Big-label mastering engineers don’t understand lossy formats, Article about Mastered for iTunes Options

[Kohlrabi]

Just found this article via twitter, circled among mastering "engineers" (in fact Heba Kadry reposted it, the girl who mastered the latest Mars Volta album, which reaches -12.79 dB on my RG scans, and is generally mastered in a horrible fashion).

This further backs my impression that most of them don't have a single clue of what they are doing. The section about the mastering practices of Rubin and Meller are especially eye-opening to me. Masterdisk "engineers" also apparently are now out to rape the Rush back catalogue. Further down they cite phase-reverse tests to prove AAC files are different from the original (wow, REALLY?).

The good thing is, I can use this article to decide which releases to avoid in the future. But I'm really at a loss what we can do beside that. I'm really fed up with mastering "engineers" destroying music releases.

This post has been edited by Kohlrabi: Mar 27 2012, 09:17

[TrustScience]

I think you're being a little harsh on the article.

Thats round the wrong way, the article said the waveform shapes were not the relevant factor it was the person skeptical of mastered for itunes that was looking at the sum of two file with one inverted.

From the Author

Thanks for sharing the article, Kohlrabi. If you have another read of it sometime, I think you'll find it's pretty even-handed.

I get the sense that you disagree with the claims made by the mastering engineers, and not with the article itself, which tries to take a neutral stance.

Since this was piece of reportage, rather than opinion, I wanted to accurately deliver the news, present both the commentary and the criticism, and to offer some dispassionate analysis of both stances. The article tries to remain skeptical of the magnitude of the claims, but I also stand by the fact that it's unfair to dismiss them entirely without proper tests.

If you do want to read an opinion piece, you'll probably find that my own feelings about blind listening are pretty congruent with the philosophy here at Hydrogen Audio. I'm a big supporter of this site and of ABX tests in general, and I even wrote a supportive opinion piece about blind listening tests in the very same issue you cited! (You can read that here):

"Can You Hear What I Hear? A Guide to Listening Blind"

I will have to disagree with the extreme stance that mastering engineers are generally snake-oil salesmen (I've never had an experience that would lead me to believe that) or that Mr. Ludwig or Ms. Kadry are incompetent (I know both, and they're definitely not).

However, as a matter of personal opinion, I would agree that Rick Rubin overstates the audible differences normally found between high-res AAC files and their original WAV masters. I'd also agree that at least one of Meller's comments was probably more figurative and expressive than it was literal.

With that said, the engineers I interviewed told me that in "Mastered for iTunes", Apple also fixed an actual issue they had in the past with creating AAC files from high-resolution masters. To be fair, even Bob Ludwig agrees that 256kbps files won't necessarily sound any worse than traditional CD files - Just so long as they're created properly, and the engineer can verify there were no issues with the transfer. (He says that this is something that they're now able to do.)

The other real development appears to be that MEs can now easily and effectively hear the differences between their original master and the file that the iTunes store's proprietary encoder will create. (I'm told that, for better or worse, it's not the same encoder used in the consumer version of iTunes.)

Who knows? Even if the AACs sound identical in 95% of cases, this new ability to actually listen and check can't be a bad thing

Personally, I think the new tools are a good idea, but I'm not about to replace my music library with new "Mastered for iTunes" versions anytime soon.

I definitely appreciate the healthy dose of skepticism here on Hydrogen Audio, and would agree that people who have a service or product to sell sometimes dramatize their claims... But show me a salesman who doesn't, and I'll show you a salesman who's out of a job!

Thanks again for the share and the comments. Keep doing what you guys do!

Very best,

Justin Colletti
Recordist, Journalist
http://justincolletti.com
http://trustmeimascientist.com
http://sonicscoop.com

[Canar]

show me a salesman who doesn't, and I'll show you a salesman who's out of a job

If that's the case, I fully support sales staff unemployment.

[TrustScience]

show me a salesman who doesn't, and I'll show you a salesman who's out of a job

If that's the case, I fully support sales staff unemployment.

Haha - Fair enough.

That Bob Ludwig quote where he seemingly does not understand what audio encoding is or why it might cause clipping is particularly galling from someone doing any sort of audio work. Particularly to people like me who consider the huge amount of clipping accidentally introduced by incompetent engineers to be one of the worst aspects of modern music.

Hi Saratoga,

I think you and Ludwig would agree that most masters shouldn't be as hot as they are. From what I remember, he's come recommending that masters peak at a maximum of a full -1db below 0dbfs to reduce unintended clipping, which is an idea that was balked at by many of the mastering engineers I spoke with. Most of them seemed to believe their clients expect peaks at at least -0.5 or -0.4 dbfs, and would continue shooting for those peaks as long as they could verify that the encoder could handle it.

First everything that clown (Shepherd) says is irrelevant since you can't show that one encoding is closer to CD by subtraction. [/b] Its not that his sample size is too small its that his test is meaningless. Its not that he took a song mastered at the wrong time, its that he doesn't know what hes doing. Why are you talking about EQ when nothing he says could possibly make sense? Actually, why are you even addressing this guy aside from to say "no I'm sorry, thats not how audio works"?

What you seem to take issue with here is that I presented the most damning argument last rather than first (because that's how mounting arguments work!) Otherwise, I think we're on the same page.

Shepherd's faulty analysis was ranked high in Google and getting a ton of reads, so I felt it was only appropriate to address it in the article. Please remember that not everyone has the same level of understanding, and that a nuanced breakdown is extremely valuable to many readers.

I read the paper by Apple back when we first discussed it. I don't believe it said anything about "a healthy dose of additive EQ to restore frequencies that are lost during the AAC conversion process," or did it?
...
EDIT: I read the paper again. It says nothing about additive EQ to restore frequencies that are lost during the AAC conversion process. Not a big surprise.

Correct greynol, that fact is not from the Apple paper, but from my interviews with over a half dozen of the busiest M.E.s in New York City.

They say that in practice, they use the new toolset to A/B before-and-after files, and then use additive EQ to compensate for changes in tone if they find it necessary to do so.

In theory, that's not much different than having two slightly different EQ settings so that a cassette and vinyl version sound closer.

Many of you might effectively argue that this process is subjective and open to error (true) or that the sonic differences between a 24-bit WAV and a 256kbps AAC tend to be much slighter than the differences between vinyl and cassette (also true).

PS: The Dynamic Range Meter results for Ms. Kadry's latest work are worse than Frances the Mute (DR7), Amputechture (DR6) and The Bedlam in Goliath (DR6).

To be fair, I believe that's a reflection of taste rather than competence. Making a master that's less hot isn't that difficult, and it doesn't take much advanced training at all.

To argue that Kadry or any other in-demand ME don't know what they're doing misses the point. I believe they know exactly what they're doing: Making very loud masters!

Whether you dig what they're doing is an entirely different question. I just don't think it makes sense to argue about that on a "technical" level.

(Unless you we were citing some scientific double-blind tests that compare specific levels of dynamic range with perceived enjoyment. Now that would be neat! If you know of such a study, I'd be the first to read it.)

I'll admit that I haven't heard the records you've cited here, but in defense of Kadry, it's important to remember that any mastering job can be deemed "good" as long as it's what it was intended to be. We just have to judge it by the creator's criteria.

For instance: I love the sound of a lot of Dave Fridmann's work with the Flaming Lips, which is often clipped-to-death. What's important to remember is that those records aren't clipped because they're loud... they're loud because they're clipped!

That's a world of difference. That style of mixing and mastering is an aesthetic choice, and not some unintended side-effect of ignorance or accident.

The truth is that mastering engineers who get repeat clients aren't making loud records by mistake. You don't have to like their work, but I still think it's important to make a fair argument. Otherwise, we just end up sounding like dimwits who trash on studio reverbs because they don't sound like actual concert halls.

It's also useful to remember that to people younger than us, taste-based arguments against hot, bombastic masters is going to sound a whole lot like "Hey you damn kids, get off of my lawn!"

In reality, that's part of the reason the kids are into those records in the first place! Don't you remember what it's like to be young? What self-respecting American teenager wants to listen exclusively to music his parents would approve of?

In the end, I think that with the right effort, we can bring healthy dynamic range back into the mainstream. I just don't think we're going to do that by making negative or unprincipled arguments.

It's what happens when the art-world and science/engineering-world meet head-on.
...
It would probably be even worse if we only got to hear songs written by artless science types*. Good songs, badly recorded, vs bad songs perfectly recorded?

Now that, I'd agree with in a heartbeat.

This post has been edited by TrustScience: Mar 29 2012, 15:16

[greynol]

Hi Saratoga,

I think you and Ludwig would agree that most masters shouldn't be as hot as they are. From what I remember, he's come recommending that masters peak at a maximum of a full -1db below 0dbfs to reduce unintended clipping, which is an idea that was balked at by many of the mastering engineers I spoke with. Most of them seemed to believe their clients expect peaks at at least -0.5 or -0.4 dbfs, and would continue shooting for those peaks as long as they could verify that the encoder could handle it.

I don't think Saratoga will mind me speaking to this. The issue is not peak levels, the issue is audible distortion that is caused by aggressive use of dynamic range compression. Reducing the volume by a "full" decibel below full-scale does very little to cool a hot master.

For instance: I love the sound of a lot of Dave Fridmann's work with the Flaming Lips, which is often clipped-to-death. What's important to remember is that those records aren't clipped because they're loud... they're loud because they're clipped!

They're loud because they're compressed (and sometimes given mid-range emphasis) and often this compression can cause clipping regardless of whether the final master reaches full-scale.

That's a world of difference.

Yes, especially when you actually focus on the offending mechanism.

You might find this educational:
http://www.hydrogenaudio.org/forums/index....showtopic=91909

This post has been edited by greynol: Mar 29 2012, 15:53

[TrustScience]

Hi Saratoga,

I think you and Ludwig would agree that most masters shouldn't be as hot as they are. From what I remember, he's come recommending that masters peak at a maximum of a full -1db below 0dbfs to reduce unintended clipping, which is an idea that was balked at by many of the mastering engineers I spoke with. Most of them seemed to believe their clients expect peaks at at least -0.5 or -0.4 dbfs, and would continue shooting for those peaks as long as they could verify that the encoder could handle it.

I don't think Saratoga will mind me speaking to this. The issue is not peak levels, the issue is audible distortion that is caused by aggressive use of dynamic range compression. Reducing the volume by a "full" decibel below full-scale does very little to cool a hot master.

I'm sorry greynol, but I think you're confused about Ludwig's point here. From what I understand, he suggests that lowering the peak level reduces the number of additional clipping errors introduced by the AAC encoder itself. According to his argument, and the Apple whitepaper, this is a separate issue from aggressive use of dynamic range compression.

For instance: I love the sound of a lot of Dave Fridmann's work with the Flaming Lips, which is often clipped-to-death. What's important to remember is that those records aren't clipped because they're loud... they're loud because they're clipped!

They're loud because they're compressed (and sometimes given mid-range emphasis) and often this compression can cause clipping regardless of whether the final master reaches full-scale.

Compression is clipping - In a very strict and literal sense. Whenever a compressor or limiter acts on peaks, it is inducing "clipping" by definition.

You might find this educational:
http://www.hydrogenaudio.org/forums/index....showtopic=91909

I've read Bob Katz's book a couple of times and seen him speak on several occasions. I don't have the time to check out another of his videos at the moment, but maybe later.

If your desire is to school me on the basics of how all this works, there may be better uses of your time. But if you honestly think it's a neat video that adds something new to the conversation, I'd love to check it out in the future!

Have a good one,

Justin

This post has been edited by TrustScience: Mar 29 2012, 17:41

[2Bdecided]

Compression is clipping - In a very strict and literal sense. Whenever a compressor or limiter acts on peaks, it is inducing "clipping" by definition.

Oh for goodness sake - you haven't a clue what you're talking about with this one!

(there's no nicer way of putting it - sorry!).

Cheers,
David.

[TrustScience]

Compression is clipping - In a very strict and literal sense. Whenever a compressor or limiter acts on peaks, it is inducing "clipping" by definition.

Oh for goodness sake - you haven't a clue what you're talking about with this one!

(there's no nicer way of putting it - sorry!).

Cheers,
David.

David, care to back that statement up with a widely accepted definition of "clipping" that would exclude an aggressive dynamic-range limiter?

To be honest, I'm really surprised and disappointed by the few posters who are taking extreme positions and resorting to anonymous ad hominem attacks, especially when we already agree on so much.

Disagreement I'm fine with. The childishness of a few of your peers, not so much.

For the record, I am ethically and legally obligated not to call anyone "a liar" or "incompetent" without direct proof. That would be called "libel".

There are related studies that can be used to suggest the MEs and Apple folks could be overstating their claims. (The article makes direct mention of that - I'm surprised I have to keep mentioning that like it hasn't been said already.) But without direct before-and-after evidence, saying that their claims are conclusively false would not only be illegal and unethical, but factually incorrect.

That's all I have to say on that. I'm sure the vast majority of readers here are smart enough and unbiased enough to understand that. For the few who aren't - I'm sorry to hear that. Good luck and be well. I wish you a long life of yelling at strangers on the interwebs.


This post has been edited by TrustScience: Mar 29 2012, 19:00

[2Bdecided]

Compression is clipping - In a very strict and literal sense. Whenever a compressor or limiter acts on peaks, it is inducing "clipping" by definition.

Oh for goodness sake - you haven't a clue what you're talking about with this one!

(there's no nicer way of putting it - sorry!).

Cheers,
David.

David, care to back that statement up with a widely accepted definition of "clipping" that would exclude an aggressive dynamic-range limiter?

You are really confused. The whole point of compressing or limiting peaks is to avoid clipping. If you increase the amplitude without squashing the peaks down, they will be clipped once their amplitude exceeds digital full scale. If you "squash" them (e.g. by momentarily reducing the amplitude), you can keep their shape, avoid clipping distortion, raise the overall level of the rest of the signal - and (up to a point) the momentary reduction in amplitude is inaudible. That's exactly what classic peak limiting does. It's almost the opposite of clipping!

It's true that some DRC processors can also introduce clipping. Sometimes clipping is introduced intentionally - good old Orban has used a module called a "clipper" for decades (though FM processing means this acts in a slightly different way to what we're discussing here). Sometimes clipping is introduced as a by-product of poorly chosen settings. Usually you find clipping on CDs because, even after the various stages including multi-band compression and peak limiting, the engineer still wants it louder, and at some point simply clipping the signal sounds less bad that even more peak limiting.

I hope this clears things up.

btw, I don't think the 84260 members of HA count as each other's peers.

Cheers,
David.

[TrustScience]

It's true that some DRC processors can also introduce clipping. Sometimes clipping is introduced intentionally.

Thank you David. That was exactly the point I was trying to make when I wrote that Dave Fridmann's work is 'not clipped because it's loud. It's loud because it's clipped. That's a world of difference.'

The rest of the discussion on the definition of clipping seems a little ego-driven to me. The main point remains that extreme compression and limiting are forms of clipping, and that they are often used in that way today for aesthetic effect.

I don't see why we should be arguing about that, unless our desire is to show off for an audience by battling over semantics and further derailing the conversation.

As far as I can tell, you seem to be the primary person launching ad hom attacks.

Please find a quote where I made an ad hominem attack directed at any specific individual. If I did, I'd be happy to apologize.

I know that I expressed my disappointment with the tone of the discussion after two anonymous ad hominem attacks were made against me. I'm doubtful that any objective reader would construe that as an ad hominem attack on my part.

...since you are an American, libel is not actually illegal.

That's technically correct. Libel is not a criminal offense - Only something you can be sued for. My sincere apologies for the semantic misstep.

Once again, the central point still stands: I am ethically obligated to state that related ABX tests invite skepticism of some mastering engineers' claims. So I did.

I am also ethically obligated to state that no conclusive proof has been presented to show that these particular mastering engineers were unable to hear differences between two similar audio files and adjust accordingly. So I did that too.

If you feel that I failed at expressing these two points clearly, I'll accept that as your personal critique. However, if you feel I have some other obligation that runs contrary to this, or that I made no concerted effort to fulfill both of these obligations, then we'll just have to disagree.

Heh, so basically you came back to post some personal attacks against people who are trying to help explain things to you. I wonder what an unbiased person would think of that.

Please see above!

Unfortunately, I'm non-anonymous, so it's necessary for me to correct the record when inaccurate portrayals of my positions are made in a public conversation.

At this point, I'd ask that the moderators lock this thread, so I won't have to monitor it for additional inaccurate portrayals like the one you made in the quote above.

I think that all essential points have been presented on both sides of this "conversation", and at this time, any further discussion in this thread is likely to be pretty darn silly.

On a more positive note, a few HA members wrote me personally to express their regret about the direction this conversation took.

I'm very happy about that, so thanks. I'll try not to let the manners and off-topic arguments of a few anonymous posters color my feelings about the site too much.

I'm especially surprised by the comments of the few, because I've already made it clear that I'm a big fan of the philosophy behind HA and of ABX tests in general. As mentioned, I've even sung their praises to thousands upon thousands of readers. I guess I'll just have to accept that some people don't get that alienating your allies is lousy strategy.

To the rest of you, thanks for reading and have a good one,

Justin

This post has been edited by TrustScience: Mar 30 2012, 17:21

[saratoga]

As far as I can tell, you seem to be the primary person launching ad hom attacks.

Please find a quote where I made an ad hominem attack directed at any specific individual. If I did, I'd be happy to apologize.

Perhaps, you missed it, but I pointed an example out before right here:

http://www.hydrogenaudio.org/forums/index....st&p=790942

Trying to discredit an argument by (falsely) claiming the author is "resorting to knee-jerk ad hominem attacks" is an ad hom attack, because it attempts to discredit the author and not the idea. Most telling, you then failed to respond to the actual substance of the argument (e.g. that you were negligent in presenting information that was not true), which IMO implies bad faith.

I know that I expressed my disappointment with the tone of the discussion after two anonymous ad hominem attacks were made against me. I'm doubtful that any objective reader would construe that as an ad hominem attack on my part.

Hmm, it seems to me that you're a little unclear on what an ad hom attack actually is.

Check wikipedia:

An ad hominem (Latin for "to the man" or "to the person"), short for argumentum ad hominem, is an attempt to negate the truth of a claim by pointing out a negative characteristic or belief of the person supporting it.

So as far as I can tell, no one here has done this to you. Its pretty much just been you doing it to other people (e.g. greynol).

I am also ethically obligated to state that no conclusive proof has been presented to show that these particular mastering engineers were unable to hear differences between two similar audio files and adjust accordingly. So I did that too.

If you feel that I failed at expressing these two points clearly, I'll accept that as your personal critique. However, if you feel I have some other obligation that runs contrary to this, or that I made no concerted effort to fulfill both of these obligations, then we'll just have to disagree.

I see that you have corrected the false statements greynol pointed out, and the incorrect stuff about frequency changes from lossy encoding. Failing to admit that you had made a mistake and correct those mistakes were my main complaints. That said, I'm still not impressed with your handling of this, nor that you edited the article without making it clear that you were retracting quite a bit of it.

On a more positive note, a few HA members wrote me personally to express their regret about the direction this conversation took.

As am I. However, seeing as you eventually made the edits you needed to, perhaps you could have saved all of us this grief by being more open feedback in the first place.

I'm especially surprised by the comments of the few, because I've already made it clear that I'm a big fan of the philosophy behind HA and of ABX tests in general. As mentioned, I've even sung their praises to thousands upon thousands of readers. I guess I'll just have to accept that some people don't get that alienating your allies is lousy strategy.

You shouldn't be surprised. ABX isn't a cult. We don't worship at it. The reason we push ABX is that its a good way to avoid being wrong. Its basically an end to that means. But the goal is to be correct and avoid spreading misinformation. If you put up an article implying that AAC encoding involves "notching" out frequencies, people are going to be angry. Using ABX here would have been great, since it would have prevented someone from mistakenly believing it, but thats not really the point.

Does that make sense to you?

[TrustScience]

That's true. I did make a couple clarifying edits and fixed a typo too.

I took out the part of Mr. Meller's quote where he said that the "highs, lows and mids" were "filtered out completely". Once I realized that some of you took him literally rather than figuratively, I agreed it was a valid critique and felt his statements were clearer without it.

There were also two minor edits for style:

1) I got some good feedback in an email saying that the section about Shepherd's tests and the alternative tests was wordy and confusing in parts. I rephrased a couple of sentences to make it clearer and less verbose. The substance of the points did not change.

2) I acknowledged the criticism that it was unclear whether Apple's whitepaper explicitly recommends the use of equalizers in the mastering process. I still feel that's a semantic distinction, but I decided to add a couple words to make misinterpretation impossible. Again, the substance of the statement remains the same.

I'll take any valid criticisms to heart if they're specific. I also routinely clean up the copy in older articles if I'm alerted to punctuation errors or unclear sentences. If something is factually inaccurate, I'll even issue retractions and do follow-ups in future issues.

Since the above were not factual changes, I have no plans to do that in this case. Of course, I'm always happy to write more on the subject in the future, and I'm always looking for contributing writers with real expertise.

Chances are that I'm not going to look to anonymous axe-grinders to find valuable sources, or to publicly thank curmudgeons for their copy-editing recommendations and spell-checking services. That's just human nature. C'mon, you guys know all about "selection bias"*

If you have an issue with a specific sentence, statement, or see a typo, feel free to shoot me an email. All that I ask is that you be human about it.

And as I said in the beginning, thanks for reading, and for the feedback.

-Justin

(* Yes, I know what "selection bias" actually means. This is what normal people call a "pun".)