/D. This is very interesting, and I may have to re-calibrate all my profiles based on Artemio's color bars.

[copy]

That said, comparing a SHVC-CPU-01 to a 1CHIP-0* should give a noticeable picture level difference.

I wish I had a SHVC-CPU-01 instead of that APU-01 console. At least fagin's CPU-01 has amazing video quality:

Spoiler

I have to say, I'm also quite pleased with my SHVC-CPU-01's quality. I just hooked it up through RGB to my XRGB-Mini for the first time, and it looks so much better than I was expecting. The slight blurriness is there, but it doesn't seem anywhere near as bad as I feared. I still hope to add a 1CHIP to my collection some day, but now I don't feel such an urgent need for it.

Related to this: FBX, I noticed in your new N64 profiles page that you optionally recommend setting the XRGB-Mini's sharpness to +1 to counteract the N64 blurring. That's a great tip. Similarly, I think it also does a pretty good job of removing the blurriness from my SHVC-CPU-01. Perhaps you could add a mention of that to your SNES page too, if you find the same results.

[FBX]

Related to this: FBX, I noticed in your new N64 profiles page that you optionally recommend setting the XRGB-Mini's sharpness to +1 to counteract the N64 blurring. That's a great tip. Similarly, I think it also does a pretty good job of removing the blurriness from my SHVC-CPU-01. Perhaps you could add a mention of that to your SNES page too, if you find the same results.

It's quite nifty, isn't it? For the longest time I thought the sharpness function was pretty much useless, but it turns out to be pretty good at deconvolution.

[FBX]

RetroRGB sent me a SNES mini with his RGB and digital audio mod work. I have to say it even outshines my 1CHIP-03 original model in clarity and sharpness. In fact, I have to say it's the sharpest RGB I've yet seen hooked into my Framemeister. I'll post some sample screenshots later on tonight.

[FBX]

Just did a definitive zoom capture and discovered I was being fooled by dynamic color range into thinking the SNES Jr. had a sharper image. In the comparison image below, I used the same focus and AR settings for both consoles. You can see there's a bit of ringing/banding on the SNES Jr. image, and it looks to be just a hair less sharp. However, these are microscopic difference most people won't notice unless they have a keen eye or a giant LCD screen. Probably can't even pic up this detail in a CRT anyway for those that use those.

[nissling]

I'll be getting a 1chip SFC within the next couple of weeks so I can make some comparisons to the CPU-01.

So finally it looks like the package has arrived to Sweden. "Only" took three months via surface mail from Japan, you could probably get there faster by bike. Either way, once I got the console here I can do some comparisons for you guys. Any specific games of interest?

[borti4938]

RetroRGB sent me a SNES mini with his RGB and digital audio mod work. I have to say it even outshines my 1CHIP-03 original model in clarity and sharpness. In fact, I have to say it's the sharpest RGB I've yet seen hooked into my Framemeister. I'll post some sample screenshots later on tonight.

You are comparing the RGB amplification by a S-RGB with the one by a THS7314. This is not fair in terms of concluding that the SNES mini has a sharper image than a 1Chip-01/02/03.
You have to build a THS7314 into the 1Chip-0x too to make a fair comparison.

Btw: at the weekend I looked into the ghosting issue of the S-CPUN. The user rama (not registered here) hints me into the direction that the issue is related by overshoots in the RGB signals. This can be seen here: yellow line is the signal coming out of the S-CPUN and the blue one after the THS7314 with load. Shown is a jump from 0 to max.



The overshoots can be reduced by adding a resistor between the AVCC pin of the S-CPUN (pin 155) and Vcc. A resistor of 10 ohms seems to be enough, higher values just further reduce the amplitude levels. However, the overshoots cannot be completely avoided (at least not in my SNES setup) with the simple resistor fix.
As mentioned the resistor also reduces the brightness. In my SNES (SNSP-CPU-1CHIP-01) 20 ohms reduced the Vpp at R, G and B to 0.7V as wanted. Also all Vpp were equal with this resistor, which was not the case before. I tested with a potentiometer.

Here is how my SNES looks at the moment. (I haven't a 20 ohm resistor, so I took a 30 ohm one)

Spoiler

Just as a note
A warning: lifting pin 155 at the S-CPUN can be a tricky and one can easily broke something. The cartridge slot is right in the way.

[FBX]

RetroRGB sent me a SNES mini with his RGB and digital audio mod work. I have to say it even outshines my 1CHIP-03 original model in clarity and sharpness. In fact, I have to say it's the sharpest RGB I've yet seen hooked into my Framemeister. I'll post some sample screenshots later on tonight.

You are comparing the RGB amplification by a S-RGB with the one by a THS7314. This is not fair in terms of concluding that the SNES mini has a sharper image than a 1Chip-01/02/03.
You have to build a THS7314 into the 1Chip-0x too to make a fair comparison.

See my previous post with the pic comparison. Turns out my 1CHIP-03 was cleaner and sharper after all. And while I agree it wouldn't be fair to compare against say a digital mod like HDMI for the N64, in this case, the amp used in the SNES mini is still analog RGB. I see no problem comparing it with others, especially considering the SNES mini can't even output RGB without being modded in some form or another. If you wanted to be ultimately fair, the mini would have to be in S-Video.

[Pasky]

RetroRGB sent me a SNES mini with his RGB and digital audio mod work. I have to say it even outshines my 1CHIP-03 original model in clarity and sharpness. In fact, I have to say it's the sharpest RGB I've yet seen hooked into my Framemeister. I'll post some sample screenshots later on tonight.

You are comparing the RGB amplification by a S-RGB with the one by a THS7314. This is not fair in terms of concluding that the SNES mini has a sharper image than a 1Chip-01/02/03.
You have to build a THS7314 into the 1Chip-0x too to make a fair comparison.

See my previous post with the pic comparison. Turns out my 1CHIP-03 was cleaner and sharper after all. And while I agree it wouldn't be fair to compare against say a digital mod like HDMI for the N64, in this case, the amp used in the SNES mini is still analog RGB. I see no problem comparing it with others, especially considering the SNES mini can't even output RGB without being modded in some form or another. If you wanted to be ultimately fair, the mini would have to be in S-Video.

There's a low pass filter in the THS7314, you aren't comparing Apples to Apples when you're comparing it against a different encoder even though it's RGB vs RGB. I'm not sure if the S-RGB is doing any filtering, didn't look at the datasheet much.

[FBX]

There's a low pass filter in the THS7314, you aren't comparing Apples to Apples when you're comparing it against a different encoder even though it's RGB vs RGB. I'm not sure if the S-RGB is doing any filtering, didn't look at the datasheet much.

Does the low-pass filter make the image sharper? Just curious how it works.

[RGB0b]

I put Borti's new board in FBX's SNES Mini, which has a THS7374. I'm looking into this now...

[Pasky]

There's a low pass filter in the THS7314, you aren't comparing Apples to Apples when you're comparing it against a different encoder even though it's RGB vs RGB. I'm not sure if the S-RGB is doing any filtering, didn't look at the datasheet much.

Does the low-pass filter make the image sharper? Just curious how it works.

Not necessarily. The point is, you're not comparing the same things. The low pass filtering can actually make things a bit blurry but helps alleviate any noise on the signal, however what if the S-RGB has an even steeper curve on its filter, causing additional blurring because the filter is stronger. The comparison isn't fair if that's the case. That's the point I'm trying to make.

[Pasky]

EDIT:

Phase and aspect were wrong, will try this again later. Want as far a comparison as possible.

[FBX]

I enlisted Bob's help on the PQ issue I posted the pic of concerning the SNES mini's "borti" board mod. We're pretty certain now that it's an issue with the board and not the SNES mini. Bob sent me a screencap under the same zoom settings on a 1CHIP-03 modded with the same board, and it looked to have the exact same PQ issues as the SNES mini pic I posted.