Ace Engineers Share Tips on Mastering for iTunes

Mastered for iTunes” is out. We explore the tools, the best practices, and the controversy.

Last week, iTunes announced a new set of tools and best practices that would allow engineers to master albums specifically for release on the iTunes Music Store.

Mastered for iTunesAlthough the application is novel, the basic concept is nothing new. Mastering engineers have long made separate masters to account for the quirks of different mediums like CD, vinyl and even tape cassette. 2012 marks the first time that engineers have had the tools, and the economic incentives, to tailor separate masters to compensate specifically for the idiosyncrasies of iTunes’ data-compressed AAC format.

Apple’s 256 kbps AAC files are supposed to sound pretty close to CD-quality and they routinely fool listeners in double-blind listening tests. But when record-producer/living-legend Rick Rubin heard the iTunes version of his new Red Hot Chili Peppers production I’m With You, he was reportedly appalled by how its sound changed during the conversion process.

“He was horrified,” Grammy-winning mastering engineer Vlado Meller told me when I visited him at Masterdisk.

“It was as if they had notched out certain frequencies in order to compress the file. When we did the A/B test with the original and the iTunes release it was like it was two different masters. If it wasn’t for [Rubin] making a stink and putting his weight behind it, we wouldn’t have this today. He deserves the credit for that.”

The Original Process

Rubin and RHCP had already spent months at the end of the production process dialing in the final sound of the record. Rubin told MTV’s Hive  that “it took several weeks of additional experimentation and mastering to reach the final iTunes master,” and that in the end, the difference between the two versions was “night and day.”

When Meller and his assistant Mark Santangelo were given the task of making the iTunes AAC versions more closely match the original masters, iTunes’ new suite of tools was not yet available and the process was arduous.

The two had to run the masters through a specialized replica of the iTunes Store’s proprietary AAC encoder using command-line code, as the store’s codec is different than the consumer-grade iTunes converter. The pair would then A/B this new file with the original master, tweak their audio processors to compensate for any changes in sound, and then run the new master through the specialized AAC converter again to hear if these changes were sufficient.

Meller likens it to “polishing a car in total darkness.” “It was like if someone gave you a beautiful BMW, handed you some polish and then shut out the light and said ‘Okay, go ahead!’ At the end you’d turn the light back on and realize you missed the whole entire roof.”

The New Procedure

While the process of creating new iTunes masters for I’m With You was painstaking, Rubin and Apple ultimately declared it a success, and a new protocol and tool-set were eventually released to streamline the procedure in the future.

Apple’s new “Mastered for iTunes” toolkit includes one applet that allows engineers to hear the affects of the AAC encoder with minimal fuss, and another mini-app that helps them measure and locate any digital clipping created by the conversion process itself. The new protocol also allows artists to submit higher-resolution 24-bit files to iTunes, leading to better conversion.

Meller and Santangelo’s colleague at Masterdisk, Andy VanDette, recently remastered much of the Rush catalog for iTunes and weighed in with some thoughts of his own. He told me that even with the improvements, the process can still be complicated because “the iTunes AAC algorithm is a quirky beast.”

“I can tell you what doesn’t work,” he said. “One of my initial trial-and-error methods was to take digital fingerprints of both versions of the song, and then try to apply a [compensating EQ] to the CD version and pump that back through the AAC encoder.”

“Well, the problem with that is that input does not equal output. It’s highly program dependent, and you rarely get the same thing twice in a row. On one record, I might have three songs that would sound very similar and need the same kind of treatment – But then on the fourth song, it would be completely different and you’d need to figure out a new setting to make them match.”

As for what those differences are, VanDette says “It can be anything. It can be level or it can be imaging. It can be top end, it can be low-end. On the live album that I did, I noticed the spatial difference much more than on the studio albums, especially with the crowds and the ambiance being picked up by the stereo mics. ”

Santangelo agrees, saying that “It’s almost like an organism — It’s got a life of its own. Sometimes I’m concerned by whether the conversion is even consistent, but this is what we have and we do the best we can with it. You have to take it song-by-song and case-by-case.”

The Skeptics

Since the new tools have become available, more than 100 new and classic albums have been released in the “Mastered for iTunes” section of the iTunes store and sales have been brisk. But not everyone is convinced of the merits of the new program.

Many blind listening tests conducted by designers of audio codecs suggest that the majority of listeners have difficulty noticing any differences between high-resolution AACs and their source files when they are properly converted.

Even segments of the mastering community are divided. One mastering engineer in Britain named Ian Shepherd has written several posts on his “Production Advice ” blog that describe the  “Mastered for iTunes” process  as “B.S.”

In a critical YouTube video, he uses phase-reverse null tests that he believes should prove that in at least one instance, a Mastered for iTunes version of a song sounded less like the original version than his own consumer-grade iTunes encode.

While I don’t doubt Sheperd’s central argument that Mastering Engineers are human, and therefore fallible, there are flaws in his methods. The engineers who believe in the merits of iTunes mastering have at least three good counter-arguments at their disposal:

First is that Sheperd’s sample size of one song is far too small to be conclusive — especially with a manual process like mastering for iTunes.

Second is that Sheperd’s files are irrelevant. He takes a song that was mastered before the new protocols were put in place, and then uses both a lower bit-depth source-file and a different AAC encoder than was used to create the “Mastered for iTunes” version. He then compares his own custom file to a down-sampled CD version rather than the original high-resolution master.

Even if Shepherd fixes this part of his  methodology, there’s still a third argument: Since the original engineers used a healthy dose of additive EQ to restore frequencies they claim were lost during the AAC conversion process, it’s plausible that the phase-shift inherent in non-linear EQs could cause a phase-based null test to report additional cancellation differences — even if the use of that EQ was successful in restoring the original frequency balance.

With all that in mind, it’s no surprise that Shepherd’s results are different.

But all of this goes without even mentioning that null tests are only useful in establishing whether or not any difference exists between two files. Null tests have never been established as a reliable indicator of how audible those differences are.

Even using a simple frequency analyzer would have been a step in the right direction. But ultimately, a blind ABX test would be necessary to get any sense of which file  sounds closer to the original when heard by real-world listeners.

Of course to average listeners, arguments on both sides may sound like a bunch of hot air as they consistently fail to hear any differences in the sound of either file-type.

Still, Sheperd presents a thoughtful case and he raises fair questions, even if his test is flawed. His call for skepticism is healthy — But he might benefit from applying that same skepticism to his own claims in order to help design a new and more conclusive study.

The End Result

Most major mastering engineers seem to disagree with Shepherd so far, and have heralded “Mastered for iTunes” as a step in the right direction. Bob Ludwig of Gateway Mastering said in a recent interview with Postive Feedback that the results of iTunes mastering “can be so dramatic you can easily hear the difference between the new and old technology on your little laptop speakers.”

“Instead of ingesting the music from a CD rip or 16-bit file, the new system uses 24-bit master files for the encode. The AAC encoder can make use of bits 17-24.

“…[Another] important addition is the realization that the act of AAC encoding can cause clipping where there was none on the original PCM .wav or .aiff file … Apple has created tools to log the number, severity and time of each clip so the mastering engineer can lower the level of the 24-bit master by fractions of a dB and the clips and resulting distortion from them is eliminated.

It is a complicated answer, but a 24-bit AAC encoded file can thus sound better and measure better in certain cases than a normal 16-bit Compact Disc, which unfortunately has been regarded as the gold standard for sound in these comparisons.”

Mark Santangelo of Masterdisk agrees that iTunes’ new-found ability to convert reliably from 24-bit source files is a central feature of the new iTunes mastering process. The other major development is that mastering engineers can finally A/B their original masters with the AAC versions, allowing them to make better choices than before.

Santangelo adds that there’s a human component at work as well, and just like mastering for any medium, there’s as much art as science involved in the process.

“When you’re going through an album song-by-song and choosing the right EQs, you’ll never [match the iTunes version] 100%. That’s just impossible with a lossy file. You’re just trying to get as close as you possibly can.”

“But when you consider the file size and fidelity of your original file and then do the math to compare that with what you’re getting [back from the encoder ] you realize ‘Gosh, I’m losing maybe 65, 75% of the information that was there.’ To be honest, I’m pretty impressed that with that kind of loss, it’s even able to do what it’s doing in the first place.”

“I guess when you think of it that way, our ears are really getting fooled quite beautifully.”

Justin Colletti is an engineer and journalist. He is a staff writer for SonicScoop and the managing editor of Trust Me, I’m A Scientist.

This entry was posted in All Stories, Featured Stories, March 2012, Most Popular, Techniques, Technology. Bookmark the permalink. Both comments and trackbacks are currently closed.
  • Sign up for our Email Newsletter


  • E-news and Updates

    Subscribe to Trust Me, I'm a Scientist Find me on Facebook Find me on Twitter Email Trust Me, I'm a Scientist