Hi,
I would like to know what is the bit-exactness (up to how many bits the
output of the fixed point decoder output is matching with the floating
point decoder output) for the fixed point implementation of the iLBC
code.
Any information in this regard would be highly appreciated.
Thanks in advance,
bells.
Bit - exactness for the Fixed point iLBC code
Started by ●June 19, 2006
Reply by ●June 21, 20062006-06-21
Bells-
> I would like to know what is the bit-exactness (up to how many bits the
> output of the fixed point decoder output is matching with the floating
> point decoder output) for the fixed point implementation of the iLBC
> code.
There is no issue of precision. If the codec input is 16-bit linear, then
after compression and decompression you again have 16-bit linear values,
for example stored as a .wav file. At that point, the 16-bit output
values must exactly match the test vector.
There is no meaning to 16088.324 in a .wav file.
-Jeff
> I would like to know what is the bit-exactness (up to how many bits the
> output of the fixed point decoder output is matching with the floating
> point decoder output) for the fixed point implementation of the iLBC
> code.
There is no issue of precision. If the codec input is 16-bit linear, then
after compression and decompression you again have 16-bit linear values,
for example stored as a .wav file. At that point, the 16-bit output
values must exactly match the test vector.
There is no meaning to 16088.324 in a .wav file.
-Jeff