sound quality, mp3 vs cd's

[Cpt.America]

And obviously, you have never engaged in such a test. That makes your comment just noise with no substance.

And I suppose the fact that YOU can't hear a difference means that NOBODY can hear a difference? That is just as much substanceless noise as anything else. Besides, I don't need to... I can hear the difference with my eyes open. The difference is TINY, I admit.. but there is a difference. I have very sensitive hearing... I can actually hear the super high pitch frequency that comes out of an older CRT television, from OUTSIDE of the house its in. I have yet to meet anybody that can also hear that. Does that mean its inaudible?

 

[fmw]

And I suppose the fact that YOU can't hear a difference means that NOBODY can hear a difference? That is just as much substanceless noise as anything else. Besides, I don't need to... I can hear the difference with my eyes open. The difference is TINY, I admit.. but there is a difference. I have very sensitive hearing... I can actually hear the super high pitch frequency that comes out of an older CRT television, from OUTSIDE of the house its in. I have yet to meet anybody that can also hear that. Does that mean its inaudible?

No, actually no human being has ever been able to distinguish between a 320 mp3 and a wav file in a bias controlled listening test. It isn't my hearing that is an issue. It is human hearing in general.

Stop for a minute and think about what you are saying. You say you can hear the difference in a bias controlled listening but you have never conducted the test to know if that is true or not. If you can distinguish a 320 mp3 from a wav file in a properly conducted bias controlled listening test, then you would be the only person on the planet that can do so. Conducting such a test is very easy and not particularly time consuming for this particular comparison. I can provide the details of what you need to do if you need them. Get someone to help you with the test, conduct it and then explain the methodology and the scores you acheived.

Why continue to tell us you can do the impossible without ever even having tried to do it? Not much credibility in that. I'm not questioning what you hear. I'm trying to explain to you that you probably don't understand fully why you hear it. What you hear does not arise from the differences in the music files. It arises from your own notions and biases. I can prove it to you but it won't be cheap. You can prove it to yourself for free. Go get the job done and report back with the scores that show you can outhear everyone else on the planet.

 

[fmw]

I think the major difference between the sound of recordings made in the vinyl era and those made in the digital era have everything to do with recording styles and little to do with the medium itself. There were some analog recordings made in the 1950's that would make you cry with the realism and presence they provide. There wasn't as much signal processing in those days simply because signal processing was more difficult. Microphones in those days were as good as the mics we have today and the tape recorders were equally competent.

There is no question that digital provides a lower noise floor and wider dynamic range with the same fidelity as vinyl. Modern pop recordings make little use of that dynamic range, as you know, but the lower noise floor is certainly there. Modern recordings also involve quite a lot of processing because using it in the digital domain is really easy.

If you make a recording either to tape or a hard drive and then master it similarly to CD and vinyl, you should get about the same sound with the CD having less noise and more dynamic range assuming the engineer allowed it.

 

[Cpt.America]

Edited for stooping to your level.


Ok... PM me all the details on the test, and I will do it.

 

[fmw]

Hardly stooping. It should be a revelation to you. No need for a PM. I'll put it here so anyone who wants to do it can do so.

What we want to do is to eliminate bias, placebo effect - all those things you read about on the forums. In other words we want to go back and forth between two musical samples in a random manner in such a way that the listener has no idea which sample they are hearing at the moment. That's how we eliminate bias.

First pick a track from CD - any track, any CD and rip it twice to separate files. Make one a WAV file which will duplicate the original and make a 320 kbps mp3 which will duplicate it but remove most of the inaudible data resulting in a file less than half the size of the WAV. You can use Windows Media Player, if you like. It is good enough for the purpose. If want to do the test with the computer as the playback source then name one file A and one file B and put them together in the same directory. If you want to use something else for playback then make a CD-r of each and mark them A and B. It doesn't matter which is A and which is B. We aren't worried about identifying which is WAV and which is MP3. We're just looking for an audible difference.

Find someone to help you and have that person work out a list of 20 A's and B's that are randomly selected. You can use a pair of dice for the purpose or flip a coin, as an example. The listener - you - can't see the list.

Let's assume you are using something other than the computer as a playback device. Put yourself in the listening position and have the assistant play 30 seconds or so of the track following the a's and b's on the random list. They must do the same things each time. In other words, if the list calls for playing A twice in a row, the assistant still needs to remove the A disc and reinsert it for the next iteration. Obviously we want it so that you, the listener can't tell what the assistant is doing so find a way to shield yourself from the process.

To start the test, the assistant should play the A sample and identify it as so and then play the B sample and id it as well. Then the assistant should play the same segment from the samples according to the random list ask to state whether you heard A or B for each iteration of the test. He or she should then score your answer right or wrong.

At the end of the test you can tally the rights and wrongs. If you had 10 right and 10 wrong, as an example, the test would be purely random. You would do the same thing by guessing. No audible difference. If you score 20/0 then the samples clearly show an audible difference. There are some rules about statistics and the experts will tell you that 10/10, 11/9, 12/8 are the same thing as 10/10. 13/7 to 15/5 would be a subtle audible difference and anything between there and 20/0 would be an indisputable audible difference. Obviously we are simply trying to eliminate successful guessing in this manner.

If you use the computer, then the easiest thing is to blindfold yourself and listen either through speakers or headphones and have the the assistant make the sample changes with the mouse.

If you stick to 30 second samples, you can complete the test in about 1/2 an hour to 45 minutes.

Just understand that the slightest opportunity for you to know which sample is playing at a given time ruins the test so be careful about how you set that up. That's all there is to it. I look forward to reading your report.

 

[TLS Guy]

Hardly stooping. It should be a revelation to you. No need for a PM. I'll put it here so anyone who wants to do it can do so.

What we want to do is to eliminate bias, placebo effect - all those things you read about on the forums. In other words we want to go back and forth between two musical samples in a random manner in such a way that the listener has no idea which sample they are hearing at the moment. That's how we eliminate bias.

First pick a track from CD - any track, any CD and rip it twice to separate files. Make one a WAV file which will duplicate the original and make a 320 kbps mp3 which will duplicate it but remove most of the inaudible data resulting in a file less than half the size of the WAV. You can use Windows Media Player, if you like. It is good enough for the purpose. If want to do the test with the computer as the playback source then name one file A and one file B and put them together in the same directory. If you want to use something else for playback then make a CD-r of each and mark them A and B. It doesn't matter which is A and which is B. We aren't worried about identifying which is WAV and which is MP3. We're just looking for an audible difference.

Find someone to help you and have that person work out a list of 20 A's and B's that are randomly selected. You can use a pair of dice for the purpose or flip a coin, as an example. The listener - you - can't see the list.

Let's assume you are using something other than the computer as a playback device. Put yourself in the listening position and have the assistant play 30 seconds or so of the track following the a's and b's on the random list. They must do the same things each time. In other words, if the list calls for playing A twice in a row, the assistant still needs to remove the A disc and reinsert it for the next iteration. Obviously we want it so that you, the listener can't tell what the assistant is doing so find a way to shield yourself from the process.

To start the test, the assistant should play the A sample and identify it as so and then play the B sample and id it as well. Then the assistant should play the same segment from the samples according to the random list ask to state whether you heard A or B for each iteration of the test. He or she should then score your answer right or wrong.

At the end of the test you can tally the rights and wrongs. If you had 10 right and 10 wrong, as an example, the test would be purely random. You would do the same thing by guessing. No audible difference. If you score 20/0 then the samples clearly show an audible difference. There are some rules about statistics and the experts will tell you that 10/10, 11/9, 12/8 are the same thing as 10/10. 13/7 to 15/5 would be a subtle audible difference and anything between there and 20/0 would be an indisputable audible difference. Obviously we are simply trying to eliminate successful guessing in this manner.

If you use the computer, then the easiest thing is to blindfold yourself and listen either through speakers or headphones and have the the assistant make the sample changes with the mouse.

If you stick to 30 second samples, you can complete the test in about 1/2 an hour to 45 minutes.

Just understand that the slightest opportunity for you to know which sample is playing at a given time ruins the test so be careful about how you set that up. That's all there is to it. I look forward to reading your report.

I have already done it. You have to pick your music. I did this with my son and musicians, with the research recordings I did on a number of rare and expensive violins and recordings of solo trumpet. The compressed files were easy to identify. My sons complex mathematical analysis of the wave forms showed why.

What is really revealing though is English Cathedral boys choristers and the huge romantic cathedral organs. That puts lossy codecs into total disarray.

The bass is drastically altered. The sense of space collapses and there is gross twinking on the treble descant lines. It never fails. It is not even subtle.

I grant on a lot of program it is impossible to tell. I have never been able to tell on any pop music selections, nor any one else I know.

Opera is unlistenable in a lossy coedec, and it is not just me who has noticed that. Pretty much anybody who has a collection of DVD operas knows you have to select the two channel loosless PCM option to enjoy the opera.

This is nonsense that lossy codecs are acceptable for all purposes. For critical applications and listening to classical productions on high resolution speakers, no loosy codec I have tried gets in the ball park. Long time reviewers for Gramophone magazine, who are not technocrats, but largely musicologists, have noted the same thing, since the inception of lossy codecs. Lossy codecs are for speech, pop and rock, and no one else period. Unless you want to listen on an ipod, and who wants to listen to that music on an ipod with earbuds!

Lossy codecs were developed by and for the pop industry, and for transmission and broadcast. The road to ruin started with the Philips mini disc player.

 

[fmw]

Compressed to what level and how? Describe your methodology and scores for the test.

 

[OttoMatic]

Yawn..........

 

[WmAx]

No, actually no human being has ever been able to distinguish between a 320 mp3 and a wav file in a bias controlled listening test. It isn't my hearing that is an issue. It is human hearing in general.

That is not true. There are codec 'buster' passages some music, primarily in classical with solos of specific instruments such as harpsichord, constanets, etc., with isolated examples of high frequency transient information without much else surrounding the incidental transient or masking it. In addition, guruboolez from hydrogenaudio.org, one of the primary codec testing coordinators, has trained his hearing over years to identify artifacts at a far higher incidence level as compared to the average trained listener, based on his ABX scores. Though, this hypersensitiveness is certainly not an ordinary situation.

The best statement one can make about a high quality encoder like Lame, is that with the right encoding setting and bitrate, it will be overwhelmingly transparent to most people on most music most of the time.

-Chris

 

[OttoMatic]

No, actually no human being has ever been able to distinguish between a 320 mp3 and a wav file in a bias controlled listening test

That is not true.

So, fmw is wrong again. Big surprise. I don't mean to be a jerk on calling you out on this, but this is at least the third or fourth time that I can think of off the top of my head that you've been factually wrong while stating those "facts" as if they were absolute truth. You berate people as if they are idiots when you are the one that's factually incorrect and doesn't know what he's talking about. Before you post your next "truth", please take some time to make sure the info is correct.

 

[zhimbo]

Before you post your next "truth", please take some time to make sure the info is correct.

Actually, no one on this thread, with the borderline exception of WmAx, has really put forth proper data regarding this issue, unless I'm forgetting something upthread.

If TLS Guy answers fmw's question, that would help present at least one first-person, properly documented example.

It rather frustrates me that the information so commonly bandied about in this type of discussion is seldom-to-never given a citation. I know this forum isn't a formal debate forum, or a scientific conference, but when threads like this get down to this level of detail, it would really help if people treated it so.

Meaning: if you make a specific factual claim, try to link or cite in print the original source.

I find it frustratingly difficult to find the actual data from well-described, well-controlled studies. Hydrogenaudio, although not exactly up to the standards of peer-reviewed science journals, seems to be the most productive site for this kind of thing (i.e.; LINK - see - there's a link!)

The best I can summarize of my current opinion is that I've seen no-one who can tell the difference between 320kps mpeg and WAV perfectly, so any claims of "night and day" differences are pretty suspect, but on at least certain recordings, at least some people can consistently reliably score better than chance (more or less what WmAx said, I think). Also, people who listen to lots of castanet music should probably [strike]get a life[/strike] use only lossless codecs.

 

[OttoMatic]

Yes, Zhimbo, you're right. There is little evidence proffered (i.e., links, citations) for the big, bold statements that are made.

WmAx is an exception and posts links regularly. He also's proven himself well enought to me for me to respect his opionion where there might be no link.

Others, though, seem to jump on the Audioholics bandwagon (no difference in sound, science can measure everything) when it seems like they're probably right, with arms waving and lights flashing.

I just like people to have the right information, and not be misled.

I think that the point at AH is frequently to help people from being led astray by thoughts such as cables having a "sound" or little stones "opening up" your CD player, we should, at the same time, not lead people astray in the other direction. That is, "we" need to maintain nuance and truth in our discussions, without making things black and white to the point where the argument becomes false. That's my problem with these types of factually incorrect, over the top, in your face posts about what's "right" and "wrong".

Just my $0.02 on the approaches on this forum.

 

[adwilk]

I'm actually with FMW... I really dont care what the programme is... it is what it is... anybody that thinks that they can distinguish between 320 and lossless should do the test and post up honest results... until then, I call BS. FWIW, a spectrum analyzer cant tell the difference...

 

[OttoMatic]

I'm actually with FMW... I really dont care what the programme is... it is what it is... anybody that thinks that they can distinguish between 320 and lossless should do the test and post up honest results... until then, I call BS. FWIW, a spectrum analyzer cant tell the difference...

Did you not read the post by WmAx four or so above? He cites references. Of course, you can believe whatever you want...

 

[fmw]

My comments don't come from other people's tests. They come from my own tests. There are all kinds of citations on the internet about the sonic effects of compression but, frankly, I don't really care about doing someone else's homework to track them down. I don't necessarily assign much credibility to what I read on the internet anyway.

I understand TLS's concerns but he has a tendency to put all compressed files into the same category and they aren't all the same. 128 kbps mp3's as an example are quite audibly different from WAV's. I've scored 100% identifying them in a bias controlled test. 320 kbps mp3's are a different matter entirely. Nobody sells 320 kbps mp3 files. You need to make them yourself. The ones you buy and download on line, as far as I know, are all audibly different from wav files. I'm talking only about 320 kbps files when I say you can't hear an audible difference between them and wav files. I'm very specific about that.

You don't need links. You just need someone to help you and you can do the silly test in about 1/2 an hour. Then you'll know. It isn't rocket science. It isn't expensive or time consuming. Why argue about who is right or wrong? Just do the test for yourself. Then you can come back and speak from experience instead of arguing credibility.

 

[zhimbo]

FWIW, a spectrum analyzer cant tell the difference...

See, this would be a great place for a link or citation. Just sayin'.

 

[fmw]

So, fmw is wrong again. Big surprise. I don't mean to be a jerk on calling you out on this, but this is at least the third or fourth time that I can think of off the top of my head that you've been factually wrong while stating those "facts" as if they were absolute truth. You berate people as if they are idiots when you are the one that's factually incorrect and doesn't know what he's talking about. Before you post your next "truth", please take some time to make sure the info is correct.

No problem, Otto. I'm really sick and tired of posting truths. If you would like to wager enough to make it worth my while, I'll come to your location and prove to you that you can't hear the difference. I'll conduct the test I outlined above with your music and your system.

Easier yet, Go do the test like I have. Don't tell me about people trained to to hear differences in codecs listening to things that aren't music. It is meaningless. Go do the tests. Then bring back your results and then you can insult me with a little more going for you. How's that? WHAT IS THE FEAR OF THESE TESTS? It is exasperating.

 

[zhimbo]

but, frankly, I don't really care about doing someone else's homework to track them down.

Actually, if YOU put forward a factual claim, it's YOUR homework. Your tests are one thing (and thanks for the details). The claim that nobody has ever detected a difference is not supported by your individual tests, however.

As I said, this isn't a formal debate board, with formal rules about this sort of thing, but, just sayin;: if you make a factual claim, it's a more productive conversation if you SHOW it to be a factual claim.

You just need someone to help you and you can do the silly test in about 1/2 an hour. Then you'll know. It isn't rocket science. It isn't expensive or time consuming. Why argue about who is right or wrong? Just do the test for yourself. Then you can come back and speak from experience instead of arguing credibility.

I have done informal versions of these tests and fully believe that, for now, 320 is good enough for me. Non-blind testing revealed no differences to me at 320. Doing the tests yourself is good advice, and when I get a better overall setup I'll probably do a more formal test on myself.

Saying that no one can hear a difference is a far stronger claim. Likewise, saying that it's easy for (nearly) anyone to tell the difference on X kind of music is also a strong claim, also not supported by mere individual testing. Or any of several other claims made by various people people on this thread.

 

[adwilk]

See, this would be a great place for a link or citation. Just sayin'.

Using voiceprint, I ripped a staind acoustic song in WAV, converted to MP3, played both and the graphs are virtually identical. I say virtually, there was one instance of the mp3 graphing roughly 1db lower at a given frequency. Hey, if thats audible, its audible. As far as this program outputting data that would correlate to actual sound reproduction is probably a bit over my head, but it tells me that I can use the mp3 format with NO worry of deficiency.

 

[adwilk]

Actually, it seems that the mp3 does show varying "weakness" above 16khz... which actually makes sense as this is where the bandwidth slims a bit for file size conservation. hmmmm... keep in mind the differences are very subtle. I'm very tempted to post some screen images.... hmmmm.... Its still hard to imagine definitive blind testing results in favor of wav superiority but its certainly plausible i suppose.... hmmmm...