MP3 Playback Improvement - Page 3 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #61 of 65 Old 11-01-2013, 01:35 PM
AVS Special Member
 
Aleron Ives's Avatar
 
Join Date: Oct 2009
Posts: 2,832
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 321 Post(s)
Liked: 373
Quote:
Originally Posted by FMW View Post

I think we are into opinion here and not test results. If you can make an MP3 that is audibly indistinguishable from the original CD - and you certainly can - what would make FLAC or AAC or anything else better. Different, perhaps, but better? I think people tend to equate MP3 with low bit rate MP3. High bit rate MP3's are audibly indistinguishable in a blind test from WAV.

The main benefit of FLAC, aside from ensuring that your music has no lossy artifacts, is that it saves you the trouble of re-ripping your CDs if you want to encode them to multiple lossy formats over the course of time. People often seem to rip their CDs to 320 kbps MP3, only to decide later that they want to save more space. Since transcoding MP3 -> MP3 is a sub-optimal solution, encoding MP3s from FLAC files is more convenient than extracting the lossless audio from your CDs a second time.

As for AAC, in a very general sense, MP3 can become transparent at ~192 kbps, whereas the improvements in the AAC format allow it to reach transparency at a lower bitrate of perhaps ~150 kbps. As such, AAC can save even more space than MP3 without sacrificing sound quality, and depending on the individual listener's tolerance for artifacts, he might be able to use an even lower bitrate without hearing any artifacts.

It's absolutely correct that if you're using a sufficiently high VBR quality setting that generates bitrates in excess of 200 kbps, there is no difference between the fidelity of MP3 and AAC, and that's exactly why AAC has failed to supplant MP3, despite being the technically superior format. The fact that AAC doesn't even gain a 2x compression improvement over MP3 means that its extra efficiency fails to outweigh the benefits of MP3's wider compatibility for many people, since everything but your toaster can play MP3 files.

As for double-blind testing, there is more than one type of test you can conduct to evaluate the quality of lossy codecs. An ABX test lets you compare a lossy file to the lossless original to determine if the lossy file is transparent, whereas an ABC/HR test allows you to score multiple non-transparent lossy codecs against each other to objectively rank their sound quality. For example, this ABC/HR test compares Nero AAC, Apple, AAC, FAAC, and LAME MP3 at 64, 96, 128, and 192 kbps. None of the lower bitrates reached transparency on the test samples, but the test ranked Apple AAC > Nero AAC > LAME MP3 > FAAC in terms of quality, and AAC showed a clear advantage over MP3 in terms of quality at mid-range bitrates.
MarkHotchkiss likes this.
Aleron Ives is offline  
Sponsored Links
Advertisement
 
post #62 of 65 Old 11-01-2013, 04:28 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,239
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 874 Post(s)
Liked: 518
Quote:
Originally Posted by Aleron Ives View Post

For example, this ABC/HR test compares Nero AAC, Apple, AAC, FAAC, and LAME MP3 at 64, 96, 128, and 192 kbps. None of the lower bitrates reached transparency on the test samples, but the test ranked Apple AAC > Nero AAC > LAME MP3 > FAAC in terms of quality, and AAC showed a clear advantage over MP3 in terms of quality at mid-range bitrates.
Just a kind note that those tests are not for the specific bit rates they specified. The encoding modes are variable with the target for the whole file specified. This is different than the typical true fixed bit rate encoding. As an example, the encoder in these tests could run at 300 or even higher bit rate for all the difficult portions and then dial down to substantially lower bit rate for the easier stuff. So while the file size on the average will maintain the numbers you list, it does not represent the quality of the codec at the specific bit rates.

In addition, encoders vary substantially in how they utilize VBR and hence, you are not really testing the codec anyway. An MP3 encoder that allocates say, 250 kbps to a few frames of audio, vs AAC encoder using 120 cannot compared to each other. This is why industry tests are done in CBR mode (which itself is VBR but over a very short window). And because of these effects, you need to test across vast amount of content and not just a dozen to see how the encoder allocates its bits.

In general, it can be shown that at least for difficult content and expert listeners, MP3 never achieves transparency at any bit rate (again, using true CBR mode) while AAC comes darn close. And all of this is useless information since we should be only using lossless ripping anyway smile.gif.
tubetwister likes this.

Amir
Retired Technology Insider
Founder, Madrona Digital
"Insist on Quality Engineering"
amirm is online now  
post #63 of 65 Old 11-01-2013, 05:26 PM
AVS Special Member
 
Aleron Ives's Avatar
 
Join Date: Oct 2009
Posts: 2,832
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 321 Post(s)
Liked: 373
It's true that they're using ABR instead of CBR (although Apple's encoder was testing when using a so-called "constrained" VBR mode that is supposed to offer slightly more flexibility than normal ABR while still adhering to a target bitrate), but the limitations on bitrate variability imposed by mandating a target bitrate still apply in ABR mode. ABR still targets a specific bitrate instead of a specific quality, just as CBR does, except that it allows for some improvement in quality by permitting a minimal amount of variance. A TVBR test might allow for 300 kbps during the most difficult portions, but ABR modes will disallow that if doing so prevents the encoder from achieving the target bitrate (since it would require a disproportionately low bitrate to balance out the average back to the target). As such, ABR's quality is probably more closely correlated to the quality of CBR, even though ABR is technically a VBR encoding technique.

As such, I don't think the use of ABR in the test by any means invalidates its usefulness, as abstract comparisons of a codec are neither useful nor even truly possible, since there is no such thing as an ideal encoder that utilises a codec to its utmost potential. Since we only have encoders that vary in the quality of their implementations, it makes sense to compare those encoders to see which of them can achieve the highest quality at specific bitrates, and using an ABR mode does allow for that. The test is as much a comparison of the encoders' VBR implementations as it is a comparison of their quality : bitrate ratio, and there is nothing inherently wrong with that, since people should be using VBR modes, anyway. A single test of course does not provide concrete evidence of one encoder's superiority over another's, but individual tests can be informative, nonetheless.

As for the ability of lossy codecs to achieve transparency, both MP3 and AAC have killer samples, and focusing on outliers is not conducive to gathering data on how encoders perform on real-world samples that are representative of the content normal users would try to encode.

We're getting pretty serious now, but hopefully we'll educate either ourselves or others in the process. biggrin.gif
Aleron Ives is offline  
post #64 of 65 Old 11-01-2013, 06:18 PM
AVS Special Member
 
kiwi2's Avatar
 
Join Date: Nov 2010
Location: New Zealand
Posts: 1,630
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 6 Post(s)
Liked: 144
I just use FLAC.
67jason likes this.
kiwi2 is offline  
post #65 of 65 Old 11-02-2013, 10:34 AM
AVS Special Member
 
jdcrox's Avatar
 
Join Date: Jan 2003
Location: Cape Cod, MA
Posts: 2,010
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 30 Post(s)
Liked: 63
Quote:
Originally Posted by kiwi2 View Post

I just use FLAC.
+1. Storage is sooooo cheap these days.
jdcrox is offline  
Reply Audio theory, Setup and Chat

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off