Skip to content

Decoder: High MLD for ParamISM3, 24.4 kbps, DTX on

Basic info

Bug description

High MLD value (> 30) for 2nd channel in ParamISM 24.4 kbps, DTX on, selection test item bm6aa1s02:

wav-diff --no-fail --print-ctest-measurement out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.flt.wav out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.fx.wav
wav files differ
  ref: out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.flt.wav
  cut: out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.fx.wav
channel 0 is different:
  non BE frames: 446 (44.6%)
  first diff: frame 0, sample 437
  MLD: avg: 0.1, max: 1.5 @10.064s, louder: 33.5%, quieter: 4.9%
channel 1 is different:
  non BE frames: 416 (41.6%)
  first diff: frame 143, sample 126
  MLD: avg: 0.1, max: 30.1 @9.240s, louder: 32.3%, quieter: 5.1%
channel 2 is different:
  non BE frames: 922 (92.2%)
  first diff: frame 0, sample 322
  MLD: avg: 0.1, max: 1.5 @10.064s, louder: 68.7%, quieter: 15.7%
<CTestMeasurement type="numeric/double" name="MLD">30.076202</CTestMeasurement>

Float: Bildschirmfoto_2024-10-16_um_16.40.52

BASOP: Bildschirmfoto_2024-10-16_um_16.41.00

There is an audible "plop" in the 2nd channel at approx. 9.2s. In addition, the noise level seems significant higher in BASOP.

Ways to reproduce

  • ./IVAS_cod_flt_ref -dtx -ism 3 ./experiments/basop-verification/paramISM3/proc_input/bm6aa1s02.wav.0.csv ./experiments/basop-verification/paramISM3/proc_input/bm6aa1s02.wav.1.csv ./experiments/basop-verification/paramISM3/proc_input/bm6aa1s02.wav.2.csv -max_band FB 24400 48 ./experiments/basop-verification/paramISM3/proc_input/bm6aa1s02.wav out/paramISM3--bm6aa1s02--24400--dtx.192
  • ./IVAS_dec_fx EXT 48 out/paramISM3--bm6aa1s02--24400--dtx.192 out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.fx.wa
  • ./IVAS_dec_flt_ref EXT 48 out/paramISM3--bm6aa1s02--24400--dtx.192 out/paramISM3--bm6aa1s02--24400--dtx.clean.dec.flt.wav

Items can be shared upon request.

Edited by multrus