Re: Fwd: [Ardour-Cvs] Ardour source branch meterbridge updated. 3.2-153-gf885622

On Sun, Jul 07, 2013 at 01:55:13PM +0200, Jörn Nettingsmeier wrote:
 
> ah, thanks for the explanation. so the threshold for the red light
> should probably be lowered to "more than N samples >= (float)(2^15 -
> 1)".

(2^15-1) / (2^15) actually.

Which would indicate clipping before it really happens if the 
input is 24 bit. On way to avoid this would be for Jack to
convert 2^(N-1)-1 to 1.0f exactly, for N the number of HW bits. 
The differences are purely of academic interest of course, we
are talking less than 1/1000 dB here. 

While recording, I'd want to see red lights flashing well
below 0.0 dBFS - there is no reason to try and use the last
half bit. Certainly not if you're using 24 of them. Having 
some 'digital gain' during mixing won't affect there result,
as the analog noise floor will be around 20 or 21 bits anyway.

While mixing that could be different. But even then EBU-R128 
recommonds a peak indication at -1 dBFS. And that's with 
four times oversampling and worst case of signal and same
with DC removed.

Ciao,

-- 
FA

A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)