Skip to content

Audio capture noisy with WM8960 SoC codec #4384

Closed

Description

@tewarid

I'm using the WM8960 SoC codec as part of a custom kernel build at version 5.10.31.

Audio playback works with Seeed's ReSpeaker 2-Mics Pi HAT after enabling a few switches that are off, and adjusting the volume

amixer -c1 sset 'Left Input Mixer Boost' toggle
amixer -c1 sset 'Left Output Mixer PCM' toggle
amixer -c1 sset 'Right Input Mixer Boost' toggle
amixer -c1 sset 'Right Output Mixer PCM' toggle
amixer -c1 sset 'Headphone',0 100%,100%

Sound capture however is very noisy, see test.wav.zip produced with

arecord -f cd -Dhw:1 -d5 > test.wav

Seeed's soundcard driver works without issue.

Found no meaningful differences in alsa settings. See asound.state.txt vs seeed-asound.state.txt.

Activity

  1. added a commit that references this issue on Aug 17, 2021
  2. tewarid commented on Feb 25, 2022

    @tewarid
    Author

    Cannot reproduce issue with kernel version 5.15.24.

  3. strix-technica commented on Mar 28, 2025

    @strix-technica

    I know this is a rather old Issue and my hardware is a little different but I noticed, when closely examining the schematic for the Adafruit Voice Bonnet and the WM8960's datasheet, that the codec has three pairs of inputs two of which are floating and will be prone to noise pick-up, depending on ambient conditions.

    As of this writing, the overlays in the Seeed and Waveshare repositories reference these floating inputs in routing:

                    simple-audio-card,routing =
                            "Headphone Jack", "HP_L",
                            "Headphone Jack", "HP_R",
                            "Speaker", "SPK_LP",
                            "Speaker", "SPK_LN",
                            "LINPUT1", "Mic Jack",
                            "LINPUT3", "Mic Jack",
                            "RINPUT1", "Mic Jack",
                            "RINPUT2", "Mic Jack";
    

    If this has the effect of enabling those inputs, this could be one possible cause of the noise mentioned in this Issue. I haven't messed with device trees before and exactly how that line will be interpreted will anyway depend on the respective version of the WM8960 driver, but it seems worth flagging anyway.

    In due course, I will need to modify the DT I'm using for my own purposes. I will update if I find anything relevant in the course of doing that.

  4. meiser79 commented on Sep 21, 2026

    @meiser79

    Hi,

    this overlay needs an update, see jxlarrea/kiosk-satellite#563 (comment)

    I was able to improve silence-to-speech SNR from ~1.5dB (original, unpatched) to ~8.7dB.

    Regards,
    meiser79

  5. pelwell commented on Sep 21, 2026

    @pelwell
    Contributor

    Why not submit a Pull Request?

  6. meiser79 commented on Sep 21, 2026

    @meiser79

    @pelwell I'd like to get one more confirmation before merging it.

  7. pelwell commented on Sep 21, 2026

    @pelwell
    Contributor

    Of course, in your own time.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions