Looking for some expert guidance on this one as I can't figure out where I went wrong.

What I did was take 16 songs, ripped them directly from the CD hardcopy in .wav format and saved them to my HD.

Then I took each .wav and compressed them using 320kpbs LAME, except for one which I went down to 192kbs.

Then I put the two different versions of each track in consecutive order and burned them to CD using EAC. Here's an example:

Track 1: Pink Floyd - Money (.wav)
Track 2: Pink Floyd - Money (320kbps)

I gave a copy of the burned CD to 20 of my friends and asked them if they could correctly identify which track was compressed and which one was not, or to tell me that they couldn't tell.

The results I got were that absolutely no one could tell the two versions apart. I listened to them myself many times and they sounded exactly the same.

My conclusion is that EAC somehow normalized the playback to make each song sound the same. BUT, I specifically DID NOT check "Normalization" in EAC Options.

The other possibility is that the sound differences between .wav and 320kbps are beyond human senstivity.

What do you think?
Where did I go wrong with this "experiment"?

Thanks!

Well, your experiment didn't have the correct procedure. You should have downloaded foobar2000 and installed its ABX add-on. Conduct a blind ABX test between a mp3 file and lossless file (FLAC, WAV, AIFF, WavPack, etc.). Odds are pretty high that you won't be able to distinguish between the mp3 file (many people think that -V 3, which is around 170kbps VBR, produces adequate results) even when using high end headphones.

EAC didn't normalize the tracks or anything like that. The tracks on the CD form the mp3 files have the exact same quality as the source mp3 files. This means that the mp3 files did their jobs, they were transparent. Compare the mp3 files to the uncompressed WAV files and you will see that not only were they transparent for you but they take up a lot less space. This is the whole purpose behind lossy formats. Their main goal is to produce transparent files so that you can squeeze more content onto your portable player, hard drive, optical disc, etc.

So even though your test wasn't accurate and isn't recommended, it shows that you and your friends were perfectly acceptable with 192kbps mp3. You will have no need to carry around WAV files (why WAV? why not FLAC or some other lossless format?) solely use WAV files with your music library. The Lame mp3 encoder did its job, nothing went wrong.

Wrong conclusion. EAC certainly did not normalize them, if you didn't check NORMALIZE.



Not 'beyond human sensitivity', but quite possibly beyond your or your friends' current ability to distinguish them.





There's nothing startling about your findings. There's already plenty of data indicating that 320 kbps LAME MP3 is 'transparent' to most people with most music.

you did not do anything wrong!?

Your not supposed to hear a difference especially at such a high bitrate.

please read this:
http://en.wikipedia.org/wiki/MP3

You don't have to use foobar2000 to conduct a proper ABX test. Having a friend randomly change songs on a CD player can be just as valid if done correctly.

Ok, let me clear this up a bit.

My intention in doing this was to make some compilation CD's with the highest quality possible (wav). Rip with EAC, make a playlist, burn with EAC. Simple. No loss.

Then I figured I could compress to 320kbps as I felt, for me personally in my experience (so shoot me!), that this was the point of no return for quality audio and get more songs on to a single CD as the files are smaller.

Well, this is not how it works, obviously, because the burner looks at the time of the song, not the size. In other words, no matter if it's a full exact copy (wav) or a compressed version (192kbps) I'd still be limited to the number of songs and 80 minutes max. Right?

My mistake. Live and learn.

Now, the questions remaining for me, are...

1) why compress if you're going to burn to CD anyway? It's a waste of time, right?

2) Is the burner using some sort of error correction and therefore the two versions of the same song are indistinguishable on a CD?

3) Is ECA using some sort of codec in the outward stream of data to the CD that makes all the bits of data the same?

Thanks much for the feedback. I'm not an expert at all. I'd like to test human hearing and perception and not be limited to hardware or software. Get it??

i've done the CD-test myself with friends before. Usually they don't care enough to perform a proper ABX, so by making a CD, you're doing to hard work for them.

There are two different ways of putting things on a CD-R.

If you burn the disk as a normal audio CD then, by the specification, it has to have a certain number of bits for a given length, as you worked out, and you won't get more music, even if you have used file compression at some point in the procedure.

If, however, your CD player can also read MP3s, then you can put the lossy compressed files onto the CD-R as a data CD. That way, you *will* get a lot more music on the disk, but not every CD player will be able to deal with it.

(What happens when you burn an ordinary audio CD from a compressed file is that the software first expands it into the WAV format, and then burns that. You don't have all the information that was in the original WAV, but as you have discovered, a good lossy encoder knows how to throw out only the stuff that humans can't/don't hear. Compressed encoding is mostly used for playing on computers, and especially portable MP3 players, but some CD players, especially in units for use in cars, and portable units, know how to read MP3, and sometimes other compressed formats. I think a lot of DVD players can do the trick, too.)

HTH

So, basically what I did was gave my listeners the same WAV file twice in a row? (see above Pink Floyd ex).

This is the crux of the whole thing. Were the participants actually listening to compressed music or not? Sounds like the answer is a big fat "NO."

I totally understand that once you take that lossy file and play it with something that reads it vs burning it to a CD, then it's a different ball game.

What's interesting in my group is there were a few that were very confident that they could tell the two songs apart, as I had them rate their confidence level between 0-100%.

Hello FoxPro5, welcome to Hydrogenaudio.

From these two posts, it looks like you lack some knowledge about lossy codecs, CD-Audio, EAC and maybe some other audio-related topics.

You've come to a good forum to get the correct information about most of those topics.


Now, onto your "problem".


A) Audio normalization:
Audio normalization is a technique that changes the amplitude (read, volume) of the audio, so that it fits a specific amount (usually, the value 98% is used) of the available range (read, maximum volume).

This process, first of all, does not set the same loudness (perceived volume) to the different files, since amplitude is not the same as loudness.

Normalization does not mean, at all, to make a file similar to another one. It only changes the amplitude (again, this is like changing the volume knob).


B) Lossy encoding:
You were told this already: lossy encoding has the goal to reduce the bitrate (file size) needed to store an audio recording, while preserving a quality similar enough to the original one.
In other words, mp3 and other codecs try to fool your ear into thinking that what you hear is the same than the original, while it is not an exact copy.
And they do succeed.

In the main page of this site you can still see the recent listening test that has been done by members of this site evaluationg lossy encoders at around 130kbps.
This is lower than 192, or your 320kbps files.
The conclusion is that for most of the time, the encoders do a good job making the files not annoying (this translates to that they are nearing transparency, in many of these samples).


So stop trying to find phantoms or weird explanations to a natural effect, which is: lossy encoding can be transparent, and it actually is many times.


C) Audio CD's:
As you found out, an audio CD always have the same size. It doesn't matter if you use the wavs ripped from another CD, or something recorded from a chat in Messenger.
Edit: as you seem you didn't get it right: A low quality file, burned onto a CDAudio will still be a low quality file.
There is no magical way that an MP3 burned onto a CDAudio becomes the same than the original. You only get the lossy wav (this process is called decoding, and is done by all the players when playing them).
You and your friends heard the original and the lossy encoded file. (at least with the information you give us)

But as you were told, Audio CD's are not the only type of CD's that exist. If you record a CD as data, you can keep the size small (as any of the CD's or DVD's that you may have, with games, magazine programs, computer drivers, ADSL configuration software....).

Also, the place where lossy files are being used the most nowadays are in downloads from Internet, and storing in flash or hard-drive based players (not CD's). There, the file is exactly the compressed one, with the reduced size.

Ok, I got it. I was using the term incorrectly. I understand it better now. Thanks.



Not trying to find anything except answers to the observations for which I don't understand. I may not be able to speak Hydrogen Forum correct lingo, but I'm not an idiot. I know we can only hear so much; I just want to be sure that I'm actually testing human hearing - not what burning software does.



But, when burned back to back to a CD, will sound exactly the same? Just to clarify.



Understood. The file sizes on my computer clearly show that.

Thanks for your help, JAZ.

No. MP3 burned to a CD will sound as MP3, not original CD.


I doubt that.

added:

After decoding to WAV any MP3 file will have the same filesize as original WAV file, but quality of MP3 file.

Hi

No, they weren't listening to the same WAV file, they were listening to two *different* WAVs. And, in all important senses, they were listening to compressed music, it's just that the compressed music was now coded in a very inefficient way.

A CD requires a certain bit-rate. When you compress a file, you reduce the bit rate; lossy compression throws away some (hopefully unimportant) information. When you burn that file to a CD, the CD player wants the same bit rate as always, so the burner has to expand it to a WAV: BUT the information that has been thrown away just can't be recovered -- it's gone for good; so the burner expands the file with the sounds of silence ™ or some such. So what you hear is (unless something has got broken somewhere) something that sounds just the same as the MP3, though it's on the disk in an inefficient format, taking up a lot more space than it needs to.

(BTW there is a form of compression that doesn't throw away information, but just packs the file tighter, like ZIP for ordinary data files. This is called lossless compression: from a lossless file, you can reconstitute the CD exactly as it was before. You can't compress all that far losslessly--60% seems typical, and it looks like 50% is a practical limit for ordinary music. This is worth knowing for putting stuff on your puter, but isn't relevant to your present question).

The exact missing piece that I needed for my n00b brain. Thank you for explaining that to me.

When you say that a CD requires a certain bit-rate, what bit rate is that? How far from 1411 kbps can you get, actually?





Baby steps.

It's called "Psychoacoustics". Do a Google search or a buy a book and read about it sometime you will learn a lot.




It's called an ABX test. Read up about or ask more questions about it if you are going to be performing a test again. It's a very simple procedure once you get the hang of it. You shouldn't be able to ABX at that bitrate, although I have seen stranger things happen.

Yea, I'm up to speed on that.

What really propelled me into doing this was reading Dr Levitin's "This Is Your Brain On Music." Really changed how I listened to my ears.

I know what ABX comparator is and does, but just haven't used it as suggested. Didn't know FB2K had a plugin for it, so I'll check that out.

But I have taken the Klippel Listening Test enough times to understand the proper place of my ego in all of this. Turns out, my ears aren't too shabby:



Again, thanks to those who pointed out errors of my ways. I'm off to run some more scans of the whining and groaning Axl has been perfecting since '94. So far, it's pretty damn good....

AFAIK, even silence needs to be encoded at 1411 kbps on CD.

Every CD is EXACTLY 1411200 bits per second.

I have just visited your listening test website. They use MP3 for testing. I doubt that the test wasn't so accurate.

???

That depends on the bitrate, the sensitivity of the listener, what they are listening with, the transparency of the codec, and the music on the original .wav.

Pedantic comment: Absolute silence is identically the same, when coded at any bitrate. Now some coders may have figured out how to screw that up, but... ;-)