[Ffmpeg-devel] ac3 encoder volume correction

[Bill O'Shaughnessy]

Hello Michael,
  Maybe my volume patch isn't real pretty, but it works well.
The current code scales down the wav by a factor two.
I assume you have cooledit or audacity to look at these
waves.  I have tried two differenent professional ac3 decoders
and both indicate that the current code scales down the
amplitude by a factor of 2. (So does my ear.)
 My submitted code works and
does a very accurate 1 to 1 reproduction of the original wave.

Both ffmpeg and aften scale down by a factor of two when tested
with two different professional ac3 decoders.

Maybe you should try taking a wav encoding it to ac3 and
taking the ac3 and decoding it to a wav and look at the
original wav and the decoded wav.  Big difference.

Good Luck,
  Bill O.


On 11/21/06, Michael Niedermayer <michaelni at gmx.at> wrote:
>
> Hi
>
> On Mon, Nov 20, 2006 at 09:20:48PM -0800, Bill O'Shaughnessy wrote:
> > Hello all,
> >  after encoding a normalized wave file, ac3enc always encoded it a half
> > the volume of the original wave.  After some experimentation I found
> > that Justing Ruggles fix did not go far enough.
> > I also fixed some multiplies that should be 32 bit multiplies before
> > being shifted right 15 bits.
> >
> > With this fix ffmpeg encodes an ac3 file to identical volume of the
> > input source .wav file.
> >
> > Good Luck,
> >  Bill O.
> >
> > Here it is:
> > --- ac3enc.corg    2006-11-20 20:31:04.500000000 -0800
> > +++ ac3enc.c    2006-11-20 20:46:52.125000000 -0800
> > @@ -330,11 +330,12 @@
> > }
> >
> > #define MUL16(a,b) ((a) * (b))
> > +#define MUL32(a,b) (int32_t)((int32_t)(a) * (int32_t)(b))
> >
> > #define CMUL(pre, pim, are, aim, bre, bim) \
> > {\
> > -   pre = (MUL16(are, bre) - MUL16(aim, bim)) >> 15;\
> > -   pim = (MUL16(are, bim) + MUL16(bre, aim)) >> 15;\
> > +   pre = (MUL32(are, bre) - MUL32(aim, bim)) >> 15;\
> > +   pim = (MUL32(are, bim) + MUL32(bre, aim)) >> 15;\
> > }
>
> rejected this change makes no sense
> also its not indented by 4 spaces
>
>
> >
> >
> > @@ -1286,10 +1287,14 @@
> > static void lshift_tab(int16_t *tab, int n, int lshift)
> > {
> >     int i;
> > +    int32_t it;
>
> wrong indention, also int32_t is unaccpetable unless required which is not
> the case here
>
>
> >
> >     if (lshift > 0) {
> >         for(i=0;i<n;i++) {
> > -            tab[i] <<= lshift;
> > +            it = ((int32_t)tab[i]) << lshift;
> > +        if( it > 32767 ) it = 32767;
> > +        if( it < -32767 ) it = -32767;
> > +        tab[i] = it;
>
> this is just wrong
> the code is
>             /* Normalize the samples to use the maximum available
>                precision */
>             v = 14 - log2_tab(input_samples, N);
>             if (v < 0)
>                 v = 0;
>             exp_samples[i][ch] = v - 9;
>             lshift_tab(input_samples, N, v);
>
>
> this by definition scales to the max so cliping per definition is not
> needed
>
>
> >         }
> >     } else if (lshift < 0) {
> >         lshift = -lshift;
> > @@ -1368,9 +1373,9 @@
> >
> >             /* apply the MDCT window */
> >             for(j=0;j<N/2;j++) {
> > -                input_samples[j] = MUL16(input_samples[j],
> > +                input_samples[j] = MUL32(input_samples[j],
> >                                          ac3_window[j]) >> 15;
> > -                input_samples[N-j-1] = MUL16(input_samples[N-j-1],
> > +                input_samples[N-j-1] = MUL32(input_samples[N-j-1],
> >                                              ac3_window[j]) >> 15;
>
> more nonsense
>
>
> >             }
> >
> > @@ -1379,7 +1384,7 @@
> >             v = 14 - log2_tab(input_samples, N);
> >             if (v < 0)
> >                 v = 0;
> > -            exp_samples[i][ch] = v - 9;
> > +            exp_samples[i][ch] = v - 10;
>
> this may or may not be correct (=requires quotation from the ac3 spec
> saying that this is correct)
>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> In the past you could go to a library and read, borrow or copy any book
> Today you'd get arrested for mere telling someone where the library is
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> http://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>