Lossless Vs Lossy (ABX Comparator)

Audio Hardware & Software Sound Cards & DACs & Other Hardware
21

None of us are denying that on paper, a lossless source certainly does contain more information, and SHOULD sound better than a lossy file of the same recording.  What we're arguing here is whether or not you can tell the difference in a double blind test.  You can do this in less than 10 minutes with foobar.  Just download foobar and the ABX comparator plugin (I'll provide links below) and use dbpoweramp to encode an mp3 of whatever song you want from a FLAC source.  Then, load them both up into a playlist in foobar, highlight them both, right click them, and go to utility >  ABX two tracks. (If you want to make things a little more transparent, put the flac source higher in the playlist than the MP3.  This will ensure that the FLAC source is "A" each time.)

Track A and B will be FLAC and MP3, and then you have to match up either X or Y to A.  Again, X and Y will be the same tracks, and you'll have to differentiate between them.  You should repeat the trial 11 times to ensure accurate results, and if you want to prove something to us, take a screenshot after you finish the 11th trial and post it here.

22

So far I'm doing better than my first trial, but I'm still within margin of error.  We'll see how things progress from here.

23

I know that the audio will be changed but the only thing you don't want is downscaling, difference between sampling frequencies and bits is a debate worth having, but telling the difference between original and upscaled sample rates is taking is a bit far.

24

+1 to kuchuck's reply. It's not about showing on a spectrogram that there is a difference - We all know there is. What this is about is if there's an audible difference, and, unfortunately, non-blind tests just aren't good enough.

personally do not like headphones, simply because headphones can't be calibrated - to be as flat frequency response as possible. There always going to be comb-filtering in the upper-mid-to-high frequency (on the driver it self). 

I think what you meant to say was doppler distortion and not comb filtering. Doppler distortion is a claimed tendency of loudspeaker drivers producing more than one frequency to modulate the higher frequency tone because of the cone movement, producing the familiar doppler effect. Comb filtering, on the other hand, is the interaction of direct sound from the sound source and a delayed sound from a reflection (say a wall) to variously add up or subtract depending on the phase of the signals, producing a comb like frequency response. It is a very real phenomenon in loudspeaker - room interaction, but does not happen in headphone listening since there's only direct sound.

Doppler distortion does not affect headphones much either. It's true that since headphone drivers are usually full range, unlike loudspeaker drivers, there's more tendency of doppler phase shift in headphones, but this is more than offset by the fact that headphone drivers have very low excursion - They just don't move fast enough for Doppler effect to become a problem (if it even ever is a problem, but that's another discussion).




25

Ironically, these videos have inspired me to try to get more FLACs into my desktop listening experience, which involves alot of time trying to find albums again.  Tiring process, especially since a good chunk of what I listen to has never had any sort of CD release (ala independent electro not on bandcamp, not mainstream Vocaloid producers)

26

Well, to better illustrate the same point, I used photoshop to subtract image of the lossy audio from lossless.

http://imgur.com/a/CtEwf

Here are conversion settings and filesizes:

OGG:
Desired Max: 500
Desired Avg: 320
Desired Min: 160
12.9 MB

MP3:
V0 (245kbps)
11.0 MB

27

Yup, you don't even need all the audio hardware to clearly hear the difference, all you need is to not use Windows.

Great setup by the way, I love Adam monitors too, those ribbons are just incredible.

I don't want to argue with the people that say that there is no difference between flac and mp3 on a Windows machine, because it's also true, well there is, but it's minimal.

I do object to the constant avoiding of the core of the problem, especially when it's done with unfait meta-arguments like "people that hear things the average joe doesn't" or "people that listen to Beethoven"... that's not scientific, that's populist bullshit that helps nobody.

The basic message stays the same: don't buy expensive audio gear for a windows PC, you'll keep buying stuff without ever being satisfied.

Thing is, people that have never heard a more open and more acoustic sounding audio reproduction, have no clue what it's all about. They really need to open their minds before anything else, and analyze the problem scientifically.

Fact: since Microsoft took over Skype, there has been a constant problem with Skype and linux audio stacks. In the latest iteration, you have to artificially add 30 to 120 ms of latency to the skype audio stream in order to prevent buffer overrun. That's because Skype is now targeted at the DirectAudio API, which is conceived to resample everything. Resampling is satan, especially if it's done in a poor way, like from 44.1 or 48 to 96, that's pretty ok, but to resample and requantize from 16/44.1 to 32/floating, or even worse, from 24/48 or 24/96 to 32/floating, that's just about the worst thing you can do, it will make everything sound dull, even if you don't think it sounds dull because you've never heard it when it didn't sound dull, to those that don't have the same problem, it will still sound like a turd.

This entire discussion is - just like Logan said it would be - a test about opening your mind more than it is anything else.

28

What do I need to do to play my music without using windows. I appreciate your input, but you can't just say "flac is better, just not on windows" with out telling us the proper way of playing flacs to our systems. Id love to be able to play flacs and hear the difference I just dont know how according to your comment.

Do I need to run Linux, burn my whole library to a multiple discs? Stream my music to a music streaming reciever? Does playing my flacs off my android phone do anything different over windows?

I created this thread to find if I was doing something wrong and what equipment or methods I need to play them properly. I wouldn't be buying CDs at all if I thought they weren't higher quality .

 

29

Do an ABX test and post your results or shut up.  It's as simple as that.

We are ALL very heavily influenced by confirmation bias.  I thought that I could hear a difference between FLAC and MP3 320 until I did my first ABX test the other day.  It proved to me that I was dead wrong.  You're probably no different. 

Until you can prove to us (with evidence) that you can hear a difference, your testimony is only anecdotal, and can be disregarded without evidence to back it up.

As for the "don't play your music in windows" argument, there are simple ways to get around this, such as using ASIO or WASAPI.  I always use WASAPI in foobar, so the windows mixer isn't an issue.

Seriously, it's not like I'm asking for much.  Just do the test.  It'll take less than 10 minutes.  If you can prove me wrong using the ABX test, I'll gladly accept that you can hear a difference.  However, if you can't prove anything, then you're just wasting everyone's time.  Now either man up and take the test or admit that you're afraid you might be wrong and stop spouting nonsense.

30

Once again, we realize that on paper there IS a difference.  No one is denying that.  What we're discussing is whether or not that difference is audible.  I suggest you do an ABX test like we've described in this thread.

31

I never claimed to be able to hear the difference and I don't have equipment good enough to hear it anyway, therefore I myself always use lossy audio (prefer OGG because of efficiency).

32

Well, try the ABX test anyways, and if you get nicer gear down the line, try it again.  You're not going to get better results.

 

33

Zoltan, you are claiming that it's impossible to have quality audio output with windows without giving any reasons for that.

It would be better if your hatred of proprietary software was helpful in some way.

There are easy ways to bypass windows audio mixer, you know.

34

I'd also like clarification on this. Zoltan, you say that,

On Windows, everything is forced into 32-bit floating but requantified in the most broken and stupid way imaginable to mankind 

Resampling is satan, especially if it's done in a poor way, like from 44.1 or 48 to 96, that's pretty ok, but to resample and requantize from 16/44.1 to 32/floating, or even worse, from 24/48 or 24/96 to 32/floating, that's just about the worst thing you can do, it will make everything sound dull, even if you don't think it sounds dull because you've never heard it when it didn't sound dull, to those that don't have the same problem, it will still sound like a turd.

Windows mixer doesn't do sample rate conversion, unless the set output sample rate is different from the source material (then it does it, of course). Also I am not sure what you mean by resampling from say 16bit / 44.1 kHz  to 32 bit / floating. It seems like you think "floating" is a sample rate, which it of course isn't. In any case, converting bit depth from 16 or 24 bit integer to 32 bit floating point is easily done with transparency, what is it that Windows mixer manages to do so terribly wrong? The fact the mixer works at 32 bit floating point is, in fact, excellent news, since this is the preferred method for digital volume adjustment/mixing, being very resilient to loss in bit depth and to avoid clipping.

 

And of course, you can bypass the mixer completely if you so choose by using wasabi or asio.

35

I actually typed the whole thing out in reply, then Drupal said the post I was replying to didn't exist, but it's still there.

I'll post separately about this later, the basics are that ASIO native prevents the problem, but that implies also using an application that uses the ASIO API natively, so no plugin in Foobar, but Cubase for instance.

Resampling and requantizing from 16/44.1 to 32/floating, is not nice on the audio, the math behind this has been posted here before.

Anyone that has used a DAW and an ASIO sound card to record stuff with his PC, using Cubase or Reaper or any other DAW, knows that there is a difference, just like anyone that has used the experimental media player based on the Mantle audio API (that isn't released yet and might not get released, we'll have to see, but it's available out there for people with an AMD graphics card).

I find it absurd that people bring out the big guns to shoot at people that testify that there is a huge difference between the quite frankly horrible MP3 files (yes, I'm done with euphemisms, it's OK to tell the truth) and flac or wav. A simple ddg.gg !g search would lead you to many documents (including MS-KB) of the problems with the directaudio API and windows audio stack. Those that know, and have opened up their mind enough to be able to experience the difference, owe nothing to those that are still ignorant, I salute those that share their experiences on how shitty MP3 is in comparison to lossless formats. Those that have the guts to break through the circle of ignorance and experience the truth, are invited to do so, but this is by no means a theoretical or elitist discussion, and it's much easier for those that don't know yet to open their minds and throw off their tunnel-vision goggles, than for those that know to spend hours looking for a bunch of technical documents nobody reads anyway.

Let me give you a practical example: since Skype has been modified by Microsoft, not only has the sound quality deteriorated, the service deteriorated, but there have also been problems with linux sound stack integration on every iteration of Skype. Most linux users use PulseAudio on ALSA. PulseAudio follows ALSA in terms of quantification and sampling frequency, and the linux sound stack has added the Windows 32/floating format long ago to ensure compatibility, however stupid that format is in relation to the formats that are used to distribute and record music/media. But PulseAudio also has a function to automatically adjust the buffers to reduce latency while avoiding artefacts. Now in the latest iteration of Skype, you actually have to start Skype up with a 60 ms minimum (some PC's require 120 ms) latency, in order to avoid constant buffer overrun which leads to choppy distorted audio or audio loss. The reason is that the Skype output doesn't provide enough constant stream to fill up standard buffers in time, it needs 30 to 120 ms of pre-buffering to ensure a constant stream. Now those familiar with digital audio, know that 30 to 120 ms is huge, that's a lot of pre-buffering that needs to be done. Skype is made for the Windows audio stack, and is timed perfectly to that, it fills out all the data that streams in the windows audio stack. That means that the stream just contains much less data than a regular stream, nothing more, nothing less. How much more proof do you need, guys? The windows audio stack isn't even made for a standard VOIP stream... That also means that when music is played through windows, whatever you quantize or sample it at, it will not go unscathed through the windows audio stack, it will be reduced.

I've posted about this in the past, well before the audio hype on the forum.

Foobar and Audacious, and other XMMS forks, are all EXACTLY the same application. Foobar is the full-featured version of Audacious for Windows. The thing that separates them, is that they target other APIs/sound stacks: Audacious targets all POSIX-compliant stacks, Foobar targets the windows stack. By using ASIO plugins in Foobar, you add a compatibility layer, you don't solve the problem at the source.

So to avoid the audio stack in Windows, there are only two ways: 1. use a pro audio interface with native ASIO drivers made by the manufacturer, or use a standards compliant audio card with ASIO4ALL, but combine it with a native ASIO application to play back the files, like Reaper for instance (don't forget to switch to ASIO output in the settings, this will block the windows audio stack from being able to access the sound card, so you won't be able to hear other sounds from applications sadly); 2. Use the experimental (and still very buggy) media playback application that targets AMD Mantle's TrueAudio, or for people that have a Creative sound card, there is a Creative application that is packaged with it, and that anyone can also download, that will bypass the Windows stack and use the creative native API of certain Creative cards, but frankly, the coding on the Creative software is so bad, you'll shoot yourself through the head after trying to use it for half an hour, because it freezes and crashes constantly.

It's not that I want to hide solutions for Windows users, I don't care, it's just about music, if there were a simple solution, I would post it for sure. I've posted other Windows fixes in the past, this isn't about windows vs *nix. It's just that there isn't any out-of-the-box solution right now for playing music in Windows that allows for users to hear the difference between bad and better quality audio.

Now for those that think that audio doesn't get any better than MP3 quality on their Windows box, I can't help you, I've given all the info you need to open your minds and solve the problem. And that's it, people. I've posted in the past about simple tricks to patch some of the horrific compression and frequency limitation of MP3's in audacious in linux. The reason why I didn't post that for Foobar, is very simple: because it serves no purpose, you can't hear the difference. I know that Foobar has similar plugins to Audacious, duh, they're pretty much the same application, and I like all people to enjoy better sounding music, not just linux users, duh, it's music... so cut the crap about this, if you want to know, it's as easy as trying it out, it doesn't cost you anything to download Ubuntu or Mint or Manjaro or Sabayon or OpenSuSE or Fedora, to install it, and to try the difference out. Most of you will have a spare computer of some kind laying around, maybe a RasPi, maybe a laptop, maybe your mom's old laptop with a Pentium IV CPU. That's enough to try out the difference, listen to it on your Windows box, then AB with the linux box. If you have a CD drive in your computer, rip it in windows in different formats, then install linux on your box and rip it in linux in different formats (ogg, mp3, flac, wav), and compare the sound difference between the different formats. After you've done that, then we'll talk again. You'll all be hit by the obvious difference in quality between wav/flac and mp3 in the face with the impact of a reentering space shuttle, and you'll all be doing yourselves a big favour. People on this forum that use macs and linux boxes have done you all the favour of sharing their experience, that's more than they have to, now it's your turn to open your eyes to the truth. It really is very simple. In fact, you don't even have to install linux on any machine, just grab a linux live iso and put it on a USB drive... how easy it is to discover the truth, amazing...

36

You know people might actually listen to you if you weren't so patronising. I'm installing an abx comparator on a linux os atm so I'm giving that a shot.

37

So those that share knowledge inspite being shot at with the big meta-argument-BB-guns, are patronising, that's perfectly fine with me. I don't need anyone to listen to me, fact of life: not even your spouse or kids listen to you lolz, hell not even the people I pay to listen to me listen to me all that well at times. I don't care if someone is listening to me or not, as I said before on multiple occasions, sharing information is nothing but an invitation, everybody's free to take the invitation or to discard it, no damage is done, noone got killed, sit back and enjoy your coffee, there are more important things to go haywire about.

I salute you for your explorative attitude though, there is nothing like finding out for yourself. There is great benefit to be had in exploration instead of theoretical discussion not backed up by experience, in the end, you never know if and how it feels different to kiss a person of a different race or phenotype until you try it out, and I don't regret my explorative attitude, even if the difference is often smaller than the difference between mp3 and flac lolz...

38

No saying stuff like 

"Those that know, and have opened up their mind enough to be able to experience the difference, owe nothing to those that are still ignorant, I salute those that share their experiences on how shitty MP3 is in comparison to lossless formats"

Is patronising, you could of removed a large portion of your comment and have it as simple helpful information and suggestions to get the most out of our audio. Were all willing to open our minds we just need to know how.

39

Resampling and requantizing from 16/44.1 to 32/floating, is not nice on the audio, the math behind this has been posted here before.

That is absolutely not true, conversion of the bit depth is very simple. I don't put many things past Microsoft, but I can't imagine they'd fuck up simple conversion like this. Also, if it was so bad on the audio, why do most professional audio software use 32 bit floating point internally (as do many stand-alone DSP's)?

40

Oh, the intention Microsoft had was very good, in fact the idea to go by the 32-bit floating standard was inspired by a couple of high profile audio professionals Microsoft hired to do just that. The problem is that the implementation was, and is still, faulty, and that Microsoft didn't get to pull it off without reducing the data stream to a point where it's ridiculous. The audio professionals afterwards also declined to have Vista marketed with their names. It was obviously Microsoft's intention to build a very high standard media-centric experience with Vista, they came out with Media Center, they did a lot to jazz the graphics up with screen gadgets, etc..., but it isn't a secret that they screwed up bigtime. My guess is that they wanted to create a market that wasn't there yet, with an unseen media quality point and a unique interface experience, and that they had the best possible intentions, but it failed because they didn't push the right buttons, because Ballmer isn't a tech guy but rather a grocer with an accountant's degree. Ballmer has said himself that he thinks that he failed in managing the Vista project, and I think that is actually an honest thing to come out of Ballmer for a change. It's a typical case of capitalistitis/corporatitis, or as the ancient romans said: there is no better way to make sure a patient dies than by hiring all of the best doctors at the same time to take care of him. Microsoft didn't feel like it was under pressure to produce Vista, they ruled the market with XP already, but they took a long look at KDE and MacOS, and saws the opportunities of monopolistic media distribution over the Internet to set new goals, that were technically overambitious. Microsoft should have stuck to just copying Apple and providing compatibility. As it turned out, Microsoft actually invested first in trying to provide a system that would allow them to monopolise media distribution over the Internet, but they screwed up so bad that Apple was able to snag it away from them with iTunes, which is just a very simple application, not half as ambitious as what Microsoft had planned with Media Center in Vista, but with the benefit of a good audio stack. In fact, Microsoft didn't learn much from the Vista experience, as Windows 8.0 demonstrated. Windows 8.1 is to Windows 8.0 what Windows 7 was to Vista, but the main problems from the Vista adventure are still there. Microsoft seems to be looking for the consumer added value since XP, but to be honest, they didn't do a very good job, there seems to be no vision. This is something that big corporations do, RedHat is screwing up Gnome in the same exact way, it's a consequence of that thing in big corporations where marketing departments start to take over the development instead of engineers, or where engineers don't grow into COOs, but external people with a 30 year old engineer's degree and an 5 year old MBA and great networking skills and a great track record are hauled in for big bucks to lead projects. It will never change, it's been like that in the car industry, the pharmaceutical industry, the military industry, the finance sector, etc etc etc...

Now why is CoreAudio so great... well, that was made by engineers gathered by Steve Jobs after he was kicked out of Apple by the Intel corporate guy. They were not high profile engineers, because Jobs couldn't afford those, they weren't lead by marketing people, because there wasn't any market for their product, and there never was, the market developed after Apple took Jobs back and his operating system with it (well, his version of of OpenBSD that is, let's be serious about this). In linux, the same thing happened, but with much more diversity, as is typical for open source, and with the connotation that audio is always the last thing that is developed by computer geeks, because it's not high on the priority list, as it involves copyrighted and licensed stuff in an open source world and things like that. That changed when linux was being used in multimedia hardware, which started about a decade ago with those satellite hack boxes that were hugely popular in Europe in the early 2000's, that did all the decrypting of scrambled broadcasts and were universally standards compliant, from mp3 to laserdisc, and that gave linux a huge boost in multimedia. Right now, high profile stuff from broadcast consoles to professional recording gear, as well as low profile stuff like mp3-players and TV's, are all on linux, and all have audio applications, and only the linux based stuff, whether embedded or not, can handle all of the commercial and non-commercial media formats, and the development path required to be able to make it do that, has kept the audio stacks and video stacks relatively straight forward and modular. You can get about the same low latency with ALSA as you can with CoreAudio, or you can put PulseAudio on top of ALSA for more operating comfort and easy-to-use functionality, but the entire system always adapts to any format and any standard. The focus of many media distribution and marketing corporations is therefore not on Microsoft to provide the best possible media experience, and again, Microsoft is not under pressure to fix things on Windows, particularly because they have launched their own media distribution platform that is hardware-based instead of being linked to MS-Windows. The implementation on the XBox One speaks for itself: the XBox one runs on a linux-like hypervisor, that virtualizes two operating systems, one of which is a slim version of Windows, including a standard version of DirectAudio, so that applications like Skype can run on the XBox, and the XBox OS, which is a relatively low-level games-centric operating system, that connects the APIs to the paravirtualized hardware, and is not hampered by the PC-Windows audio stack limitations. So Microsoft sure makes a hell of a detour in the Xbone to provide compatibility with PC-Windows applications, which again demonstrates the unorthodox methods of the Windows audio stack. So they can't rewrite Skype because of license issues, and Skype is a major pillar of their Internet business, so they actually have to use a Hypervisor to solve the historical problems...

So there you have it, that's pretty much the bottom line, I could go on forever about all of this, there are so many problems due to the historical Vista-era mismanagement, that Microsoft has missed more opportunities than they had deserved to miss, but that doesn't make the product any better. I wish Microsoft hadn't fucked up with the audio stack and other things (like the filesystem, to which they keep on clinging because they have software patents on them that now turn out to be null and void in the first place... like wow, speaking of mismanagement, they actually fucked up core performance and functionality of their operating system for that through the years...), because at the time, I had a complete Windows-based recording setup for advertisement overdubs in my company and in my home studio for my band to record music, and Vista broke compatibility with a lot of pro audio gear that I had invested in, even though it was ASIO native gear, and didn't care about the Windows audio stack..., and I really liked Cubase, had been using it since the early 90's, and I didn't feel the need to change it because I could use it blindfoldedly, it was the most efficient solution for me. I would still use Cubase right now if Windows would work with the 50k worth of pro audio gear, because I think it's still more efficient to use than Ardour, but Microsoft decided otherwise, they decided to break compatibility for all ASIO gear that didn't also comply with the crappy DirectAudio implementation they introduced with Vista, even when that gear was standard compliant. So this is not a discussion about resolutions or sample frequencies or the mathematics behind it, it's just practical knowledge about the Windows audio implementation since Vista that I've experienced first hand, and that I'm very frustrated with. In particular, I have audio interfaces that are not capable of 32/floating natively, but that are capable of 24/96, 24/48, 16/48 and 16/44.1, and they just won't work with Windows Vista and later versions, even though they actually sound better at 16/44.1 than audio interfaces that I have that are compatible with Win7 and that get 24/96 from the 32/floating windows audio stack, that generates the stream from a 24/48 source, when using the windows audio stack that is. So even though the idea of 32/floating fixed audio processing is really good, the implementation is seriously faulty, and the commercial goal that initiated the decision to go with that standard, was never realized, so the fault was never fixed.