Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
How's about presenting a Matlab example... feed it in a train of 16-bit Linear PCM samples taken from a "typical" sound source... Run it through a well-implemented AAC encoding scheme... Run that through the standard AAC decoding algorithm...

Compare the output set of samples against the input set... What is the typical numeric error comparing sample to sample from input to output?

That type of comparison doesn't factor in the way we, on average, hear and process sound. So it wouldn't be a fair comparison from the point of view of what a human would/could detect.

Audio compression algorithms (like image and video compression algorithms) are designed with humans in mind. They throw out aspects of the audio signal the human ear either cannot detect and/or will synthesize.
 
In a real music player there must be an analog low pass filter after the D/A converter. So say the sample rate is 44Khz it would be low pass filtered to about 20Khz before being sent to the headphones. A proper test would hve to compare the waveforms after low pass filtering.

Precisely.

The reconstructed analogue waveforms are what must be compared for a reference sample in order to predict what, if any, artifaction should be detectable.

The ignorance of audiophiles to the most basic principles of digital audio encoding and systems design is really beginning to show when something as basic as a low-pass (antialiasing) filter at the Nyquist limit is missed.
 
EMI could have made an announcement on their new DRM-free music without playing favourites and tying the announcement in, specifically, with iTunes.

But I wonder if this wasn't a nod of appreciation to Jobs and Apple who, through iPod and iTunes, stepped into the morass of music 'sharing' that existed and offered those who wanted to pay for their music (be if for reasons of moral conviction, convenience, or assurance of a certain quality of the song file) the first real and legit option to do so.
 
Precisely.

The reconstructed analogue waveforms are what must be compared for a reference sample in order to predict what, if any, artifaction should be detectable.

The ignorance of audiophiles to the most basic principles of digital audio encoding and systems design is really beginning to show when something as basic as a low-pass (antialiasing) filter at the Nyquist limit is missed.

http://en.wikipedia.org/wiki/Psychoacoustic_model

Is a good introduction why a lossy compression / expansion will lead to a totally different waveform. This also just barely scratches the surface of the science behind the algorithms.
 
http://en.wikipedia.org/wiki/Psychoacoustic_model

Is a good introduction why a lossy compression / expansion will lead to a totally different waveform. This also just barely scratches the surface of the science behind the algorithms.

Your last sentence is totally correct. This article is worthless to someone who already has a cursory understanding of the fundamentals of digital audio. A better reference is Ken Pohlmann's Principles of Digital Audio which dives heavily into the technical design requirements of digital audio encoding systems.

What that Wikipedia article does underscore is no different from my point. Even if differences in the overall waveform DO exist, the question is which of those differences are actually within the A-weighted spectrum and/or otherwise perceptible. Knowing what we know about psychoacoustics, we should be able to predict what, if any, artifaction is perceptible in an analogue soundwave reconstructed from a particular perceptual coding algorithm.
 
Seriously what is so great about the removal of DRM? Was it really stopping you from doing ANYTHING? I didn't even know that it was on there....and neither does the average user....so it's not like sales are going to skyrocket...this is a stupid topic
 
I can't wait to hear the official response from Microsoft.

I wonder if they will have a scare campaign for the record lables to show how Apple's move is a threat... Blah, blah, blah...:rolleyes:
 
Seriously what is so great about the removal of DRM? Was it really stopping you from doing ANYTHING? I didn't even know that it was on there....and neither does the average user....so it's not like sales are going to skyrocket...this is a stupid topic

It stopped people from playing the tracks on their mobile phones, on their DAPs (that were not ipods), on their TIVOs, on linux, on their car stereos (unless you burnt an audio cd), basically on any non apple device.
 
I think this is an iTunes-wide option and not just for Shuffle. But either way, anyone can just downconvert the files to a suitable size, format, whatever since... hello... there's no DRM!

So if the non-DRM alone is reason enough to buy, then you pay your 30 cents difference, get the file, make a 128kbps copy to pop on your ipod and keep the original 256 for whatever else... AppleTV perhaps.

You'll have to pardon me... I just find it oddly amusing that people have known that non-DRM digital music files are not tangibly-fixed product and yet people seem to be thinking that what they're getting for $1.29 is somehow a fixed product that they can't themselves modify.

You'd think it were obvious, but wasn't the entire point of removing DRM to allow you to do whatever you wanted to your copy of the file?

:)

Yeah, but that's messy. It takes your actively doing this to some set of files that you choose, and then you have two of each file, etc. On the other hand, if you could simply specify that you wanted all music at 128kbps on your ipod b/c that's what you can hear on your headphones, and all music in itunes at it's max, b/c it doesn't really matter and you have a nice stereo system or headphones, it would be amazingly elegant. Hope we get this.
 
Couple quick thoughts here on this...

First, I'm glad that Apple has put it's money where it's mouth is. It really helps to shut up the nay-sayers (here and elsewhere) with regard to Apple's espousal of DRM in their product.

Second, I think most of the general public are a bunch of ignorant, lazy sheep who (by and large) would have been incapable of getting this achieved on their own, so basically even though some of you folks want to treat this as a non-event, you should really give Apple their due for basically taking the bull by the horns and representing not solely their own interests, but those of the general public as well.

Third, and this is something I've pointed out to people I know (friends and acquaintances) regarding HD-TV, but the entertainment industry has now caught the "features and upgrades" bug that's so long plagued the computer industry, and so even though your discussions here on digital audio processing, digitizing and the cumulative effects on spectra and waveform due to lossy compression are well taken, they're well beyond the reach of the average person using one of these devices or, for that matter, the average consumer using ANY modern audio or video entertainment device.

So, by all means I'm not trying to shut you folks up (honest, go for it, it's very informative!) but just try to understand that the average person couldn't give a toss about relative signal quality, so long as it doesn't significantly intrude upon the human-perceptible range.
 
Second, I think most of the general public are a bunch of ignorant, lazy sheep...

What an arrogant thing to say. I think most of the general public might not care about bit rates or obsess about the evils of DRM, but that doesn't make them ignorant, lazy sheep - it makes them people who have interests that differ from yours.
 
The average use can understand the idea of re-encoding and why it is bad. Just do a simple test:
Take a lossless file(from CD), encode to 256kbps AAC name it "songAAC", encode that AAC to a 128kbps AAC and name it "songAAC2", encode the 2nd AAC at 128 and name it "songAAC3", go further if you want, but depending on the music you should be able to hear the quite annoying difference between songAAC3 and songAAC, while hearing the difference between songAAC and the original lossless file is not so easy and in fact might be impossible even if the original AAC was made at 128kbps.

My point wasn't really to argue whether 128kbps sounded identical to LPCM though, but rather that iTunes users who want(or require on their device) 128kbps shouldn't be forced to re-encode a 256kbps file because of the artifacts introduced. If 128kbps AAc was as good as you're saying it is, these artifacts wouldn't be amplified so greatly even after 1 encode. Go ahead and try it...even the "ignorant lazy sheep" can hear the problems that occur with re-encoding.
 
Seen that many times before. Several problems...

1. It is not a randomized, double-blind test. The subject pool is not picked at random. The subject pool consists ostensibly of people who tend to go to the site which would include, largely, people who have an interest in vindicating their audio snobbery with these sorts of listening tests. Even in a blind test, this lack of randomization will skew the results one way or another. This is not scientific. ...

You're a riot! :p Post you theory about how AAC@128 is "indistinguishable" on a pro sound/music forum and you'd be laughed off it.

You really can't accept or don't understand the fact that many people can hear much better than you. Not in terms of frequency response, but it terms of being able to discern elements of sound that you'll never be able to hear.

Can you detect if two tones are 1 cent out of tune without using a tuner? Do you have perfect pitch? Can you tell the difference between a violin and a voila playing the same note? If I give you the starting note, can you fill in the rest of the notes on a staff if I play a simple melody for you? Can you hear the difference between an open fifth tuned to equal tuning and one tuned to harmonic tuning?

You probably can't (do any of the above.) Why? Because you don't have "an ear." Anyone who's developed "an ear" for music can hear some/all of the things above. That's how it works in the real world. Accept it and move on.

Your knowledge of acoustic theory might help to build a room with less echo, but your "it sounds exactly the same" theories would be of little or no use in the real world of professional musicians or other sound professionals who work daily with sound for a living (other than giving them a good laugh.) :p

This sort of "ear" (for sound/music) I'm talking about isn't something some "snobby audiophile" made up. Sound is realized in the brain (not in a lab, e.g., your machines can't actually "hear."):

Musicians have been found to have more developed anterior portions of the corpus callosum in a study by Cowell et al. in 1992 (Strickland, 2001). This was confirmed by a study by Schlaug et al in 1995 who found that classical musicians between the ages of 21 and 36 have significantly greater anterior corpora callosa than the non-musical control.

You can read some slightly dumbed down information regarding this at the Wikipedia article, music and the brain... If you really desire to understand how hearing (vs sound) works in the real world, you should back away from your slide rule and pocket calculator and put your pencil back in your pocket protector and do some research on the stuff of the real world. ("Hearing" occurs in the real world.)
 
I will wait and listen to one that is bought from a friend. I don't really care about the DRM thing, since I only use an iPod
 
You're a riot! :p Post you theory about how AAC@128 is "indistinguishable" on a pro sound/music forum and you'd be laughed off it.

You really can't accept or don't understand the fact that many people can hear much better than you. Not in terms of frequency response, but it terms of being able to discern elements of sound that you'll never be able to hear.

Can you detect if two tones are 1 cent out of tune without using a tuner? Do you have perfect pitch? Can you tell the difference between a violin and a voila playing the same note? If I give you the starting note, can you fill in the rest of the notes on a staff if I play a simple melody for you? Can you hear the difference between an open fifth tuned to equal tuning and one tuned to harmonic tuning?

In 1997, a series of tests were conducted jointly by the BBC, NHK (Japan Broadcasting Co.) and at the CRC Signal Processing and Psychoacoustics Audio Perception Labartifacts. Using 24 participants in double-blind procedures, including 7 musicians and 6 recording engineers, they were fed a series of samples, AAC 128Kbps tested higher than any of the other codecs including AC-3.

The panel also concluded that AAC 128Kbps met the requirements for perceptual transparency, and stated:

The AAC codec operating at 128 kbps per stereo pair was the only codec tested which met the audio quality requirement outlined in the ITU-R Recommendation BS.1115 for perceptual audio codecs for broadcast.


These findings were published in the Journal of the Audio Engineering Society in 1998.

Source:

G. A. Soulodre, T. Grusec, M. Lavoie, and L. Thibault, Signal Processing and Psychoacoustics/Communications Research Centre, Ottawa, Ont.: Subjective Evaluation of State-of-the-Art 2-Channel Audio Codecs, Paper presented at the AES 104th Convention, 1998.

This paper was edited and reprinted in the Journal of the Audio Engineering Society:
G. A. Soulodre, T. Grusec, M. Lavoie, and L. Thibault. Subjective evaluation of state-of-the-art 2-channel audio codecs. J. Audio Eng. Soc., 46(3):164 – 176, March 1998.
 
The AAC codec operating at 128 kbps per stereo pair was the only codec tested which met the audio quality requirement outlined in the ITU-R Recommendation BS.1115 for perceptual audio codecs for broadcast.
Isn't that study trying to determine the best codec/bitrate for a radio broadcast? This was at a time when AAC was only first being developed and was far from the quality of today's codecs. I'm sure the study found AAC the best one at 128kbps, but the title of that paper includes the phrase "low bitrate", so they were not looking at transparent audio, rather at what should be used for broadcast at lower bitrates.

You can site all the papers you want, I'm not even sure what you're trying to prove anymore though and if you test a re-encode from 256 to 128kbps AAC, you will see the quality loss due to artifacts and compression that I am trying to demonstrate. I'm sure you're taking a very interesting course in DSP right now and you think you know it all, but as localoid said, you need to get your nose out of the books and train your ears to hear imperfections so you can start to question these decade-old studies and start forming an opinion of your own based on experience.

The requirements of the ITU for audio quality even have considerations for artifacts:
"For emission, the most critical material for the codecs must be such that the degradation may be 'perceptible but not annoying' (grade 4)"
which is fine for the radio communication sector, and I agree with their findings: AAC @ 128kbps is a better choice than mp3 @ 128 or mp2 @ 192kbps. They also conclude that AAC (Main Profile) @ 96kbps gives BETTER results than mp3@128kbps. The point of the ITU, however, is to ensure efficient use of the radio communications, so their goal is ultimately to find the lowest bitrate that can be considered (by them) good enough (allowing perceptible but not annoying artifacts) for radio communications.

What I don't like about their study is that they give no details about the encoders used or the specific settings. One can easily pick out a "bad" mp3 made by iTunes for example, even at 160kbps, but will have more trouble with a more developed encoder like LAME @ 160kbps.
 
I'm not one to promote that tired old mythology about the ongoing Microsoft-Apple rivalry because, let's face it, they're both big companies with their own concerns and goals and they're definitely not sitting around plotting against each other like some Spy Vs. Spy strip.
I don't know about Apple, but Microsoft has been behaving like the black Spy somewhat over the years, plotting the overthrow and destruction of its enemies by various nefarious means, like WordPerfect and Netscape. For instance, they tried repeatedly to kill Quicktime, by threats (Knife the Baby or Else), sabotage (fake error messages in Windows blaming QT, forcing OEMs to remove QuickTime, forcing 3rd party developers to cease QT support, buying hardware manufacturers solely to remove Quicktime drivers, etc), and theft (having an Apple partner provide QT sourcecode and putting it into their own product, for which they were found guilty of software piracy). So there is a certain delicious irony in seeing how Apple foiled MS' attempts at making WM the international MPEG-4 standard by having QT accepted instead, and go on to trounce MS in the "MP3"-player and Download marketplace. Removing DRM could help other stores adopt AAC (or at least go back to MP3), since over 70% of players (iPods) don't support WMA. I hope they will go for AAC, though many current WMA-players may not be compatible.
 
Isn't that study trying to determine the best codec/bitrate for a radio broadcast?

Where do you derive this? The study was to determine standards for television broadcast... unless by "radio broadcast" you are referring to RF transmission.

This was at a time when AAC was only first being developed and was far from the quality of today's codecs.

Hardly. The foundations for AAC actually lay in Dolby SR-D/AC-3 which was developed in the late 1980's-early 1990's.

I'm sure the study found AAC the best one at 128kbps, but the title of that paper includes the phrase "low bitrate", so they were not looking at transparent audio, rather at what should be used for broadcast at lower bitrates.

AES specifically commented on transparency, however.

You can site all the papers you want, I'm not even sure what you're trying to prove anymore though and if you test a re-encode from 256 to 128kbps AAC, you will see the quality loss due to artifacts and compression that I am trying to demonstrate.

I already discussed this... I said it depends on the transcoder being used. Run of the mill transcoders do a crappy direct conversion instead of resampling the decompressed output.

I'm sure you're taking a very interesting course in DSP right now and you think you know it all, but as localoid said, you need to get your nose out of the books and train your ears to hear imperfections so you can start to question these decade-old studies and start forming an opinion of your own based on experience.

No actually I've been involved in professional audio and video for quite some time. I've produced professionally-encoded CD's and DVD's, and I mastered a nationally-released jazz album. I wrote a research paper on internet distribution of music in 1996. I have 15 years of experience with audio and video and while you're jumping to asinine conclusions in your ad hominem attacks (which is no way to substantiate an argument in a debate) I'll point out that during a mass media course in college, sitting in an auditorium of about 350 students I arose and went to the professor and asked him if he had switched on the video projector. He said he did. How did I know? I could hear the 60Hz refresh cycle of the projector over the noise of 350 students talking before class.

What irritates me is the assumption being made that I have no field experience and that I haven't used my own ears to sort things out. But look at it from my point of view... everyone wants to believe that their hearing is impeccable. You can't trust statements like that because they're loaded with confirmation bias. Scientific approaches filter out bias especially where perception is concerned.

What irritates me further is the confusing of Socratic/logical/academic theoretical principles in a book versus the hard empirical observation of science... Science does not rest idle on the pages of a book... Science takes the book and puts it to a practical, systematic examination to see if the hypothesis holds in the real world. But what I hear some people saying is that "No no... ignore the science. ignore the facts observed in controlled settings and instead take my word for it."

Why the hell should I? How do I know I can trust the ears of a bunch of self-affirming audiophiles who sit around and massage their Rotel monoblocks all day over professional engineers who have the ears AND the education to understand what causes what.

Maybe you'll tell me some BS like "controlled settings don't exist in the real world." This betrays a total ignorance of what a controlled setting means. It doesn't mean an environment so sterile and impractical that the results would not be replicable in the real world... it means an environment where other possible phenomena are isolated out so as to not allow any confusion as to what the root cause is.

As good as my own ears are, I don't trust them to tell the whole story... and neither should you. You can goad me with your stories of how audiophiles glued to their self-congratulatory message boards would laugh at me because of my insistence on book knowledge...

Do you mean to tell me I should not listen to SMPTE and AES engineers and instead entertain the opinions of those who speak without having even the cursory/fundamental knowledge of digital encoding and system design to actually know what it was they thought they were hearing and why?

If opinions based on experience are not rooted in fundamental academic knowledge, those opinions can lead down all kinds of corridors of aspersion and syllogistic nonsense.
 
You would argue that the 128kbps AAC from 1980 is "transparent", but even advances in quicktime's AAC implementation in the last few years has greatly improved quality. Plus you're talking about sophisticated decoders, when we are all talking about iTunes(wasn't that the original point of this news post?), even so, can you point us to such a decoder that magically recreated the original perceived audio and is immune to re-encoding artifacts amplified by the iTunes encoder? Like I said, you can read all the studies you want, but until you actually see(hear) it in practice, you shouldn't place all your faith in them. So show me an example of multiple re-encoding steps(as I have above) that produces a file indistinguishable from the first and I'll concede. But to my knowledge, no such process exists for lossy encoders, regardless of bitrate each re-encoding step is losing information, whether perceivable or not, the amplification of this information in later steps is perceivable, that no amount of resampling can ever completely recover. If it could, these codecs would not be considered lossy.

ITU-R deals with radiocommunications, you might be thinking of ITU-T, but the paper you are siting is an ITU-R one and as I said above, it states that some amount of degradation is allowed as long as it is not annoying, and would still receive their recommendation as AAC @128 did.

hey, congrats on the large vocabulary though, makes you look smart
 
So? 20,000 songs is not some sort of guarantee....They always put the disclaimer in the literature for an iPod that the "song count" is based on an average of 4 minute songs at 128kbps.
True, but that's the bitrate, and presumably a pretty average song length, of the songs on iTMS, which is why that disclaimer always made sense. It would not have made sense if Apple had said, "Fits 80,000 songs*...*note: based on 2 minute average song length at 64kbps", even though it would still be technically correct.

Now that they are offering 256kbps files, the 128kbps part of the disclaimer loses a little of its justification.
 
ITU-R deals with radiocommunications, you might be thinking of ITU-T, but the paper you are siting is an ITU-R one and as I said above, it states that some amount of degradation is allowed as long as it is not annoying, and would still receive their recommendation as AAC @128 did.

Yes, ITU-R deals with radio specs, including the standards for broadcast transmission (e.g. ITU-R 601/CCIR 601). Are you attempting to imply that because they deal with radio frequencies (including HDTV quadrature) that their fidelity standards are somehow inferior to your personal standards.

And no, the paper I'm citing is not an ITU-R paper. It's an AES paper (indicated in the original citation) which, among other things, references an ITU-R spec... but the paper in its entirety goes beyond discussion of that spec into the overall fidelity of AAC.

hey, congrats on the large vocabulary though, makes you look smart

Must you repeatedly insist on ad hominem as though the text of your argument cannot stand on the substance of its own merits?
 
Skip the middle men...

Why not skip the middle man, i.e. the record labels, and have Apple produce music instead. At first they'll only have U2, but when most artists realize their music isn't worth anything at EMI they'll sign over to Apple Music instead.
 
Why not skip the middle man, i.e. the record labels, and have Apple produce music instead. At first they'll only have U2, but when most artists realize their music isn't worth anything at EMI they'll sign over to Apple Music instead.

While that sounds great, I doubt Apple wants to deal directly with tens of thousands of artists directly. Apple probably has a minimum collection size to get involved with iTunes. Labels provide a service of providing Apple with a finished product and some like CDBaby don't take that much for the privilege.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.