Why does the track meter display +2.1 for a mp3 file that I've imported?

[jonespnice]

I have the fader set to zero and all I've done was import an instrumental of the song "Bossy" by Kelis. I'm not understanding how it can show up on the display higher than zero, can someone explain what I must be missing?

[ashcat_lt]

It's an mp3?

[jonespnice]

Yes it's a mp3 file that I imported from my iTunes library.

[DVDdoug]

Simple... MP3 is not limited to 0dB.

And since MP3 is lossy compression, some peaks get higher and other's get lower so it's not unusual for some peaks to go over 0dB even if the original was only hitting 0dB, especially if the original file was compressed/limited or clipped. But, +2dB is not that common (1dB or less is more common).

Many (most?) of the MP3s I've ripped from CD go over 0dB.

Your DAC is hard-limited to 0dB so it will clip if you play it at full-digital volume into your DAC. However, I've never heard of a case where the clipping caused by Mp3 compression was audible... In other words, if there are audible compression artifacts, I don't think lowering the volume to prevent clipping will remove the artifacts.

[jonespnice]

Ok I see, I never knew that this was common for mp3 files. It doesn’t sound distorted or anything so I was able to normalize it within Reaper and now it playback without going over zero. Thanks guys for your replies.

[Bri1]

Ok I see, I never knew that this was common for mp3 files. It doesn’t sound distorted or anything so I was able to normalize it within Reaper and now it playback without going over zero.

^Aye-- ya see what happens is... when a file gets limited at say -6db with a limiter- but gets exported out @ say -0.5db --- when that mp3 gets reaimported and converted back to float-- that 'limited+compressed energy@ is then realeased again. =)
The peaks are contained to a fixed point on exports with mp3/ lossy formats and under 'fixed' 24bit.. when converted back to analog energy- sound can become 'non fixed' again.

[bigjoe]

They are Intersample peaks, nothing more, nothing less.
They are the reason why recently is very important to export a master with at least -1db of headroom (max peak value -1db true peak) when exporting for lossy format (mp3, aac etc...).

[ashcat_lt]

They can't be intersample peaks because Reaper's meters can only show us the level of the individual samples. They may represent what might have been intersample peaks in the original wav file, but after its converted to MP3 and back, the peaks are actually real sample peaks.

The thing is that any filter you apply to a signal will change the peak values in unpredictable ways. You'd think that a low pass filter just attenuating high frequencies, so what comes out should never be louder than what comes in. You'd be wrong. The overall integrated energy may be less, but some individual samples probably will be louder than they were before the filter.

It's like if you try to take a corner at 60MPH. In a smaller lighter vehicle you might make it, but in a big heavy truck - even if you try to slow down a bit before the curve - you'll probably still end up going past the turn.

So any filter can and will undo any hard limit that you try to set. This includes things like the oversampling on your favorite fancy limiter thing. The oversample filters mean that the limit is actually a limit. The only way to be sure that no sample will ever be louder than some maximum is hard, unfiltered clipping with all the nasty aliasing that entails.

[Bri1]

Ey-i think the other thing to this is where the actual samples land when reaconstructed--- if they are closer to peak value then they may be more likely to clip on reasampling.
When people are pushing the rms values of a track they are increasing the energy ---so engineers are going for a limiter to squash the "peaks"-- when reaconstructed-- this high rms can push some samples over sometimes because the energy is already high! =)
Loudness wars are over--there's really not much need to ever go near 0db if wanting to preserve full dynamics.. when people limit,they gain no more than the 0db limit--all they do is push the rms/vu/lu higher n higher==loosing d.range in that process. sounds loud-but can be fatiguing over long periods.. ears need loud n soft--'energy' also comes n goes hard+softer naturally__ so replicate that eh?...
Even with 32bit-we are stuck @0db limits--we just gain extra signal range and integer values of that range! right?
The problem with this increased range-is the softwares faders do not really register microscopic changes like that on the faders we use!! they mainly only go 0.00< 2 places.