BASOP encoder for 80kbps Stereo does not trigger dtx when it should

Basic info

  • Float reference: 5067d70d
  • Fixed point: b4a8a48f

Bug description

The basop encoder in stereo mode at 80kbps does not trigger dtx for this testcase while the float reference encoder does. In the below spectrograms, one can easily spot the dtx portions via the band limitation in the float -> float output.

Float enc -> float dec output:

Screenshot_2025-02-21_at_11.28.15

FX enc -> float dec output:

Screenshot_2025-02-21_at_11.29.09

Also, when running the fx encoder without -dtx, the result is bitexact to the version with -dtx.

Ways to reproduce

Box folder: ...\Box_EXTERNAL_IVAS_BASOP_VERIFICATION\issues\issue-1311

# flt version
./IVAS_cod_flt_ref -q -dtx -stereo -max_band FB 80000 48 stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated_48.wav out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-flt].192
./IVAS_dec_flt_ref -q stereo 48 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-flt].192 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-flt].dec[clean-48kHz->stereo-flt].wav

# fx enc version with dtx
./IVAS_cod_fx -q -dtx -stereo -max_band FB 80000 48 stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated_48.wav out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-fx].192
./IVAS_dec_flt_ref -q stereo 48 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-fx].192 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-dtx-fx].dec[clean-48kHz->stereo-flt].wav

# fx enc version without dtx
./IVAS_cod_fx -q -stereo -max_band FB 80000 48 stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated_48.wav out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-fx].192
./IVAS_dec_flt_ref -q stereo 48 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-fx].192 out/stereo-noisy-speech.stereo-noisy_speech-p800-2-cat1-lab_b-20ms_aligned_files_concatenated[0dB].enc[48kHz-FB-80000-fx].dec[clean-48kHz->stereo-flt].wav
Edited Feb 21, 2025 by Jan Kiene
Assignee Loading
Time tracking Loading