EffusionStudios wrote:The distortion and 'digital overs' from encoding mp3's of a well mastered (major label or other well done) release may not be as obvious or offensive to the ear, due to the style and nature of the material.
Absolutely true. Overs will cause aliasing, and while that is very obvious in material with simple harmonic structures, it is almost inaudible on noise-like sounds such as snare hits and cymbal crashes -- and in pop and rock music, snare hits and cymbal crashes tend to be the loudest things and tesnt to clip before anything else.
As a result, you can often get away with occasional minor clipping in that kind of material without anyone noticing. But do the same with a pinao recording, and everyone will hear it!
Technical Editor, Sound On Sound...
(But generally posting my own personal views and not necessarily those of SOS, the company or the magazine!)
In my world, things get less strange when I read the manual...
On my blog I recently did a little reverse polarity test between a wav and mp3 of the same material. The wav was normal polarity. I time aligned the mp3 and reversed its polarity. The leftover material was quite interesting. You can check it out if you click the link to Speaking Audio in my sig. Feel free to tear it up if I happened to get something wrong here... we're all here to learn and experience.
MaTrix, nice test on your blog. Lots of audio info left over in the polarity reversed result file. interesting. Am I correct in assuming it was a 128 Kbps MP3 you inverted? Depending on the resolution and encoding settings you select for the MP3 it will have a big effect in how it handles stereo information. (Joint Stereo Encoding and other settings) it might be fun to try an MP3 at 256 Kbps with true stereo as opposed to the M/S type processing.
MaTr1x2051 wrote:On my blog I recently did a little reverse polarity test between a wav and mp3 of the same material. The wav was normal polarity. I time aligned the mp3 and reversed its polarity. The leftover material was quite interesting. You can check it out if you click the link to Speaking Audio in my sig. Feel free to tear it up if I happened to get something wrong here... we're all here to learn and experience.
I've found a similar effect when cutting parts out of a stereo track (make both parts mono and invert one of them... change relative gains to cut out different parts). I'll usually get most of the track sounding fine, but then a few funny leftovers from the bit I cut out.
There's something wrong there -- I suspect the time alignment is well off. You shouldn't be hearing all that mid to high frequency material (the pronounced piano harmonics, the cymbals and all those vocals).
A small timing error between the original and MP3 file would produce exactly that result.
What you should hear is a something that has very little harmonic content at all -- really just a relatively low level, filtered and heavily modulated noise. It's usually enough to understand dialogue or lyrics, but only hint at the musical arrangement! It tends to sound like the voice of the devil when he is suffering a very sore throat!
Because of the complex filtering that goes on in MP3 encoding, aligning the original and encoded versions accurately is tricky. One easy solution is to add a single sample click about a second ahead of the original music track to act as a reference point. You'll find the mp3 encoded version has a stretched click, but it you align to the centre of that you should find that you get a better cancellation and the true artefacts of the mp3 process are revealed far more clearly.
Technical Editor, Sound On Sound...
(But generally posting my own personal views and not necessarily those of SOS, the company or the magazine!)
In my world, things get less strange when I read the manual...
Hugh you were right on... the time alignment was off by 1 sample! The new material still presents the distortion I was looking for, but like you mention its mostly on the vocals. I'm updating the clip as I write this... it should be ready for viewing in 10 minutes or so. -- Mp3 samples
MaTr1x, the new file sounds great....well...greatly awful actually, in the way I would hope...like mp3 space monkeys had a baby and that was the result.
Hugh Robjohns wrote: You need to give yourself something like 0.5dB (or more) of additional headroom before trying to convert to MP3 (or any other lossy codec). Experiment with levels to find the optimum.
Hugh
This is new to me and very interesting. I wondered why 'master masterer' Greg Calbi has set a ceiling of -0.3db to a recent album project!
More from him in an interview: 'It's assumed that MP3s will be heard in shuffle mode, competing against unknown music." Right, and that leads to extreme dynamic range compression; so all of the music's natural soft-to-loud dynamics are squashed flat; MP3s have to be loud all the time because with MP3s everyone is screaming for attention in a crowded market.'
This is almost a new thread - should mastering for mp3 be different not only in levels but also in other things?
Hugh's talking about the peak level, and suggesting a digital ceiling of -0.5dB to avoid clipping distortion during the MP3 conversion.
Setting it to -0.3dB will probably be OK, and I'm sure Greg Calbi has the gear to hear any resulting distortion. However, it's safer to allow a slightly bigger margin just in case, and an extra 0.2dB really isn't going to make any noticeable difference to average loudness.
MP3 limiting is another matter altogether, and if you want to have a really loud MP3 to compete with other tracks that have also been squashed to smithereens then it's the number of dBs of limiting that determines the average loudness - you can still set a digital ceiling of -0.5dB and get a really loud track.
Personally I leave my MP3 dynamics the same as my WAV files, but then I end up with more dynamic but quieter tracks
I think Greg Calbi's logic is false. I understand the marketing department's aim to get 'impact' on radio play etc, hence loudness wars, but to suggest that music needs to vy for attention once its already in someone's collection is just plain barmy IMO. People, especially young people, are quite ecclectic in their tastes nowadays. Certainly in my experience. So they are listening to Bob Marley and Hendrix next to Lady Gaga, and they have discovered the controls on the MP3 player that alter the volume.
For this reason I will always ignore any advice that depends on an assumption of the end user's listening habits. Because nobody actually knows what people's listening habits are, and anyone who believes they know is talking nonsense.
