Why does "hiss" roughen when fading out?

Discussion forum for Amadeus users

Moderator: Martin Hairer

Sir Cecil
Posts: 46
Joined: Sat Jul 30, 2011 4:05 am
Location: San Francisco

Re: Why does "hiss" roughen when fading out?

Post by Sir Cecil »

[quote="cdjonah"]Not to be obvious but a sixteen bit number that is transferred to 32 bit
will translate back exactly to the same 16-bit number. Why would you
want to add noise?


I don't want to add "add noise". Perhaps I'm not expressing my question properly. I'm just trying to understand what goes on.
If one applies dither after adding (say) three or four smart edits and a long end fade to a track of music via Amadeus, does that whole file then get dithered, or does the dither apply itself only to the actual momentary points of the crossfades that form the smart edits and the end fade itself, leaving the remainder and vast majority of the file untouched?

CDJonah
Posts: 119
Joined: Sat Nov 11, 2006 12:10 am

Why does "hiss" roughen when fading out?

Post by CDJonah »

I am confused -- Martin says he doesn't add dither so what dither are we
talking about?

I have only dealt with dither in data acquisition systems and in those
systems I always felt that one should really just take the data right.

Lets make sure we are talking about the same thing. Say we have sound
values from 0 to 100. What we do is then add a white noise to the
values so that when we have a 1, we could end up with a value between
0.1 say and 1.9 and we can then instead of getting a flat value of 1, we
will get a random value between 0 and 2. This will then mean as we come
down from 3 to 2 to 1, we won't have sharp steps but sort of a noisy
value that drifts down. In general, in any recording I do, I have enough
noise in the background that this effect would be totally invisible.

Chuck
On 4/27/13 1:16 PM, Sir Cecil wrote:
cdjonah wrote:Not to be obvious but a sixteen bit number that is transferred to 32 bit
will translate back exactly to the same 16-bit number. Why would you
want to add noise?


I don't want to add anything. Perhaps I'm not expressing my question properly.
If one applies dither to a single file that has three or four smart edits and an end fade along its length, does the whole file then dithered, or is the dithering applied only to the actual momentary points of the crossfades that form the smart edits and the end fade itself, leaving the remainder of the file untouched?




_______________________________________________
Amadeus forum mailing list
Unsubscribe / change settings at http://two.pairlist.net/mailman/listinfo/forum_list

_______________________________________________
Amadeus forum mailing list
Unsubscribe / change settings at http://two.pairlist.net/mailman/listinfo/forum_list

JEGG
Posts: 122
Joined: Sat Feb 24, 2007 6:33 pm

Post by JEGG »

Perhaps I can renew my suggestion, once again, that you look at some resources.

One resource, already mentioned, is a book titled Mastering Audio by Bob Katz. Why not read it?

Or there's this, intended only to be a brief introduction to the subject, so there's a limit on the specificity of the information.

http://izotope.fileburst.com/guides/Dit ... _Ozone.pdf


The Katz book is a much more complete presentation of the topic, as might be expected. It's an easy read for everyone.

A confusion seems to be that numbers change when expressed as 16 bit or 24 bit or 32 bit. They don't. But numbers change when processes are done which involve calculations, and 32 bits allow for many more numbers (greater precision) than 16 bits.

And the calculations are being done at 32 bit.

Dither is **not** intended to mask sounds! That is why there are different kinds of dither shapes-so as to **not** mask the audio.

It is a very small amount "special" noise intended to randomly toggle the least significant bit(s) to:

1. Prevent truncation distortion, which is additive and can become audible as colourization and other artifacts.

2. Extend the program material dynamic range (and quality) by permitting material to be heard well under the noise floor, and-relatedly- enable quieter sounds (fades, reverb tales, ambience) to be heard far more smoothly and with greater fidelity, **not because they are masked (you'd not hear them),** but because the gravelly breakup is not occurring at low levels due to dither applied at much, much, much, lower levels. These aspects are clearly heard at normal listening levels.

Consider how much dynamic range is available in a 16 bit file, the level of the quietest elements in such a file, and the level of a two or three levels of truncation distortion. Clearly there is a problem at these and higher levels.

Again: Dither smooths out material which runs out of dynamic range, and it prevents truncation distortion. The two together are a huge factor.

I mentioned the subject of dither as it has an obvious relationship to the OP's initial question. I apologize for my contribution to steering this thread off course and on to more general issues.

Finally, just so there is no misunderstanding, my reservations about not having dither included in AP are not indicative of any lack of appreciation or affection for AP. I use it all the time, and have used the previous version for many years. I'm a fan!
Last edited by JEGG on Sun Apr 28, 2013 7:03 pm, edited 12 times in total.

JEGG
Posts: 122
Joined: Sat Feb 24, 2007 6:33 pm

Re: Why does "hiss" roughen when fading out?

Post by JEGG »

cdjonah wrote: In general, in any recording I do, I have enough
noise in the background that this effect would be totally invisible.
Not trying no to extend this discussion further, but your point that self noise essentially makes the material "self dithered" is a good one. A one shot normalization and short fades should be fine if your noise is truly random. You might run into problems if your fades were quite long, or you did additional processing and multiple resaves w/o dither.

Post Reply