| Author
|
Peak + 0.86 db - not clipping / cubase / protools
|
frisbeehead
IsraTrance Junior Member
Started Topics :
10
Posts :
1352
Posted : Jul 5, 2013 00:11
|
Assuming that some form of noise is applied in the dithering process, if any further changes in level and bit depth are made, chances are the shaped noise could get loud enough to sum with the levels of the mix signal at some points, no?
Just trying to make sure if they did it all correctly and got this problem anyway...
mentioned you don't use ozone, do you mind saying what are your favourite tools for mastering? thanks
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Jul 5, 2013 00:42
|
it s not necessary to do it once , it depends ,noise can still be below your converter noise floor or gear noise floor depending the amount and shaping and bit reduction.
32 bit float to 24 files then 24 to 16 can be no problem
personally i avoid shaped noise like many and use TDPF dither.. |
|
|
|
frisbeehead
IsraTrance Junior Member
Started Topics :
10
Posts :
1352
Posted : Jul 5, 2013 04:36
|
Quote:
|
On 2013-07-05 00:42, PoM wrote:
it s not necessary to do it once , it depends ,noise can still be below your converter noise floor or gear noise floor depending the amount and shaping and bit reduction.
32 bit float to 24 files then 24 to 16 can be no problem
personally i avoid shaped noise like many and use TDPF dither..
|
|
that's different. you're changing the bit depth twice. a good dithering zeros out the extra bits that are to be left out on the truncation process. if you're changing twice, then it's ok. what I meant was, it's not a good thing to have one on a limiter, or Ozone (or whatever chain really) and have the host application doing another one on top of that. most specially if we're talking noise shaping which always differs and it will sum and stress the peaks of the audio a bit - not saying that it will necessarily be heard, but it messes up the audio and can generate nasty artifacts.
the fact it's within the noise floor of the gear isn't really the issue here, but the effect it can have on stressed enough signals. because, if Ozone's brickwall limiter, for example, is showing the audio within decent levels and you're introducing more noise shaping afterwards, in a stressed signal this can mean that the extra shaping will be enough to make the audio clip - and I was suggesting this could be the case to begin with here XD
maybe it's not the case at all, but even so If your reducing the bit depth once, then dither once seems to be the rule of gold to me.
p.s. I don't like noise shaping either, specially for EDM. 
UV22HR is the one I like best btw... no noticeable coloration for my ears |
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Jul 5, 2013 16:58
|
ah man i didnt get what you mean, if i got it yes it s not good.. reducing the bit depth once, then dither once ..
it could maybe still happen in some plugins that convert to 64 and apply their own dithering back to 32 |
|
|
|
MuckyPuh
IsraTrance Junior Member
Started Topics :
14
Posts :
86
Posted : Jul 5, 2013 22:27
|
wow ... sorry for the delay - i was not here the last days xD
so yes the comparisment between the cubase and protools stats was made with an mp3 file a friend bounced in pro tools, which i checked then in cubase ...
and with masterd mp3 files i ment wave files that where proabably propperly masterted and pressed to a cd once... and then ripped by somebody and encoded to mp3
Quote:
|
On 2013-06-29 22:32, Colin OOOD wrote:
Quote:
|
On 2013-06-29 22:07, Melange5738 wrote:
Actually, if you read the Cubase manual it says something like because it uses floating integers to handle the peaks it should not necessarily clip just because you go above 0 dBFS and that it can handle significant values above 0 dBFS before clipping occurs.
|
|
Yes, that's because it works at 32-bit float internally. Virtually unclippable. However whether or not a decoded mp3 file can contain these overs depends on the bitdepth of the WAV file created by the decoding process (which is what happens when you drag an mp3 file into Cubase); the only way an audio file will give stats showing a maximum sample value >0dB is if that audio file is 32-bit float. If the mp3 we're talking about is decoded into a fixed-point audio file, that file may well clip if the inter-sample peaks are large.
|
|
ah ok .. i was not thinking that the mp3 is no mp3 anymore the moment i see the stats from cubase ...
so if i would set the internal mode to any other bit length then 32 float and repeat the process i would see -0.00 db as Peak in the stats?
 http://www.muckymusic.blogspot.de/
http://soundcloud.com/muckypuh-muckymusic |
|
|
|
Colin OOOD
Moderator
Started Topics :
95
Posts :
5380
Posted : Jul 6, 2013 13:32
|
|
