digital audio loudness question - volume exceeding 0.00db ???

hello.

i have a question!!

not sure if it is relating to digital audio mastering ? or specifically audacity as a program ?? or mp3s ??? 

 

but okay. 

let's say i open an mp3 in audacity, because i feel like doing my daily analysis of waveforms in audacity ... 

a lot of mp3s in my collection seem to exceed volume of 0.00db. 

 

continued in spoilers because biggish pics .. 

 

 

 

can this be explained ? even just links or an explanation of what the terminology for this is ?? 

i assumed digital audio files couldn't store info beyond 0.00db ??

i for eg. can't render a file beyond 0.00 without that additional info just clipping and being cut off.. i thought ??

 

or is audacity just a glitchy program??

i'm confused. pls halp. 

 

**edit** sorry for the dumb terminology. 

where i'm saying, beyond 0.00, i'm actually meaning beyond 1db. ???

Edited October 11, 2016 by bitroast

does it just mean audio files can exceed +1db and that's normal ?

halp

Welcome to the mysterious world of loudness measurement, a place full of 'dBu's, 'dBfs's, LUFS, K-Weighting, PPMs etc. etc. - As you say digital 0db is the 'top' value in the digital realm but many DAWs have different measurements in the calibration of their metering.

 

Where you see the value being ABOVE 0db in the digital world they're no doubt using a different measurement than dbfs, possible dbu. In the above example of audacity the +/- 1.0 is just the full linear scale of the waveform, think of it as the current position of the speaker cone, with 1 being fully 'in' and -1 being fully 'out' (or t'other way round!). The audacity manual actually has a good explanation and will show you how to change it to the db scale: http://manual.audacityteam.org/man/audacity_waveform.html

yaaaas. thx dudes. will analyse that audacity link.

 

i get the feeling this might be something that'll exceed my basic understanding of how the world works. but even just a basic awareness will help :^)

  On 10/11/2016 at 11:02 AM, mcbpete said:

Welcome to the mysterious world of loudness measurement, a place full of 'dBu's, 'dBfs's, LUFS, K-Weighting, PPMs etc. etc. - As you say digital 0db is the 'top' value in the digital realm but many DAWs have different measurements in the calibration of their metering.

 

Where you see the value being ABOVE 0db in the digital world they're no doubt using a different measurement than dbfs, possible dbu. In the above example of audacity the +/- 1.0 is just the full linear scale of the waveform, think of it as the current position of the speaker cone, with 1 being fully 'in' and -1 being fully 'out' (or t'other way round!). The audacity manual actually has a good explanation and will show you how to change it to the db scale: http://manual.audacityteam.org/man/audacity_waveform.html

 

Couldn't it also have something to do with the intersample peaks?

 

@bitroast

If you were on a Mac I would have adviced you to download the AU plugin called AU Lab and then to make it all a bit more simple download a script called "Kosmos afclip Droplet". This script analysis your files for intersamples and gives you a txt file with some data for you to look at.

 

I'm gonna give you some examples and show you some screenshots... just a sec.

  On 10/11/2016 at 11:38 AM, Squee said:

Couldn't it also have something to do with the intersample peaks?

Yeah that's definitely the case above where normalising an already normalised track caused the track to be reduced in amplification a likkle bit

@bitroast

 

Fucking typical... the new MacOS won't let me run the Kosmos afclip Droplet script anymore. Oh well, fuck it. I'll just do something else then:

 

So in the following screenshot you'll see a so-called "amplitude statistics" from Adobe Audition.

 

 

As you can see it says "0 possibly clipped samples".

Alright, so everything is good to go, right? Nope! So there are samples within samples and they're a pain in the ass.

Here's the exact same track but here I've analyzed it with another meter and as you can see, it tells me that there are 824 intersample peaks on the left channel and 1021 intersample peaks on the right channel.

 

  On 10/11/2016 at 11:53 AM, Mesh Gear Fox said:

does anyone know anything about how mp3 files are formatted, or more specifically how the data in the file represents the audio signal? always something i've been curious about. a lot of "how mp3s work" articles will explain the way they delete frequencies but not necessarily how the file itself is compiled.

Having looked at it for a while but seem to remember it going through multiple passes of chucking out data before being crunched up into frames of information (kinda like the blocks you get in jpeg files)

FrameLengthInBytes = (12 * BitRate / SampleRate + Padding) * 4

 

Full info here - http://mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm (reverse engineered as remember it's still a copyrighted codec)

Oops sorry yep you're right! mp3 = mpeg layer 3

  On 10/11/2016 at 12:16 PM, mcbpete said:

FrameLengthInBytes = (12 * BitRate / SampleRate + Padding) * 4

 

Full info here - http://mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm (reverse engineered as remember it's still a copyrighted codec)

 

I understood everything right up until it said, "The frame header is constituted by the very first four bytes (32bits) in a frame....". I hate numbers.

Just means every frame chunk (likkle block of data) has 4 bytes (8 'bits' make up 1 'byte' hence the '32 bit' reference) of information at the start giving the specification of the data to follow.

 

So it'd look something like: "Hey you mister 'puter, this bit coming up is 320kbps and stereo and at 44.1khz ... here you go : .... ..... Hey you mister 'puter again, this bit coming up is 256kbps and stereo and at 44.1khz ... here you go : .... "

  On 10/11/2016 at 2:01 PM, mcbpete said:

Just means every frame chunk (likkle block of data) has 4 bytes (8 'bits' make up 1 'byte' hence the '32 bit' reference) of information at the start giving the specification of the data to follow.

 

So it'd look something like: "Hey you mister 'puter, this bit coming up is 320kbps and stereo and at 44.1khz ... here you go : .... ..... Hey you mister 'puter again, this bit coming up is 256kbps and stereo and at 44.1khz ... here you go : .... "

 

Mister 'Puter sounds like a pretty cool guy to me.

But yeah, I've always had a hard time understanding how any kind of data processing or programming works unless someone tells it in a way that I can visualize.