[regression-merge] Decoder: Loudness difference ivas-float-update vs. float-pc for MC BINAURAL_ROOM_IR + Headrotation output

Basic info

• Float reference (ivas-float-update):
  • Encoder (float): c2135116
  • Decoder (float): c2135116
• Branch float-pc:
  • Encoder (fixed): n/a
  • Decoder (fixed): aa9a6370

Bug description

When decoding the attached bitstream with head-rotation, there is a significant loudness difference between the floating-point decoders in ivas-float-update and float-pc: The decoder from float-pc is significantly louder. The fixed-point decoders (main, main-pc) are loudness-wise inline with ivas-float-update. Also, when we switch off head-rotation, this loudness increment disappears again.

Ways to reproduce

IVAS_dec -t scripts/testv/headrot_case00_3000_q.csv BINAURAL_ROOM_IR 48 stv51MC48c.wav_Multi_channel_5_1_at_64_kbps_48kHz_in_48kHz_out_BINAURAL_ROOM_IR_out_HR.192 out.wav

Bitstream: stv51MC48c.wav_Multi_channel_5_1_at_64_kbps_48kHz_in_48kHz_out_BINAURAL_ROOM_IR_out_HR.zip