Menu

#2116 Quantization errors

v3.x
open-need-info
nobody
None
GTK3
Sound
2025-08-06
2024-12-29
DJ_Dancer
No

It has been fixed,
https://github.com/libsidplayfp/libsidplayfp/issues/156

However, it is still clearly audible in Vice. Or is it something else?

Related

Patches: #396

Discussion

1 2 > >> (Page 1 of 2)
  • gpz

    gpz - 2024-12-29

    No idea what you refer to to. That bug report also seems to refer to residFP - which isnt part of VICE

     
  • DJ_Dancer

    DJ_Dancer - 2024-12-29

    Doesn't the code originate directly from libsidplayfp?

    Please refer to the examples in the link above, which clearly illustrate the issue.

     
  • Leandro Nini

    Leandro Nini - 2024-12-30

    There was quantization noise due to float to integer conversion, as mentioned here:
    https://github.com/libsidplayfp/libsidplayfp/issues/156#issuecomment-2408899548

    But the issue in resid must be somewhere else, the code is quite different. The attached test by OP should show it, if you crank the volume very loud.

     
  • DJ_Dancer

    DJ_Dancer - 2025-01-01

    Compared to earlier versions of Vice, the effect appears to have worsened somewhat, becoming noticeably louder.

     
    • Leandro Nini

      Leandro Nini - 2025-01-04

      If you can find the exact revision where things got worse it would help narrowing down the issue.

       
  • DJ_Dancer

    DJ_Dancer - 2025-01-11

    Vice 3.5 OK, Sounds fine
    Vice 3.6 BAD

    So the issue must have been introduced between 24/12-2020 - 24/12-2021

     

    Last edit: DJ_Dancer 2025-01-11
  • gpz

    gpz - 2025-01-11

    That seems weird - did you use the exact same settings? The only change affecting reSID output in that time period was https://sourceforge.net/p/vice-emu/code/39905/ - which should not affect this

     
    • DJ_Dancer

      DJ_Dancer - 2025-01-11

      I briefly tested with 48kHz, resampling, 8580 SID.

      Could it possibly be due to the filter8580new?

      Notably, in the Denise emulator, you can choose between the latest reSID filter, the old 2.4 filter m, and Chamberlin. When selecting one of the latter two, the "quantization" noise disappears.

       

      Last edit: DJ_Dancer 2025-01-11
      • Leandro Nini

        Leandro Nini - 2025-01-12

        It likely is, https://sourceforge.net/p/vice-emu/code/40418/
        Does it only affect the 8580?

         
        • DJ_Dancer

          DJ_Dancer - 2025-01-12

          On the 6581 chip, it appears only faintly. You need to listen very carefully to notice it.

           
  • Querino

    Querino - 2025-01-11

    i can hear a difference after https://sourceforge.net/p/vice-emu/code/45158/

    but i guess that's not surprising?

     
    • Leandro Nini

      Leandro Nini - 2025-03-02

      In fact the aliasing noise was there since before that but the change amplified it, more so for the 8580.
      The problem lies in the filter when the input voices are scaled, adding some rough dithering noise saves the day.

       
  • DJ_Dancer

    DJ_Dancer - 2025-01-27

    Any news on whether the whole thing can be solved easily?
    Unfortunately, Vice reSID cannot be used for composing and recording in this state, even if the general emulation is otherwise first class.

     
  • gpz

    gpz - 2025-01-27

    I am unable to hear any problem even when cranking up the volume so much that my neighbours complain :)

    You should probably make a ticket on Leandros github repository - as that is where the reSID development is happening these days

     
    • DJ_Dancer

      DJ_Dancer - 2025-01-28

      Listen again :)
      Completely noisy. Recorded in Vice 3.9. Introduced between Vice 3.5 - Vice 3.6

       

      Last edit: DJ_Dancer 2025-01-28
  • gpz

    gpz - 2025-01-28

    Still no idea what the problem is, sorry. Sounds like a random SID kickdrum to me. (How am i supposed to know whats wrong anyway? - We'll need a reference recorded from a real SID. Ideally from several different ones :))

     
    • DJ_Dancer

      DJ_Dancer - 2025-01-28

      I currently have no way to record and am looking for volunteers.

      But don’t you hear the noisy, distorted SID output? My 8580 SID has never sounded like this—it’s usually crystal clear. The kick drum has strange background noise (buzzing overtones), which last longer than the actual sound. And even the filtered sounds have artifacts.

       

      Last edit: DJ_Dancer 2025-01-28
  • gpz

    gpz - 2025-01-28

    Sounds just fine to me. As said, without a reference from the real thing its very much impossible to tell if this is wrong or not, and of so how

     
    • Ingo Korb

      Ingo Korb - 2025-01-29

      Recorded on a 250469 with real 8580. Not sure if the very faint falling noises after the initial kick sounds from the VICE recording are present, the background noise on a real machine is too loud.

       
      • DJ_Dancer

        DJ_Dancer - 2025-01-29

        Thank you. So far, the test file on the real SID behaves as expected. The filter and sound are crystal clear, with no distortion/noise from the emulation. Naturally, the SID's characteristic background noise is present on the real chip.

        Regarding test files, it is important to listen with headphones if possible and as loud as possible to hear the anomalies in the reSID.

         

        Last edit: DJ_Dancer 2025-01-29
  • gpz

    gpz - 2025-03-04

    try r45514

     
    • DJ_Dancer

      DJ_Dancer - 2025-03-04

      Does not seem to be fixed in this build. Noise is still audible.

       
  • Leandro Nini

    Leandro Nini - 2025-03-09

    Well, some aliasing is still there but it's less than before.
    Amplified recordings attached for reference.

     
    • DJ_Dancer

      DJ_Dancer - 2025-03-11

      Ah, that's why I couldn't tell the difference
      reSID-Fp is the cleanest. Too bad that code is not used in the emulators.

       
1 2 > >> (Page 1 of 2)

Log in to post a comment.

MongoDB Logo MongoDB