audio questions

[L233]

Adding a fraction at the beginning IS a delay, what's wrong with that?

What are you talking about? Of course it's a delay and the point is that you completely missed that fact and claimed that the zxlpw shows signs of "additional signal processing" while, in fact, you merely looked at the wrong sample position and then convinced yourself that the zxlpw sample is the uncompressed original.

I posted a frigging screenshot showing the correct position.

As for the rest, you can only hear the missing cymbals if they're actually being played. In case of Amon Amarth he merely plays stupid 4/4 with zero subtleties to begin with.

You didn't do any better with the Nevermore sample, did you? And you didn't even bother with the other two I posted. Instead, you're whining that the production of the Amon Amarth track isn't to your liking. Way to miss the point.

It's always the same old story with the deluded audiophiles. Excuses and then some more excuses. It's the wrong type of music. The music has a shitty production. Something must be wrong with the encode. There's been foul play. The test sucked to begin with. Whatever.

For goodness' sake, you couldn't even correctly identify the 128 kbit CBR one.

And whatever your opinion, it is simply impossible that the compressed audio produces more highs than the uncompressed one without any additional processing. That's a simple fact, no need to even talk about it. That is clearly audiable and that's why I choose what I chose.

Where are these "more highs"? Again, your screenshot from Sound Forge shows the wrong sample position. I provided you with the correct sample position and I don't see "more highs". So where is this "clearly audible" stuff that you supposedly heard?

Sorry, but this is getting tiresome. You tried to cheat, you looked at the wrong sample position and then you convinced yourself that the second most compressed audio sample has "clearly audible" differences and thus must be the uncompressed one. That's the classical placebo effect - it sounds differently to you because you think that it has to sound differently.

I'm not saying that you manually processed anything, but probably the software you used did it automatically. That is not a simple conversion, that is processing by definition.

Sigh, AGAIN where is that supposed sound processing? Look at the screenshot I posted. And are you seriously claiming that the zxlpw sample sounds different to you? The music merely starts 0.025 seconds later, that's it.

 

[N00b]

And whatever your opinion, it is simply impossible that the compressed audio produces more highs and more "air" than the uncompressed one without any additional processing. That's a simple fact, no need to even talk about it. That is clearly audiable and that's why I chose what I chose. Before I viewed it in SoundForge as mentioned, not that it has any relevance.

That's actually wrong shows you know nothing about audio compression. MP3 compression works by filtering out inaudible (or barely audible) frequencies, so it's totally possible that the remaining frequencies become more pronounced and thus clearer. Audio compression does not produce more highs, it removes some frequencies that mask the highs and make the sound muddier. That's why people frequently favour compressed audio over non-compressed audio in double blind tests. Maybe you should gather some basic knowledge about audio compression before making dumb accusations.

 

[_xxx_]

Stop riding on that sample position bull, I'm talking about the amount of visible ripple at any given position.

I'm not arguing that I "failed", not at all. But the fact that one "likes" something better still doesn't make it more truthful to the original. And that fact is definitely audiable.

Noob: where did I claim to "know something" about compression? I was talking about compressed music not sounding entirely like the original and that is obvious in all of the cases, compressed sounding more pleasing to the uninformed listener or not.

You can only discern the original exactly if you exactly know it's sound, otherwise you can only talk about the sample being "better" or "worse" to your ears. For comparison, take a guitar and plug it in with a flat EQ, it won't sound very enticing ususally. Add a good EQ-setting to that and there you have a ripping sound. But it's a product of processing and doesn't represent the correct original sound, regardless of which one you like better.

 

[Albuquerque]

Epic LOL

I think you already said what you need to say: you can't tell, and you like the compressed one better.

Done.

 

[N00b]

Noob: where did I claim to "know something" about compression?

Right here (or how do you know it's impossible without any knowledge of audio compression?):

And whatever your opinion, it is simply impossible that the compressed audio produces more highs and more "air" than the uncompressed one without any additional processing. That's a simple fact, no need to even talk about it.

But this is really my favourite part of your post:

You can only discern the original exactly if you exactly know it's sound, otherwise you can only talk about the sample being "better" or "worse" to your ears. For comparison, take a guitar and plug it in with a flat EQ, it won't sound very enticing ususally. Add a good EQ-setting to that and there you have a ripping sound. But it's a product of processing and doesn't represent the correct original sound, regardless of which one you like better.

O RLY? I remember you writing something different just a couple of days ago:

Pretty much any musician will tell you the difference between any mp3 and the original with ease. Especially music with lots of dynamics is an easy candidate. Again, try Death's "Human" or "Individual Thoughts Patterns" or Nevermore "Dreaming Neon Black". It's day and night, even with some $20 headphones.

LOL

 

[_xxx_]

What is "different" about that sentence? Of course it implies knowing the original, anything else makes no sense.

Albuquerque: take a classical guitar and put it through a distortion effect and I'll like that better too. It says absolutely nothing about it being truthfull to the original or not. We're obviously talking about two different things here to begin with. The original question was "Does mp3 compression alter the original signal audiably" and the answer to that is a clear yes. Liking it or not has zero meaning.

 

[thehulk]

Never gets old no matter how many times I've seen this.
I was a member of r3mix forums years ago, followed by HA forums.

Overconfident/misguided sap claims all kinds of BS, then gets shot down by
not backing it up with ABX results or failing a listening test.

Hilarious.

 

[Albuquerque]

The original question was "Does mp3 compression alter the original signal audiably" and the answer to that is a clear yes. Liking it or not has zero meaning.

Actually no, the original question was "does lossy compression make for bad audio?" and the answer has been an emphatic NO as proven by the two people who said they could unequivocally tell the difference: 2008 and you.

Here, let's give you a nice reminder from the very first post:

5. PC games don't currently use wma lossless do they? Just making sure it wasn't my imagination b/c to me games from back in the day that used Redbook sound a hell of a lot clearer and have much better dynamic range. Dynamic range in recent pc games sounds very compressed (highs and lows are very limited), so I don't know how it they couldn't be using MP3's.

Dynamic range is being negatively affected and it's due to lossy compression. Nowhere does he say that he got to listen to the uncompressed raw original file and knows the difference, the only statement here (which started this entire blind comparison of audio files) is that he knew that lossy compression was causing the problems and he would immediately know the difference if it wasn't compressed.

You made the same claim, and you both failed miserably.

 

[L233]

Stop riding on that sample position bull, I'm talking about the amount of visible ripple at any given position.

No, I won't stop. I know that it must be embarrassing for you but you made a huge and revealing blunder here and you're still maintaining that the waveforms show something that would amount to an audible difference.

You're simply changing your story. You looked at the wrong sample position, called foul play and claimed that there was some sort of signal processing done ("boosting", "upconverting", "added FX") and that the three audio samples that look near identical must be the compressed ones because they have "less information" and there is a difference in the "level of detail". And, of course, you claimed that you can clearly hear this "additional sound processing" that doesn't even exist.

And now you're talking about ripple. What? Look at my screenshot and point out to me where this different amount of "ripple" is supposed to be when looking at the correct sample position. And tell me please how you deduced that there was some boosting or upconverting or whatever done.

I'm not arguing that I "failed", not at all. But the fact that one "likes" something better still doesn't make it more truthful to the original. And that fact is definitely audiable.

Right, so you're moving the goal posts now. That's the last resort after all excuses fail.

Sure, MP3 compressed audio is not 100% identical to uncompressed audio. It's a lossy compression. This has never been the debate, everyone knows that. The question is: is the difference audible to an extend that makes compressed audio sound bad?

Even if there are a few tiny instances over the course of a track where there are discernible artifacts (when you really, really, really pay attention), does it matter? Does it make the compressed audio sound worse (which was the initial claim) or just very, very slightly different in some rare instances?

This was never about 100% "truthful" replication, it's about compression resulting in audibly worse sound quality. But let's review some of the claims you made, shall we?

- you claimed that you can hear missing cymbals and guitar harmonics when listening to MP3s in your car (presumably while driving)
- you claimed that there's a notable loss of "highs and higher order harmonics"
- you claimed that one can "literally hear some stuff missing" even with 320 kbit MP3s
- you claimed that "pretty much any musician will tell you the difference between any mp3 and the original with ease"

No one expected you to be 100% correct but considering your claims, you should have at least been able to correctly identify the most compressed one (which certainly must miss the largest amount of cymbals, harmonics and highs) and the uncompressed one (which has all these cymbals, harmocis and highs that should be missing from the compressed samples).

I think at this point it's fair to call bullshit on all of your claims.

You can only discern the original exactly if you exactly know it's sound, otherwise you can only talk about the sample being "better" or "worse" to your ears.

No one has ever claimed otherwise. The point is that you cannot do a proper ABX test over the web because some people might just use tools like, say, Sound Forge and if they have an uncompressed source file, they can cheat easily and identify the uncompressed test sample in a second.

Furthermore, the point remains that, despite your grandiose claims, you couldn't even correctly identify the most compressed audio - you know, the one from which all the cymbals, highs and harmonics are missing. That's weird, because you supposedly can hear all that missing stuff when listening to MP3s in your car with no uncompressed source audio for direct comparison anywhere in sight.

Let's face it, even 192 kbit VBR compressed MP3s have become so good with modern encoders that even a trained ear like yours can rarely hear discernible differences (at least not with Hard Rock and Metal music). Even if very slight differences can be heard here and there (when listening very intensely and closely on high-quality audio equipment), it rarely affects the audio quality negatively in the sense that there's stuff "missing" like you claim.

 

[_xxx_]

Actually no, the original question was "does lossy compression make for bad audio?" and the answer has been an emphatic NO as proven by the two people who said they could unequivocally tell the difference: 2008 and you.
...
You made the same claim, and you both failed miserably.

Only that "bad" for me means "not true to the original". And of course the often given artefacts desribed in the thread, which in the case of L233 samples were not audiable.

But frankly, can't you tell any difference between those samples? Regardless of those being for the better or worse.


L233: the compression makes mp3 sound bad sometimes, depending on the source material and the compression algorithm used. I give you that I didn't state that clearly, thus I wanted some material where I actually know the difference since I've been listening to those records for many years and know pretty much every single note on them by heart.

- you claimed that you can hear missing cymbals and guitar harmonics when listening to MP3s in your car (presumably while driving)
- you claimed that there's a notable loss of "highs and higher order harmonics"
- you claimed that one can "literally hear some stuff missing" even with 320 kbit MP3s
- you claimed that "pretty much any musician will tell you the difference between any mp3 and the original with ease"

And that still stands for the material I'm referring to. Which I own on CD as well as mp3 CD for the car usage. I did the A/B test in the car and surely that's the case. BTW the car stereo also works when the motor is not running

Let's face it, even 192 kbit VBR compressed MP3s have become so good with modern encoders that even a trained ear like yours can rarely hear discernible differences

Well here I'll have to give in, obviously. I'm not ashamed to admit a failure, but then again I've been using the same old SW for conversion since 3-4 years, so I guess I just missed later developments.

 

[Scott_Arm]

I somewhat agree with _xxx_. If you had a completely uncompressed sample of the original to listen to first, and then you had all of the evaluation samples, you might get a different result. You have a reference point to work with.

But this does defeat the point that 128/192 are so horrendously bad that they're immediately identifiable. The audio compression is obviously very close to the original, even if not 100% accurate.

I still stand by my point that putting money into sources is a waste and the best place to put your money is speakers. You just need a basic, well-designed amp, rudimentary interconnects and wiring, and kick ass speakers.

I'm looking at building either the Orion+ or the Pluto-2, if I can get the money gathered (and after having done some more research). http://www.linkwitzlab.com/

 

[Albuquerque]

I somewhat agree with _xxx_. If you had a completely uncompressed sample of the original to listen to first, and then you had all of the evaluation samples, you might get a different result. You have a reference point to work with.

Generally agreed, but when do any of us get to listen to a pure original sample? None of us do, unless we are somehow allowed into the recording session. CD's aren't a pure original sample, uncompressed samples are not pure original samples, etc.

The discussion in this thread, again, was "Does lossy compression make for bad audio?" It makes for different audio, but it's now been shown multiple times to be anything but bad audio.

I still stand by my point that putting money into sources is a waste and the best place to put your money is speakers. You just need a basic, well-designed amp, rudimentary interconnects and wiring, and kick ass speakers.

I'm looking at building either the Orion+ or the Pluto-2, if I can get the money gathered (and after having done some more research). http://www.linkwitzlab.com/

Yes indeed, and I have NO WAY to ever spend that much money on an audio system Every audio system I own is sub-optimal by several orders of magnitude compared to the equipment you linked, and you know what? Sometimes, ignorance truly is bliss

 

[N00b]

If you had a completely uncompressed sample of the original to listen to first, and then you had all of the evaluation samples, you might get a different result. You have a reference point to work with.

I know that sounds kinda logical but it is much harder than you think, next to impossible. Our ears and brain are simply not good enough for that (we really have bad audio memory). The only way to do with some certainty it is A/B comparisson. And even with A/B comparisson when asked for the better sounding sample people tend to choose the compressed sample in double blind tests, claming it has superior audio quality.

 

[L233]

But frankly, can't you tell any difference between those samples? Regardless of those being for the better or worse.

I can't, at least not to a degree of certainty that would exclude the placebo effect. I'm not a musician but I've been listening to that kind of music for almost 20 years now. Listening to the samples, I simply can't pinpoint any differences. I sometimes "feel" like they sound differently but that's so unspecific that it's most likely just my imagination, i.e. the placebo effect.

What surprised even me (and I've long believed that the "crappy MP3 quality" case is hugely exaggerated) is how good the 128 kbit CBR files sound. I hear barely any of the distorions and artifacts that used to be so common.

I have some old 128 kbit MP3s on my HDD and most of them sound utterly horrible. Pre-Echos, all sorts of distortions and artifacts, cymbals sound like farts and so on. Completely unacceptable.

I know what compression artifacts sound like but the output that LAME 3.98.2 produces even at 128 kbit/s with Rock music is surprisingly good. I'm sure I could easily identify it in an ABX test when juxtaposed with the uncompressed source but still, I wouldn't call the 128 kbit one "bad audio" by any stretch of imagination. It sounds more than acceptable for mobile use but I'll stick to 192 kbit VBR just to be safe (there are probably pathological cases in which 128 kbit will fail horribly).

When it comes to higher bitrate MP3 (and most certainly 320 kbit), I think virtually all people will have problems discerning even a difference between the compressed audio and the uncompressed audio in an ABX test where they can always switch to the source. The c't test (linked elsewhere in this thread) demonstrated that.

L233: the compression makes mp3 sound bad sometimes, depending on the source material and the compression algorithm used. I give you that I didn't state that clearly, thus I wanted some material where I actually know the difference since I've been listening to those records for many years and know pretty much every single note on them by heart.

I tried to provide material that is as close as possible to the stuff you mentioned, since I don't own any of the tracks you specifically mentioned (I'm a bit of a Nevermore hater, I never quite understood what all the fuss is about. Their debut album is the only one I can stand.).

I think what has been demonstrated here is that it's pretty damn hard to even identify heavily compressed audio nowadays by simply trying to hear artifacts and missing stuff.

And that still stands for the material I'm referring to. Which I own on CD as well as mp3 CD for the car usage. I did the A/B test in the car and surely that's the case.

It wasn't a blind test and that's really relevant. If you know which one is supposed to sound differently (i.e. "miss" stuff) then that's what you'll hear. Your preconceptions will always affect the subjective result and that's why such tests have to be blind.

There's a legion of audiophiles out there who claim to hear the difference between €2000 per meter speaker cables and lamp cord from the home improvement store. They can always "hear" the difference in self-conducted, non-blind A/B-tests but, oddly enough, they always fail in blind ABX-tests.

The placebo effect is extremely powerful and if the test isn't blind the result is worthless. It's very possible that you never really heard any differences on your car stereo, you just imagined you did.

 

[Scott_Arm]

Generally agreed, but when do any of us get to listen to a pure original sample? None of us do, unless we are somehow allowed into the recording session. CD's aren't a pure original sample, uncompressed samples are not pure original samples, etc.

The discussion in this thread, again, was "Does lossy compression make for bad audio?" It makes for different audio, but it's now been shown multiple times to be anything but bad audio.

Yes indeed, and I have NO WAY to ever spend that much money on an audio system Every audio system I own is sub-optimal by several orders of magnitude compared to the equipment you linked, and you know what? Sometimes, ignorance truly is bliss

I agree with everything you've written. I just meant that if you're trying to compare whether compressing audio will degrade from the source (CD -> MP3), having the reference CD to listen to first, and then see if you can match it to audio samples which include mp3, and uncompressed might help the listener pick the more "accurate" sample.

I know that sounds kinda logical but it is much harder than you think, next to impossible. Our ears and brain are simply not good enough for that (we really have bad audio memory). The only way to do with some certainty it is A/B comparisson. And even with A/B comparisson when asked for the better sounding sample people tend to choose the compressed sample in double blind tests, claming it has superior audio quality.

I agree. I said might, because I think there's a chance some people might be able to do it, but for the vast vast majority of cases, I think what you've said is true.

 

[_xxx_]

It wasn't a blind test and that's really relevant. If you know which one is supposed to sound differently (i.e. "miss" stuff) then that's what you'll hear. Your preconceptions will always affect the subjective result and that's why such tests have to be blind.

As I realize now, the problem was due to the crappy compression preduced by my not-so-new software. I've been using SoundForge (this version is about 4 years old) or the Creative SW delivered with my Zen. And both produce way worse results than what you delivered here. There IS stuff missing for sure, but now I know why.

Oh well, at least I learned something new. The wows of not being a geek anymore, I missed some developments

 

[L233]

As I realize now, the problem was due to the crappy compression preduced by my not-so-new software. I've been using SoundForge (this version is about 4 years old) or the Creative SW delivered with my Zen. And both produce way worse results than what you delivered here. There IS stuff missing for sure, but now I know why.

LAME hugely improved over the years and the version numbers can be slightly misleading. For example, between 3.90 (released in 2001) and 3.97 (released in 2006) there has been a huge increase in encoding speed and compression artifacts have been reduced to a point where there are barely any left, at least not at the bitrate and with the type of music I encode. And going back even further, the improvements that have been made over the years are nothing short of staggering, especially at lower bitrates.

Using "-v 2" with 3.97 or later produces a 170-210 kbit/s VBR (192 kbit/s on average) that is IMO utterly transparent except in maybe a few extreme cases (and I've never encountered one). Hell, even "-v 3" (~175 kbit/s vbr) ist probably completely transparent to all but the most trained ears listening to "problematic" music.

It has been a (very long) while since I've been lurking at audio forums but I remember people ABX'ing their asses off and most people can't even hear much of a difference between 192 VBR and 320 CBR, or at least the statistical difference in the ABX test results is so tiny that it hardly justifies the increased file size (~60% larger).

Anyway, if you would like to put yourself to tests, here's a little free ABX program (doesn't even require installation):
http://www.kikeg.arrakis.es/winabx/

It only takes WAVs and you can do ABX or ABXY tests. You input 2 source files, and they're randomly assigned to A or B. X is set to either A or B. You have to match either A and B to X. Repeat it a couple of times to achieve statistical significance (the program does the counting) and see how you'll do with the WAVs I provided.

You can even do an AXY test (you know what A is and you have to match Y or X to A), which is the kind of tests I think you talked about earlier (i.e. have the uncompressed original as reference).

Start with 128 CBR vs the uncompressed - it's harder than you think (I did very well with the Shadow Gallery sample, though). Here are the solutions for all samples:

Amon Amarth:
pcm: yjkdh
320: iuzst
192: zxlpw
128: wtsnc

Destruction:
pcm: twhgv
320: xlkgw
192: ygtha
128: zwqsf

Shadow Gallery:
pcm: vtrzx
320: xlkgw
192: ygtha
128: zwqsf

Nevermore:
pcm: wrtkn
320: ybqlf
192: xcipw
128: zdqnr

 

[Zaphod]

I used foobar (edit: as in that it has a flexible built in ABX comparator) when deciding on my own tolerances for compression[*]. Nice and easy.

[*]Which incidentally turned out to be 96 kbit/s OGG Vorbis (aoTuV) for portable use. I could reliably ABX some of the music I ordinarily listen to at that rate, but I had to concentrate to the point of getting a headache.

 

[Bambers]

<results>

Hmm I had xcipw as the 128, rest I couldn't tell apart.

the Destruction sample I have twhgv as the 128, again rest I can't pick between.

Overall for both those samples though they essentially all sound the same

The shadow gallery one:

ygtha is the 128, bit more confident on this one than the other samples

and on a much less confident note and some (lots ) guessing I'd put zwqsf as the 192, xlkgw as the 320 and vtrzx as the original

Now that's quite interesting in a way.

My methodology in the destruction and nevermore tests was basically to pick out the one that sounded the most different from the other 3 to my ears and assume that must be the 128kbps. Apparently that assumption is balls , and anyway picking out a difference is one thing, deciding which is 'best' is another, not to mention the differences anyway were minute and I was questioning myself if I was really hearing a difference or not.

The shadow gallery track was certainly alot easier and there is a difference between the 128 and the rest though it was still much more faithful to the original than the earlier piano 128 sample, better encoding I would assume.

 

[2008 IQ is unacceptable]

The only other version of music I had heard that I listen to, other than uncompressed, or lossless, was on myspace, so I had thought that was representative of all lossy compression. I've never ripped audio in anything "lossy". Just wma lossless and wav.

I was just told that myspace was in 22khz at [h]ard forum, so now I know why i could hear such a huge difference in quality between that and lossless/uncompressed.

Of course, i'm sure there are times when bad lossy methods are used, which i believe to be the case in dmc4, but many games sound great that used lossy compression, such as max payne 1 and 2.

Regarding textures, I think that, the s3tc format is decent, but the dreamcast uses a better method, b/c i prefer blurriness rather than blotchiness/tiling/banding. So I'd really prefer a method like the dreamcast uses to return.

I think it's ok to use lossy, but the best lossy format should be used, rather than a bad one, when lossy is used.

I now rest my case against music with lossy compression. DMC4 still could've sounded fine had it used a better lossy method, so DMC4 doesn't even need to be in lossless, imo, it just needs to use a much better lossy method than it did.

I haven't played a whole hell of a lot of games in the past few years, so recent games could very well have excellent draw distances. I've been too busy bitchin' about he flaws since late 2006, that may not still be there.