Preprocessing time-domain issues
# Basic info <!--- Add commit SHA used to reproduce --> - Fixed point: - Encoder (fixed): 20cb10087c533033d2cfcca2266cf4df8bac73a4 # Bug description For the time domain stereo, the ICA is systematically under-estimated the time shift as seen in the figure below. from top to bottom : - input signal - time shift (input to quantized) float - time shift (input to quantized) fixed point - difference between float and fixed point input to the quantizer - output of the quantized float - output of the quantized fixed point - difference between float and fixed of the quantized output ![image](/uploads/fc2d0c8e65efe4e55b43ed1103ab6cad/image.png){width=2318 height=1029} The next figure shows the same plot as above, but with the correction proposed in !1135 . It can be seen that when the input to the quantize is the same between float and fixed point, the output is the same as well ![image](/uploads/fc0e892121f012aa5f3bc5e77a062930/image.png){width=2309 height=1014} # Ways to reproduce Box folder: ...\Box_EXTERNAL_IVAS_BASOP_VERIFICATION\issues\issue-xxxx <!-- Commandline or script --> ![input_32k_short](/uploads/fe41da607f53f0ec4c480ce776f48803/input_32k_short.wav) ```bash IVAS_cod -STEREO 2 32000 32 input_32k_short bit ``` <!--- Below are labels that will be added but are not shown in description. This is a template to help fill them. Add further information to the first row and remove and add labels as necessary. -->
issue