[Ffmpeg-devel] Re: BUG: AC3 encode volume is low

[Michael Niedermayer]

Hi

On Sun, Apr 09, 2006 at 07:18:54PM -0400, Justin Ruggles wrote:
> Michael Niedermayer wrote:
> > On Sat, Apr 08, 2006 at 08:59:26AM -0400, Justin Ruggles wrote:
> >>Why does the MDCT output to a 32-bit int instead of 16-bit?  Isn't it
> >>just a signed 15-bit fixed-point implementation of the formula in the
> >>spec?  I can't find where the output coefficients would need to be more
> >>than a 16-bit signed integer.
> > 
> > 
> > and thats what scares me, i dont know ac3 very well, but i do know that a
> > N point decorrelation style transform will generally need log2(sqrt(N)) 
> > bits more at the output to maintain the same precission
> > 
> > 
> > 
> >>>From what I can tell from the MDCT code, and from the fix15() function,
> >>the output coefficients should have a range of -32767 to 32767.  Thus,
> > 
> > 
> > i dont think this is true
> 
> I've looked into this a little more now.
> 
> Using floating-point math in the 512-point MDCT reference equation, and
> taking the window function into account, it seems that the maximum range
> of the coefficients is +/- ~0.752.  This would give a fixed-point range
> of +/- 24641.
> 
> I've attached the test program I used to come to this conclusion.  If
> there is something I'm missing just let me know.

ok, you are right, the mdct implementation in ac3enc is inaccurate and
needlessly scales things down, so the patch is ok and can be applied


[...]
-- 
Michael

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