|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
99.185.61.185
I'm no Apple zealot. But I like scientific rigor and I question sweeping pronoucements from dubious data.
Production Advice's Ian Shepherd posted a YouTube video in which he tested whether a Mastered for iTunes AAC file 'sounded more like the CD' than a homemade AAC of the same song created by standard iTunes. He declared his homemade file the winner and labeled Mastered for iTunes a failure. Check it out. His "I Say BS" headline has been re-blogged and re-tweeted more than 2000 times without challenge. Carumba!
I believe there are fundamental flaws in his test method that made his conclusion inevitable, which also makes it neither surprising not significant. (Can you say, "tautology?") And I think I have the data to back it up. Judge for yourself.
Follow Ups:
And the blurb shoud read "Butchered for iTunes".
.....does not mean "sounded like the CD"............I was never impressed with iTunes-sourced recordings. And I believe if sourced from CD, it is not overly difficult to construct music files of similar format and compression which are noticeably superior from a sonic standpoint.
Edits: 03/15/12
"Hypothesis #2. To get an objective measure of correlation, I mixed aligned versions of all WAV/AAC pairs (one in-phase, the other phase-inverted) and calculated the RMS volume of the resulting difference wave. That is, on average, how audible the difference is between the WAV file and AAC file tested. Results below."
The use of RMS differeneces to evaluate errors produced by complex encoders that are based on psycho-acoustic principles is invalid. This is akin to comparing distortion in amplifiers by a single "THD" number.
There are two main suggestions in Apple's recommendation: keep levels safely below maximum and use 24 bit input to the encoder. Both of these have been known for some time, at least by those engineers who would rather try and avoid having anything to do with AAC or MP3. The errors caused by overloaded encoders and decoders can be gross. The differences gained by encoding with 24 bit audio are subtle, more akin to the differences between 24 bit and 16 bit uncompressed PCM at 44 kHz. Even if both suggestions are followed the results will still be rotten, it's just a question of the degree of stench. If Apple were really interested in sound quality they would be offering lossless downloads in the iTunes store.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
I downloaded the Master for iTunes tools and procesed my test set of 24-96 WAV masters. In all cases, the afconvert program in the MfI toolkit produced an AAC file virtually identical to the one produced by standard iTunes. They were not exactly identical, but the audible difference amounted to occasional burps and murmurs. You have to conclude that the tools temselves are nothing new.
So it seems MfI is a concept, a carrot for the industry to create unique masters optimized for compression using existing tools, because there wasn't the incentive before. But, does certification a priori imply improvement?
There's a huge thread on the Gearslutz Mastering Forum. My conclusion is that this the Mastering for iTunes program is marketing hype. Compressing CD quality (or higher music) into 256 kbps AAC is like trying to put 5 or 10 pounds of music into a 2 pound bag. Using the best possible procedures may get you the full 2 pounds instead of 1 3/4 pounds.
If iTunes sold music encoded with ALAC at the same price then this issue simply wouldn't arise. Anyone who cared about sound quality could just download the ALAC version. If they wanted to store their entire library in a portable device they could use their computer to compress to lossy. It makes no sense to encode downloads in a lossy format today with bandwidth so cheap. The only possible reason for using lossy encoding today is for streaming applications over wireless networks (or to people who live in the third world when it comes to Internet connectivity).
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
They also recommend keeping 32-bit floating point intermediate representations at every point until final exit. I can't argue with starting from the highest resolution possible and preserving mathematical precision every step of the way.
I agree with the suggestion to keep 32 bits for as long as possible. I had been in the practice of keeping intermediate files in 24 bits, but then a critical listener pointed out that portions of a file that I hadn't gotten to editing as yet did not sound as good as the original 24 bit file. I'd been removing clicks and pops and had already sunk several hours into a few minutes of a 30 minute recording and the portions that I'd fixed were obviously better, but I was shocked at the comment that the other portions sounded different. So I did a null test between the original file and the partially edited file, looking only at the end of the file that I hadn't gotten to yet. The file nulled out at -135 dB, i.e. all of the 24 bit samples were either identical or differed by +- 1 in the low order bit. This difference produced a different sound. Now I keep all of my master files at the highest available resolution, which is 32 bit floating point for most of my software. (This difference was caused by the edition of dither noise when converting from 32fp to 24int. Successive unnecessary conversions will add additional amounts of dither noise.)
Actually, 32 bit floating point is not good enough either for any serious processing such as equalization or sample rate conversion. At least 48 bits of working precision are needed and this means using 64 bit floating point in most computer systems. I suspect in a few years we will see workstations keeping files in 64 bits. One can't keep track of a million dollar bank balance down to the last cent if one uses 32 bit floating point. :-)
In the article you linked there is an incorrect reference. "The map is not the territory," was said by Alfred Korzybski, not Marshall McLuhan. Marshall McLuhan is famous for the phrase, "The medium is the message." With AAC or MP3 the medium is definitely the message, and the message isn't music. (Some say the same when it comes to 44/16 digital.)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
.
Thanks for the catch. You are correct re: the quote. The great thing about pucblishing online is that errata can be addressed immediately, and not in the "next edition."
If you haven't read it, you might enjoy "The World of Null-A".
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Interesting premise. I'll add it to my reading queue.
Thanks!
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: