John Atkinson STEREOPHILE article: "Sound quality of MP3 vs AAC vs FLAC vs CD"

Read that last night. I want to show those graphs to some people who claim the music is not being corrupted by these psychoacoustic algorithms.

 

Thanks for that! Good reading

-s1m0n-

 

Great reading and it's about time an industry pro took this on. Thanks Steve!

From what I've heard, my opinion only, I'm not in agreement with this part of the statement.

"Lossless compression is benign in its effect on the music. It is akin to LHA or WinZip computer data crunchers in packing the data more efficiently on the disk, but the data you read out are the same as went in."

 

I think that if it weren't for the space restrictions on portable players (especially the flash memory ones), 128kbs MP3 files - which were once considered the standard for music downloads - would already be a thing of the past. They sound too awful.

Download speed is no longer an issue. It takes longer to download a complete album in high quality or lossless format, but nobody buys so much music that his connection would be too slow.

As soon as 256kbs MP3 or AAC become the standard for music downloads, it will be difficult for hifi fans to argue that lossy compression means poor sound. Blindfold tests anyone? These files are mostly used with iPods, not high end equipment, so they don't need to be in perfect CD quality. For the audiophiles who play downloads at home, a lossless option would be fine.

The graphs are interesting, but they are irrelevant for someone who doesn't hear a difference between a lossless file and a 256 or 320kbs MP3. Would you upgrade your hifi cables only because of some measurement graphs, when you can't hear a difference?

 

Great link, Steve - essential reading!

I ran some tests like theirs at home a while back and posted the results in a thread. Like the article demonstrates, the lossy encoding, even at 320kbps discards quite a bit of info (mostly ambience and upper frequency content). The audible difference at 320, however, was fairly negligible. Like 5% even when I was aware of what file I was playing.

And I learned something! -

Hadn't thought about transient smearing in the encoding process before.

 

as a musician, i want the listener to hear the sounds as i intended when the track was recorded. if the basic file is compromised for any reason, the listener is being cheated

 

Dave,

I tend to agree but I think Mr. Atkinson covered this in the below footnote. You don't get anything for free and maybe what we notice even with lossless compression is the higher error rate resulting in some loss of fidelity. Interesting thought, but your ears are better than mine.

"Footnote 1: Something I have rarely seen discussed is the fact is that because all compressed file formats, both lossless and lossy, effectively have zero data redundancy, they are much more vulnerable than uncompressed files to bit errors in transmission."

 

Very interesting. Thank you for posting the link, Steve. It partly answers my question in the Beatles downloads thread.

Robert

 

Steve, good point. I missed the correlation between the two as John was kind of soft-shoeing it in your quote.

 

It is a good read, with objective testing, but doesn't really seem to accomplish all that much, all things considered.

If one is a Stereophile reader, either in print or online, you probably already know that lossy compression is a bad thing. And that opinion was probably formed from listening, or you already understand the concepts which JA related via the testing graphs.

I'm perfectly fine with lossy for use in my iPod. But I have several friends (even those who like good LPs ) who don't mind storing their digitized music collection in MP3 format on their computer. This is typically because MP3 is the most convenient lossy format to the average user. Remember, "convenience is king"! And iTunes is an easy-to-use interface for the majority of users, and it of course defaults to lossy encoding for CD rips.

Good effort by JA, but he's basically preaching to the choir IMO.

 

Pretty much agree with Sam. There is a lot of preaching to the choir going on in Atkinson's article (which was very interesting btw). Glad he acknowledged that lossless is lossless; far too many people still don't seem to get that. As for compressed mp3 files, those who care about the quality of the music know it already: compressing is a compromise between sound quality and storage space. Just as i find listening to 128kps on my home system unacceptable, I also find listening to wave or flac files on my iPod - over cheap headphones and often in an already noisy environment - a waste of perfectly good storage space. Both formats have their uses, depending on ones' needs and listening conditions. Mp3 or AAC is a lossy process - it's intended to be - but if done properly (and by "properly" I don't mean 128kps) it can sound perfectly fine under all but the most critical listening.

 

Good read, IMO MP3 is still crap.

 

Why no mention of VBR? The only tests he used were CBR.

 

Exactly-there's nothing new here or anything we didn't already know. It all depends upon how you use the files. I have 2 iPods-one of which is a 4-gig Nano. At home, neither is used as a primary music source in my system except listening to various podcasts. But it's a question of quantity or quality, quantity wins out because of the places where I use the 'pods-in my car, outside exercise, lawnwork, flights, etc.

 

I don't think that matters, he does a test at 320kbps which basically generates the highest possible files with MP3 and AAC codecs.

I don't really see the point in this article myself, I agree he's preaching to the choir, and there is nothing new revealed in this article.

 

I wouldn't go so far as to say he's preaching to the choir. There's plenty of people around this place alone who don't even know the difference between lossy file compression and dynamic range compression.

Yes, plenty of articles and white sheets have been done about this, but some folks have never read them. It just takes one to raise awareness.

 

Good article.

I guess I'm odd because I consider CD to be a lossy format.

 

I was just going to mention that. Most records, even homemade albums, are recorded at the 24 bit level. CDs really are a lossy format.

I see all this lossy data compression eventually going away to a large degree as storage becomes less of an issue. These are the dark days like when cassettes ruled the world. It'll get better.

 

I'm a big fan of MP3s and believe I cannot tell the difference between a LAME -V2 encoded file and its source WAV, but that's just me. I repect that many can tell the difference and strongly prefer lossless.

I did not learn much from the article, but it helped to reinforce a few things that really irritate me: One is that 128kbs has been successfully marketed as "CD Quality" and 96kbs as "FM Quality". I would rather listen to a beat-up garage sale 45 than a 96kbs MP3 music file. 10 years ago it was great to get a mix tape or CD, but now I dread getting a mix CD from most friends because they source them from 128kbs AAC files.

My big question is: Can the masses really not tell the difference between 128kbs MP3/AAC and lossless or are they just apethetic toward the difference?

 

We just played my new violin recording back in 24/176 yesterday and it completely creams the 16/44.1 downrez. Absolutely destroys it.

 

Yes. Sad to think that 'we' (audiophiles) are at the point of fighting for something that is not what we really want. I want at least high-res PCM stereo, 96khz 24bit. Downloads will be tempting to me only when if I can download flacs from PCM 96/24. But there will be sometime until home internet connections will be fast enough to do that in a reasonable time.

 

So true. Boy have I ever slaved in studios going for nuance, room sounds and seperation.

 

Lossless is LOSSLESS. I've used the FLAC encoder (both 1.1.4 and 1.2.1) to encode upwards of 20K files, always with the "-v" (verify) option, which runs a decoder in parallel with the encoder, comparing the decoder's output with the encoder's input. It's never reported even a single bit of difference. On several occasions, I've loaded FLAC files into Cool Edit to monauralize monaural tracks extracted from CDs, and Cool Edit shows the left and right tracks of the decoded FLAC to be numerically identical (except for dither), giving me further confidence that FLAC is indeed lossless. If you're hearing different results, chances are there's some sort of flaw in your encoding/decoding system (yes, it does happen, according to the FLAC Sourceforge page; that's what the "-v" flag is for).

 

I would never hear lossy music at home, but...

320 kb and lossless files are fine for me, when I'm on the way.... can't live/walk without music!!!!!!!