Digital Audio Formats: A Guide

[mlts22]

From what people have said, I'm pretty sure they use the Fraunhofer's encoder. (The MP3 version, not the MP3Pro version which is compatible with MP3, but allows for better low end, though I have not messed around with the MP3Pro format.)

[dragfree]

T-logik--This is an excellent introduction to audio formats, but there are a few opinions presented as if they are facts.

"MP3 can achieve transparent, or indistinguishable from CD, quality at around 192kbps-256kbps."

Even at 320kbps MP3 is distinguishable from CD with good source material and playback equipment, though less so with typical pop recordings played through an iPod or other DAP with standard earbuds.

"When using the LAME encoder, the --alt-preset standard setting is recommended. It provides bit rates usually between 192 and 256 kbps depending on the complexity of the music and provides transparent quality on most music."

Recommended by whom? I use LAME alt-preset insane 320kbps. It sounds better than alt-preset extreme, which in turn is clearly superior to alt-preset standand. The 320kbps setting comes closest to "transparency."

"AAC has 30% more encoding power than MP3."

30%? This figure is meaningless, yet presented as if it were a fact based on some objectively measurable criterion.

"At any bit rate, AAC will sound better than MP3."

I cannot agree completely. I rather like AAC at 256 and 320. At these bit rates it sounds pretty darned good to me. But so do LAME MP3s. And they all sound similar enough to one another and to the original digital CD files that I would have a difficult time choosing one as superior to another.

"a 128 kbps AAC file sounds closer to CD quality than a 128 kbps MP3 file."

I certainly agree that 128kbps AAC files sound better than 128kbps MP3s. AAC was developed to provide more acceptable quality than MP3s at the low bitrates suitable for streaming transmission over the web. The developers claim that AAC@128 sounds as good as MP3@160. But not even they claim that it's very close to CD quality.

"Ogg Vorbis provides better quality than MP3 and is about equivalent to AAC in terms of quality at the moment."

The listening tests you cite rate Ogg superior to AAC--and that's at 128kbps, the rate which AAC is engineered for, whereas Ogg is designed for high quality instead of high compression and really shines at high bit rates. Interesting that you cite the statistically insignificant difference between LAME and AAC as evidence that AAC is superior, but disregard the statistically significant difference confirming Ogg's superiority to AAC.

For a pretty good basic introduction to lossy compression and competing formats, I recommend this Wikipedia article: http://en.wikipedia.org/wiki/Lossy_data_compression

For an interesting comparison of MP3 and AAC, I recommend this paper on the subject by a Fraunhofer engineer (one of the developers of both): http://www.aes.org/publications/down...03162000122117

[Teqnilogik]

Thanks for pointing some things out, however, without starting a debate, I'd like to ask you a few questions.

You say Preset Insane sounds better than Preset Extreme. Did you do an ABX listening test to determine this? I'm asking because if you didn't, the placebo effect could be in play here.

And I would say that not ALL lossy codecs are easily picked out at high enough bit rates. If you take say, MusePack, and encode a track at the Braindead setting, I would bet that the majority of the people ABXing that track would not be able to tell a difference between it and the original CD quality.

I did make some modifications to my guide, thanks for pointing out potential areas where people may disagree.

[dragfree]

Hi, TL. I commend you for the modifications and for your humility in making them. And I apologize for what in retrospect seems an overly critical tone. Yes, I've done ABX testing (using Foobar) of LAME MP3s and AAC at different bitrates, and using good classical & jazz recordings which are generally more revealing than typical pop fare. I would agree that LAME extreme sounds very close to LAME insane, yet in critical passages I can distinguish between them enough of the time (around 7 out of 10) to back up my general impressions. Note they are certainly much closer than AAC at 128 and 256, between which I scored 100% accuracy without effort the only time I tried. And my testing between LAME and AAC at 320--as with AAC at 256 and 320--was inconclusive, which isn't surprising since both sound very similar to CD to me as well, at least when played back through my PC.

If I were extremely concerned about storage space, I'm sure I could get along happily with LAME extreme or AAC@256 strictly for iPod playback, but the extra 20 or 25% storage space the larger files require strikes me as cheap insurance against feeling the need to re-rip in the future, as does using the LAME MP3 format instead of AAC. As for MusePack, I'll take your word for it, especially since I have a hard enough time distinguishing between LAME and CD as it is!

[fallenipod]

ipod 6g needs to be way bigger memory and better battery so we can use lossless format. i listened to it and it sounds good and would really sound good wired to a car. i want that quality but it was 1064kbps lol.

[kornchild2002]

Wow, after 9 months I never thought a topic was going to be started in this thread again. Yes, I think the next generation of iPods need a much larger capacity. However, Apple would have to introduce a 200GB iPod in order to the average person to fit all their lossless files on it. Additionally, Apple would have to put a massive battery and a lot of buffer RAM (along the lines of 512MB to 1GB) in order to get decent battery life. Right now the technology really isn't there for portable lossless playback. I think Apple is moving towards this direction though.

One day someone will come out with a lossless encoder that allows for large volumes of compression. I see a mpeg-4 based lossless encoder coming out in which songs take up the same amount of space as if they were encoded at 320kbps mp3.

So, right now there are two things that need to be worked on. A more efficient lossless encoder in which a 4 minute song would take up about 8MB (instead of the standard 22MB or even 40MB) and we need larger capacity iPods with longer battery life.

Still, I chalenge anyone who does CASUAL listening to differentiate betweeen a -V 2 --vbr-new Lame 3.97b2 encoded mp3 or a 192kbps VBR mpeg-4 AAC file and the lossless source. Again, just talking about causal listening. I am not talking about you sitting down in a quiet room with your high end headphones or audio equipment where you are listening for flaws and trying to pick which song sounds more like the original.

Lossy encoders have come a long way though even after this thread kinda died down. Recent tests by hydrogen audio show that 128kbps VBR mpeg-4 AAC (iTunes) is very listenable and that soon the standard bitrate for near CD quality will probably change. Lossy encoders are becoming for efficient. One day you will probably be able to achieve CD quality at 96kbps or even 64 kbps (WMA9 doesn't count those lying #######s).

[shabbs]

Originally posted by fallenipod
ipod 6g needs to be way bigger memory and better battery so we can use lossless format. i listened to it and it sounds good and would really sound good wired to a car. i want that quality but it was 1064kbps lol.

Well, of course it will sound good, it's lossless. Heh heh. It will sound as if you have the original source in there.

But, does it sound any better than a high qualiy LAME MP3 or AAC encoded file? Can you really tell the difference? I doubt the quality can be differentiated, especially using the stock Apple iPod ear buds while your walking around downtown...

Cheers.

[oagersnap]

Originally posted by Teqnilogik It [Monkey's Audio] generally compresses better than FLAC using the highest setting, however, having the best compression isn't the best thing when talking about a lossless codec.

Why not? It's lossless, so it will be the same quality as the original sound; to me it would seem that having a higher compression would be better.

[Teqnilogik]

Odds are, the quality different you hear is placebo effect in play. You think it sounds better because you know the quality is techncially superior than the other file. Do a blind ABX test with the files and that will tell you whether you can truly tell a difference between the two audio files. You will most likely be surprised by the results. I have a hard time distinguishing most samples from 128 kbps AAC and the original files in ABX tests. Try it and find out if you really can tell a difference.

[yinyang]

Originally posted by Teqnilogik
Odds are, the quality different you hear is placebo effect in play. You think it sounds better because you know the quality is techncially superior than the other file. Do a blind ABX test with the files and that will tell you whether you can truly tell a difference between the two audio files. You will most likely be surprised by the results. I have a hard time distinguishing most samples from 128 kbps AAC and the original files in ABX tests. Try it and find out if you really can tell a difference.

just to support that last comment, this is from a recording engineer who 'used to be' very skeptical of anything lossy...

http://www.kenrockwell.com/apple/itunes.htm

very interesting read

[shabbs]

Originally posted by yinyang
just to support that last comment, this is from a recording engineer who 'used to be' very skeptical of anything lossy...

http://www.kenrockwell.com/apple/itunes.htm

very interesting read

The one thing that bothers me about that article is this:

MP3, actually MPEG-2 layer 3, was OK several years ago and is has established itself as a popular format for illegally copied music.

I wonder if he's given EAC/LAME a look? I think he'd be quite surprised at how far LAME has come and how powerful EAC is.

Cheers.

[kornchild2002]

Originally posted by shabbs
The one thing that bothers me about that article is this:





I wonder if he's given EAC/LAME a look? I think he'd be quite surprised at how far LAME has come and how powerful EAC is.

Cheers.

I e-mailed the guy, seems nice enough. The anal person in me asked him to change the mp3 format from being mpeg-2 layer III to mpeg-1 layer III the way it should be. He did do this. However, I asked him about trying EAC and Lame at -V 5 --vbr-new, -V 4 --vbr-new, and/or -V 2 --vbr-new and he said he wouldn't do it because he didn't have the time to test all those "formats." I am just a little perturbed that he would make such a bold statement about the quality of the mp3 format without actually testing the BEST mp3 encoder out there. You simply cannot make a statement about the mp3 format without testing Lame, it is sacreligious. For example, it would be like me critisizing the Mandraxe Linux OS when I haven't even seen it run on my computer. You can't critisize something without even looking at it.

There, that was my rant, I got it out of the way. The guy did seem nice though but until he tests the Lame mp3 format, I won't take him seriously. Now, if he were to conduct a blind ABX test with the Lame mp3 format and the iTunes AAC format then I would have more respect for the article. But audio encoding is not the main subject of his website.

[juancho]

I'm surprised not to see any mention of wav files here. Are not wav files 'as close to the original' cd as we can get?

Also, is there any difference in qulaity between the wav encoder in I Tunes and that in EAC?
Thanks
juancho

[kornchild2002]

Originally posted by juancho
I'm surprised not to see any mention of wav files here. Are not wav files 'as close to the original' cd as we can get?

Also, is there any difference in qulaity between the wav encoder in I Tunes and that in EAC?
Thanks
juancho

The reason you don't see wav being talked about is because it sucks when archiving your music. Also, here is a little equation for you: WAV=Lossless=CD

A wav file has the same quality as the CD and a lossless file has the same exact quality as the wav file and the CD. The reason wav sucks when backing up your music is because the files are larger that lossless files and wav files cannot retain tag information. A lossless file will take up about 10% less space than a wav file and the lossless file will contain the trag tack information (things like artist, genre, album, year, etc.).

As far a the wav encoder used in iTunes and one used in EAC, they are technically the same. The only thing that is different about iTunes and EAC is the way they rip CD's. If you want a secure rip of the CD, then use EAC. If you want a decent rip of the CD, then use iTunes.

I cannot say this enough, stay away from wav when archiving your music. There is no need for you to have to plug-in all your ID tag info everytime you transcode from a wav, that is why you choose lossless files. A lossless file has the exact same quality as the original CD. There is no difference between a lossless file and source wav file in terms of audio quality.

Hence, that is why wav really isn't talked about that much.

[Teqnilogik]

I have updated the MP3 and AAC sections of this guide to reflect some changes that have occurred for both formats since the last update.

For example, instead of referring to presets in LAME I now reference to the Q modes with the --vbr-new command that are recommended for use now.