24th February 2012
|
#1 | | Gear nut
Joined: Oct 2009 Location: Englewood, CO Thread Starter | Mp3 Distortion
My mixes sound distorted when I convert from wav to a low quality mp3 (128 kbps), but fine at higher quality (320 kbps). "turning_good.mp3" is at 320 and "turning_distorted.mp3" is at 128. I've done some searching, and it sounds like this is because the song needs some headroom or the encoding process to mp3 will cause intersample peaks.
1. I never see sound clips posted with the discussions on mp3 distortion though, so I wanted to post mine and see if this is what people are talking about.
2. Why does distortion only occur at a lower bit rate?
3. If the case is that the encoding process is making stuff go over 0, is there a way to reduce the volume, convert to mp3, and then normalize the file back to 0?
| 
|  | |
24th February 2012
|
#2 | | Lives for gear
Joined: Mar 2010
Posts: 1,388
|
Three things:
*The 128 kbps MP3 sounds exactly as I would expect it to sound. They're just kinda' naturally bad.
*No, you can't perform a level change on an encoded MP3.
*That dude's voice sounds like Weird Al in the best way possible.
__________________
- Mike Tate
Live sound guy
Wilmington De
| |
| | |
24th February 2012
|
#3 | | Lives for gear
Joined: Dec 2007
Posts: 876
Verified Member  |
1. There are several topics on this forum discussing mp3 distortion.
2. Every lossy codec produces more distortion at lower bitrates, be it an audio or a video codec. If you are talking specifically about clipping distortion, low bitrates change the waveform more, so more clipping may result.
3. This is actually a recommended practice! If you are mastering for mp3, keep your peak levels below -0.3 dB FS (or even lower if you are targeting low bitrates). If your mp3 has already been clipped, you can completely undo the clipping by unpacking the mp3 with a 32-bit-float decoder, like iZotope RX. It will correctly decode all peaks above 0 dB FS and let you turn the gain down.
| 
| | |
24th February 2012
|
#4 | | Gear Head
Joined: Oct 2009
Posts: 57
| Quote:
Originally Posted by Cheebs Goat  *No, you can't perform a level change on an encoded MP3. | Luckily we all have mp3directcut, and it's free  | |
| | |
24th February 2012
|
#5 | | Gear nut
Joined: Oct 2009 Location: Englewood, CO Thread Starter |
I've read about the -0.3 db thing, but I never see that in modern professionally recorded music. I always import a few pro mixes while I'm mixing my own stuff to make sure I'm on target, and they are usually in mp3 format since I just grab them from my iTunes library. All of them are still nailing 0db, and I hear none of the distortion that shows up in my mp3s.
mp3directcut looks pretty sweet. I'm going to have to try that and see if it solves my problem.
| |
| | |
24th February 2012
|
#6 | | Gear nut
Joined: Oct 2009 Location: Englewood, CO Thread Starter |
I just tried mixing down at -1db, and then -3db, and the mp3 still sounds just as bad, so it must not be that the song is too loud. I guess the LAME encoder is exactly that. I have only noticed this on cleaner stuff like this song, so I guess I just have to avoid 128kbps mp3s for those.
| |
| | |
24th February 2012
|
#7 | | Lives for gear
Joined: May 2007 Location: Los Angeles
Posts: 1,091
|
Lame is fine, a 128bitrate mp3 is what it is. At 128 I recommend Joint stereo + 2pass mp3 encoding. Post the wav if you want to hear the difference.
__________________
"Any experiment of interest in life will be carried out at your own expense." | |
| | |
24th February 2012
|
#8 | | Lives for gear
Joined: Jan 2007 Location: Melbourne - Australia's music capital. Verified Member |
Or... Quote:
Originally Posted by AnnihilationRob My mixes sound distorted when I convert from wav to a low quality mp3 (128 kbps), but fine at higher quality (320 kbps). | two words, courtesy of Bob Newhart.
| |
| | |
25th February 2012
|
#9 | | Gear addict
Joined: Dec 2006 Location: Ayr, Scotland
Posts: 434
|
I've never seen any technical proof to say that wav to mp3 conversion can produce digital clipping. I would say that any distortion heard is from the artifacts of low conversion and not from the mp3 clipping.
Just my thoughts, nothing set in stone. 
G
| |
| | |
25th February 2012
|
#10 | | Gear addict
Joined: Mar 2008
Posts: 420
| Quote:
Originally Posted by Ecktronic I've never seen any technical proof to say that wav to mp3 conversion can produce digital clipping. | You can easily verify by yourself that the peak level of a mp3 is very often higher than the one of the original .wav file.
| |
| | |
25th February 2012
|
#11 | | Gear nut
Joined: Oct 2009 Location: Englewood, CO Thread Starter | Quote:
Originally Posted by sanddigger1 You can easily verify by yourself that the peak level of a mp3 is very often higher than the one of the original .wav file. | Ran this test, and found that you are correct. Here's a screen shot from Sonar. You can see that the mp3s made from a wav mastered at -3.0dB are shooting up to 1.5dB higher.
Files "D" and "E" are the original files I posted "turning_good.mp3" and "turning_distorted.mp3". They were made from a wav mastered to 0.0dB. Strange that sonar does not indicate any clipping at all in those files.
I'm not sure what conclusion to draw from all this. It seems like everyone is partially correct in some way.
| |
| | |
26th February 2012
|
#12 | | Gear addict
Joined: Dec 2006 Location: Ayr, Scotland
Posts: 434
| Quote:
Originally Posted by AnnihilationRob Ran this test, and found that you are correct. Here's a screen shot from Sonar. You can see that the mp3s made from a wav mastered at -3.0dB are shooting up to 1.5dB higher.
Files "D" and "E" are the original files I posted "turning_good.mp3" and "turning_distorted.mp3". They were made from a wav mastered to 0.0dB. Strange that sonar does not indicate any clipping at all in those files.
I'm not sure what conclusion to draw from all this. It seems like everyone is partially correct in some way. | Interesting. I'm now thinking that mp3 encoders may have a built in limiter on them? Or maybe I'm totally wrong and the mp3s are digitally clipped due to the higher peaks.
G
| |
| | |
27th February 2012
|
#13 | | Gear interested
Joined: Feb 2012
Posts: 10
|
I have the same problems. The wave had no clipping. The 320 kbps mp3 clipped in a few places. Took care of that. But when loaded into Soundcloud, the 128kbps conversion produced audible artifacts in the bass. I can't understand why they don't use aiff. Probably can't afford the Apple royalties. But I converted a couple of Kontakt libraries to .aiff from wave and I can't tell the difference.
| |
| | |
27th February 2012
|
#14 | | Gear addict
Joined: Dec 2006 Location: Ayr, Scotland
Posts: 434
| Quote:
Originally Posted by TheoIris But I converted a couple of Kontakt libraries to .aiff from wave and I can't tell the difference. | .aif file sounds exactly the same as .wav file.
G
| |
| | |
27th February 2012
|
#15 | | Gear addict
Joined: Oct 2010 Location: Bucharest, Romania
Posts: 315
| Quote:
Originally Posted by Ecktronic .aif file sounds exactly the same as .wav file. | Because it IS the same, just in a different container.
| |
| | |
28th February 2012
|
#16 | | Gear interested
Joined: Feb 2012
Posts: 10
|
But it IS smaller in size, or am I just imagining this?
| |
| | |
28th February 2012
|
#17 | | Gear addict
Joined: Oct 2010 Location: Bucharest, Romania
Posts: 315
| Quote: |
The audio data in a standard AIFF file is uncompressed pulse-code modulation (PCM). There is also a compressed variant of AIFF known as AIFF-C or AIFC, with various defined compression codecs.
| By default, it's the same thing as a WAV. But just like WAV, you can use different compressors within the container.
__________________
We all have the same problem, it's how you handle it.
Everybody uses the same loop, it's how you sample it.
(The Terrorists - Terrorize Tracks) | |
| | |
29th February 2012
|
#18 | | Retired old guy
Joined: Jul 2009 Location: Sunny Seattle
Posts: 473
Verified Member | Quote:
Originally Posted by Ecktronic I've never seen any technical proof to say that wav to mp3 conversion can produce digital clipping. | Um, MP3 modifies the waveform by removing signal, which, paradoxically, means you are adding noise to the original signal. This changes the wave shape, and very well can increase the peak levels. This is not in any dispute at all.
If you have something that goes to full scale, and you run it through any perceptual encoder, you WILL get clipping.
If what you run through the coder that goes to full scale has a very dense spectrum, you will get more clipping, and more annoying trash.
__________________
Fourier and Shannon were right!
| |
| | |
2nd March 2012
|
#19 | | Locked away
Joined: Jan 2008 Location: In the shed
Posts: 1,004
|
Question: Why do you want to use 128kbps? Is your HDD space somehow more expensive than mine?
The input peak of mp3 encoding and the output peak differ. The output will always be higher in places.
When an audio file is digitally clipped and then inserted onto a track, the DAW won't know that it was clipped before. So it won't show overs.
128 kbps sound bad, very bad. The lowend gets quirky and there will be a very steep rolloff somewhere at 16 kHz.
In the end, if 320 kbps sound fine to you, then use it.
| |
| | |
2nd March 2012
|
#20 | | Gear addict
Joined: Dec 2006 Location: Ayr, Scotland
Posts: 434
| Quote:
Originally Posted by Th3_uN1Qu3 Because it IS the same, just in a different container. | I meant to add that also.
G
| |
| | | |
New Reply 
Facebook
Twitter
LinkedIn
