The reason it's not possible to RGB mod an NES without replacing the PPU is because the chip outputs composite NTSC directly. There is no place on the board that has a raw RGB signal. What if custom circuitry were made to "reverse" the PPU's encoding? Conceptually, it might be possible to analyze the composite output in real time to determine what colors the PPU was trying to display, and re-generate RGB output based on that. If information isn't lost during the composite encoding process, this should be possible. By information loss in the NES PPU's case, I mean there existing two distinct PPU attempted outputs that have electrically indistinguishable composite output streams. If this is possible, then it would be possible to make a converter box you could attach to an NES to generate SCART and/or YPbPr output. I'm entirely theorizing here, because I'm not an electrical engineer; I'm a software reverse engineer. =^_^=
Oh, and while it's at it, such a device could fix the faint vertical line problem with NES2 and Famicom2. This would make Famicom2 by far the best device in the series - able to have external sound games, disk system, American and European controllers, and Japanese custom controllers. Only thing it couldn't use is the microphone, though tearing up a broken Famicom1 to get its controllers would work.
Well, you could build a device that would decode the signal, make a table out of the decoded pixels and then display the video via vga. With some simple digital signal processing you could remove most of the interference on screen.
This be madness talk. Just be happy we don't have to live with RFU like in the olden days thanks to composite...
What you are describing is basically what happens in your TV, the encoded Signal is decoded back to RGB, because all CRT tubes work with RGB. Sadly the quality loss is in part pretty much irreversible. Composite is analogue and the interferences between all the different parts of the signal will have already degraded the quality. Some TVs with better filtering ( such as 3d comb filter) will do a better job of Decoding composite, which will result in a better image but you simply cannot resurrect the original RGB quality while working with Composite. At least not without some pretty marvelous engineering, ( or advanced digital processing as Xmog pointed. ) If you need RGB for the sake of connectivity, there is a French version of the NES which has a So called RGB output. But it gets it from decoding the Composite the way you suggested. From what i read about it the image quality is no better than a regular NES. Im no expert but i hope what im saying makes sense. Otherwise hopefully someone will add more details and correct if im wrong.
This. If you want better quality output you need to accept the messed up colors from an Arcade or Sharp Titler RGB PPU or reverse engineer the PPU and make your own replacement that outputs RGB. Or just accept Composite Video. It really doesn't look that bad! Last alternative, emulate it.
It's most likely possible (barring insurmountable low pass filtering), but it probably won't be happening any time soon. Why this kind of question here? Seems more appropriate for NESdev (it's actually been talked about before). Basically you first need a narrow flash ADC properly biased for the video. The ADC has to be very fast because the composite generator shifts at both edges of the 21 MHz clock. Once the levels are cleanly detected logic can be constructed to isolate colorburst and lock to it/multiply and deskew it with a PLL/delay loop in order to approximately recover the master clock. Since all that is really difficult someone'd sooner recover the phase forcefully with ridiculously fast logic running asynchronous to the pixel/subpixels. Once the phase is recoverable the pixels can be decoded however sloppily then crossed into a separate clean 21 MHz clock domain with its own local oscillator for processing and output.
