[feature] sound interpolation options (specifically 'off')

[miteluode]

as in "retain sound interpolation options (specifically off) in v2.0 with the new core if that is at all reasonable". because oh right, forgot to remake this post. please see here first.

right, so... blah blah interpolation sucks ass blah blah zsnes lets you screw with this other stuff anyway blah blah real crts don't allow even native res blah blah options are cool blah blah i don't remember what i originally said aaaa

but yes actually i'm curious as to what that so-called 'no' interpolation is... surely it's simply a 'clearer' scheme itself, like the other schemes?

[grinvader]

That was to squeeze more speed out of a slow game, but seriously nowadays there's absolutely no reason to leave that stuff in.
Some samples are supposed to be left not-interpolated (method 0), but the others have to, else it doesn't even sound close to the original.

There's also the issue of mixing the low-level interpolation (needed for proper sample decoding) and sound output interpolation (resample to fit your sound hardware). From what I gathered, we currently apply the one choice to both routines, which doesn't fit the right bill.

The next ver will have proper core code (the delicious blargg stuff) and a separate output resampler. However the plan is to leave some hooks in the core for other people to add video/audio filters, so killing various low-level steps should be feasible.

[joe_devore]

That was to squeeze more speed out of a slow game, but seriously nowadays there's absolutely no reason to leave that stuff in.
Some samples are supposed to be left not-interpolated (method 0), but the others have to, else it doesn't even sound close to the original.

There's also the issue of mixing the low-level interpolation (needed for proper sample decoding) and sound output interpolation (resample to fit your sound hardware). From what I gathered, we currently apply the one choice to both routines, which doesn't fit the right bill.

The next ver will have proper core code (the delicious blargg stuff) and a separate output resampler. However the plan is to leave some hooks in the core for other people to add video/audio filters, so killing various low-level steps should be feasible.

GREAT NEWS everyone! ^_^ loi (thx for the info GrinVader)

[kode54]

I think they're doddering about interpolated resampling. The actual hardware employs some Gaussian like bell curve to resample from whatever arbitrary rates to the ~32KHz of the native output rate. Using zero order hold or nearest neighbor resampling sounds like crap, and is what the original releases did when they produced sound. To include that into future releases would be silly, but some people seem to want that nonsense, which is why I ended up adding it to foo_gep for SPC and SFM playback.