Clicks in MC5.1 output at 160kbps with FER
Basic info
Bug description
There are several clicks for passthrough MCT-MC 5.1 decoding at 160kbps. The problem can most easily be seen by comparing the STEREO decoder output. This causes a high MLD of 8.34.
There is also an entire signal segment between 5.5 - 6.5s where the concealment goes wrong and a tone is produced, this seems like a bug and not a numerical difference.
Diff signal for STEREO output (replace 5_1 in decoder commandline below):
Ways to reproduce
Bitstream has been uploaded.
./IVAS_cod_flt_ref -mc 5_1 -max_band FB 160000 48 experiments/basop-verification/MC-5_1/proc_input/bm2aa1s02.wav out/mc-5_1--bm2aa1s02--160000.192
eid-xor -vbr -fer out/mc-5_1--bm2aa1s02--160000.192 error_patterns/mc-5_1--bm2aa1s02--160000_ep.192 out/mc-5_1--bm2aa1s02--160000.fer.192
./IVAS_dec_fx 5_1 48 out/mc-5_1--bm2aa1s02--160000.fer.192 out/mc-5_1--bm2aa1s02--160000.fer.dec.5_1.fx.wav
./IVAS_dec_flt_ref 5_1 48 out/mc-5_1--bm2aa1s02--160000.fer.192 out/mc-5_1--bm2aa1s02--160000.fer.dec.5_1.flt.wav
Edited by Archit Tamarapu