Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[cerberus]

 http://recforums.prosoundweb.com/index.php/m/281284/2119/#ms g_281178

please help me with facts and opinions on the issue of mp3s that
are converted to pcm...  and they are all clipping!  

does it really happen?

t.i.a.

jeff dinces

[jdg]

isn't it being converted to PCM by the codec anyway for playing from the DAC?

i know its  fact that converting to mp3 from a pcm source close to 0dBFS will clip when transcoded (played) back

[Tomas Danko]

cerberus wrote on Mon, 15 October 2007 04:40

http://recforums.prosoundweb.com/index.php/m/281284/2119/#ms g_281178 please help me with facts and opinions on the issue of mp3s that are converted to pcm...  and they are all clipping!   does it really happen? t.i.a. jeff dinces

Yes, the way the audio is encoded dynamically it's bound to clip upon decoding.

It creates it's own approximate scale trying to hit the original dynamical targets. Obviously it will sometimes go below, sometimes above. Trying to iterate until it's almost spot on takes a lot more time when encoding, and in reality it's not going to hit the spot anyway.

Which is why we should top out at -0.3 dBfs to avoid the worst and obvious clipping of mp3's.

[cerberus]

pretty eye opening stuff!   thanks guys.  

Tomas Danko wrote on Mon, 15 October 2007 04:12

... in reality it's not going to hit the spot anyway.

i feel bad knowing i have sent mp3s to clients that clipped. though  i suppose that
most of the stuff for sale at the itunes store clips too.

like, where is craig anderton when we need him?
imo, we've got to make people aware, or this situation won't improve.

jeff dinces

[Fede]

Not that I buy a ton from the iTunes store (would much rather buy a CD and encode at lossless, as I can rarely stand MP3) and yes I have probably ranged towards singer/songwriter (although I have bought some hip-hop) rather than truly loud music, but to my knowledge nothing I have bought from there has notably clipped.

[Tomas Danko]

Fede wrote on Mon, 15 October 2007 16:51

Not that I buy a ton from the iTunes store (would much rather buy a CD and encode at lossless, as I can rarely stand MP3) and yes I have probably ranged towards singer/songwriter (although I have bought some hip-hop) rather than truly loud music, but to my knowledge nothing I have bought from there has notably clipped.

Since the files bought from iTunes are protected, how do you open them into an editor to see the real audio and if it's clipping or not, instead of just a flat line?

Just being curious.

[bblackwood]

Tomas Danko wrote on Mon, 15 October 2007 14:49

Fede wrote on Mon, 15 October 2007 16:51 Not that I buy a ton from the iTunes store (would much rather buy a CD and encode at lossless, as I can rarely stand MP3) and yes I have probably ranged towards singer/songwriter (although I have bought some hip-hop) rather than truly loud music, but to my knowledge nothing I have bought from there has notably clipped. Since the files bought from iTunes are protected, how do you open them into an editor to see the real audio and if it's clipping or not, instead of just a flat line? Just being curious.

Burn CD, import into editor?

[Tomas Danko]

bblackwood wrote on Mon, 15 October 2007 20:51

Burn CD, import into editor?

Sssshhhh, I asked him not you. Now he will know one way to answer the question!

Just wondering if Fede had actually checked those files, or relying solely on listening tests. (Which in the end is what matter, but we're talking mastering geek stuff right now)

[bblackwood]

Tomas Danko wrote on Mon, 15 October 2007 15:04

bblackwood wrote on Mon, 15 October 2007 20:51 Burn CD, import into editor? Sssshhhh, I asked him not you. Now he will know one way to answer the question!

Sorry.

*bows head and stands in corner*

[Nigel Jopson]

cerberus wrote on Sun, 14 October 2007 22:40

mp3s that are converted to pcm...  and they are all clipping!  jeff dinces

... well, maybe it has to do with distortion introduced when converting from PCM to MP3 in the first place!

At last year's AES Paris conference I chaired a seminar in which we isolated distortion components from MP3s. First we played the audience the MP3, then the distortion. Some people took their IR headphones off when the distorted extracts were played. The tunes were encoded from a "name and shame" list of hot CD masters Thomas Lund (of TC Electronic) had compiled in an article (Distortion To The People) he wrote for Resolution magazine:
http://resolutionmag.com/pdfs/TECHNO~1/DISTOR~1.PDF

Some work which the perceptual codec distortion investigations were based on is included in this paper:
http://www.tcelectronic.com/media/nielsen_lund_2003_overload .pdf

Just as interesting, to me, is the distortion generated by loud-but-legal levels in consumer players. Modern upsampling players can generate 3-10% distortion from perfectly "legal" levels of test signals. Thomas had tested a range of players owned by TC Electronic staff. This part of his talk is in this excerpt:
http://www.jop.myzen.co.uk/arc/2006_aes_Lund.mov

When I'm mastering I always check the 'test burn' on a range of ordinary players, and I think there's definitely a good case to be made for data-reducing encoding to be done carefully by the mastering engineer ...

[compasspnt]

Nigel Jopson wrote on Tue, 16 October 2007 13:48

Just as interesting, to me, is the distortion generated by loud-but-legal levels in consumer players.

YES!!!

Thank you.

[Alexey Lukin]

cerberus wrote on Sun, 14 October 2007 23:40

please help me with facts and opinions on the issue of mp3s that are converted to pcm...  and they are all clipping!

Yes, they are clipped if levels of the original PCM were close to 0 dBFS. This clipping is not introduced during encoding, it's introduced during decoding. The reason is that MP3 format, just as all other lossy formats, distorts the waveform. And any waveform distortion near 0 dBFS has a chance of going over 0 dBFS. Since most programs decode to fixed-point PCM, the result will be clipped.

If you are using RX on a mac, it will decode MP3 to floating-point PCM, and you can get the non-clipped decoding, suitable for normalization or limiting. This doesn't work on a PC yet! (the Windows WMA decoder that we are currently using for MP3 can only decode to fixed-point).

[Matt_G]

What sounds worse the clipping on decode to PCM or the Mp3 compression? If consumers don't care about listening to uncompressed audio why worry whether it's clipping? You can't really do too much unless you want to drop your masters down by 6db to prevent the clipping. But then clipping is still fashionable anyway isn't it?

Matt

[cerberus]

thanks alex for the technical explanation, it is exactly what i thought was the case.

i did indeed use rx (mac) in the manner you've outlined. once i learned the
true reconstructed signal values,  it was an eye opening and educational
experience for me.  the values were way outside of my expectations.

in fact it was reminiscent of  figuring out that santa claus does not exist.
most kids play along for a while after they discover the truth because
it is to their benefit,  in the end though, poor kids get nothing.
that was the hardest part to admit.  well.. there may be no
santa... but  -they- have a "santa", don't they?  sadly, no.

jeff dinces

[Ben F]

Just master to -0.3dBFS and you should be covered for most codecs. I've noticed audio with intersample clips distort far easier, easily removed by Sequoia before making a master.