Video probs w/ SNES RGB (Vertical Bands, etc)

[Voultar]

@Borti,

Your S-RGB data is very interesting. I've analyzed gobs of those and always came to a 6.5dB conlcusion. Ste of HD Retrovision and I were talking about how odd it is that the S-RGB has such a wonky gain setting. Maybe there are swing variances in the encoder, too?








That's very interesting, indeed.

So, if people are using Voultar's board, they should probably be removing the 1k2 resistors on that PCB and replacing R3 with the appropriate value on the console PCB (and adding a series cap before the THS7374).

No. The result is the same.

[borti4938]

I don't see any reason for replacing this, too

It's just another option if
a) you build your own DIY RGB bypass (or buy a ready to install board without the 1k2 resistors, or
b) you want to use the stock S-RGB and want to keep the system looking as much as possible 'stock' like

[Syntax]

I don't see any reason for replacing this, too

It's just another option if
a) you build your own DIY RGB bypass (or buy a ready to install board without the 1k2 resistors, or
b) you want to use the stock S-RGB and want to keep the system looking as much as possible 'stock' like

I know its anal but it evenly attenuates the colours, and is easier than using 3 through holes.

[Voultar]

I don't see any reason for replacing this, too

It's just another option if
a) you build your own DIY RGB bypass (or buy a ready to install board without the 1k2 resistors, or
b) you want to use the stock S-RGB and want to keep the system looking as much as possible 'stock' like

I know its anal but it evenly attenuates the colours, and is easier than using 3 through holes.

Totally. If you're building something up, you can current limit the DAC with a resistor.

[borti4938]

@Borti,

Your S-RGB data is very interesting. I've analyzed gobs of those and always came to a 6.5dB conlcusion. Ste of HD Retrovision and I were talking about how odd it is that the S-RGB has such a wonky gain setting. Maybe there are swing variances in the encoder, too?

[...] some pictures [...]

That's very interesting, indeed.

Yes, I was extremely surprised, too. Today I made some measurements with my second SNES: SNSP-CPU-1CHIP-02 with a S-RGB A.
This time it looks close to your findings.
Also there is a massive improvement in retaining the rising and falling times.

I will sort the pictures tomorrow (or the day after tomorrow) and present them here, too
(For me it's time to go asleep for today)

[Voultar]

@Borti,

Your S-RGB data is very interesting. I've analyzed gobs of those and always came to a 6.5dB conlcusion. Ste of HD Retrovision and I were talking about how odd it is that the S-RGB has such a wonky gain setting. Maybe there are swing variances in the encoder, too?

[...] some pictures [...]

That's very interesting, indeed.

Yes, I was extremely surprised, too. Today I made some measurements with my second SNES: SNSP-CPU-1CHIP-02 with a S-RGB A.
This time it looks close to your findings.
Also there is a massive improvement in retaining the rising and falling times.

I will sort the pictures tomorrow (or the day after tomorrow) and present them here, too
(For me it's time to go asleep for today)

The SRGB-A could simply have a different fixed gain setting. Something to definitely further analyze.

[FBX]

Got my 0603 caps in from Digi-key, and happy to report the ghosting is removed as expected!

This is on my mini (my 1CHIP-03 already has the cap replaced):





And here's raw the output from Voultar's board to the OSSC at optimal timing x4 mode, and into the Datapath Vision E1s:



CLEAN!!

-FBX

[mikejmoffitt]

I picked up some ~300nF caps from the local electronics store. I'll give this a try when I get home.

[nakedarthur]

I replaced the cap on my SNES Mini with the one FBX linked above and unfortunately still having a visual issue. I don't think it's the same thing you guys are looking at, its more like a streaking to the right side of certain objects. On Super Mario World its noticeable for me on Donut Plains 3 platforms, or in the water with fire suit. I'm using an RGB modded TV, so who knows, could be something with my RGB mod and I just haven't noticed it on other consoles. Here's a couple crappy pictures, sorry it was very hard to capture since it's a subtle effect.

[Voultar]

I replaced the cap on my SNES Mini with the one FBX linked above and unfortunately still having a visual issue. I don't think it's the same thing you guys are looking at, its more like a streaking to the right side of certain objects. On Super Mario World its noticeable for me on Donut Plains 3 platforms, or in the water with fire suit. I'm using an RGB modded TV, so who knows, could be something with my RGB mod and I just haven't noticed it on other consoles. Here's a couple crappy pictures, sorry it was very hard to capture since it's a subtle effect.

Oh yes, I see exactly what you're talking about.

That looks like a clamping error on the device end of things. Unfortunately swapping the cap won't fix that.

[nakedarthur]

That looks like a clamping error on the device end of things. Unfortunately swapping the cap won't fix that.

Thank you, m'Lord. Without derailing the thread too far is there anything I can try to fix it? I have a KV-27S42 with 75ohm termination and 0.1uf caps running into the jungle.

[Voultar]

You're better off posting that in the TVRGB thread, I'll take a look at it in a bit.

[paulb_nl]

I encountered an issue after replacing C11 on my consoles. Some capcom games like Street fighter and Aladdin have a big black bar at the top. This bar now protrudes into the top of the image and it gets worse when it fades out. I confirmed that this does not happen with my unmodified Super Famicom Jr.

Here is a video of it in action. Take a look at top of the player select screen when it fades in and out. There is also a somewhat lighter horizontal bar at the bottom but I dont know if that is related.

https://youtu.be/UjuWdeT5Uq4

This screenshot was taken during fade in and brightened.

[FBX]

I encountered an issue after replacing C11 on my consoles. Some capcom games like Street fighter and Aladdin have a big black bar at the top. This bar now protrudes into the top of the image and it gets worse when it fades out. I confirmed that this does not happen with my unmodified Super Famicom Jr.

Confirmed these issues on both my 1CHIP-03 and Junior, which both have the 470nF cap installed. Looks like we need to find out the capacitance range of when this starts happening to these Capcom games.

Edit:

I ordered a blanket spread of caps starting at 100nF and going all the way up to 330nF. When they arrive, I'm going to test each one and see where the perfect balance is of anti-ghosting without causing issues with these Capcom games. I'm sure Voultar already has plenty on hand to test with, but I need a good spread of these anyway just in case something else comes up.

[rama]

Wow, what an off effect. I get it on my SNES as well.
It looks like their soft fade in/out routine doesn't work properly anymore.
It doesn't seem like an analog artefact at all!

Update:
Yeah, it is only affecting that shadow mask effect. Capcom games like to use it a lot.
I tried the improved Breath of Fire 2 hack, which uses it on the overworld map.

[Voultar]

Holy cow that is super bizarre!

I'll take a look at it shortly. But yes, maybe if we reduce the charge (100nF-200nF) this will remedy that.

Very, very strange!

[borti4938]

From what I've seen for the moment is that the effect is also present with the stock C11, but much less noticeable and hard to see.

I haven't opened my SNES yet, but here are some spontaneous ideas:

• Leaving the stock C11 in place and use just another 300nF-470nF cap in parallel
• Assembly one or two decoupling capacitor to the AVDD pin 155 (e.g. a 0.1uF and a 0.01uF)

[mikejmoffitt]

In case anyone wanted to know what having no capacitor on C11 looks like...

[borti4938]

^

By the way, here are my measurements from my second 1Chip SNES. This time I measured different values.

• R3 = 1.74kohm for S-RGB output with 150ohm load (as like as in a NTSC big model 1Chip),
• R3= 1.79kohm for S-RGB with post transistor and 150ohm load at the transistor output (mostly as like as in a stock PAL 1Chip-SNES), and
• R3 = 1.72kohm for THS7314, und THS7374 with 150ohm load (THS7374 with filter off is state of the art RGB bypass method)

@Voultar: here you can see that I measured a 6.22dB gain of the S-RGB A.

Spoiler

Just a small note:
Finally I reduced the value of R3 down to 1.69k (I use the THS7374) due to the fact that the RGB additionally attenuates the signal. Assuming 75Ohm output, 1Ohm cable resistance and 75Ohm load at the TV or scaler, one get from 720mV output at the console ~700mV at the TV or scaler.



And finally some measurements on the rise and fall times. Note that the S-RGB A behaves a bit better that the S-RGB.

S-RGB A:

Spoiler

S-RGB A with post transistor:

Spoiler

THS7374 in bypassmode:

Spoiler

[rama]

Thanks again, borti4938

Those slight S-RGB gain variations almost seem like an external factor is causing them.

Regarding C11: What would the effect of a higher / lower ESR capacitor (of the same size) be?

[paulb_nl]

Thanks borti4938 for your measurements.

I have been trying to find the right R3 value for my consoles with the 240p suite white screen test with an OSSC and Datapath VisionRGB E1.

US 1-CHIP-01: R3 measured 1585 Ohm. Needs around 1620 Ohm but red is weaker than green and blue.

PAL 1-CHIP-02: R3 measured 1593 Ohm. Needs around 1800 Ohm but red is stronger than green and blue.

US Jr/mini S-RGB mod to multiout 75 Ohm: R3 measured 1586 Ohm. Needs around 1680 Ohm but red is weaker than green and green is weaker than blue. That is the exactly the same as GigaBoots.

As you can see none of my consoles have equal values for red, green and blue.

[FBX]

As you can see none of my consoles have equal values for red, green and blue.

I use the OSSC to correct RGB levels using the Gain function. What I do is pipe the OSSC into my E1s, then I load up a white screen in the Test Suit. From there, I screencap and load up a paint program to check the RGB values. Then I adjust the Gain for the channels that are lower, screengrab, and compare in the paint program again. Once they are all perfectly aligned, I save the OSSC profile. Unfortunately this doesn't work for component input. Only RGB.

[borti4938]

@paulb_nl:
Thanks for the feedback. This helps to find an average value to use. An average is the best option for the 'general' user as like as for the three resistor method, where the 'common' used values 1.2k and 750 were the result by measurements of just a single console.

BTW:
From my measurements, the first console has equal outputs for red, green and blue and the second one is weaker on green. Not with mentioning.
The best option is to attenuate each channel for each console individually at the output. But that's by far the most effortful method.

[paulb_nl]

@FBX
I used more or less the same method to find the correct resistor value for R3. I removed R3 and soldered a 10k trimpot and then adjusted the trimpot until the RGB values on the capture were close to the wanted values. I captured the white screen set to 240 and 255 in the 240p suite.

@paulb_nl:
Thanks for the feedback. This helps to find an average value to use. An average is the best option for the 'general' user as like as for the three resistor method, where the 'common' used values 1.2k and 750 were the result by measurements of just a single console.

Is it possible to find a good average with such a big variance between consoles? I recommend buying a range of resistors and a trimpot and find the best value.

[FBX]

My package of caps arrived today (fast shipping for just a few bucks)!

Gonna immediately get started testing different values to see what the happy balance is for Capcom games. Now if it turns out we can 'get away' with a little adverse effects on the fade-in sequences, I won't mind that at all, but the main thing is to get that top line of graphics fixed on the fully revealed arena.

[FBX]

Bad news:

Street Fighter Capcom games are so sensitive that even 100nF screws with the top of the graphics:



And here's how much ghosting is reduced at 100nF (image put in negative colors and over-sharpened):



So there's no wiggle room. However, I'm thinking of a compromise in increasing the capacitance until that top line of graphics is uniformly dim without cropping into the next line. Hopefully one of the caps I have will work perfectly for this. I'll report back my progress.

[Voultar]

I'm seeing this phenomenon (to a lesser extent) with an unaltered C11. Stabilizing the compensating node just exacerbates this effect. I'll do some more testing when I have time, but I believe by the evidence that this is just a quirk of the ASIC. On the digital side that's carried over through the DAC. Can't say for sure though until I dig a little deeper.

I conducted several bandwidth tests and saw virtually no degradation to the slew/transitional response time. So this is something else entirely.

I really hate the 1CHIP SNES.

[Kez]

I'm seeing this phenomenon (to a lesser extent) with an unaltered C11.

Can corroborate this. I have a stock 1CHIP-01 SFC and SF2 looks very reminiscent of FBX's 100nF screenshot.

[Voultar]

Here's a quirky little idea.

Modify the value of R3 and make the video output hotter. Then attenuate accordingly on the analog outputs.

We can try manipulating the amplitude inside the chip (by playing with R3) and always attenuate on the outputs. Just an idea to float with.

[FBX]

Sorry for the delay in showing more results, but the Super NES's Pico fuse decided now was the time to crap out, so I had to stop everything and take care of that. And not to worry, it wasn't me messing with C11 that blew the fuse. It was just one too many power cycles testing Street Fighter and Mega Man X.

At any rate here are more negative image tests with Mega Man X's health meter:


220nF:


330nF:


470nF:


I'm of the opinion to just stick with 470nF since it really does the best job of anti-ghosting. Like Voultar was suggesting, the focus should be elsewhere now. I was half-jokingly thinking we need coders to make hack ROMS that fix the line rendering on these Capcom games

Below are the effects on Street Fighter II:

220nF:


330nF:


470nF: