So as the dust starts to settle on this year’s Dynamic Range Day, I have a moment to revisit “Mastered for iTunes” – hopefully for the last time !
Why do I need to ? Because since my recent posts about this subject, I’ve been criticised for them. Some of it’s justified, but much of it isn’t, and I wanted to answer some of the comments.
The story so far
So first of all, if you haven’t already read my summary of what “Mastered for iTunes” is – and what I think of it – please read this post:
(Short version – I applaud Apple’s initiative; I’m excited by the implication that lossless audio from Apple may be just around the corner; and I fervently hope that MFi will ultimately lead to an improvement in audio quality – but the early signs aren’t enouraging.)
Now, the other two posts I’ve made about MFi are both more critical of the way the idea is being publicised, and they focus particularly around the suggestion made by some engineers that we should be making special “tweaked” masters to compensate for way that AAC encoding changes the sound of our audio.
(Bear in mind – the hype about this process was how we first found out about “Mastered for iTunes”, not via the recently released Apple guidelines.)
If you want to read the posts, here they are:
These are the two posts that have attracted the most criticism. I’ll go through some of these points and address them.
1 – I only demonstrates my test on one song – that doesn’t prove anything
Hands up – absolutely, and in the heat of the moment when making my “null test” video I completely ignored this, and used words like “proof” – I should have known better. Small sample sizes can never prove anything.
On the other hand, I’ve yet to hear an example of “tweaked” MFi audio that lives up to the hype…
2 – Apple never said MFi sounded “closer to the CD”, they said it sounded “closer to the source”
Well yes… and I never claimed they did.
The “closer to the CD” comments were made by the engineers who worked on the RHCP album, after my original post.
Many sites picked up on my “null test” post, and some people interpreted it as if I were saying that Apple were trying to deceive us – but that was never my intention. I added updates to the original post to make my point clearer.
On the other hand, Apple do have a section in the iTunes store called “Mastered for iTunes”, featuring albums like “I’m With You” and “Beyond Magnetic” which actually disregard Apple’s own guidelines. These clearly state that “clipped” samples should be avoided – but here’s the output summary of Apple’s own “afclip” utility from the song “Hate Train” from “Beyond Magnetic”, developed to help mastering engineers avoid this pitfall:
total clipped samples Left on-sample: 105510 inter-sample: 286969
total clipped samples Right on-sample: 130525 inter-sample: 360890
Whereas the whole second half of “Bredan’s Death Song” from “I’m With You” reads an intersample peak level of +1dBFS.
Until very recently this release was one of the first you’d come across in the MFi section of the store.
And even Coldplay’s “Mylo Xloto”, mastered by the legendary Bob Ludwig – who has often spoken out against the loudness wars – fails to take maximum advantage of the extra quality offered by the new 24/96 masters requested by Apple. It’s DR values are indisinguishable from the CD, sadly – although they are at least more reasonable than many recent releases.
Bob has even confirmed in an interview that unlike the engineers I mention above, he has no special approach to mastering for an AAC encode:
The creative part of the mastering process hasn’t changed one bit with the Mastered for iTunes program. Ludwig said he masters his music exactly the same way. The change comes after the creative process is finished — that’s where Mastered for iTunes comes in.
This makes perfect sense to me, but confirms my instinct that the benefits of Apple’s new MFi guidelines at this stage will be minimal on most music.
3 – Null tests don’t tell you anything about the way things sound
True. (Although they can give you clues.)
But again, I never said they did. I used a null test because my ears told me something didn’t sound right first, and the test simply confirmed this – ie. that there was a bigger difference between one pair of files than the other.
4 – My null-test was invalid – I don’t have the original files
The suggestion here is that to do a “scientific” comparison, I need access to the original 24/96 masters.
Actually, no. Because again, I wasn’t testing Apple’s claim that their new AAC encoder sounds closer to the source, I was testing the engineer’s claims that their tweaked masters sounded closer to the CD. So all I need to make that comparison is the CD, and their final files – which I have.
But while we’re on the subject:
DO Apple’s new encodes sound “closer to the source” ?
Well, perhaps theoretically – but in my tests, not much. Not in their current 16-bit, 44.1 kHz form, anyway.
I took a few 24/96 masters from my own archive and encoded them using Apple’s new “droplet” utility, and the standard iTunes encoder from my own CD down-sampled and dithered versions.
There was a difference, yes – but I doubt if most people will hear it. So how do I know there really was a difference ? Because the null test shows it. But unlike the EQ tweaks revealed by testing the MFi “tweaked” masters, this null revealed differences so small I could barely hear them at normal listening levels.
Of course, it may be that the files I used just happened not to reveal any differences – but they all had plenty of >20 kHz frequency content, which is the only area that could make a significant difference.
More likely is that whereas as mastering engineer, I’ve chosen the very best sample-rate conversion and dithering software I can find, there’s plenty out there that isn’t as good – and Apple’s version may well sound better than these options.
That in itself is enough justification for supplying hi-res files as Apple suggest – but won’t make much difference to professionally mastered audio.
Apple’s “Mastered for iTunes” initiative is great, and I’m behind it all the way. Sadly, I’ve yet to hear an MFi release that even begins to realise the potential of the idea, and several that change the sound in pointless ways – in my opinion.
Null tests and clipping-detection utilites are far less important than what our ears tell us of course, but so far they are backing up what my ears tell me – that “tweaking” masters for AAC encoding is misguided and ineffective, and despite the hype and it’s genuine potential, “Mastered for iTunes” is a concept that has so far failed to make any real impact on the quality of the audio we’re hearing on mainstream releases.