Audio levels

[iriverdude]

I've spoken to a hifi person and he mentioned on a sonos it seems to sound like it's clipping. I quickly explained about replaygain & altering mp3 levels. So I'm going to give him a couple of tracks with different alterations. Makes sense to give him original, with replaygain, and one with alteration in the mp3 gain levels. The track is from Shaina Twain Come On Over. Seems quite high in levels, Replaygain shows -10.92 for track gain.

I assume to determine if the Sonos is clipping, I should give him a track with a high gain level (from the original CD) then attempt to alter mp3 gain levels and this should hopefully give a clipping free track?

Here's a picture. The top one is mp3 with just replygain (scan not alter) and the bottom is mp3 with Media Money "level volume" When I replaygain scan the same track it now shows -0.38.

Is this correct? Or should I reduce it not so much?

 

[ph0rk]

Looks like the original track has clipping, so replaygain probably won't help much.

 

[cswann1]

Yeah, that looks like too big a swing, and I agree that it looks like their may be some clipping of the original track.

Adobe Audition 2.0 (formerly Cool Edit Pro) has a great clip restoration tool. The current Adobe sound editing software, Soundbooth, sucks imo (no clip restoration that I've found). I have the CS4 Master Collection, but I do any and all sound file editing in Audition 2.0.

 

[iriverdude]

Well if that's clipping, then so must all other tracks with -9dB replaygain or higher reduction!

 

[iriverdude]

delete

 

[iriverdude]

ahh wrong picture

 

[cswann1]

Quote:

Originally Posted by iriverdude /img/forum/go_quote.gif Well if that's clipping, then so must all other tracks with -9dB replaygain or higher reduction!

It's hard to tell from the image but anytime the peaks of the wave go all the way to the top of a graph, their is usually at least a bit of clipping going on.

I think studio engineers do that on purpose. It's part of the loudness war. They want their artists tracks to sound as loud as possible without noticeable clipping. Just a theory

Edit: Oh yeah! on that image you just posted, their is very noticeable clipping.

 

[iriverdude]

I'm also encoding flac-ogg for my iriver I may as well alter the oggs so I don't need to use replaygain which I believe uses more CPU power. What settings should I use in foobar and on the sliders?

 

[gregorio]

Quote:

Originally Posted by cswann1 /img/forum/go_quote.gif It's hard to tell from the image but anytime the peaks of the wave go all the way to the top of a graph, their is usually at least a bit of clipping going on. I think studio engineers do that on purpose. It's part of the loudness war. They want their artists tracks to sound as loud as possible without noticeable clipping. Just a theory

I don't think many engineers will deliberately clip the audio, compress the hell out of it maybe but not deliberately clip. It can be very easy to clip by mistake, digital meters are notoriously inaccurate because they are based on sample values rather than on the amplitude of the waveform once it is reconstructed. This means in theory you could be as much as 6dB over peak limit without any clipping being registered by the software or system. Of course, there are some engineers who just aren't very good at their job and clip/distort out of ignorance.

G

 

[ROBSCIX]

The odd clip of a peak may be considered acceptable when mixing but making the audio clip on purpose by amplifying it up to the digital ceiling is a no,no in recording.
Maybe there is somthing else at work here.

 

[Rempert]

Keep in mind that a hot CD track will almost always have peaks above full scale after it has been converted to mp3 (or virtually anything else is done to it for that matter). This is not actually clipping in so far as mp3 uses a floating point format, but it defies convention and if played at 100% volume now it will be clipping. That case is the easy fix, since you just need to lower the gain of the track. Any distortion on the original CD from pushing the loudness too hard will be almost impossible to fix.

 

[gregorio]

Quote:

Originally Posted by ROBSCIX /img/forum/go_quote.gif The odd clip of a peak may be considered acceptable when mixing but making the audio clip on purpose by amplifying it up to the digital ceiling is a no,no in recording. Maybe there is somthing else at work here.

Sorry, I'll have to disagree, the odd clip may NOT be considered acceptable when recording, mixing or doing any other digital audio process! Sometimes when there is an indicated clip (peak over) it maybe difficult or even impossible to hear any artifacts. However, during the mastering process and often duing broadcast, compression (and/or multi-band compression) is usually applied and what was imperceivable before may suddenly become noticable. If you are ever involved in recording or mixing anything, NEVER clip!!

Quote:

Originally Posted by Rempert /img/forum/go_quote.gif Keep in mind that a hot CD track will almost always have peaks above full scale after it has been converted to mp3 (or virtually anything else is done to it for that matter).

I'm not sure what you mean by this? Full scale means all bits set to "1", the maximum value possible in a digital audio system or file so, by definition there cannot be peaks above full scale.

G

 

[Rempert]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif I'm not sure what you mean by this? Full scale means all bits set to "1", the maximum value possible in a digital audio system or file so, by definition there cannot be peaks above full scale.

Here is an example for you. Christina Aguilera's "Fighter" was meant to be the high impact point of a commercial pop album. Not surprisingly then, scanning the original track shows it's maximum peak level to be 1.000000, which I'm sure is hit or approached more than a few times. I then encode the track to mp3 using LAME -V2, which will then be decoded by the standard library in foobar2000. While the average levels of the track remain the same, the highest track peak is now 1.224119. Since albums now seem to have every moment of every track as the high impact point, I'd bet these examples are more the rule than the exception.

The reason I brought this up is because the original post compared an mp3 without gain reduction to one with gain reduction, and then there was some discussion about the original track appearing to clip. My point is that the mp3 will not appear the same as the original lossless track with regards to the maximum peak values you would see in editing software. I suppose it could also be a cause of audible distortion if you are playing back at 100% volume to a 24 bit or 16 bit system, if the clipping will last long enough to notice...

 

[gregorio]

Quote:

Originally Posted by Rempert /img/forum/go_quote.gif Here is an example for you. Christina Aguilera's "Fighter" was meant to be the high impact point of a commercial pop album. Not surprisingly then, scanning the original track shows it's maximum peak level to be 1.000000, which I'm sure is hit or approached more than a few times. I then encode the track to mp3 using LAME -V2, which will then be decoded by the standard library in foobar2000. While the average levels of the track remain the same, the highest track peak is now 1.224119. Since albums now seem to have every moment of every track as the high impact point, I'd bet these examples are more the rule than the exception.

Ah, I see. It's some strange measurement system employed by that software. Not quite sure why it measures like that but there you go ....

JFYI, digital audio is always measured (well, not always apparently!) using the dBFS scale. The highest value possible when using the dBFS scale is 0dB (Full Scale), which indicates all bits set to "1" and therefore measurements in dBFS are always negative values. IE., some value lower than 0dB.

G

 

[Rempert]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif Ah, I see. It's some strange measurement system employed by that software. Not quite sure why it measures like that but there you go .... JFYI, digital audio is always measured (well, not always apparently!) using the dBFS scale. The highest value possible when using the dBFS scale is 0dB (Full Scale), which indicates all bits set to "1" and therefore measurements in dBFS are always negative values. IE., some value lower than 0dB.

Well it's a result of the software working in floating point. Convert the same mp3 to a 32 bit wave file and load it in a 32 bit editor, and it peaks at +1.9 db. Convert the same mp3 to a 16/24 bit wave file and load it, and it peaks at 0 db (having already been clipped).

I don't know if lossy files sold on itunes are the same way, but I'd be surprised if they aren't.