The default 8580 filter sounds off
Versatile Commodore Emulator
Brought to you by:
blackystardust,
gpz
Menu
▾
▴
#2004 The default 8580 filter sounds off
The default 8580 filter bias is 0 mV but heavily filtered tunes sound much closer to real 8580 recordings if set to -2mV or so. For example listen to /MUSICIANS/L/Linus/64_Forever.sid at 0 & -2 mV and compare to the C64 recordings:
http://oakvalley.textfiles.com/oakvalley/soasc/hvsc/062/MP3/MUSICIANS/L/Linus/64_Forever_T001.sid_CSG8580R5.mp3
https://sourceforge.net/p/vice-emu/bugs/_discuss/thread/b758ddbb54/783b/attachment/64_Forever.flac
Another fine example of 8580 filters being off in VICE is the bass in /MUSICIANS/D/Daf/Opium_tune_2.sid (0:02): http://81.166.255.142/files/soasc/hvsc/049/MP3/MUSICIANS/D/Daf/Opium_2_T001.sid_CSG8580R5.mp3
Do you have any 8580 chip that sounds closer to the bias of 0mV when measured? There's some variance in how different 8580 chips sound, unfortunately.
Since it is mostly Leandro who is looking into reSID these days, it would be good if you'd made a ticket here as well: https://github.com/libsidplayfp/libsidplayfp/issues
I don't think changing the default based on a single recording is a good idea.
And it would be off for NTSC anyway.
Can we close this?
No, https://csdb.dk/release/?id=76056 with resid-fp v0.16 still sounds much closer to real 8580 recordings than VICE
As said, make a ticket on https://github.com/libsidplayfp/libsidplayfp/issues - there is no resid development going on in the VICE repo anymore
I had issues with Github asking for a phone verification, if you can copy the report on my behalf then it'd be great but Leandro responded above in this ticket so I think he knows about the bug anyway.
I cant make a report for you, i cant describe what the problem would be, nor would i be able to answer any questions.
Leandro already answered indeed, but that doesnt mean he is aware of any bug - it has yet to be determined if there even is one (comparing with soasc by ear isnt very useful in most cases)
What's useful for determining if the 8580 filter is emulated correctly if comparing the emulated sound to recordings from real hardware isn't the right approach?
There are two real 8580 recordings of the beginning of /MUSICIANS/L/Linus/64_Forever.sid , the filter in the intro is clearly much weaker in VICE and both real C64 recordings closely match the output of https://csdb.dk/release/?id=760561 . There's something that the 16 year old resid-fp v0.16 does differently that makes its output closer to a real 8580 than resid in VICE 3.9
VICE 2.4 still had resid-fp, i wonder if that one did a good job too?
It did.
For debugging its important to have good reproducable test cases. We don't even know how exactly this soasc recording was created, and we don't know if different chips wont sound different (there differences between 8580 too). Ideally we have a test case that only plays a single note, on a single channel - and then record it on different real C64s.
In any case, however, Leandros repo is the place to report this, thats where the development happens.
resid is planned for removal in next major version, residfp is mature enough to supersede it.
Anyway, the parameter is configurable exactly for this reason, if you don't like how it sound just change it. Maybe if there are enough people preferring a different value or someone with a solid proof the default can be updated.
Also check NTSC tunes, the 8580 filter depends on CPU clock.
The thing here is, to me it seems resid also sounds fine in VICE 2.4.
Last edit: Querino 2025-11-23
Oooh. Damn. Guess we have some work to do in VICE then as well.