Volume Normalization question

[Freelancer]

I'm encoding many audio files these days, and tried foobar and dMC, the reason why I choose dMC finally is that it has a better remove silence between track function than foobar's gap killer plugin. But I found that the volume normalization feature in dMC is not so functional as foobar's encode with replaygain, the problem is even I encode a file with that DSP enabled, the louder is still louder and the lighter is still lighter, and when I use foobar to analyze those encoded files, I found that the normalization effect is not so obvious, the result is shown as below: (all use default setting, dMC with Simple normalization)
before encode after encode(dMC) after encode(foobar)
gain peak gain peak gain peak
file1 -5.17db 0.945954 -5.23db 1.065336 -0.02db 0.531334
file2 -7.93db 1.125331 -7.51db 1.303613 -0.02db 0.454055

As you can see, foobar is really try to get the track gain close to 0 db, but dMC doesn't seem to do anything reasonable. And the output of foobar is really seem to be standardize the volume.....

Can anyone help me to explain this ?

[ChristinaS]

Re: Volume Normalization question

Volume normalize raises the volume of quieter passages, it does not lower the volume of loud passages.

[Freelancer]

Re: Volume Normalization question

Thanks for your explanation. So I've used another file to have a try the result is:
before: +3.57db peak 0.358367
after: -4.89db peak 0.976459

Is this the expected result? But the gain is still not 0db just like foobar, however, it seems that dMC is trying to make the peak close to 1.0. So I've the concern that which parameter shall we consider in Volume Normalization, foobar focus on track gain, but dMC focus on peak.... Require further explanation.... Thanks.

[ChristinaS]

Re: Volume Normalization question

You might want to read a bit on normalization issues: http://forum.dbpoweramp.com/showthread.php?t=5351

[Spoon]

Re: Volume Normalization question

Replay gain works by making the majority of tracks quieter, then trying to match other tracks to that average volume. If your entire system is replaygained then it works well, but ideally volume normalizing should be done by the device playing back (then it can make advanced of going to 24 bit when scalling down the volume so there is no effective quality loss).

If your playback device is something like a normal audio cd or protable player then the normalize has to be done beforfe buring / uploading. dMC has an 'adaptive normalize' which works better than replay gain in noisy environments.

[Freelancer]

Re: Volume Normalization question

If your playback device is something like a normal audio cd or protable player then the normalize has to be done beforfe buring / uploading. dMC has an 'adaptive normalize' which works better than replay gain in noisy environments.

But I found that by using 'Adaptive normalize', the volume of song is divided into several pieces, and some pieces will be tuned to louder than before... Is this the right behavior of 'Adaptive'? But this really break the original volume ratio of the song, right? What I need is just make one song louder and another lighter, not tune the volume of a piece of song. So I suppose I should use 'Simple normalize', right? But the 'Simple normalize' doesn't seem to work perfectly, though ChristinaS explained that normalize is only used to make quieter sound louder, not louder sound quieter, so I follow the step of another thread, which use a 'Volume Quietn' before 'Volume Normalize', it did work, but the output is not so satisfactory, a louder song is still that louder, but really others which are quieter than that song are normalized to a nearly standard volume.
And another test I've made is that, use that louder song, run dMC with only 'Volume Quietn' to make it quieter, and then run 'Volume Normalize' to normalize it, the result is much quieter than the previous one.
Below is some data: all use foobar's replaygain analyze
before: -9.68db peak 1.298282
dMC(1pass: quietn+normalize): -7.62db peak 1.015296
dMC(2pass: quietn, normalize): -9.68db peak 1.138048
others are in the range of +3db to -4db, and after 1pass dMC, the output is in the range of -4db to -5.xdb which I think is acceptable.

[Spoon]

Re: Volume Normalization question

R12 of dBpowerAMP will include replaygain.

[Freelancer]

Re: Volume Normalization question

So could you explain what's the difference between replaygain and volume normalization? To my understanding, they're both used to make several audios listened at a same volume level, right? And my target audio player is a mobile phone which not support replaygain tags, so I think I need the volume tuning to the real content of audio not just put a tag....
Another question is, could you tell when will R12 be probably out? Thanks.

[LtData]

Re: Volume Normalization question

R12 is about 4-6 months out.

[Spoon]

Re: Volume Normalization question

Standard dMC normalize is simply a peak to peak volume detection, not average loudness (using a filter array).

[gameplaya15143]

Re: Volume Normalization question

R12 of dBpowerAMP will include replaygain.

yay :smile2:

normalizing modifies the original PCM waveform, usually to use the maximum range the bitdepth offers.
replaygain is a tag which tells the decoder how loud playback should be so that no samples are clipped.

*compression often induces clipping, so normalizing isn't as efficient.