Slashdot Mirror
Are iTMS's 128kbps Songs Worth Collecting?
pinchhazard writes "Randall Stross of the New York Times offers his opinion on iTunes Music Store's decision to offer downloads at only 128 kbps, and that decision's potential to affect collectibility of the songs. The article says that Apple makes the claim on its web site that "you'll get the full quality of uncompressed CD audio using about half the storage space."
Rhapsody, which offers encoding at 192 kbps, is compared."
The "Half the size" bit is about Apple Lossless
That what was all this school was for... to teach us how to solve our own problems. -- janeowit
Bookmark BugMeNot or get the BugMeNot Firefox plugin. Use it. Love it.
Apple's site uses the "you'll get the full quality of uncompressed CD audio using about half the storage space" in reference to the Apple Lossless codec, not the 128 kbps compression in iTMS songs.
Just check out CDs from the Library and rip them with abcde to flac and archive the .flacs on DVD-R (you can fit about 11 "CDs" per DVD), then make .ogg copies or whatever for your devices.
Yes, unquestionably. Actually, the poor overall quality of low-bitrate lossy encoding is what deters me from iTunes et al.
It should be noted that the defects of inferior recordings become increasingly apparent with better playback hardware. Limitations of consumer-grade hardware is a key limiting factor to the widespread adoption of higher quality audio recording formats (both physical media and encoding schemes).
A few points:
1. The Nyquist Theorem states the maximum possible encoded frequency in a digitized waveform. It says nothing about how the waveform may or may not suffer aliasing as the frequency approaches half the sample rate. I.e. a rate of 44.1khz is necessary (but may not be sufficient) to encode a 22.05khz tone. I'm not sure this was clear in your reply.
2. "Human ears listen up to about 16kHz." Leaving aside the variance between ears (which is huge- some can hear above 20khz), nigh-subconscious overtones depend on these frequencies. Even if you can't hear these high frequencies alone very well, they do (measurably, and meaningfully) add something to music. Just crop everything above 16khz on a song and listen critically.
3. "A CD delivers audio at 1411.2kbps. The CD audio format was created to conform to what is the best that human ears need." Yes, based on 1980s research. We've come a long way in audio theory, though. Also, all bits are not created equal- I guarantee you that a DVD-A stream compressed into 1411.2kbps would sound better than a CD.
I think my points still stand.
Best,
RD
Whatever your opinions on AAC vs. MP3 vs. Ogg and so on, anybody reading this article should know that Randall Stross has an extreme bias against Steve Jobs. Stross is the author of "Steve Jobs and the Next Big Thing", a historical piece on NeXT Computer. You can't go two pages in that book without running across Stross editorializing (negatively) about Jobs' personality or intelligence. Not very professional for somebody calling himself a "historian".
So, aside from the fact that Stross is a completely non-technical writer, take his views on Apple strategy and products with a grain of salt the size of Gibraltar.
1. The Nyquist Theorem states the maximum possible encoded frequency in a digitized waveform. It says nothing about how the waveform may or may not suffer aliasing as the frequency approaches half the sample rate. I.e. a rate of 44.1khz is necessary (but may not be sufficient) to encode a 22.05khz tone. I'm not sure this was clear in your reply.
Sorry, this is simply not true. The Nyquist theorem states that you can completely reproduce a band-limited signal if you evenly sample at twice the bandwidth of the signal. If the signal is not band-limited, the content at frequencies above the half-Nyquist rate will be aliased back into the lower frequency spectrum.
A tone, by definition, is a sinusoid and has no higher harmonic content. Therefore, by your example, a 22.05kHz tone (i.e. a sinusoid) can be reproduced with no aliasing by a 44.1kHz sampling rate. However, when your signal is very near the half-Nyquist rate, the phase of the signal becomes important. In practical terms, that is why you usually slightly over-sample your band-limited signal if phase information (i.e. exact reproduction of the signal) is important. For CD audio, the goal was to accurately reproduce 20kHz signals, therefore, the slight oversampling to 44.1kHz (Note: low-pass filter responses also contributed to the need to slightly oversample).
Bottom line, the aliasing in your example comes about because you are not talking about a band-limited signal since you have a non-sinusoidal waveform with its fundamental at 22.05kHz but higher harmonic content at integer multiples of 22.05kHz.
Being the skeptical type, when iTMS was launched, and having not been impressed w/ 128k mp3s I did my own quick A/B test of 128k AAC / AIFF from the CD. First, Rip a song (preferably one w/ a significant dynamic range to 128k AAC in itunes. Then select the song and choose "Show song file" from the file menu. Right/Control click (thats either right -or- control, yes, you _can_ use a multi-button mouse w/ a mac.) and choose "Open With" then select Quicktime Player from the submenu that appears. Open the song track from the CD in a similar manner. Now you have the original and your 128k AAC both open in QT Player. Then select "Play all Movies" from the Movie menu. both will start simultaneously. Now you can option-tab to switch between which as focus (and thus which is heard) and do a real-time AB test. It put me at ease. Once you have your hand in place you can close your eyes and randomly switch back and forth a bit to loose track, then try to guess which you're listening to.
MSRP - Tax, Title & Licence Extra Your Milage May Vary
It's not a /. contributor that really misleads, but the column (and Stross, it's author) itself.
The scary thing is that to the uninitiated you will sound like you know what you are talking about even though you haven't the slightest clue.
First you say that 320 kbps is lossless because the waveform is "pretty much the same". Then you demonstrate your extreme ignorance of the very process you try to sound knowledgeable about by implying that all 256 kbps compression does is filter frequencies above 20Khz. This is wrong. Completely wrong. Not even close to how it works.
Your assertions about what people can hear or not hear are just that: assertions. It's totally meaningless. Just because audiophiles have not bothered to produce some kind of test to your satisfaction that proves beyond any doubt whatsoever that they can hear the difference does not mean that they cannot hear the difference.
I have done so called abx testing of my own using the software from hydrogenaudio, and I was not only able to correctly distinguish which tracks were compressed, but also which tracks were encoded with which codec (although not as accurately).
I am confident that I could personally pass any legitimate test you could come up with comparing 256 kbps material with the CDs. I will admit that 320 kbps would be more difficult, but given a sufficient amount of time and high enough quality source material, I could blindly identify those as well. So much for your 320 kbps is lossless theory.
There is one codec that I am not completely confident I can identify though: MPC. At a high vbr this codec tends to sound really good, even to me. The author of that proggy really did his homework when it came to psycho-acoustic compression. I haven't done a lot of testing with it. Maybe it really is transparent. I don't know.
I am poor and I can assure you that I do not buy CDs out of the goodness of my heart or because I feel sorry for the rich record company executives and "artists". I buy CDs because I can hear a very significant (varies based on source quality and bitrate) difference.
Quite an experience to live in fear, isn't it? That's what it is to be a slave.
You don't want to be hooking the headphone socket of your iPod to your stereo. Use the dock instead, which has a real line-level output.
Expensive studio mics reproduce the full range 20-20,000 Hz and leave it to the sound engineer to filter out high frequencies if necessary. Here's a real studio mic, a Neumann U89, -4 dB at 20 kHz (see PDFs under "Documents"). Good for about $3000. Or, an order of magnitude cheaper, Audio-Technica AT853a.
I sometimes use the latter type for making live recordings of chorus performances on minidisc and apparently, the white-noise background also extends to 20 kHz. It seems that the Atrac-compression (350 kbit/sec) has a hard time with the noise because you don't need golden ears to hear the compression artifacts.
Avantslash: low-bandwidth mobile slashdot.