A Moogle  🏳️‍🌈 💻 🎮 🎶 🇸🇪

@furrtek @plgDavid Yes, it probably would help, but my suspicion is that it would be at the cost of frequency response. What you more likely want is a notch filter for the frequency which is intruding on the final video signal.

But how would the notch filter distinguish between the parasitic signal and graphics that produce the frequency ?

Sorry, I only saw the initial screenshot in the thread, not the subsequent tweets. That said: The jailbars seem to be about 1 cycle per tile. That's far lower than any signal that'd be getting thrown around the Famicom's internal buses. Tuned properly, it could work.

I was thinking more along the lines of something that could be connected in-line with the Famicom's video output in order to mute the jailbar effect. Given the NTSC color burst frequency, notching out a frequency of 1 cycle per tile shouldn't affect the eventual output.

From what I can tell, the SRAM seems to be parasitically pulling a couple milliamps off of the the bus, which leads to cycles of voltage dropping (increased luma) across the raster, one per nametable square at the beginning - corresponding with nametable CIRAM fetch by the PPU.

Perhaps the most noninvasive strategy involving the output signal would be to introduce a comb filter instead of a notch filter, tuned in such a way to repeat the low-frequency luma component in the pre-bar zone in the raster but not the high-frequency chroma subcarrier.

I believe in fixing the issue at the source, by providing the ICs all the current they need even when they pump like crazy using carefully chosen capacitor values and placement?

Same opinion. I don't get how a filter on the final output would hide the problem without affecting the useful signal in some way or the other. What if a game wants to display something similar to the jailbars ?