How come my mp3's exceed 0dBfs?

M

Maor

New member
Hey guys,
I got involved in mastering lately, working on my own, as well as other band's material. iv'e encountered something quite weird, and maybe one of you can shed a light on the subject.

I'm using Cubase5, and when i start a mastering session usually i import some songs from library to use as reference for loudness and harmonic balance.
I've got quite a loud song in now, 'Get Over It' by OK GO, and the meters are peaking. not just reaching 0dB, also exceeding it. on the beat i constantly see values varying from 0 to +2dB.
usually i don't trust the cubase meters, so i popped out iZotope's OZONE plugin, and the meters indicate the same thing. on the track, and on the stereo out bus.
needles to say, i didnt do ANY processing to this song, and there are no plugins running in the entire project. didn't touch the volume, gain, phase, pan or anything else. its just, as is.
this 'distortion is not audible in my opinion, and as i zoomed in on the waveform, indeed most "clipping" regions are just 5-15 samples in length. the song still sounds great.

Correct me if im wrong, but with digital audio we're working with dBfs- as in - full scale - as in - no way any audio is ever gonna go over the 0db mark, and if it does, the program will just flatten out the sound wave, making it "clipped". how is it that some songs go over that? WHATS GOING ON HERE?!
 
Intersample peaks -- It's probably just acting different because it isn't a MP3 anymore -- It's PCM data in a floating-point environment.
 
It's very common for slammed (heavily limited) WAVs that are converted to MP3 to have overshoots. This, I assume, is a result of what Massive pointed out regarding inter sample peaks. That being said, once you alter an original WAV file that peaks at 0dBfs - in this case encode it to mp3 - it will create artifacts due to the encoding process which in turn gives you overs. I'm not 100% sure how this works because I'm not a data compression expert, but I have witnessed this phenomenon often.

Cheers :)
 
When an MP3 is created it will be applying filtering as part of it's perceptual encoding, these filters are capable of increasing the peak amplitude of the resultant encoded file. There is a plug in just released that might interest you...

http://www.sonnoxplugins.com/pub/plugins/products/codec/codectoolbox.html

Worth noting there are some limitations to this tool maxing out at 48kHz, no LAME MP3 and no Apple AAC (Mastered for iTunes+) for PC users.
 
Last edited:
SafeandSound said:
When an MP3 is created it will be applying filtering as part of it's perceptual encoding, these filters are capable of increasing the peak amplitude of the resultant encoded file. There is a plug in just released that might interest you...

http://www.sonnoxplugins.com/pub/plugins/products/codec/codectoolbox.html

Worth noting there are some limitations to this tool maxing out at 48kHz, no LAME MP3 and no Apple AAC (Mastered for iTunes+) for PC users.
Click to expand...

That's the answer I was looking for. I used to know this but it got lost in my head somewhere. Thanks for the recap!

Cheers :)
 
Massive Master said:
Intersample peaks -- It's probably just acting different because it isn't a MP3 anymore -- It's PCM data in a floating-point environment.
Click to expand...

strictly speaking, the peaks generated due to mp3 coding aren't intersample peaks (=peaks between two samples when the digital data is DA converted), but real sample peaks due to the filters & the quantization used in the mp3 coding.
 
You must log in or register to reply here.
Back
Top