Question about mp3gain -- exactly what does clipping mean?

[statik213]

So I analyzed my mp3s and most of 'em are in the 95~100 db range, and it suggests a normal volume of 89 db.
1. It also has a red Y under clipping for most of em. What does this mean?
2. Some tacks have a red Y under clip(Track), what does that mean?

 

[gorcorps]

clipping is bad... it creates the little clicking sound in the background that you hear from bad rips as well as using mp3gain too high. Sometimes it's worse than others... but are you telling me that you haven't heard the distortion in the songs after you raised the vol?

 

[statik213]

Originally posted by: gorcorps
clipping is bad... it creates the little clicking sound in the background that you hear from bad rips as well as using mp3gain too high. Sometimes it's worse than others... but are you telling me that you haven't heard the distortion in the songs after you raised the vol?

Yes, haven't really noticed the clipping... hmmm.. so does it mean that there already is clipping or there will be clipping after normalizing?

 

[Concillian]

Originally posted by: statik213
So I analyzed my mp3s and most of 'em are in the 95~100 db range, and it suggests a normal volume of 89 db.
1. It also has a red Y under clipping for most of em. What does this mean?
2. Some tacks have a red Y under clip(Track), what does that mean?

Basically clipping is that you have a volume knob that goes from 0 to 10, but you are playing them with the knob turned to 11.

It's trying to use >100db but 100 db is the max. This will result in lost information. That lost information can sound really awful depending on the severity. It can also cause speakers to do some really unnatural things that can potentially damage them if played at high enough volume levels.

 

[statik213]

I normalized all my mp3s and things seem fine, I didn't notice any clipping in most of the mp3s before anyway.

Thanks!

 

[nineball9]

Draw a sine wave, erase the positive and negative peaks and replace them with flat horizontal lines. That's clipping!
(Easy to see clipping with a wave editor, or in electronics, with a scope.)

 

[statik213]

Originally posted by: nineball9
Draw a sine wave, erase the positive and negative peaks and replace them with flat horizontal lines. That's clipping!
(Easy to see clipping with a wave editor, or in electronics, with a scope.)

cool! that's easy to understand. So does clipping occur 'cos it was encoded at too high a volume?

 

[gibbyman]

The term clipping as used here is bad, what MP3 Gain does is NOT what we understand as clipping.
Clipping is what an amplifier does when over-driven, beyond the power/voltage available from the power supply, it, as described above, flattens the top and bottom of the waveform, this produces DC spikes which pushes the speaker cones to extremes and produces heat in the speech coil, thereby frying the speakers, this is very common especially with cheaper audio equipment that uses a bad progression on the Volume Control Pot'.

In the case of MP3 gain, which uses an extremely clever algorithm, it simply means that the program has modified the file, either up or down in level.
BTW what is 100DB the max of?
A decibel is a ratio, not an actual amount of anything when related to Gain and it's logarithmic, that is the least change the human ear can detect is 3db which is a doubling of power when related to wattage.
Some examples,
A Library - about 40db
Traffic 30 mtrs - 70db
Threshold of pain - 112db
Rock band crescendo - 120db
Instantaneous irreversible hearing damage - about 145db

Try this test if you like, rip an audio CD to 320kbs MP3, drop the results into MP3 Gain and do a track analasis, with the target set at 89db, you will probably see it saying the lot is clipping, then look see what levels the analysis says and set the target to the most common level, now it's not going to process those ones and the clipping thingy has gone.
The guy who wrote this program should have used a different term, using clipping in this context is misleading, due to the algorithm used MP3 gain will never introduce distortion.

Sorry just noticed the date on this thread, but nevertheless it may help someone.

 

[oynaz]

Sorry just noticed the date on this thread, but nevertheless it may help someone.

Except for that the fact that you are wrong. Applying too much gain anywhere, including mp3gain, will exceed the dynamic range, thus resulting in clipping.

You are right that analog and digital clipping are two different things, but they are still clipping, and you want to avoid it.

Unless you are a modern music producer, where the loudness war has been taken taken to a ridiculous extreme - listen to Metallica's Death Magnetic for a glaring example.

 

[gibbyman]

Except for that the fact that you are wrong. Applying too much gain anywhere, including mp3gain, will exceed the dynamic range, thus resulting in clipping.

You are right that analog and digital clipping are two different things, but they are still clipping, and you want to avoid it.

Unless you are a modern music producer, where the loudness war has been taken taken to a ridiculous extreme - listen to Metallica's Death Magnetic for a glaring example.

Thing is with a floating number of bits, I.E. not limited to 16 as a conventional CD is, clipping in the digital sense should never occur, even if it does it's not distortion as such it's simply no increase in volume similar to compression-(limiting), without the analogue "breathing" effect.
What we are in fact talking about is the limitation of Dynamic Range of 16 bit 44Khz CD standard, but we are not dealing with this criteria, we are dealing digitally with something which is already digital, so providing the replay equipment can cope there is no audible problem.
Mind you I agree much recent "Pop"/Chart music is way over-modulated, trying to take advantage of the physiological effect of something louder sounding better to the untrained ear.
Truth is we could debate this back and forth for an eternity, simply because when something is subjective there can be no right or wrong.

 

[AndreyT]

Except for that the fact that you are wrong. Applying too much gain anywhere, including mp3gain, will exceed the dynamic range, thus resulting in clipping.

No, he is not wrong.

Modifications made by MP3Gain are completely lossless. MP3Gain does not modify the actual audio information stored in MP3 data blocks. It simply changes the gain coefficient stored in MP3 file separately. (This is the reason any changes made by MP3Gain are always reversible.) For this reason, by itself MP3Gain cannot really introduce any actual physical clipping. MP3Gain does not even touch the audio data with scissors.

The only reason MP3Gain warns about clipping is that it sees that at some points in the track the signal level (at the current gain value) will exceed the formal standard specification. However, this is no more than "educated guess" made y MP3Gain based on the idea of some imaginary theoretical standard playback equipment. In reality, all playback equipment is different. You might indeed run into clipping on some Dollar-store pocket MP3 player (you might even run into clipping on poor quality equipment when MP3Gain said everything was OK.) Meanwhile, higher grade equipment might be able to handle your high-gain MP3 file without any issues at all.

In fact, most of the MP3 music tracks sold from Internet stores like Amazon today will have their volume in 96-99 dB range with clipping reported by MP3Gain in almost every track.

 

[TheRyuu]

Isn't it actually safer to just use Replaygain[1] analysis in a program like foobar2000[2] to handle this sort of stuff? That way there really is no modification of the original file going on other than adding a value to the metadata.

mp3gain isn't always lossless[3] although it seems to be rare that it isn't. It's still preferable to just change the metadata though and not have it modify the mp3 values if possible. So if your player (like foobar2000) supports Replaygain then just use that instead.

Clipping really isn't as evil as people make it out to be unless your doing some heavy modification of the audio or something like that. In general I'd say it's safe to just not even worry about it unless it's somehow producing noticeable audio distortions. It's also entirely possible that the clipping may not even occur if your Windows volume slider isn't at 100% (since I believe it may "soft clip" in that case).

[1] http://wiki.hydrogenaud.io/index.php?title=ReplayGain
[2] https://www.foobar2000.org/
[3] http://wiki.hydrogenaud.io/index.php?title=ReplayGain#MP3Gain

 

[AndreyT]

Isn't it actually safer to just use Replaygain[1] analysis in a program like foobar2000[2] to handle this sort of stuff? That way there really is no modification of the original file going on other than adding a value to the metadata.

That's exactly how MP3Gain works. There really is no modification of the original file other than adding a value to the metadata.

mp3gain isn't always lossless[3] although it seems to be rare that it isn't. It's still preferable to just change the metadata though and not have it modify the mp3 values if possible.

"Change the metadata" is exactly how MP3Gain works.

Your own link clearly explains when and how MP3Gain can become "lossy": when you attempt to push the metadata gain value beyond its range. If you do that, the original gain value is formally "lost" (although no audio frame data is modified). But no program in existence can overcome this limitation. It is an inherent property of MP3 format. Every program will run into this issue, if you force it to adjust gain that far.

 

[DigDog]

i'm gonna ignore the previous replies and just give you a description of what clipping is.

in digital audio, the volume is a multiplier, it makes your sound louder. sound exists within a container, and this container has a ceiling. if you make things too loud, they will hit the ceiling. not ALL of the sound will hit the ceiling, as you probably know, sound waves are "spiky" and "wavey", meaning a sound is made of several different components, some louder than others.
again, in digital audio, anything which is louder then 0 (which means the ceiling) cannot exist - it simply vanishes. however, only the portion thus cut vanishes.

so if you have a very spiky sound, and you make it clip, the tip of that spike vanishes, but the truncated cone still remains. you still hear that sound, but it's missing the tip.
Now, this "tip" is very important, it gives a sound its particular timbre, or what scientists call the audio spectrum.

In the end, if you make things too loud, you lose some information, and the sound becomes worse. On analog systems, it's quite more complicated, as natural compression ... you know what, never mind. Just dont make things clip.


oh btw..
i cannot really know what your mp3 sound like (maybe they were re-encoded by some retard to make them louder, it would not surprise me), but in commercial audio, it's not uncommon to have the clipping light come on - most systems show you the red clipping warning a touch before the system actually clips, so you can use that as a guideline that you are near clipping.
(people do this because they want their records to sound loud, but yet do not want clipping)

if you can blast the music loud enough, you should be able to hear wheter a song clips or not (even if yours are probably badly encoded 128kbps mp3s from the space invaders era), but in case not, you can look at this
https://www.youtube.com/watch?v=-v6ML2DsBfA
and it should give you a good idea of what clipping is.

i also recommend you watch Mastered By Muppets, by Clipping Death
https://www.youtube.com/watch?v=QPu0DKyGgZI


for the tech heads, clipping in digital occurs when the audio parameters of the encoder are exceeded - i.e., you want to enter a number which is outside the numeric scale you are using (kinda like writing 65 bit code), which in this case is called dBFS, or decibel full scale, where 0 represents 100% of the allowed amplitude.
decibels over dBFS represent the noise floor ratio, and it's a number which varies according to the equipment; for example, a yamaha OD2 might have 126dB of SNR(Signal/noise ratio, the noise floor),so 126dB is the equivalent of 0dBFS.

analogue audio is technically the same, for most aspects, although the physical medium behaves in weird ways when subject to near-zero amplitudes.

regarding playback of clipped sounds, you can do it without any ill effects, as clipped audio is essentially just a square wave. you are (likely) listening to clipped audio whenever you play a chiptune. speaker cones haven't suffered from square waves since just after they were invented, although in unusual circumstances (coupled with an excessive amplifier) yeah, they would wear out faster when reproducing clipped audio... but a regular waveform is just as capable of breaking said speaker, given the necessary amplitude.

now, i have no problem believing that WinAmp or something like that can manage no more than 100dB SNR, and that's how it counts clipping. FYI, the occasional screeching sound you hear from bad mp3s from the 90s is because there is no anti-clipping in those softwares. the additional, truncated information is still fed though the system, which tries to reproduce it as if it were a sound. modern digital audio does not do that, it understands that a sound has clipped and omits entirely the clipped portion.

imho, if you get that digital blip-blip from your mp3s, you should consider throwing them out and re-encoding the songs yourself.

source: nearly 15 years of audio tech work, acoustics, studying with audio professionals, music production major in college, working with protools and logicpro, aduio installations, humping at concerts, mixing, soldering, PAs, you name it i've done it.

i really, really hope this post stops the arguing, and if it doesn't i'm gonna go grab my 1400 pages Glen Ballou' Sound Enginnering handbook and throw it at your heads.

 

[AndreyT]

i'm gonna ignore the previous replies and just give you a description of what clipping is.
...
i really, really hope this post stops the arguing, and if it doesn't i'm gonna go grab my 1400 pages Glen Ballou' Sound Enginnering handbook and throw it at your heads.

Er... This is all nice, but unfortunately has absolutely zero relevance to the topic of this thread. With the same degree of success you could've dumped here the entire "MPEG-2 Audio Layer III" specification under the premise the we are talking about MP3 files.

So, I suggest you read the original post(s) in this thread and get yourself familiarized with what the topic here is. I will give you a hint to get you started: we are not talking about the definition or manifestations of clipping in audio world (about which most of us are educated as well, or better, than you). We are talking about a related, but still very different thing: what the term "clipping" means in MP3Gain program and whether this "clipping" is lossy or lossless with regard to the original MP3 file.

I'm sure the wall of text you just pompously dumped on us is perfectly correct in itself, but since it does not even attempt to answer the original question, it has zero value in the context of this thread.

And no, we are not arguing about anything here. When I stated that MP3Gain does not physically (lossily, irreversibly) clip audio data in MP3 file, the question was answered conclusively and the discussion was over. Nobody's arguing anymore.

P.S. And in the future, make sure you aren't "gonna ignore the previous replies" if you want to be welcome here, as opposed to being sent flying out of here like a cork out of a champagne bottle.