Re: Lost residue modifications

Thanks a lot, xiphmont, everything makes sense now.

[email protected] wrote:
>>> for(count=0; count<psy_look->n; count++) {
>>>   vb->pcm[i][count] += 1;
>>> }
>>>       
>
> Hi,
>
> You're updating the wrong vector, or rather, the wrong part of it.
> The residue vector at that point is in two halves; elements [0-n/2)
> are pre-quant/norm and elements [n/2-n) are post-quant/norm. (my 'n'
> is the blocksize, psy_look->n is blocksize/2).  The lossless coupling
> exclusively uses the second half.
>
> If you stick to lossless coupling only (-q 6+ like you said), you can
> change vb->pcm[i][count] += 1; to vb->pcm[i][count+psy_look->n] += 1;
> and your changes will stick.
>
> Cheers,
> Monty
>
>