I've been using the search function on this forum and browsed a lot of topics on MP3Gain but I'm afraid I'm still confused. What I want is this:

• set my entire collection to one level so I don't have to adjust volume any other song when listening to a mixed playlist
• keep the internal dynamics of any song.

I'm planning to do so using 'Track Gain' on my songs setting them to 95 db (from a previous test I discovered 89 to be far too low for my portable player). I've just ripped one of my alltime favourites 'The Cult - Electric (1987)' with EAC and LAME @ -aps and analyzed it with MP3Gain 1.2.5, the range goes from 91,3 to 93,3 db. Two songs (at 92,0 and 92,3) are marked as clipping.

Being a non-native speaker I seem to have trouble grasping the subtle differences and meanings of 'loudness', 'volume', 'clipping'. Doesn't clipping mean that the max volume on my system is exceeded? So how can it clip at 92 instead of going over 99,9?

Also my standalone does seem to have problems with APE tags so I'd like to remove the tags which means I can't go back to the file's original volume. Which on the other hand isn't that bad cause I could always go back to another level losslessly so even when the volume intended by the artist (or producer or whoever) can't be accomplished anymore the way the loud and low pieces of the song, it's dynamics, are to be perceived is still preserved, right?

Volume: amplitude of the samples in the song
Loudness: how loud the song actually sounds (determined by psychoacoustic analysis)
Clipping: a sample that is too large (too high volume/amplitude) to fit, and in the case of an mp3, to be correctly decoded

The loudness is expressed in decibels, not a percentage. You can in theory have clipping at any loudness, it depends on the nature and the dynamics of the music. But generally at 89dB there will be none.

MP3Gain/Replaygain never affects the internal dynamics of a song, or, in album mode, of an entire album. It only corrects loudness differences between songs (or in album mode, between albums).

I don't understand the last paragraph.

On second thought it was the stuff I said before repeated all over again I just wanted to be sure that even when having used MP3Gain to alter the songs without keeping tags (so that there's no way to go back to the original except by guessing the original value) it basically still is the same song since both it's quality and internal dynamics is kept, I only make it sound louder.

So a way to go would be to turn up all songs to 92 or 95 and then tone down those (hopefully) few that show clipping. Contrary to the two clipped songs I ripped from the CD I do not loose any information, the information on those two songs however is lost (i.e. wasn't there to begin with) because it's out of scale, right?

I would advise to turn down all songs to you reach a level where none clip.

The information isn't necessarily lost. If the volume is lowered, it's possible that those peaks that didn't fit before now start fitting.

You shouldn't do that if this is what you want: "set my entire collection to one level so I don't have to adjust volume any other song when listening to a mixed playlist"

You can individually maximize the volume (to the max level without clipping) if you like, but it spoils the variations that should exist between songs within an album and also between different albums. In other words there some songs that are inherently louder or softer than others and this is part of an albums dynamic range.

Having said that I must confess that I sometimes like to make my mp3's a little louder than replay gains 89dB reference (though never at the expense of clipping). In my case it is to compensate for some sub-standard audio equipment that has a bit of noise and interference which is reduced when running at lower settings on the volume knob.


Just curious, but what is the specific problem with 89dB RPG'ed tracks on your portable, does the volume control have insufficient range or is it a noise/interference issue similar to the problem I have?

mp3gain shows you the perceived loudness in SPL, not dBFS. it doesn't have to be 100 to clip. what i also found with mp3gain is if the levels hit 0 dBFS but don't necessarily clip, it may still report that the track is clipping. the only way to know for sure is to open it in a wave editor. either way, if clipping is indicated, it means the track is either clipping or right on the verge of clipping and you should lower the volume only for that track if you use track gain.

A lot depends on the degree of clipping. For many people, clipping that only needs a -1.5 dB adjustment to remedy is tolerable and listenable, especially on a portable player. As with so many other audio issues, your own ears will need to be the judge.

Regards,
Madrigal

Granted, I do not use the best headphones on my Nex IIe but the signal/noise ratio when tuning up to a sufficient volume on the player isn't very good. Apart from that I happen to forget to mp3gain new songs before putting them on the player so being at high volume and stumbling across a new song (usually much louder than 89 db) can be quite painful. Last but not least I'm not entirely convinced that running on almost max volume doesn't degrade battery life faster than playback on modest levels.

Clipping can additionally mean that two or more consecutive samples show the same value (which should never happen in music since we are dealing with waves but can be very likely introduced during false mastering) ... these ranges of samples will then be displayed (e.g. in a wav editor) like the ones that are too loud after decoding (higher than allowable) and are therefore being displayed as "clipped" ...

Some pieces of music do contain clipped passages at "normal" RMS volumes (not talking today's overcompressed pop music) that cannot be restored with ReplayGain or MP3 Gain ... I found that I don't even care about that since I can't hear it.

After all, clipping is nothing more than "straight lines" in a wave ...

I don't agree. There is no technical reason why even a properly mastered wave file might not have two consecetive samples sitting at the same level.

Clipping is not as easy to define as one would first think. First, clipping never means exceeding the max volume allowed by the given digital format, since that is not possible - per definition. (i.e. If it is louder than maximum then what you think is maximum really isn't!)

What clipping means is that a sequence of sample data is either pegged to full minus or full value negative (or both, as I get to). The tricky part is defining how long that sequence has to be before it is called clipping. Two consecutive samples is NOT usually considered clipping. More common definitions are 5 or 7 samples, but it varies from program to program and engineer to engineer. By and large, your ears should tell you if there is a problem. If something sounds fine, even though some software is telling you it's clipped, then it's probably OK.

Audio with 4-5 samples at the same + or - level usually sounds OK, but there is a version of clipping where the wave flips from full positive to full negative and this is both aweful and unmissable. It usually sounds like a very distinct click or worse. This is uncommon on modern audio editors, but used to be popular since it made clipping during multi-track mixdown impossible to miss.

The only caveat with this is that further operations (eg reverb, EQ...) on data that is maxed out (even if it sounds OK) is more likely to result in artifacts.