[FFBookman]

Using perceptual coding for data loss in music ----
 
Evil?
 
Necessary?
 
Not evil but no longer necessary?
 
 
Sound off.

 

[FFBookman]

Bandwidth needed per second of run time:
 
STEREO AUDIO:
lossy:
AAC ~ 256k
MP3 ~ 320k
lossless:
CD ~ 1000k
24/44 ~ 1800k
24/96 ~ 4000k
24/192 ~ 5800k
 
 VIDEO:
Netflix SD ~ 2000k
Netflix HD ~ 6500k

 

[FFBookman]

crickets....  maybe people avoid this due to past flame wars.
maybe people are sick of the debate?  not sure.
 
To me it's a critical debate, one about digital standards and consumerism.  
 
I've lived all the way through the mp3 era, I remember the realaudio era! 
I remember when "lossless" and "HD" were just fantasies for the internet, circa 1997.
I remember when I first asked a recording engineer why he records at 24/48 if everyone buys it at 16/44, circa 1998.
I remember the first time I met a producer who recorded and mixed in lossy formats, circa 2006.
 
and of course I remember ripping and filling my early iPods with this new amazing lossy format in the early 00's. I mean they sounded ok, right? Good enough for how cool digital files were.  Also there was 9/11 and a major war and other BS my country was involved in so mp3 fit right in to the new, portable, no attachments, nothing is permanent feel of the time.
 
So I guess this could be relabeled  
 
In what year did lossy become not "good enough" for you?
 

---

 
For me it was around 2011, after spending a year programming an amazing mix of music for one of the streaming systems at 192k. I worked tirelessly mixing/programming/meshing music, with every bit of DJ (club and radio) skill I had and it just wouldn't cut through to anyone.
 
I would work on my sets with vinyl and cd and really connect with the few that would hear my rehearsal. I'd be paying attention to kick drums and hi-hat patterns and so on as I was mixing. I'd then recreate it online at 192k mp3 and it would just sound like paper. I was missing a lot of the mix cues and really noticing the emotion of the music and the overall set was just killed.
 
I knew what lossy was, I understood there was degradation, but I just had accepted it as a depressing part of life, like a friend who had passed. I just gave up, thought lossless was too big and also didn't matter since all we had were phones to play it back on. Who wants to fill up a phone with a few lossless CD's? not me.
 
Now i'm almost all lossless, own prob 40 albums in 24bit, have a hi-res DAP with about 300gb of microSD that get a serious workout.  A big change in 5 years.  

 

[WraithApe]

 
Sick of this debate and I feel like you're just fishing now but I've got nothing better to do so I'll bite. It's just so much hot air. In the past, you've always refused to do any proper ABX testing of FLAC transcoded to lossy formats, preferring instead to rely on sighted evaluation, auditory memory and subjective impressions. It's pretty well established that with 320 mp3 (assuming a good encoder) or 256 aac, the point of transparency is reached and it simply isn't possible for most people to reliably distinguish them from lossless in a controlled test.
 
If your point is why bother with lossy audio in this age of huge bandwidth and cheap storage, then I might be more inclined to agree - but there are still some use cases for mp3, mainly in the form of DAPs. I want to have my entire music library on my 240gb DAP as I like to play it on shuffle and I just couldn't do that if I went for lossless - and, I still maintain, I wouldn't be able to hear the difference anyway!

 

[FFBookman]

 

 
Sick of this debate and I feel like you're just fishing now but I've got nothing better to do so I'll bite. It's just so much hot air. In the past, you've always refused to do any proper ABX testing of FLAC transcoded to lossy formats, preferring instead to rely on sighted evaluation, auditory memory and subjective impressions. It's pretty well established that with 320 mp3 (assuming a good encoder) or 256 aac, the point of transparency is reached and it simply isn't possible for most people to reliably distinguish them from lossless in a controlled test.
 
If your point is why bother with lossy audio in this age of huge bandwidth and cheap storage, then I might be more inclined to agree - but there are still some use cases for mp3, mainly in the form of DAPs. I want to have my entire music library on my 240gb DAP as I like to play it on shuffle and I just couldn't do that if I went for lossless - and, I still maintain, I wouldn't be able to hear the difference anyway!

I guess I'm fishing to know with how many people and why lossy survives. Yes based on bandwidth, of course.  I was all about MP3 15 years ago when I had a 1gb iPod and a system drive of 80gb!
 
Your use case is a solid one, but why not just get a few more MicroSD cards and go lossless?   I currently have 4 cards totaling 256gb and will move to 128gb cards next.  They are so tiny I could carry hundreds if need be.
 
So in your case it's still necessary, since you need data reduction and don't believe you are degrading the files in a way you can hear.

 

[WraithApe]

To be honest, I could probably live with less music on my DAP - it's more the hassle of going lossless at this point. I'd have to re-rip my (not inconsiderable) CD collection and re-source all of the digital files, all for what, to me, is negligible benefit. I didn't just dip my toe into ABX testing, I did a a lot at one point using Foobar's Comparator because I genuinely wanted to know whether it was worth my while going down the lossless route. I've seen it said by some that it's not viable means of testing but I could consistently and reliably pinpoint the difference between 192 mp3 and FLAC, so I think that proves it works. I was however, unable to tell the difference between even 256 mp3 and FLAC after many attempts with a lot of different genres including classical, jazz and electronic.
 
But if I understand correctly, you're less concerned with arguments about audible transparency and the veracity of that, but have a wider philosophical problem with perceptual coding on the grounds that lossy audio is basically dishonest - fooling the brain, if you like; palming off sub-standard goods on the consumer. I just don't think it's that black and white - I may be OK with lossy audio in principle but I still have standards. I happen to agree that 192 and below is audibly inferior (even though I don't accept it's night and day with the quality of today's encoders) so I will never settle for less than 320, where that is available.
 
Apologies if my first reply was somewhat confrontational, but you do have something of a rabble-rouser reputation when it comes to this subject 

 

 

[VNandor]

I go with lossy because given the type of music I listen to I don't really have a choice. Sometimes I could get lossless versions but it usually costs more for no reason at all which makes me mad. I mean it does not take more effort to publish music in lossless than in lossy, does it? They would try to charge me more because they are greedy not because they put in more effort.
 
With that said I go with lossless if it costs the same as lossy because why wouldn't I do that? Even if I totally understood how perceptual coding and human perception work and know why lossy sounds so similar (to the extent to be perceived as same in my experience) to lossless, I would still go with lossless.

 

[WraithApe]

  I go with lossy because given the type of music I listen to I don't really have a choice. Sometimes I could get lossless versions but it usually costs more for no reason at all which makes me mad. I mean it does not take more effort to publish music in lossless than in lossy, does it? They would try to charge me more because they are greedy not because they put in more effort.

 
I think this is a good point. Why do lossless files cost more? Surely they aren't putting a dollar value on bytes? It's not a commodity in any real sense. I suspect it's exactly what you say: because they can. It's just become accepted that lossless files will cost more and people rarely think to question it.

 

[Music Alchemist]

  I go with lossy because given the type of music I listen to I don't really have a choice. Sometimes I could get lossless versions but it usually costs more for no reason at all which makes me mad. I mean it does not take more effort to publish music in lossless than in lossy, does it? They would try to charge me more because they are greedy not because they put in more effort.
 
With that said I go with lossless if it costs the same as lossy because why wouldn't I do that? Even if I totally understood how perceptual coding and human perception work and know why lossy sounds so similar (to the extent to be perceived as same in my experience) to lossless, I would still go with lossless.

  I think this is a good point. Why do lossless files cost more? Surely they aren't putting a dollar value on bytes? It's not a commodity in any real sense. I suspect it's exactly what you say: because they can. It's just become accepted that lossless files will cost more and people rarely think to question it.

 
Since used CDs are available on Amazon for as low as a penny (plus shipping), lossless needn't cost more unless you're talking about a digital exclusive. If it's only available as a download, I prefer to get the highest quality version just to be safe. I mean, some downloads from iTunes, Amazon MP3, and the like have...sound quality issues. Could be due to a different master in some cases. I honestly find the high prices of lossy downloads more offensive than the slightly higher prices of lossless ones, simply because the lossy downloads sometimes have problems and end up sounding worse, even if it's not due to the lossy format itself, plus the fact that you can just get the CD at a fraction of the price. "HD" downloads, on the other hand, are far more expensive, and sometimes use the same master as CDs and consequently don't sound any better.
 
Just for kicks: free music for all!
 
http://www.head-fi.org/t/644595/official-free-flac-file-music-sharing-thread
http://www.head-fi.org/t/731523/the-free-music-thread-100-legal-downloads-of-all-formats

 

[WraithApe]

Slightly OT, but I would just say some used CDs are available on Amazon for very little money. Some, that have been OOP for a long time and had a limited run, can't be picked up so cheaply. I've got several in my collection that you wouldn't be able to pick up for less than £50 on Amazon, Discogs Marketplace or anywhere else. I once sold a CD for £250 - a very rare Icelandic version of The Sugarcubes' Here Today, Tomorrow Next Week!
 
But to be clear, I was talking about digital downloads. Like @VNandor, I just see no logical reason why there should be tiered pricing for different levels of encoding offered by sites like Bandcamp and Beatport. It doesn't take any longer to encode a lossless file, and if the argument is that it takes up more server space for the vendor, well... I would scoff at it. Like I said, I think they do it purely because they can; because people accept the tiered pricing without question - "it's just the way it is".

 

[Music Alchemist]

  Slightly OT, but I would just say some used CDs are available on Amazon for very little money. Some, that have been OOP for a long time and had a limited run, can't be picked up so cheaply. I've got several in my collection that you wouldn't be able to pick up for less than £50 on Amazon, Discogs Marketplace or anywhere else. I once sold a CD for £250 - a very rare Icelandic version of The Sugarcubes' Here Today, Tomorrow Next Week!
 
But to be clear, I was talking about digital downloads. Like @VNandor, I just see no logical reason why there should be tiered pricing for different levels of encoding offered by sites like Bandcamp and Beatport. It doesn't take any longer to encode a lossless file, and if the argument is that it takes up more server space for the vendor, well... I would scoff at it. Like I said, I think they do it purely because they can; because people accept the tiered pricing without question - "it's just the way it is".

 
That's why I used the wording "as low as".

 
I also sold a CD for over $200 once. It was a limited edition autographed version of an album by a band whose main member committed suicide. (Man, that sounds dark.)
 
Since lossless downloads generally still cost less than buying a new CD from a retail store, it doesn't bother me either way.
 
It really does cost more money to host larger files online, though, especially when they are being frequently downloaded.
 
At least Bandcamp gets it right. Almost always low prices, and you can download in any format you want, whether lossless or lossy. (I have never seen tiered pricing for digital downloads on Bandcamp. If you know of any, please link me.)

 

[WraithApe]

No, you're right - Bandcamp was a bad example - and I agree, they are one of the best out there when it comes to digital downloads; I use it a lot. Beatport definitely charge more depending on bitrate though, and they're the norm in my experience; Bandcamp is the exception I think.
 
I understand the cost of hosting - that's why I mentioned it - but I don't accept it as a valid reason. It's quite likely that companies like Beatport have a flat rate deal with the provider or even host the space themselves.

 

[FFBookman]

   
I think this is a good point. Why do lossless files cost more? Surely they aren't putting a dollar value on bytes? It's not a commodity in any real sense. I suspect it's exactly what you say: because they can. It's just become accepted that lossless files will cost more and people rarely think to question it.

Lossless files are larger but require one less step of processing. So they could charge more, they could charge less, they could charge the same. I see arguments on all sides.  That's really up to the marketplace. 
 
I know this -- audio coming from the artist/producer is 24/88 at least. That's about 3000k in mp3 terms. The mastering engineer always requests the highest resolution you are able to record\mix at. The mastering engineer then upsamples it to the highest multiple of that that his system works at (in this example 24/176).
 
After doing his work mastering (mainly EQ, volume, sequencing) he is paid to deliver the files in whatever format is needed for distribution.  From 1985-2010 or so this meant 16/44 PCM lossless. Since he was usually above 16/44 working, he has to downsample and dither to get back to 16/44.  
 
My mastering engineer would apply different dithers to different sets of files to let me preview them and choose the one which closest matched my original mix.  None of them would 'match' (dither does that) but I would pick my favorite and from there he'd lossy them down to 320k mp3, 192k mp3, and 256k AAC and then ID and tag them for distro to the various store aggregators. 
 
When Apple launched their "mastered for iTunes" program a few years back they recommended all music be submitted as 24bit and Apple would apply the lossy encoding and store both the master copy and the AAC, but only sell/stream the AAC. Apple is sitting on lots of 24bit masters and doing nothing with them.
 

---

 
Lossy coding attacks the mix of the music and since I mix music, I take that personally.   Not only are they selling us 10% of the data (art) and charging full price, they are destroying critical elements in the music and the mix. The layering and interplay of instruments and voices is what makes music tick.
 
It's called 'perceptual coding' because it is a method based on the inability to "reliably perceive" various types of sounds, and by that they mean people must describe them the same, identify them each time. I see how it works. You can't even get 3 people on head-fi to agree on terms and measurements for what they hear in mixed music. Science has to go to test tones, which don't help, since they have no mix, no interplay of instruments, no layers of sounds, no musical anything. You can't do much science with music, music is still too complicated and emotional for science.
 
Music is the magic here. Science doesn't really know how it's made or computers would have written all the amazing melodies already. They have written none. No computer has ever played a stringed instrument or sung anywhere close to a human. No computer has ever composed a good pop song, much less a symphony on it's own. Even percussion, which computers can be decent at, is still far better when played by a human than a computer. Of course they are tools to be used by humans.

 

[FFBookman]

  To be honest, I could probably live with less music on my DAP - it's more the hassle of going lossless at this point. I'd have to re-rip my (not inconsiderable) CD collection and re-source all of the digital files, all for what, to me, is negligible benefit. I didn't just dip my toe into ABX testing, I did a a lot at one point using Foobar's Comparator because I genuinely wanted to know whether it was worth my while going down the lossless route. I've seen it said by some that it's not viable means of testing but I could consistently and reliably pinpoint the difference between 192 mp3 and FLAC, so I think that proves it works. I was however, unable to tell the difference between even 256 mp3 and FLAC after many attempts with a lot of different genres including classical, jazz and electronic.
 
But if I understand correctly, you're less concerned with arguments about audible transparency and the veracity of that, but have a wider philosophical problem with perceptual coding on the grounds that lossy audio is basically dishonest - fooling the brain, if you like; palming off sub-standard goods on the consumer. I just don't think it's that black and white - I may be OK with lossy audio in principle but I still have standards. I happen to agree that 192 and below is audibly inferior (even though I don't accept it's night and day with the quality of today's encoders) so I will never settle for less than 320, where that is available.
 
Apologies if my first reply was somewhat confrontational, but you do have something of a rabble-rouser reputation when it comes to this subject 

 

 
I know - I generally lurk then when I chime in I'm considered a rabble rouser. Whatever, I say, I've been doing pro audio since before most people here owned an iPod, so I can take it.
 
I get the fact that you don't want to re-rip. I still haven't re-ripped everything 4 years after going lossless. Some discs are long gone, some I bought as MP3 and don't really want to buy another.  But there are about 30 albums I've bought again as 24bit FLAC, unlocked, as hopefully the last version I'll ever need to buy. I also scan the cheap CD bins because lossless LP for $1-$2 is a great deal. 
 
Your concerns about audio transparency and veracity make perfect sense to me. I think you were methodical and logical in your assessment, but I think you still came to the wrong conclusion. Audio is like that.  It's transparent and hard to quantify. 
 
This debate from 1998-2012 or so was more complicated -- there were concerns with storage needed and the playback ability of digital files. It just seemed ridiculous to need 100mb for 1 song. You trying to tell me that a vinyl record holds 800mb? Yeah right.  When transferred to proper 24bit digital files, it does. It seems backwards, counter intuitive to the mp3 generation, but it's a fact of life.
 
When they put out mp3 I was as amazed as everyone else. How did they get 50mb into 5mb and still make it listenable? How can the vocals and melody and beat be so clear with minimal distortions?  It's all about masking and taking away timing cues. Attacking everything in that song except the lead vocal, melody, and beat. Attacking people's perceptions, making a defense or damnation of the format nothing but emotion.

 

[WraithApe]

 
Your concerns about audio transparency and veracity make perfect sense to me. I think you were methodical and logical in your assessment, but I think you still came to the wrong conclusion. Audio is like that.  It's transparent and hard to quantify. 

 
How can you can say that though? I have plenty of ABX logs to prove that I am unable to reliably distinguish high quality lossy from lossless, so the conclusion is inescapable. I'm not questioning your professional credentials but I still think your objection is primarily a philosophical one: that mp3 and aac is selling short (or 'attacking' as you put it) the artist's original vision and palming off a kind of degraded facsimile of the real thing on the consumer. I accept that the highest resolution possible is essential for mastering and archival purposes, but for playback, if I am unable to distinguish 320 from FLAC in a controlled test, then what's the point - other than having the peace of mind that I possess a bit-perfect copy of the digital master?