[Copperears]

So, enjoying the long-standing thread by gregorio (are you Julian Hirsch reincarnated, btw? He saved me from much audio madness years ago!), I have a separate but related distortion question.

I will try to be very specific: I have an AAC 256-bitrate file purchased from iTunes ; it's one of a bunch of files in the performance of Elliott Carter's Clarinet Concerto, recorded digitally in 1999 and offered, presumably, as a well-encoded AAC on iTunes.

Here's the problem: the clarinet -- which is almost a square wave in its core sonic signature -- sounds fine soloing at mid volume level, but if there is a sustained note at either really low volume, or a sustained note at peak volume, I hear this "warble" in the note that sounds (to me) annoying and unrealistic.

It's less obvious played through a set of warmer headphones than through a set of IEMs. It is more noticeable at higher volumes, and at the louder, high-pitched sustain note.

I've since noticed, perhaps sensitized by this, similar warbliness (it's almost like old analogue tape flutter) in high-volume, high-pitched sustained notes in, say, operatic soprano singing.

It seems entirely absent in non-classical iTunes music. Which is the thing that confuses me.

It sounds a little bit as if an LFO (low frequency oscillator) had been applied to the note to modulate its volume (not pitch).

I also sometimes notice this warble in the noise floor of quiet analogue passages, what would, pre-lossy-compression, simply have shown up as hiss.

Does anyone have a good, technical explanation of what's happening here? It only seems to show up on some instruments, particularly the clarinet, and sometimes horns, and sopranos, in classical recordings purchased from iTunes.

Has there been sloppy encoding, or is this just an unavoidable artifact of lossy compression, even with 256 AAC? Or is it in the IEM itself, some kind of induced in-ear-canal resonance produced at certain frequencies by a balanced armature driver?

Any true insight would be most welcome, thanks in advance!



Sent from my iPad using Tapatalk

 

[limpidglitch]

If you could post an audio sample of the effect it would help greatly.
A control, in the form of a lossless version of the same file, would also be needed to confirm/exclude compression as the cause.

 

[Copperears]

I'm hesitant to do so as the file is copyrighted, but I know the CD version does not have this issue.
 
So, another example: Krystian Zimerman's performance of the Chopin Ballades. Deutsche Grammophon, circa 1988 original recording.
 
The first note of the first Ballade, in the lower register.  On the CD, it is a sustained "Bommmmm........"  On the iTunes download (have downloaded it multiple ways, to test), it is "Bommlmlmlmlmlmllmlml......" -- there's a quaver in the sustained note that is not there on the piano, and not there on the CD recording.
 
The Carter Clarinet concerto is a bit more difficult, as it's possible part of the "problem" might simply be the performer's use of flutter-tonguing....???  (https://en.wikipedia.org/wiki/Flutter-tonguing) -- I'd have to have the concerto's score in front of me to see if that's recommended and -- since apparently the score gives direction but provides encouragement to the performer to improvise -- even if it's not, the performer might have chosen to do so.  Hard to say, but the flutter is regular enough that it sounds more like a digital artifact than a performance technique to my ears.

 

[stv014]

You can post a 30-second sample without problems.
 
The effect you have described sounds like distortion due to clipping and/or aggressive peak limiting, which is a common problem in badly mastered music. But it could also be an artifact of the lossy compression (perhaps related to the near-clipping input signal level), although well encoded 256 kbps AAC should be transparent most of the time. It would be easier to tell if you posted samples.

 

[Copperears]

Ok will do so later.
 
It's definitely some sort of issue with the iTunes/DG processing of the file to AAC; I'm surprised it's not transparent, which is why I bring it up.
 
The actual recording has no criticisms of its mastering, in fact it's known as one of the best, if not the best, effort on the part of the Deutsche Grammophon engineers to capture piano sound.  And no, doesn't sound like clipping/limiting, I'm very familiar with what that sounds like.
 
Question is whether Apple will do anything about it, I doubt it. Ashkenazy's recording of the same Ballades have _no_ warbles at the equivalent points in the performance, so at least, fortunately, it's not my playback system!
 
I remain a bit cynical about Apple's overall quality control of classical music on its iTunes store, unfortunately.  It shocks me that such an important recording as this one has been handled so poorly.

 

[StudioSound]

There are two problems that contribute to what you are hearing:

1. The audio is too compressed. No matter what anyone claims, lossy formats are not "CD quality" and it results in this "warbling" that you hear.
2. Apple does not apply an appropriate reduction in gain before compression, which results in the codec being overloaded and clipping. (whenever you encode to a lossy format, you have to reduce the gain)

The whole talk is interesting, but there's a ten minute section on iTunes in this video which goes into detail about this very issue: https://www.youtube.com/watch?v=BhA7Vy3OPbc&t=10m15s

It's the reason I still buy CDs and rip them to a lossless format these days. (and I don't recommend using iTunes to rip your CDs either)


Something which may also interest you, is that JRiver has support for R128 analysis and volume leveling (which that video was primarily about) which will prevent inter-sample clipping on playback - though it cannot do anything for clipping which has been encoded into the file.

 

[Copperears]

Yep, this has all convinced me yet again that the claims about the adequacy of lossy are severely exaggerated.

I'll stick to CD for music with serious dynamic range, and leave the .m4as for massively-crushed current pop music. Much easier to encode music that has a 6db dynamic range, clearly.

 

[StudioSound]

I'll stick to CD for music with serious dynamic range, and leave the .m4as for massively-crushed current pop music. Much easier to encode music that has a 6db dynamic range, clearly.

 
Actually, highly compressed pop music is going to fare worse, because it's far more likely to clip when compressed to a lossy format than well-mastered music.

 

[limpidglitch]

Yep, this has all convinced me yet again that the claims about the adequacy of lossy are severely exaggerated.

I'll stick to CD for music with serious dynamic range, and leave the .m4as for massively-crushed current pop music. Much easier to encode music that has a 6db dynamic range, clearly.

 
There are limits to the transparency of lossy compression, but it shouldn't be quite that bad if applied correctly.
If you already have the CD version, why not try compressing it yourself, at various settings using various software?

 

[Copperears]

Yes that's the problem: I trusted iTunes to properly encode the CD in their store, which they didn't.

I've found this with a number of items on the store, and think I will refrain from buying any more music from there in the future.

I'll just stream radio at work.

 

[bigshot]

I have found that iTunes encodes AACs at a dB or so louder than the source file. If the track is normalized all the way to the top, it clips when it's encoded to AAC. You can lower the volume of the track before encoding, but that won't help you with the iTunes store. I would report it to the iTunes store as a defective track and give the the time of the clipping. They have been pretty good in the past about replacing defective rips.
 
This has nothing to do with the ability of AAC to encode. It's improper mastering of the original CD they ripped from. It shouldn't have been pushed right up to the edge like that.

 

[StudioSound]

The warbling is not caused by clipping - that's just compression at work - though it is often accompanied by clipping due to the lack of a gain reduction step when encoding the master to AAC.
Gain reduction needs to be performed when encoding to any lossy format - at least if the track is close to clipping, which modern tracks normally are.

 

[bigshot]

I did a lot of testing of difficult to encode sounds with AAC when I was establishing the standard for my collection. I found one track that artifacted at 192, It was fine at 256, and I added VBR. Since the sound is at peak level there, I think clipping has more to do with the encoding error than the sound itself.

 

[limpidglitch]

 
This has nothing to do with the ability of AAC to encode. It's improper mastering of the original CD they ripped from. It shouldn't have been pushed right up to the edge like that.

 
If it was an 128kbps encode I'd readily accept it as the encoders fault, but at 256 I can't help but think there's something else at play as well.

If the OP has a lossless version of any of these 'swirlie' tracks, it would be interesting to see if he could replicate the result encoding it himself, using both iTunes and some alternative software.

 

[bigshot]

Yep. That is the first thing I'd do. If anyone is interested in the CD that gave me trouble, it was the Sammy Davis Jr Decca sides. The string texture was very complex and it tended to gurgle like audience applause sometimes does. At 256 AAC it was perfect.