PDA

View Full Version : Normalization exceeds specified limit?



Barak
07-07-2002, 02:05 PM
When using adaptive normalization to 85% with a 3000ms window, there are occasional spikes that go to 100%. How do I make this not happen?

Spoon
07-08-2002, 04:29 PM
This ties in with your other post, I am quite sure that the file was already at 100%, I don't think the adaptive has increased the volume.

Barak
07-10-2002, 04:50 PM
Spoon,

Here's the original sound file:

http://www-users.cs.umn.edu/~wymore/temp/before.jpg

Here's the same file, with Adaptive Normalization @ 80%, 3000ms window, and a max normalization of 20x:

http://www-users.cs.umn.edu/~wymore/temp/after.jpg

As you can see, while the vast majority of the file is kept within 80%, there are occasional deviations which can go to 100%. How do I stop this from happening? I would expect that 80% normalization means "no audio will exceed 80%", but that is not what is happening here.

Spoon
07-11-2002, 02:34 AM
Ah I see what you mean, for the moment you will have to run a Volume Quieten and set it to 80%.

Barak
07-11-2002, 05:28 AM
Spoon,

OK, if I use 80% with the "Clip any values over maximum" checked, that would work. The problem is that I'm using quieten already to make sure all my files are at 80% or below so the normalization happens (as per your instructions from the other thread). dbpowerAmp only seems to allow me to use each effect once. Why is that? I'd like to:

Quieten to 80%
Adaptive Normalize to 80%
Quieten (using clipping) to 81%

So I'll have to run each file through encoding twice to get the second Quieten?

BTW, in what order are the filters processed - top down?

Spoon
07-11-2002, 12:12 PM
They are done in order top to bottom, so there is little point on doing the first quieten, just do:

Normalize
Quieten

Barak
07-11-2002, 01:02 PM
Spoon,

But the point is to have the whole file be at 80%. If I do Adaptive normalize on a loud file which has not been Quieten-ed first, some sections will be above 80%, so when I do the Quieten after the normalization, the loudest, above 80% sections will be reduced to 80%, and the 80% portions will also be correspondingly reduced.

In order to make the whole file 80%, I need to:

1) Quieten the entire file to 80%.
2) Adaptive normaile so the whole file is at 80%
3) Use Quieten with clipping to hack off the 100% "jaggies"

There's no way for me to do this, since the DSP effects don't seem to allow more than one Quieten filter. Is there a way to do it?

BTW, is the "jaggies" thing a bug?

Spoon
07-12-2002, 02:36 AM
It is a bug yes.

Barak
07-12-2002, 03:17 PM
Spoon,

Sorry to drag this out - so is there a way to do Quieten, then Normalize, then Quieten?

-Ben

Spoon
07-12-2002, 04:06 PM
No, I will make the fix as quick as I can, but there is a backlog of stuff straining to get out the door (codec updates etc)

Barak
10-13-2002, 02:24 PM
Spoon,

Sorry to bug you - do you know roughly when this might be fixed?

-B

Spoon
10-13-2002, 04:48 PM
Bumped to the top of my 'to do' should be next week

Spoon
10-21-2002, 02:44 PM
How many bit is the file? I see it is mono 11KHz?

Unregistered
10-22-2002, 04:48 PM
Spoon,

The files are 16 bit, 11Khz, Mono PCM .WAV files, but I seem to remember that the problem appeared with other sample rates.

-B

Barak
12-29-2002, 11:34 PM
Spoon,

Is the fix done? If so, how can it be downloaded? Thanks!

-B

Spoon
12-30-2002, 06:42 AM
Sorry Barak, I havn't been able to reproduce the error :(

Barak
01-12-2003, 11:57 AM
Spoon,

Here's the original file (http://www-users.cs.umn.edu/~wymore/temp/before.mp3). Here's the dbPowerAmp-ed file (http://www-users.cs.umn.edu/~wymore/temp/after.mp3). Here the filter configuration file (http://www-users.cs.umn.edu/~wymore/temp/standard2.dds). Listen at 23 seconds and you'll here 100+% clipping, despite the setting of 90% normalization.

In this sound file, I'm recording a lecturer, and he's speaking with people in the audience, so dbPowerAMP needs to keep switching from extreme amplification to not-so-extreme amplification. This is the scenario where I'm consistantly encountering this problem.

-B

Spoon
01-12-2003, 12:18 PM
Can you try a test (I have a theory), convert that wave to file 44Khz 16 bit stereo and run the normalize on it, does it have the problem?

Barak
01-12-2003, 11:27 PM
Spoon,

When I take the "before" file, and convert it without filters to a 44.1K 16 bit stereo wave file, and then convert that file with filters to a 32Kbps 22Khz MP3, the problem is greatly reduced, but it still exists. Over an hour lecture, this new method yields maybe 10 trouble spots, as opposed to the direct conversion which yielded maybe 40 trouble spots.

-B