Skip to main content

Topic: Big-label mastering engineers don’t understand lossy formats (Read 41302 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • julf
  • [*][*][*][*]
Big-label mastering engineers don’t understand lossy formats
Reply #100
Except psychoacoustically, small differences in 'volume' are not necessarily perceived as 'loudness' changes -- they are perceived as quality changes.  That's why the biasing effect of level mismatch is so insidious.


Absolutely. When I did a rudimentary 16-bit vs 24-bit and 44 vs 48 vs 96 k listening test over at CA, one of my references was a file that had been amplified by 1 dB. That was the one consistently picked as "best" despite not being the best sample rate or resolution...

  • ErectX
  • [*]
Big-label mastering engineers don’t understand lossy formats
Reply #101
Engineers really need to go back to the pre-mastered version and use this for a quality lossy encode.

Mastering these days is a euphemism for making the tracks as hot as possible. This means not just using typical analog-style dynamic compression to get it as close to 0dB as possible, they use digital processing on the waveform samples to essentially clip (with rounded edges) every peak to make a CD packed full of maximum level samples. You can see this when viewing the waveform on a ripped CD of something new, you would swear the signal is just clipped from the multiple max samples in a row. Even just decoding the samples back into a analog waveform (with DAC lowpass frequency filtering) will result in levels higher than 0dB. That's how you get a CD that looks like this: http://upload.wikimedia.org/wikipedia/comm...se_waveform.png

Another factor turning 0dB meant-for-CD tracks into a clipped mess is that when encoding lossy, the audio content is filtered into many passbands, and the ones deemed to be inaudible are discarded. Changing the spectral content will result in a 0dB signal going above 0dB too, as frequency components that had previously cancelled each other out are removed. One has to encode and then decode to see the resultant levels above peaks, and then process them into analog VU levels beyond that to avoid DAC clipping of the analog waveform. In other words, encode audio that gets nowhere near 0dB.

Then finally there is the fact that the psychoacoustic model can't work well on audio cranked up way beyond max. Lossy compression works on figuring out what you can't hear and discarding it, in comparison to a reference level. When the RMS level of the signal is +20dB of what an uncompressed and unprocessed version would be, lossy codecs can no longer work well as it will not be played back anything like the codec hearing model unless you like jet-engine volumes.

There is no reason to encode mp3/aac at loudness war levels. The audio is not going to be played on radio and doesn't have to compete with other tracks in the "loudness" war. The problem is that to do a proper "mastering for iTunes", you have to go back to the source before it went through the bit-cranking hard limiters, and it is doubtful that many engineers will be paid to do this. And unfortunately, "mastered for iTunes" more likely means that misguided engineers actually compress the audio more and screw with the equalization, picturing cheap speakers and ipods as the destination.
  • Last Edit: 13 April, 2012, 12:31:54 PM by ErectX

  • Thasp
  • [*]
Big-label mastering engineers don’t understand lossy formats
Reply #102
I think you're being a little harsh on the article.

One thing amazed me though...
Quote
Bob Ludwig of Gateway Mastering, who is known for his objectivity and diligence, said "…[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"
This is news to a diligent mastering engineer? Has he just stepped out of a time machine from 1996?!


I haven't looked into "mastered for iTunes" - it seems strange to "need" to EQ a track to compensate for the effects of an AAC encoder running at 256kbps.

Cheers,
David.


He was trying to be informative. Is it really necessary that he denote everything he says with "I already knew this, but just incase YOU didn't.."

When you are trying to get people to listen to you - you should not be condescending. This is a social industry, you get nowhere by giving people the "wow, you didn't already know X? Well, here's what you SHOULD know" attitude.

Most mastering engineers hate what they are forced to do by the labels. They do it because if they don't, someone else with an L2 or Ozone maximizer plugin. Mastering engineers crushing music sucks, but it is 100000x worse when some bozo with a single plugin tries to do the same thing.

It is weird that in 2013 we are still reading threads where people blame the mastering engineers for everything. If mastering dude X doesn't ruin the music, he will be fired and they will give it to mastering guy Y. Considering how few paying jobs there are left in the music industry, you can't expect them to abandon their paying clients to fight a losing battle when the crushing will happen down the hall at another studio anyway.
  • Last Edit: 20 May, 2013, 01:20:31 AM by Thasp

  • Kohlrabi
  • [*][*][*][*][*]
  • Global Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #103
Most mastering engineers hate what they are forced to do by the labels. They do it because if they don't, someone else with an L2 or Ozone maximizer plugin. Mastering engineers crushing music sucks, but it is 100000x worse when some bozo with a single plugin tries to do the same thing.
You can make a recording sound hot without showing off your technical inability by letting your signal clip or claiming FUD about AAC. I simply don't buy the "poor mastering engineers" story in these cases. How come that regardless of producer the same engineers always produce the same horrible masters?

It is weird that in 2013 we are still reading threads where people blame the mastering engineers for everything. If mastering dude X doesn't ruin the music, he will be fired and they will give it to mastering guy Y. Considering how few paying jobs there are left in the music industry, you can't expect them to abandon their paying clients to fight a losing battle when the crushing will happen down the hall at another studio anyway.
So, we should all be thankful for the great job Vlado Meller did on the last few RHCP recordings, because someone else might have done a worse job? That's a pretty low standard there.
It's only audiophile if it's inconvenient.

  • 2Bdecided
  • [*][*][*][*][*]
  • Developer
Big-label mastering engineers don’t understand lossy formats
Reply #104
He was trying to be informative.
If that's informative, it just shows you how behind the times the audience is.

FWIW some (lesser) mastering engineers absolutely despise discussions like this. Their results aren't shameful to them - they are quite proud of what they do, and are very insulted that anyone should question their "judgement" and "skill". It's the good ones that realise what crap they're pushing out.

We should remember that, even in 2013, there are still whole swathes of the industry that are largely unaffected by these trends - with many wonderful recordings issued every year.

Cheers,
David.


  • Kohlrabi
  • [*][*][*][*][*]
  • Global Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #105
We should remember that, even in 2013, there are still whole swathes of the industry that are largely unaffected by these trends - with many wonderful recordings issued every year.
Which is problematic if you are into pop, rock or metal music (and derivatives), since the mastering standards in these genres are generally exceptionally low.
  • Last Edit: 20 May, 2013, 05:23:54 AM by Kohlrabi
It's only audiophile if it's inconvenient.

  • db1989
  • [*][*][*][*][*]
  • Global Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #106
To add to the recent unearthing of this post:
One thing amazed me though...
Quote
Bob Ludwig of Gateway Mastering, who is known for his objectivity and diligence, said "…[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"
This is news to a diligent mastering engineer? Has he just stepped out of a time machine from 1996?!
Not to mention that his oversimplified statement implies that such introduced clipping is more likely to be audible/consequential than it is in reality.

Quote
I haven't looked into "mastered for iTunes" - it seems strange to "need" to EQ a track to compensate for the effects of an AAC encoder running at 256kbps.
Indeed. Many of us had the same sentiment in the thread about this, erm, service. Something is fundamentally wrong with the concept. Another amazing innovation from Apple!

  • skamp
  • [*][*][*][*][*]
  • Developer
Big-label mastering engineers don’t understand lossy formats
Reply #107
Not to mention that his oversimplified statement implies that such introduced clipping is more likely to be audible/consequential than it is in reality.


I recall that you retorted to me in the past, that I jumped the gun by drawing conclusions from certain statements. This time I think it applies to you. Simply saying that some processing can cause clipping doesn't imply anything about audibility, just like it doesn't imply anything about the amount of clipping that will occur. And as a techie who likes to do things Right™, you're probably going to aim for a result that doesn't include any clipping at all, even if SOME of it would be acceptable (inaudible). But then that only implies attenuation (or headroom), in the smallest amount that will do the job.

Quote
Indeed. Many of us had the same sentiment in the thread about this, erm, service. Something is fundamentally wrong with the concept. Another amazing innovation from Apple!


I don't recall Apple's paper saying anything about applying EQing or any other sort of processing prior to encoding. The important part of it was, "please feed our encoder your high resolution master without prior downsampling, as our encoder will do it automatically with optimized settings" (I'm paraphrasing here). The idea that a different master is needed, more likely comes from mastering engineers like Vlado Meller (and producers like Rick Rubin) trying to get more business.
  • Last Edit: 20 May, 2013, 09:29:21 AM by skamp
See my profile for measurements, tools and recommendations.

  • db1989
  • [*][*][*][*][*]
  • Global Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #108
I recall that you retorted to me in the past, that I jumped the gun by drawing conclusions from certain statements. This time I think it applies to you. Simply saying that some processing can cause clipping doesn't imply anything about audibility, just like it doesn't imply anything about the amount of clipping that will occur.
This is a good point. I should have qualified my statement with the condition that what he said was technically correct. What I meant was that it might be open to being misread by other people. But that is their problem, not his, and it is probably not very feasible to expect people always to bullet-proof their statements against misinterpretation by others.

Quote
And as a techie who likes to do things Right™, you're probably going to aim for a result that doesn't include any clipping at all, even if SOME of it would be acceptable (inaudible).
This is a nice thought, if only it were true for more engineers!

Quote
I don't recall Apple's paper saying anything about applying EQing or any other sort of processing prior to encoding. The important part of it was, "please feed our encoder your high resolution master without prior downsampling, as our encoder will do it automatically with optimized settings" (I'm paraphrasing here). The idea that a different master is needed, more likely comes from mastering engineers like Vlado Meller (and producers like Rick Rubin) trying to get more business.
Good points too. I confess that I remembered little about MfI, and not to blame David, but when I read what he said, I just defaulted to assuming it was definitely correct.

Still, looking at the official PDF about MfI, certainly not all of it is unreasonable, but very near the start, we have this:
Quote
AAC is the new standard for digital music. It only makes sense to create masters specifically for this format.
As the document goes on to explain, albeit without making it crystal-clear and thus avoiding the risk of undermining their pet format, the concepts recommended by Apple during mastering are not format-specific and, indeed, are little more than common sense. Whilst I cannot argue with many of their recommendations on the latter level, I disagree with how they are presented as revelations and points in favour of AAC. Anyway, no doubt this has all been said before.

  • smok3
  • [*][*][*][*][*]
  • Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #109
From what I observed lately mastering goes all the way from useless to quite ok, however there is that weird feeling when you enjoy youtube-bootleg version better than the final master.
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

  • Thasp
  • [*]
Big-label mastering engineers don’t understand lossy formats
Reply #110
So, we should all be thankful for the great job Vlado Meller did on the last few RHCP recordings, because someone else might have done a worse job? That's a pretty low standard there.


It's not about Vlado Meller. I'm not saying you should be happy, I'm suggesting you blame the people responsible for the wreckage. You're blaming the soldier instead of Bush, the pilot instead of Truman. Blame the decision makers, not the people who carry out the decisions, and maybe things wil actually change. I wish I had a better analogy, but it is 2 AM here and I'm pretty tired.

Quote
How come that regardless of producer the same engineers always produce the same horrible masters?


It is likely that the engineer who puts out five loud albums probably gets hired by the same label, which probably has the same ignorant staff members who demand that it be loud. Then it creates this mindset in the group members that want it to be loud because the label people want it to be loud, then they force that on the mixers & mastering engineers. Or, the label directly forces that on the mixers & mastering engineers.

Some engineers record Tori Amos today, and Opeth tomorrow. Others are known for what they do,, and get asked to work on those projects.

My favorite mastering engineers during my time intering under them long ago used to ask the artists - do you want to make music people want to turn up, or shit people want to lower? He'd show the common sense, play recordings, and was almost always ignored, and forced to crush nonsense to pacify the talking heads in the room. A lot of good mastering engineers don't really love making crap loud, but have to find ways to do so or lose paying clients. Again, in 1984, maybe they could afford to. In 2013, if someone is paying you to work on music - you hold on for dear life, and say yes sir! Thank you sir!

Quote
FWIW some (lesser) mastering engineers absolutely despise discussions like this. Their results aren't shameful to them - they are quite proud of what they do, and are very insulted that anyone should question their "judgement" and "skill". It's the good ones that realise what crap they're pushing out.


Some consider it a video game. "Well, I'm foced to make this loud, how can I do it while still maintaining some sembleance of musicality. " It's the only way people who hate what they are doing can keep doing it, try to turn it into a game for yourself where you set goals and achieve them. Same as any other job.

Look on forums like the womb forums, prosoundweb, tape op, even the smut that is gearslutz, and you'll get to hear more than one article. You can read thousands of comments from hundreds of real, working professionals. When you do - when you work in the industry, and meet the people, a lot of these generalizations go away. It's easy to keep the disdain for mastering engineers alive as long as we're looking at one article a year, taking 2 or 3 lines of it and bashing it, but it's hard when you get to see the community as a whole & their thoughts on the issue. Check out sites like http://www.turnmeup.org , where you have legions of these mastering engineers fighting to get the industry(and the public) to care about the problem.

Quote
You can make a recording sound hot without showing off your technical inability by letting your signal clip or claiming FUD about AAC. I simply don't buy the "poor mastering engineers" story in these cases. How come that regardless of producer the same engineers always produce the same horrible masters?


There are a lot of mastering engineers that master to -0.3 or -0.2 dB or so. A lot of mastering engineers will work on these perfect systems with $10,000 lavry converters and finish their work. We can argue all day about whether differences in DACs are audible, but one DAC's -0.2 dB can easily be another DAC's clipping. If you master to a slightly lower volume you avoid a ton of problems when your master leaves your system and goes to another. It is possible that one system's -0.005 dB  is another's clipping.

Making separate masters for AAC is kind of silly, if you make a master that won't clip any DAC, it probably won't clip any lossy codec either. The "big picture" point I think he was trying to make is that a master can clip on one system while it doesn't on another. I would venture to say it's best to strike a balance, because no one is changing the station based on 0.2 dB.
  • Last Edit: 22 May, 2013, 02:28:35 AM by Thasp

  • 2Bdecided
  • [*][*][*][*][*]
  • Developer
Big-label mastering engineers don’t understand lossy formats
Reply #111
In that context, you may this quaintly careful...

The BBC wouldn't let its internet streams peak above -6dB FS for fear of inducing clipping during encoding and in crappy consumer sound cards...
http://www.audiomisc.co.uk/BBC/iPlayerRulesOK/Page2.html

EBU R128 limits the "true peak" (i.e. the oversampled / inter-sample peak) to -1dB FS...
http://tech.ebu.ch/docs/r/r128.pdf
(item n)


We could discuss the "sound" of clipping due to lossy encoding, and clipping due to inter-sample overs, and clipping due to really crappy old sound cards that can't cope with 0dB FS signals - but some of these are rarely audible - and others, if audible, are far less audible than the dynamic range compression that's been applied to "smashed" masters. So dropping back to -0.2dB FS rather than 0dB FS will almost never solve the problem, and the sound isn't really that much better if the problem is solved.

Whereas at least the BBC's worry (and solution) with Radio 3 is somewhat justified: clipping, in the context of a pure piano or flute note, is easily audible and pretty horrible.

Cheers,
David.

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Big-label mastering engineers don’t understand lossy formats
Reply #112
Depending on the material, 1dB is still not enough to prevent clipping during lossy decoding. This is especially true for squashed masters.

Thanks for talking about actual audibility, David.

More on the point if audibility of different things, we don't "argue all day" unless there is actual proof of the phenomenon in the way of positive ABX results. Unfortunately I read far more crap from mastering engineers about differences in sound than I read any knowledge about ABX. Instead I hear about A/B which is usually not blind, let alone double-blind; nor does it seem to be properly level-matched or have other necessary controls.
13 February 2016: The world was blessed with the passing of a truly vile and wretched person.

Your eyes cannot hear.

  • skamp
  • [*][*][*][*][*]
  • Developer
Big-label mastering engineers don’t understand lossy formats
Reply #113
Depending on the material, 1dB is still not enough to prevent clipping during lossy decoding. This is especially true for squashed masters.


Peaks of 1.5 (+3.52dB) or more are quite common, so I guess that at least 4dB of headroom would be needed to entirely avoid clipping due to lossy decoding.
  • Last Edit: 22 May, 2013, 10:46:38 AM by skamp
See my profile for measurements, tools and recommendations.

  • 2Bdecided
  • [*][*][*][*][*]
  • Developer
Big-label mastering engineers don’t understand lossy formats
Reply #114
Depending on the material, 1dB is still not enough to prevent clipping during lossy decoding. This is especially true for squashed masters.
In the context of EBU R128 though, a squashed master adjusted to a loudness of -23LUFS couldn't have a peak anywhere near -1dB FS (unless you specially created a test signal to cause this).

If Mastered for iTunes included a reference level of -23LUFS (or even soundcheck's ~ -16.5LUFS, or ReplayGain's 89dB which is ~ -17.3LUFS), then we'd never worry about clipping of smashed masters due to encoding, and we could probably forget about lossy-induced clipping and inter-sample overs for almost all "popular" content (and for almost all "classical" content with a -23LUFS = ~ ReplayGain's 83dB reference level).

Cheers,
David.

  • tnargs
  • [*]
Big-label mastering engineers don’t understand lossy formats
Reply #115
I think the most damning thing about the mastered for itunes program ....Essentially the entire document assumes that the reader has absolutely no idea how to master a CD, and if left to their own devices, they would screw it up.


I get the impression Apple want to speak to the 'small indie' operators, whose clients might even go down to the 'home recording' musicians and bands. The first line of Apple's Technology Brief says "Whether you’re a major label or a small indie,...". If they help some of these operators to get it right(ish), good thing.