ALAC vs. FLAC

[GRUMPYOLDGUY]

1. Even the most dynamic of commercial recordings virtually never exceed a 60dB dynamic range, that's the equivalent of 10bits. So with a 16bit distribution format the last 6 or so LSBs are just noise or digital silence. A 24bit release therefore just has another 8bits worth of noise (or digital silence). Providing those additional 8 LSBs of random noise (or digital silence) can be removed transparently, which has been easily achievable for well over a decade, how can loosing them in any way be described as a compromise? And as for loosing a band of inaudible ultrasonic frequencies is concerned, how is that a compromise? On the contrary, it's actually a potential fidelity improvement as it removes the likelihood of any downstream IMD compromising the sound quality! And finally, a 192kHz sampling rate is itself already compromise compared to lower sample rates. 192k is not higher quality because it's more data, it's lower quality!

2. Of course 24/192 is spectrally different, there's another 8bits of noise plus a far larger band of inaudible, ultrasonic freqs in there. And yes, you do increase the digital noise floor; from roughly 14bits below the recording's noise floor, to only 6bits below it! And sure, it's maybe possible to foul up a downsampling/dithering process to the point that it's not audibly transparent but that would take fairly severe incompetence and exceptionally poor quality tools, for which there's no excuse as transparency was achievable many years ago even with free tools!


The cheaper version (to buy) is commonly the more expensive version to actually make! Typically, we record and mix the highest quality we can, given the constraints of budget and time. That mix is then tweaked during mastering to produce a high quality master. To produce a lower quality master then requires additional steps to be taken, costing more time and therefore money. It's usually not feasible or often even possible to take a low quality mix (or master) and turn it into a high quality master but the other way around is of course far easier.

Also, we have to be careful about what we mean when we say "an inferior version". A version which would sound "truly inferior" when listening with good equipment in our critical listening environment may actually be "truly superior" in a different situation. For example, I'm currently mastering an album for a talented guitarist, which was recorded at 24/96, even though spectral analysis shows there's virtually nothing above about 20kHz; an acoustic guitar produces very little above 20k to start with, it looks like standard studio mics were used which roll-off at about 20k anyway and the dynamic range of an acoustic guitar is relatively small. As part of my mastering I'm applying fairly heavy compression because the artist will distribute on YouTube and wants the recordings to sound good on a variety of playback equipment including laptops, tablets and even mobile phones. Although this compression compromises the SQ of the recording when listening with decent equipment/environments, the artist feels this compromise to be relatively minor compared to the fact that the recording would be un-listenable in those other situations which are important to him. So why record, mix and master in a (so called) HD format to start with, especially as there's virtually nothing there which could make even a theoretical difference, let alone an audible one? It's simply as a future proofing measure. The artist may enter a relationship with a label or his circumstances may change in some other way which may require a HD release version. A task which will now be trivial but would probably have required him to spend weeks completely re-recording the whole album if he hadn't recorded and mixed it at 24/96 to start with. This HD version would sound significantly better to the critical listener than the current version, purely due to the fact that I could significantly dial down the amount of compression. If it were up to me and if/when such a HD version were required, I would release this HD version at 16/44.1 because there would be absolutely no audible difference between this version being distributed in a HD format or at 16/44.1. Unfortunately, the world of music marketing doesn't currently work that way and it's virtually certain a HD version would have to be distributed in a HD format.

Lastly, while analogies with visual images can often be very handy, they can also quite often be entirely misleading. SD and HD video in this case is an example of the latter. Even the higher resolution limits of HD video is well within the ability of the human eye and additionally the move from SD to HD involved a fundamental change to the replay technology (bigger plasma or LED screens rather than CRT), all this added up to a very obvious and noticeable difference between SD and HD. With audio, even the resolution limits of (so called) SD audio is beyond the ability of the human ear and additionally, there has been no fundamental change in the replay technology (speakers are essentially the same tech as before even SD digital audio).

G

I get a headache everytime I read one of your posts.

Not only are you missing the point of my statement entirely, you're clearly just arguing for argument's sake. Enjoy.

 

[sonitus mirus]

  The cheaper version (to buy) is commonly the more expensive version to actually make! Typically, we record and mix the highest quality we can, given the constraints of budget and time. That mix is then tweaked during mastering to produce a high quality master. To produce a lower quality master then requires additional steps to be taken, costing more time and therefore money. It's usually not feasible or often even possible to take a low quality mix (or master) and turn it into a high quality master but the other way around is of course far easier.
 
Also, we have to be careful about what we mean when we say "an inferior version". A version which would sound "truly inferior" when listening with good equipment in our critical listening environment may actually be "truly superior" in a different situation. For example, I'm currently mastering an album for a talented guitarist, which was recorded at 24/96, even though spectral analysis shows there's virtually nothing above about 20kHz; an acoustic guitar produces very little above 20k to start with, it looks like standard studio mics were used which roll-off at about 20k anyway and the dynamic range of an acoustic guitar is relatively small. As part of my mastering I'm applying fairly heavy compression because the artist will distribute on YouTube and wants the recordings to sound good on a variety of playback equipment including laptops, tablets and even mobile phones. Although this compression compromises the SQ of the recording when listening with decent equipment/environments, the artist feels this compromise to be relatively minor compared to the fact that the recording would be un-listenable in those other situations which are important to him. So why record, mix and master in a (so called) HD format to start with, especially as there's virtually nothing there which could make even a theoretical difference, let alone an audible one? It's simply as a future proofing measure. The artist may enter a relationship with a label or his circumstances may change in some other way which may require a HD release version. A task which will now be trivial but would probably have required him to spend weeks completely re-recording the whole album if he hadn't recorded and mixed it at 24/96 to start with. This HD version would sound significantly better to the critical listener than the current version, purely due to the fact that I could significantly dial down the amount of compression. If it were up to me and if/when such a HD version were required, I would release this HD version at 16/44.1 because there would be absolutely no audible difference between this version being distributed in a HD format or at 16/44.1. Unfortunately, the world of music marketing doesn't currently work that way and it's virtually certain a HD version would have to be distributed in a HD format.
 
Lastly, while analogies with visual images can often be very handy, they can also quite often be entirely misleading. SD and HD video in this case is an example of the latter. Even the higher resolution limits of HD video is well within the ability of the human eye and additionally the move from SD to HD involved a fundamental change to the replay technology (bigger plasma or LED screens rather than CRT), all this added up to a very obvious and noticeable difference between SD and HD. With audio, even the resolution limits of (so called) SD audio is beyond the ability of the human ear and additionally, there has been no fundamental change in the replay technology (speakers are essentially the same tech as before even SD digital audio).
 
G

 
 
Good stuff, as usual.  With regards to "SD" vs "HD", I was mostly referring to the current market trends where consumer options for mainstream music are either MP3 or AAC from mega stores such as Amazon, Apple, or Google vs. the heavily marked up HD files.  If I could get a FLAC/ALAC version at CD quality for all of my music and at a similar price that I can currently buy a new CD, I'd be happy.  Mostly the lossy versions are audibly transparent, but not in every situation for every listener on every system.  For personal confidence, if I am going to purchase music, I'd want it to be available in a format with no potential compromises in sound quality.  Currently I purchase CDs, used or new, when I need to fill in gaps that a streaming music service does not support.  It would be great if lossless version were made available to purchase for entire catalogs.  It was only the foolishness of the major record labels that had for so long been able to control distribution that they failed to see any benefit in changing their methods.  The only way they allow online sales of lossless quality files is to make them available at outrageous prices in ludicrous formats that nobody has shown to be sonically superior to Red Book in any meaningful way.  I'm fine with 24/96 or any format that is established as the norm for online distribution, but I'm not going to pay $30 for an online digital version if the CD is available to buy at Amazon for $7.  

 

[gregorio]

  For personal confidence, if I am going to purchase music, I'd want it to be available in a format with no potential compromises in sound quality.

 
Me too. What annoys me is that this format should be 16/44.1 or a FLAC/ALAC equivalent. In practise though it's usually SACD/DSD or 192/96/24, purely for marketing purposes. Consumers presumably won't cough up for a huge mark-up between two different versions both at 16/44.1 but apparently will if one of the versions is only available in a different (higher data) format. While some mark up is justified for a higher quality version (be it 16/44.1 or higher), the consumer is to some degree effectively being scammed, very effectively actually because so many never even realise they are being scammed!
 
I sometimes wonder where we would be if the industry had focused on audio quality rather than on audio formats. Maybe one day it will come back round but for now the push is towards ever more ludicrous formats, formats which if anything degrade audio quality rather than enhance it.
 
G

 

[Hi-Fi'er]

Truth about Audio: https://www.youtube.com/watch?v=Z5S_DI99wd8&feature=youtu.be

 

[wnmnkh]

Truth about Audio: https://www.youtube.com/watch?v=Z5S_DI99wd8&feature=youtu.be

 
Yes. No doubt Dr. Waldrep is one of the sane voices regarding 24bit stuffs. You may also want to visit his blog site too: www.realhd-audio.com

 

[GRUMPYOLDGUY]

Truth about Audio: https://www.youtube.com/watch?v=Z5S_DI99wd8&feature=youtu.be

 
That was an overall disappointing video, though there were some good takeaways from it.
 
He spent the entire time arguing that making 24/192 versions of recordings that were recorded at far lower fidelity is pointless. Who the heck on this planet would argue that (besides Neil Young, apparently... who is biased since he needs to make sales)? As he eloquently put... You can't use a CD as a source to generate content that sounds better than a CD.
 
There is a far more involved, and quite frankly better and more important, discussion to be had on the value of recordings that captured at 24/192 in the first place vs. 44.1/16 for example. The speaker certainly seems to think there is value in that, though he himself readily admits he doesn't know why and "hopes" that the brain can use some of that extra content. He in fact does his own recordings with higher resolution (I watched the video last night and forgot the exact number, 96/24 maybe?)... 
 
As for myself, I think I have the same opinion as the speaker.... I want the exact acoustics in the studio or live performance to captured and reproduced, regardless of whether or not the content is audible... If a guitar produces some high frequency content beyond 20KHz for example, I want that in the recording and played back by my hardware despite the fact that I can't hear that content. 

 

[sonitus mirus]

   
As for myself, I think I have the same opinion as the speaker.... I want the exact acoustics in the studio or live performance to captured and reproduced, regardless of whether or not the content is audible... If a guitar produces some high frequency content beyond 20KHz for example, I want that in the recording and played back by my hardware despite the fact that I can't hear that content. 

 
I tend to agree with your agreement.  However, with higher recorded frequencies, I believe that it is possible to introduce distortion (IMD?) that might negatively impact what is being heard.  So, what may not have been an issue listening to something live might not work correctly through speakers that have to struggle to play back the wider frequency range when all instruments/voices are combined into a single audio signal.   A solution might be to simply use speakers that cannot reproduce ultrasound.  Some speaker setups using 24/192 files could potentially be making things worse for the listener.

 

[GRUMPYOLDGUY]

   
I tend to agree with your agreement.  However, with higher recorded frequencies, I believe that it is possible to introduce distortion (IMD?) that might negatively impact what is being heard.  So, what may not have been an issue listening to something live might not work correctly through speakers that have to struggle to play back the wider frequency range when all instruments/voices are combined into a single audio signal.   A solution might be to simply use speakers that cannot reproduce ultrasound.  Some speaker setups using 24/192 files could potentially be making things worse for the listener.

 
This is the greater discussion that I wish the speaker had gone into. It's an interesting topic for sure. 

 

[spruce music]

This is the greater discussion that I wish the speaker had gone into. It's an interesting topic for sure. 

On another forum last year there was circulating a 96 khz file with 30 khz and 33 khz recorded at a high level. So you could test for the 3 khz imd result. Mostly with headphones but also with speakers no one could hear it until they pushed the amps too far. Seems transducers were less of a problem than what is usually assumed.

 

[GRUMPYOLDGUY]

On another forum last year there was circulating a 96 khz file with 30 khz and 33 khz recorded at a high level. So you could test for the 3 khz imd result. Mostly with headphones but also with speakers no one could hear it until they pushed the amps too far. Seems transducers were less of a problem than what is usually assumed.

 
With 30KHz and 33KHz, IMD spurs would show up at 27KHZ and 36KHz. I wouldn't be able hear that far up in the spectrum, let alone for spurs that are 60+dB down. 
 
I suspect what likely happened is that people pushed the amps so far up it caused some periodic distortions at the maxima and minima of the tones, causing audible harmonics which is what they heard. 
 
Of course what I'm saying is all speculation, I have no way of going back in time and measuring it. 

 

[spruce music]

With 30KHz and 33KHz, IMD spurs would show up at 27KHZ and 36KHz. I wouldn't be able hear that far up in the spectrum, let alone for spurs that are 60+dB down. 

I suspect what likely happened is that people pushed the amps so far up it caused some periodic distortions at the maxima and minima of the tones, causing audible harmonics which is what they heard. 

Of course what I'm saying is all speculation, I have no way of going back in time and measuring it. 

The idea was to look for the difference tone at 3 khz. That is where imd will usually be audible.

 

[GRUMPYOLDGUY]

The idea was to look for the difference tone at 3 khz. That is where imd will usually be audible.

 
Good point. I'm used to talking about intermod in terms of 3rd order products. Those are typically closest to our signal of interest and usually loudest. 
 
The 2nd order difference tone will be the quietest usually. 

 

[pinnahertz]

The discuss of IMD as a by-product of bandwidth and system nonlinearity really only scratches the surface.  The correlation between system bandwidth and IMD is spongy at best, as there are other factors to consider.
 
A good reference paper that starts to get at the meat of the issue is "Spectral Contamination Measurement" by Deane Jensen and Gary Sokolich, AES 85th Convention, November 1988 (available at AES.org).  The went beyond two-tone IMD and created a several excitation "patterns", like energy at 120Hz intervals from 10kHz to 25kHz with a measurement window from 30Hz to 8kHz.  Their work started to get to the roof of why amplifiers with lower gain-bandwidth product sound worse than ones with higher gain-bandwidth product, etc.  They tested digital systems, analog tape recorders, all-pass filters, amps, and more.  I'd read this one before theorizing too much about ultrasonics and IMD.

 

[spruce music]

  The discuss of IMD as a by-product of bandwidth and system nonlinearity really only scratches the surface.  The correlation between system bandwidth and IMD is spongy at best, as there are other factors to consider.
 
A good reference paper that starts to get at the meat of the issue is "Spectral Contamination Measurement" by Deane Jensen and Gary Sokolich, AES 85th Convention, November 1988 (available at AES.org).  The went beyond two-tone IMD and created a several excitation "patterns", like energy at 120Hz intervals from 10kHz to 25kHz with a measurement window from 30Hz to 8kHz.  Their work started to get to the roof of why amplifiers with lower gain-bandwidth product sound worse than ones with higher gain-bandwidth product, etc.  They tested digital systems, analog tape recorders, all-pass filters, amps, and more.  I'd read this one before theorizing too much about ultrasonics and IMD.

I read that paper around 5 years ago.  Tried this test and didn't find it revealing anything new.  Maybe modern gear has such high gain bandwidth op-amps it isn't a problem.  I found spectral contamination to show me nothing more useful than two tone IMD.  I also have used a pair or three or 4 swept tones.  Equal spacing.  Like 500 hz or 1khz difference.  Run the sweep starting with 1khz and 2 khz with both sweeping to 20 khz.  Again I don't recall any of the few pieces of gear tested to show much different than just using 19 and 20 khz. 

 

[pinnahertz]

 
I read that paper around 5 years ago.  Tried this test and didn't find it revealing anything new.  Maybe modern gear has such high gain bandwidth op-amps it isn't a problem.  I found spectral contamination to show me nothing more useful than two tone IMD.  I also have used a pair or three or 4 swept tones.  Equal spacing.  Like 500 hz or 1khz difference.  Run the sweep starting with 1khz and 2 khz with both sweeping to 20 khz.  Again I don't recall any of the few pieces of gear tested to show much different than just using 19 and 20 khz. 

What did you use to generate the 125-tone test signal?  Did you apply 10kHz HPF before the DUT and the 8kHz LPF after the DUT? What devices did you test?