An actual SNES mod to sharpen the pixels of 2-CHIP consoles?

The place for all discussion on gaming hardware
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

I'd say for $10USD in parts its well worth upgrading a 2 chip SNES for. I cant wait to give it a spin, I plan on taking some detailed stills on an FD Trinitron Wega to compare the differences.
User avatar
Opatus
Posts: 39
Joined: Mon May 25, 2020 10:28 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Opatus »

Just wanted to compare the results of the 2-Chip digital RGB mod with this sharpening circuit. Unfortunately I cannot capture the image directly, I had to take a photo of the TV screen. The SNES was connected to an OSSC in Line5X Generic sampling mode to oversample the signal.

Of course the mod is not yet perfect, more complicated to install and the BOM is more expensive than $10 for sure.

Image

Image Image
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

Opatus wrote:Just wanted to compare the results of the 2-Chip digital RGB mod with this sharpening circuit. Unfortunately I cannot capture the image directly, I had to take a photo of the TV screen. The SNES was connected to an OSSC in Line5X Generic sampling mode to oversample the signal.

Of course the mod is not yet perfect, more complicated to install and the BOM is more expensive than $10 for sure.

Image

Image Image
Is there any way your digital RGB mod could include a DAC for analog out, thus providing the best of both worlds? :mrgreen:
User avatar
Opatus
Posts: 39
Joined: Mon May 25, 2020 10:28 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Opatus »

Is there any way your digital RGB mod could include a DAC for analog out, thus providing the best of both worlds? :mrgreen:
This is analog, it goes through an R2R ladder, the next version will have a 10bit DAC IC instead. All the images are done using an OSSC with SCART RGB. Personally I don't really care about an HDMI mod.
I just call it digital RGB, while it uses the digital signals coming directly (more or less) from PPU2.
User avatar
Gunstar
Posts: 617
Joined: Wed Apr 09, 2014 10:29 am
Location: UK

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Gunstar »

Opatus wrote:Just wanted to compare the results of the 2-Chip digital RGB mod with this sharpening circuit. Unfortunately I cannot capture the image directly, I had to take a photo of the TV screen. The SNES was connected to an OSSC in Line5X Generic sampling mode to oversample the signal.

Of course the mod is not yet perfect, more complicated to install and the BOM is more expensive than $10 for sure.
Thanks for the comparison pic, the digital mod is looking damn good, hopefully it does get perfected.
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

Opatus wrote:
Is there any way your digital RGB mod could include a DAC for analog out, thus providing the best of both worlds? :mrgreen:
This is analog, it goes through an R2R ladder, the next version will have a 10bit DAC IC instead. All the images are done using an OSSC with SCART RGB. Personally I don't really care about an HDMI mod.
I just call it digital RGB, while it uses the digital signals coming directly (more or less) from PPU2.
Sweet man, I could give 2 shits about HDMI myself, give me good old analog and if I need HDMI I'll pass it through an OSSC for HDMI. Great stuff. Any idea when it will be ready?
RetroSwim
Posts: 14
Joined: Tue Oct 15, 2013 4:21 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by RetroSwim »

yoshiyukiblade wrote:It looks like maybe optimal sampling wasn't used. That's the preferred option for pixel perfect output, but it could make some artifacts worse. The capture device can also throw away more information from colorspace conversions, so it's tricky to get it right. Even worse, some do real-time H.264 transcoding so it's all sorts of messed up. Overall it looks like there is a fair bit of improvement, but it's a similar result to what I had a year and a half ago.
No doubt my OSSC configuration isn't optimal, I had it dialled in to get the best picture possible with the old blurry RGB signals, I think Reverse LPF and some other things are unusually high.

When I've got some more time, I'll fiddle with the OSSC and try to get it looking as nice as possible.

With other consoles I can get a really nice, crisp image, even through my OSSC and capture setup (Elgato HD60S, meaning I can only use Line2X). It's only been the SNES that looks really soft.
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

I guess I should throw my work-in-progress mod into the mix:
Spoiler
Image
Although it looks pretty good, there is still overshoot ringing along many edges (some worse than others), and there's really no way around it at the moment. The mod is pretty costly for parts and is difficult to install. It will also never be anywhere near the quality of the digital mod, so I guess it's the worst of both worlds :P I'll also link my long test video here https://www.twitch.tv/videos/645520068 (7 hours of Zelda: ALttP)

This project is mainly for research purposes, and there is still a deeper rabbit hole to go down. I was recently inspired by the 2-piece adapter board + module board Opatus made for the digital mod and I want to steal it for my project :D. The reason is to give the PPU2 chip an isolated, clean power supply. I always ignored everything but the analog DAC section because I thought that the digital portion didn't affect the final output. But now I know that it is constructed directly from the digital pulses, which have a 21 MHz ripple coupled in. I believe that the power source is responsible for the high frequency ripple because I can see the same characteristic patterns when probing VCC around the chip. Instead of hacking bypass capacitors around the chip, I figured its better to take the whole damn thing off and feed it a dedicated, low noise, regulated supply. Hopefully this will get rid of the last bit of CPU/RAM interference in the analog signal.
RetroSwim
Posts: 14
Joined: Tue Oct 15, 2013 4:21 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by RetroSwim »

Yes, I think a replacement DAC will give the best result for sure.

For the low cost and simplicity though, this mod is a good step up!

Edit:

Getting there:

Image

The picture isn't stable, and I'm not sure whether that's the OSSC or the capture card. But this proves that the quality is there!!!

Edit2: Looks like some of the artefacts I was seeing earlier were indeed caused by my distribution amp, going straight into the OSSC gives much better results. Time to find a replacement :(
User avatar
Opatus
Posts: 39
Joined: Mon May 25, 2020 10:28 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Opatus »

Sweet man, I could give 2 shits about HDMI myself, give me good old analog and if I need HDMI I'll pass it through an OSSC for HDMI. Great stuff. Any idea when it will be ready?
I cannot tell when will it be ready. In my mind I already have everything roughly ready. I'm still looking into the FPGA CIC and IGR impementation so no external PICs will be required. So it will problably take at least one month for the new board, as I don't have much time. Btw in the original thread you can find a PVM photo as well.
Although it looks pretty good, there is still overshoot ringing along many edges (some worse than others), and there's really no way around it at the moment. The mod is pretty costly for parts and is difficult to install. It will also never be anywhere near the quality of the digital mod, so I guess it's the worst of both worlds :P I'll also link my long test video here https://www.twitch.tv/videos/645520068 (7 hours of Zelda: ALttP)

This project is mainly for research purposes, and there is still a deeper rabbit hole to go down. I was recently inspired by the 2-piece adapter board + module board Opatus made for the digital mod and I want to steal it for my project :D. The reason is to give the PPU2 chip an isolated, clean power supply. I always ignored everything but the analog DAC section because I thought that the digital portion didn't affect the final output. But now I know that it is constructed directly from the digital pulses, which have a 21 MHz ripple coupled in. I believe that the power source is responsible for the high frequency ripple because I can see the same characteristic patterns when probing VCC around the chip. Instead of hacking bypass capacitors around the chip, I figured its better to take the whole damn thing off and feed it a dedicated, low noise, regulated supply. Hopefully this will get rid of the last bit of CPU/RAM interference in the analog signal.
I think some over/undershoot will remain depending on the colors, some colors are probably off by a few mV but without measuring it is not really noticable. The 21MHz comes from the master clock. Maybe you could try to lift the analog VCC pin and feed that with a clean 5V, it should/could make a difference.
For the low cost and simplicity though, this mod is a good step up!
Yes it gives really good results for the price and ease of installation. This mod started my interest in sharpening the video signal. I did not try this mod myself but I was considering it.
User avatar
maxtherabbit
Posts: 1763
Joined: Mon Mar 05, 2018 4:03 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by maxtherabbit »

yoshiyukiblade wrote:Instead of hacking bypass capacitors around the chip, I figured its better to take the whole damn thing off and feed it a dedicated, low noise, regulated supply. Hopefully this will get rid of the last bit of CPU/RAM interference in the analog signal.
It's really hard to argue with the ease and efficacy of stacking 10uf MLCCs on top of the existing bypass caps for PPU1/2 CPU and WRAM. Removing the PPU from the board just to clean up the power seems like insane levels of overkill to me
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

maxtherabbit wrote:
yoshiyukiblade wrote:Instead of hacking bypass capacitors around the chip, I figured its better to take the whole damn thing off and feed it a dedicated, low noise, regulated supply. Hopefully this will get rid of the last bit of CPU/RAM interference in the analog signal.
It's really hard to argue with the ease and efficacy of stacking 10uf MLCCs on top of the existing bypass caps for PPU1/2 CPU and WRAM. Removing the PPU from the board just to clean up the power seems like insane levels of overkill to me

Whatever he is doing, it appears to be working as his image is the best in this thread so far. I just got my parts to populate the 2 chip mod yesterday, havent had a chance to solder it together yet though.
User avatar
maxtherabbit
Posts: 1763
Joined: Mon Mar 05, 2018 4:03 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by maxtherabbit »

Josh128 wrote:
maxtherabbit wrote:
yoshiyukiblade wrote:Instead of hacking bypass capacitors around the chip, I figured its better to take the whole damn thing off and feed it a dedicated, low noise, regulated supply. Hopefully this will get rid of the last bit of CPU/RAM interference in the analog signal.
It's really hard to argue with the ease and efficacy of stacking 10uf MLCCs on top of the existing bypass caps for PPU1/2 CPU and WRAM. Removing the PPU from the board just to clean up the power seems like insane levels of overkill to me

Whatever he is doing, it appears to be working as his image is the best in this thread so far. I just got my parts to populate the 2 chip mod yesterday, havent had a chance to solder it together yet though.
I was commenting specifically on the power filtering aspect of his post. The efficacy of his video filtering circuit is a separate issue
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

maxtherabbit wrote: I was commenting specifically on the power filtering aspect of his post. The efficacy of his video filtering circuit is a separate issue
I get that, but he feels this will further improve the output video quality and his current results seem quite nice, so I say let him go down the rabbit hole. The 10uf cap additions you are referring to are part of what mod specifically? If not to help clean up the video output, what's it for?
User avatar
maxtherabbit
Posts: 1763
Joined: Mon Mar 05, 2018 4:03 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by maxtherabbit »

Josh128 wrote:
maxtherabbit wrote: I was commenting specifically on the power filtering aspect of his post. The efficacy of his video filtering circuit is a separate issue
I get that, but he feels this will further improve the output video quality and his current results seem quite nice, so I say let him go down the rabbit hole. The 10uf cap additions you are referring to are part of what mod specifically? If not to help clean up the video output, what's it for?
it is to clean up the video, they are not part of any other mod

viewtopic.php?f=6&t=62631
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

Opatus wrote: I think some over/undershoot will remain depending on the colors, some colors are probably off by a few mV but without measuring it is not really noticable. The 21MHz comes from the master clock. Maybe you could try to lift the analog VCC pin and feed that with a clean 5V, it should/could make a difference.
I lifted the AVCC pin and fed it a clean supply, but it had little effect on the waveform. I also lifted the RGB pins too, which had a very big effect on reducing interference, namely vertical bars appearing every 8 pixels. It is still slightly visible in some places if you look for it, and it is very visible if the sampling phase is off too, so it's not fully eliminated.
maxtherabbit wrote: It's really hard to argue with the ease and efficacy of stacking 10uf MLCCs on top of the existing bypass caps for PPU1/2 CPU and WRAM.
It was ineffective in getting rid of this type of interference, but it did help with some other issues like the vertical light bar. I stacked additional ceramic caps well before diving into any of this a couple years ago. The bypass caps are on the opposite side of the board, so there's a lot of travel distance to the VCC pins, increasing parasitic inductance. Doesn't help that it has to pass through one or more vias to get to the top side either.
Removing the PPU from the board just to clean up the power seems like insane levels of overkill to me
No such thing as overkill in the name of research. :mrgreen: I'm not doing this with the intent of making a marketable product as the total cost probably exceeds a 1chip-type console. Heck, the mod board has 75 ohm coax cables running straight off of it so the return path doesn't meander its way back to the board and signal integrity is maximized. This, along with lifting the RGB pins, seems to have had the largest effect in reducing most of the interference.
User avatar
maxtherabbit
Posts: 1763
Joined: Mon Mar 05, 2018 4:03 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by maxtherabbit »

I'm guessing you have one of the 3-chip revs that's not a SHVC-CPU-01?
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

Yep, it's GPM-02 revision board.
RetroSwim
Posts: 14
Joined: Tue Oct 15, 2013 4:21 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by RetroSwim »

I'm writing a quick install guide to be included with my kits.

This is my install:

Image

I notice that while my layout matched captaineos' SNES, buttersoft's SNES is laid out differently, what model mainboard is it? Are there others that should be taken into account?

Thankfully the criss-crossing wires of captaineos' install can be avoided by using the Q3 inputs for Q7 and vice-versa. The three amp channels are identical, so this causes no issuse and makes installation easier.
User avatar
buttersoft
Posts: 383
Joined: Sun Jul 24, 2016 7:49 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by buttersoft »

RetroSwim wrote: I notice that while my layout matched captaineos' SNES, buttersoft's SNES is laid out differently, what model mainboard is it? Are there others that should be taken into account?
IIRC you mentioned PAL somewhere? Captaineos is using a PAL SNES model SNSP-CPU-02. I'm in Aus myself, but was using a cheap SFC ordered from eBay. Apparently the SHVC-CPU-01 SFC i used and NTSC-US SNES of the same model number are identical inside.
User avatar
Voultar
Posts: 550
Joined: Fri Jan 10, 2014 8:29 pm
Location: USA

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Voultar »

This analog filter mod is boosting the rising edges to modify the slew-rate of the DAC.

The problem is that it's non-linear -The quality of the visuals is going to be contingent upon what the PPU is drawing.

Because of this; Some content can look pretty good. Ie: SMW Title Screen. Whereas other video content Ie: A scene with fast or quick video transistions. Is going to have a lot of overshooting, ringing artifacts.

The OSSC tries to address this with a permutation of this filter in software. The results are mixed, and most people leave it off for a variety of games.



To summarize: The results can be very unpredictable, and can do more harm than good, in some cases.

I would wait for the digital stuff to come along. Definitely wouldn't put this into kit form in this state.
User avatar
maxtherabbit
Posts: 1763
Joined: Mon Mar 05, 2018 4:03 pm

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by maxtherabbit »

Voultar wrote:This analog filter mod is boosting the rising edges to modify the slew-rate of the DAC.

The problem is that it's non-linear -The quality of the visuals is going to be contingent upon what the PPU is drawing.

Because of this; Some content can look pretty good. Ie: SMW Title Screen. Whereas other video content Ie: A scene with fast or quick video transistions. Is going to have a lot of overshooting, ringing artifacts.

The OSSC tries to address this with a permutation of this filter in software. The results are mixed, and most people leave it off for a variety of games.



To summarize: The results can be very unpredictable, and can do more harm than good, in some cases.

I would wait for the digital stuff to come along. Definitely wouldn't put this into kit form in this state.
I leave the OSSC filter set to an intensity of 3 for this reason. Seems to be a good compromise on all content with an SHVC-CPU-01 being sampled optimally
User avatar
buttersoft
Posts: 383
Joined: Sun Jul 24, 2016 7:49 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by buttersoft »

Voultar wrote:This analog filter mod is boosting the rising edges to modify the slew-rate of the DAC.

The problem is that it's non-linear -The quality of the visuals is going to be contingent upon what the PPU is drawing.

Because of this; Some content can look pretty good. Ie: SMW Title Screen. Whereas other video content Ie: A scene with fast or quick video transistions. Is going to have a lot of overshooting, ringing artifacts.
Thank you for the information!

So, something like the turtles intros, where the city is streaking past, is going to be affected? (unless i'm thinking of the arcade intro there, but you get the image i mean). What would the effect be? And is it going to be as noticeable on a CRT?
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

I think he's referring to fast voltage transitions. In other words, pixel transitions that look fine without the RLPF will end up being overdriven with the RLPF, causing it to overshoot its mark. This is where we see ringing/overshoot in both the positive and negative direction. Quick visual example of the raw alternating black/white pixel transitions here:

Image

The white-to-black transition is actually very sharp and doesn't require filtering, but the black-to-white transition could use some help. It is so slow that it doesn't even reach full brightness before shooting back down to black. Compensating for the slow transition will overcompensate the fast one, causing the overshoot. Furthermore the black-to-white transition isn't even the slowest there is; it's the black-to-grey or white-to-grey transitions. So compensating for the worst possible case will overdrive everything else.
User avatar
buttersoft
Posts: 383
Joined: Sun Jul 24, 2016 7:49 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by buttersoft »

Yeah, cheers for that. I could feel the mod was doing... something, and this might be it. Blacker than black is not a problem though, or maybe any lowlight stuff. But would any 1/16th fall in brightness of a particular colour, say, ring to the same amplitude? I'm guessing it's dependent on how far the fall is?

I'm also curious about some other points. How big is the ringing? What is the time index on your scope relative to drawing pixels? So, in your graph, is that a single white pixel followed by part of a black one? (i realise i could probably work this out, but am admittedly being lazy and asking). And if it is a single white pixel, and the circuit overcompensates, then surely the ringing is going to damp fairly fast, leaving an average value for each pixel round about the right level? Which, given phosphors tend to smear such imperfections, isn't really going to be visible on a CRT at SD resolutions? I grant that digital upscaling is going to be interesting, but surely not insurmountable with the right filter (which may not at present exist).

Apologies if i'm coming across as ignorant or abrupt, just trying to make sure i understand the situation we're talking about :)

A digital solution with a DAC is probably going to be better, but this mod does work pretty well, despite the above. At least on CRT it does, though it may actually be making images a bit brighter than the designers intended along with much, much clearer.
User avatar
Josh128
Posts: 2123
Joined: Thu Jan 16, 2014 9:01 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by Josh128 »

A digital solution with a DAC is probably going to be better, but this mod does work pretty well, despite the above. At least on CRT it does, though it may actually be making images a bit brighter than the designers intended along with much, much clearer.
Well, you have actually used the mod ( I havent installed mine yet )-- have you noticed any negatives from it? They are saying white text on a black background might have problems? Whats your honest opinion?

As far as the problem itself, I find that white text on deep blue background is one of the worst culprits for seeing the bad IQ of the original 2 chip SHVC models.
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

buttersoft wrote:Yeah, cheers for that. I could feel the mod was doing... something, and this might be it. Blacker than black is not a problem though, or maybe any lowlight stuff. But would any 1/16th fall in brightness of a particular colour, say, ring to the same amplitude? I'm guessing it's dependent on how far the fall is?

I'm also curious about some other points. How big is the ringing? What is the time index on your scope relative to drawing pixels? So, in your graph, is that a single white pixel followed by part of a black one? (i realise i could probably work this out, but am admittedly being lazy and asking). And if it is a single white pixel, and the circuit overcompensates, then surely the ringing is going to damp fairly fast, leaving an average value for each pixel round about the right level? Which, given phosphors tend to smear such imperfections, isn't really going to be visible on a CRT at SD resolutions? I grant that digital upscaling is going to be interesting, but surely not insurmountable with the right filter (which may not at present exist).

Apologies if i'm coming across as ignorant or abrupt, just trying to make sure i understand the situation we're talking about :)

A digital solution with a DAC is probably going to be better, but this mod does work pretty well, despite the above. At least on CRT it does, though it may actually be making images a bit brighter than the designers intended along with much, much clearer.
Yes, that is a scope plot of a black-white-black pixel transition (to be more precise, it's a red pixel transition since it's only one of the RGB channels, but they all look roughly identical). Overshoot below the black level can actually cause problems with black level clamp (at least I think it did on the OSSC), especially if its nowhere near the op amp rails.

There are a couple things I can think of that can limit overshoot: the dynamic range of the op amp, and diodes in the feedback path. The original Japanese mod kind of had a mix of both. There was no DC offset before the op amp other than ~0.6V from the PNP buffer, so it probably saturates/distorts the lower range of the signal. True rail-to-rail op amps are very uncommon, and the op amp chosen swings within 0.8V of the ground rail at the output. As such, the diodes limit overshoot mainly in the positive direction. Otherwise, it will probably shoot up to the upper limit of the op amp, which could be a couple volts beyond where white should be. That is a huge amount of overshoot when you consider that the full range of unloaded video is just 1.4 Vpp. This should give you a rough idea on how strong the filtering has to be to take care of the worst case blur. Even the slow black-to-white transition will end up overshooting by some volts. Overshoot recovery is usually more than 1 pixel long (over 186 ns) in the worst cases, which is why we see ringing.
User avatar
buttersoft
Posts: 383
Joined: Sun Jul 24, 2016 7:49 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by buttersoft »

Josh128 wrote:Well, you have actually used the mod ( I havent installed mine yet )-- have you noticed any negatives from it?
Yes, it's it messing with the brightness of colours a bit - i did kinda say that from the start. Overall the result is much sharper than before, and i like the image much better, but part of that might be the increased overall brightness (or just increased contrast on sudden transitions from bright to dark and vice versa)

A digital mod is going to be superior, but whether massively so or even noticeably so remains to be seen.
yoshiyukiblade wrote: Yes, that is a scope plot of a black-white-black pixel transition (to be more precise, it's a red pixel transition since it's only one of the RGB channels, but they all look roughly identical). Overshoot below the black level can actually cause problems with black level clamp (at least I think it did on the OSSC), especially if its nowhere near the op amp rails.

There are a couple things I can think of that can limit overshoot: the dynamic range of the op amp, and diodes in the feedback path. The original Japanese mod kind of had a mix of both. There was no DC offset before the op amp other than ~0.6V from the PNP buffer, so it probably saturates/distorts the lower range of the signal. True rail-to-rail op amps are very uncommon, and the op amp chosen swings within 0.8V of the ground rail at the output. As such, the diodes limit overshoot mainly in the positive direction. Otherwise, it will probably shoot up to the upper limit of the op amp, which could be a couple volts beyond where white should be. That is a huge amount of overshoot when you consider that the full range of unloaded video is just 1.4 Vpp. This should give you a rough idea on how strong the filtering has to be to take care of the worst case blur. Even the slow black-to-white transition will end up overshooting by some volts. Overshoot recovery is usually more than 1 pixel long (over 186 ns) in the worst cases, which is why we see ringing.
Hmmm, i'd be interested to see how the modded video output is actually responding. (Your grab was from an unmodded SNES/SFC, right?) You're saying the diodes are limiting overshoot on increasing brightness, and the fall on your graph is much faster anyway - is the ringing faster too, one that end? I don't think i'm having any problems around black level, but i am using a CRT, and my understanding was that the back porch was typically where a CRT would grab black level. That said, I'll throw up some more test patterns and see what i can see.
Last edited by buttersoft on Mon Jul 27, 2020 1:50 am, edited 1 time in total.
yoshiyukiblade
Posts: 63
Joined: Fri Sep 27, 2019 3:47 am

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by yoshiyukiblade »

The scope plot is from an unmodded video signal, but the console itself had more bypass capacitors and regulator changed. The diodes limit overshoot in both directions, but the negative direction is also limited by the op amp linear range. Overshoot recovery can be fast, depending on the circuit specifics and op amps. I'm also curious to see scope plots of the Japanese mod for these details. I didn't investigate it much because, at the time, I was just starting to learn how to use a scope and wanted to try my hand at making a mod. I did notice black level and other distortion issues through OSSC capture tests. A study of the datasheets may explain the distortions more clearly.
User avatar
unmaker
Posts: 139
Joined: Sat Jan 02, 2016 7:27 am
Contact:

Re: An actual SNES mod to sharpen the pixels of 2-CHIP conso

Post by unmaker »

I gave this mod a shot to see how it looked for myself. I followed buttersoft's guide and installed this into a SHVC-CPU-01:

Image

There was a very noticeable reduction in blur. I then installed the 270 pF capacitors and looked for any change:

Image

There was no change at all with the 270 pF capacitors installed onto the transistors from what I could tell. The output did not look better or worse. I don't think it's true these caps degrade the picture quality and I'm not sure what purpose they serve. They are a part of Torapu's original mod so I left them on.

Although the blur was greatly reduced, the output was way too bright. So I went a step further and installed borti4938's RGB bypass mod:

https://github.com/borti4938/SNES_RGB_B ... 1-reported

Image


With the brightness attenuated things looked so much better. Borti's mod also appeared to help with blur reduction but this could be because it was now less bright. Hard for me to tell here.

Although the picture quality was greatly improved, it was definitely not perfect. I was having serious noise issues when using it with the OSSC. I would recommend to anyone who has tried this mod to check for noise through the OSSC with FF3/FF6 intro. I had jailbars with generic settings and wavy noise with optimal timings. So I did some more work on the console:

-I replaced C90-C95 with 10uF X7R ceramics (you could also install these on top instead of replace and that is actually better I heard. C92, C94 were already replaced as a part of the mod. I got this idea from maxtherabbit's post here)
-C81 is a 1 uF capacitor at the voltage regulator's output. I replaced this with a 22 uF ceramic (idea taken from RetroRGB's website)
-replaced the voltage regulator with a 78S05 2A
-installed a 1000 uF capacitor at C67

With this additional work done the wavy noise when using optimal timings was now gone. This was a huge improvement. There are still faint jailbars when using generic settings but I don't care about generic settings personally. It also looks great on my PVM. Here is a capture using the OSSC to my Datapath E1S to give an idea of how things look now:

https://streamable.com/lzph9r

Image

Overall I'm very pleased with the results but things are still not perfect. There is some noise that I think is a result of overshoot or ringing which Voultar mentioned in his post. I'm wondering if this mod can be further improved upon or if this is the best that it gets without resorting to say the digital RGB mod that Opatus is working on.

Big thanks to buttersoft for making this PCB!
Post Reply