Hi,

I'm mixing for a friend of mine CD's for Step-Aerobic Courses and last time I used MP3's as a source of them.

While the CD was played very loud, it seemed to me, that the hole CD was overamplyfied (what it wasn't). I disliked it very much and my pain to listen to the music was nearly physical. So I replaced the source with the original WAV's, left the mixing as it was before, and the next week everything was fine.

So I was asking myself: Does the transparency of MP3s depend on the loundness of the playing. Is it much more easy to recognice MP3s, when you play them (very) loud, like in a disco-environment?

While I'm writing this down, I remember a CD of a music magazin that demonstrates the masking effekt, that MP3 uses. Two sinus-tones where played and you could only hear the louder one. But the magazin told me to replay the sample a second time, but much louder. So this time, the masked frequence was noticeable.
So maybe, when I play MP3s very loud, I notice the absence of the masked frequencies?

Has someone similar experiences?

BTW: No ABX was done to test my experience.

Although in my experience you can get away with ABR 128kbps for djing @ clubs, pubs etc,... without anyone noticing Usually, there are so many speakers blaring in different directions that any sense of intricate time resolution is lost anyway, you certainly would find it difficult to discern any smearing unless you are right up next to one big speaker set in particular, and in which case, you are probably half deaf (or in my case, too caught up in music to notice)...

But... on some cases, whereby I have --APS encoded stuff that I deemed "never to play again", and then at a later date thought "well, maybe I will", I have noticed this with regard to songs played at the end of the night: Some songs have a long long fade out, and with the volume cranked up to ridiculous proportions in a club, when there is virtually no signal left (well, you wouldnt be able to hear it unless you are a bat, or in a bl00dy loud club), the signal goes a ^bit^ smudgy / watery / white noisy.

BTW: No ABX was done to test my experience either (due to extreme feasibility problems!)

Note: This could be complete bull, because I know that your hearing is screwed after a session in a club. But then again, it would make sense, I suppose

Edit: Piss poor construction of sentences

If there's a compressor dragging the analogue silence from an mp3 up by some stupid amount, then you could well hear tinkling sounds. VBR mp3 will not code exceptionally quiet analogue silence (i.e. very very quiet hiss) correctly, because it rightly assumes that you can't hear it. But if there's something in the signal chain that thinks "ooo - everything's gone quiet - I'll turn the volume up!", and if the system gain is very high to begin with, you might just notice.

I haven't tried this with -aps though, so it may not be an issue.

The answer is to adjust the compressor so that, below -60dB (say) it doesn't try to increase the gain too much.

The other answer is not to play silence, and to put the "end of the night fade-out" in manually yourself.


Still, if you've got a compressor that's set to drag the silence up on an mp3 to a point where it's audible, it would probably sound TERRIBLE if you ever played a vinyl record through it which contained a quiet part!



Other issues with playing mp3 loud: it doesn't matter. There's more masking in the human ear at higher levels - spectacularly more at club/disco type levels - which means that you're much less likely to hear any problems, as Lev said.

Only one issue remains: feeling the music. You "feel" a lot of bass in a club. This isn't something that you can check with ABX and a pair of headphones. I'm not sure what the psychoacoustics do (or if many codecs implement a 20Hz high-pass filter), but I've not heard of any problems, even though a lot of people DJ with mp3s.

Cheers,
David.

Thankfully, Luminar Leisure want 20 minutes of background music now, which means I get to listen to some decent ambience / william 0rbit at the end of the night

re: People DJ'ing with MP3's... Yep, some songs you just cant get hold of otherwise, so, yes, I have fallen fowl of playing Xing's, as have many people... I cant converse properly when a real duff Xing is playing in a club, no matter how drunk. It just utterly distacts me

There has been at least one thread on the ogg vorbis encoder using too low a bit rate on
quiet passages. In one of my cases a nominally 64 kb/s encode came out at 10 kb/s for
a couple of minutes and the effect was quite audible. The temporary fix is to specify a minimum
bit rate.

Its hard to say whether that is AT ALL related to this situation since the poster didn't say
what encoder, what settings, CBR or VBR/ABR etc.

From the description of sounding overamplified, you might want to try converting the mp3 file
back to a wave, then use something like Cool Edit to compare with the original with regard to
clipping, peak level, and average level.

Oh sorry, I was using LAME 3.93.1 APS.

I was using Freuro to create WAVs out of the MP3s, because at the same time, I wanted to burn an Audio-CD with those MP3s and in that very special case it was only drag & drop for me. Normaly I don't decode MP3s with my CD-Burning program.

The other thing I can clearify is the enviroment, where I was listening to both versions of my Step-Aerobic-CD. It was the fitness-studio-room (4m * 5m) with four JBL-Loudspeacer (those small black boxes you often can see in Bar's, fitness-studios or elseware). They normally can play the music very loud without any problem. The music was not so loud, that you have to worry about your ears, but the trainer must scream, to be understood.

[Edit] And: It was a continous 50min mix without any fade-outs or silent passages. Every song has a similar loudness. Music-Genre: Trance like Kai Tracid, Talla 2XLC, Kate Ryan. The effect of overamplify seemd to me during complete tracks, some tracks were more problematic than others.[/Edit]

lame VBR @ low volume tested here:

MP3 vs. MPC vs. Ogg: Low volume test

It's no problem for CBR
possible workarounds:
1. use CBR (e.g. --alt preset insane)
2. use -F (with aps) to keep bitrate from dropping below 128kbps (should work, but I haven't tried)
3.a use mp3directcut to apply fadeout manually after encoding
3.a (a pain, but) "compensate" the original fadeout with a waveeditor's fadein and do 3.a after encoding

@Vietwoojagig: The mp3s you're talking about have been mp3gained to non-clipping volume - right?

No.

WAV ---LAME APS ---> MP3 ---FEURIO---> WAV

But, do I have to MP3Gain all my freshly encoded files? But why should this be a problem only if I play them (very) loud?

. Depends on what you encode. Just have a look at some of your mp3s with mp3gain to find out if they're clipping.

You said it sounded "overamplified". In my experience, on not-so-decent equipment clipping sounds worse at high volume (or starts to become noticable while I can't hear it at low volume). I guess this is because the higher the amplitude of an imput signal the likelier an amplifyer will distort.

OK, this sounds reasonable to me.

Thanks