FD BWE issues

Several ToDo comments that point out to a missing tuning or suboptimality in the FD BWE exist in the code:

• ..\lib_dec\swb_bwe_dec.c(270): /* IVAS_fmToDo: wtda() does not support L_FRAME length; thus temporarily resample the signal */
• ..\lib_dec\swb_bwe_dec.c(271): /* IVAS_fmToDo: delay output[] by 1.25ms ? */
• ..\lib_dec\swb_bwe_dec.c(589): /* todo - wtda() does not support L_FRAME length; thus temporarily resample the signal */
• ..\lib_dec\swb_bwe_dec.c(590): /* todo - delay output[] by 1.25ms ? */
• ..\lib_dec\swb_bwe_dec.c(737): /* IVAS_fmToDo: synth[] is @internal_Fs!!! */

Huawei colleagues, please address them!

For completeness, these details were recorded in the past:

Issue 4

• Name: FD BWE signal alignments in DFT stereo
• Details: The FD BWE, both in WB and SWB, needs more work to accommodate for the IVAS framework with core-coder delay changes wrt. EVS (0.9375s shorter encoder and 0.9375s longer decoder delay). In the DFT stereo, the ACELP synthesis is resampled using the DFT and is not available for computing certain BWE parameters. Instead, an ACELP synthesis at internal sampling rate is used. There are few points to be verified:
  • Is the use of ACELP synthesis at internal sampling rate fine?
  • Should the ACELP synthesis be delayed by 1.25ms to equalize for the resampling delay?
  • Could the lerp() resampling be avoided in case of 12.8kHz internal sampling rate and run wtda() with the length of 256?
• Open: Review and fixes are needed. Corresponding code is in file swb_bwe_dec.c, lines 264-272, accompanied by "TBD" and "TBC" comments.