Hi... just wanted to know what information do the extra bits have .... actually how do they sound... roughly if the original wav sounds like X and its lossy compression to say iTunes AAC at 129 kbps is Y. how does X-Y "sound" like, if the question makes any sense... i remember seeing something like this for JPEG... hence wondering....

Why don't you experiment yourself? Make a few different AAC files (from an uncompressed WAV) at different bitrates... Maybe from 62kbps to 244kbps. You should be able to tell the difference between the 64kbps file and the original WAV file, but the 129kbps file may sound just as good as the WAV.

It's hard to describe the compression-distortion & artifacts with words. (And I don't have that much experience with it, because I try to use high bitrates.) All of these compression methods use "smart" compression... They try to throw-away the "least important" data. So, the "sound" of the compression depends on the composition/complexity of the original sound. I think you'll first notice that horns & cymbals start to sound "harsh", and at very-low bitrates it might start to sound like a cell phone (sort-of like when you talk into a tube).

Yes. JPEG is also "smart-lossy" compression. If a picture has a big area of blue sky, a lot of useless blue-sky-data will be thrown-away. And, more data will be used for other details in the picture. If the picture is full of detail, the quality loss will be more noticeable, because there is no useless-data to be thrown-away.

Don't expect to use this to get any sense of what's missing. The only reason you can hear any of that sound in the difference WAV (X-Y) is because the sound that masks it has been removed.

Jayant, N. S., Johnston, J. D. and Safranek, R. J., “Signal compression based on models of human perception,” Proc. IEEE, Oct. 1993, pp. 1385-1422.

May be extremely useful to explain how these things work, if you have access to it.

http://www.aes.org/sections/pnw/ppt/other/...tual_coding.ppt

Is also useful, but very, very old.

I know what 64 kbps sounds like comapared to the original wav ... cant make out the difference between wav and 128 .. hence i wanna hear what was "missed out"

I think benski already summed it up perfectly!

It looks like you are trying to measure "how much quality is lost" (which has already been discussed in several threads). This is technical possible using audio editors like Audition etc., but you will never be able to get a real measure for lossy files. But still, you could try this yourself just for fun, but I think you'll have a lot more fun starting to ABX the difference between lossless and compressed files

Then, as a practical matter, nothing has been lost!!!!

You'll have to read-up on the science if you want a scientific answer. But, there is no way to determine exactly what data/details were removed from your particular file. (You can't mathematically "subtract" the two files, because of phase-shifts.)

...If you were a computer programmer, you could probably modify the LAME encoder to make an audio file of everythng that was lost while encoding an MP3. It would sound like noise, but the "character" of the noise might be interesting.



I'll give you a crude example of the kind of detail that's lost:

Imagine that you are recording a very-loud electric guitar. If you stand in front of he speaker and whisper or speak softly, the guitar will drown-out your voice. Nobody will hear you, and you will not be able hear it on the recording. But, the whisper/voice is there, both live and on tape... Drowning-out the sound doesn't eliminate the sound, it makes it impossible for you to perceive it.

Now, if you AAC (or MP3) encode the recording, that whisper/voice will be eliminated along with a lot of other details you can't hear. Most of the removed-details will be from the guitar-sound itself. (You cannot isolate and "recover" the voice from the removed-details, because the all of the removed-details would be jumbled-together in one big noisy mess.)

okay ... instruments may sound "different" ... but lets pick up the binary case... gimme a sample where you can hear something in the wav but not in the AAC.. like the whisper you mentioned.. can i get such a sample... ? I can easily figure out that things sound different at 64 kbps and 96.. but dont have ears to test the 128 case... cant ABX out if you like at 128... so some sample where its binary.. like a whisper present in the wav but not in the 128 encode...

Samples that are especially difficult to encode are called killer samples. There are "MP3 killer samples" and "AAC killer samples" that bring-out the weaknesses in a particular format. I'm not an expert, so perhaps someone else can suggest some songs. Or, I'm sure you can find some examples if you search the forum (or the Net) for "killer sample".

My point is that you cannot hear the whisper live or on the WAV file. So, the AAC compression can throw it away and it doesn't matter because you can't hear it anyway. (Even without the whisper, there's a lot of other "things" in the acoustic sound wave that you can't hear.)

Well... That's the whole goal of (high fidelity) compression. The AAC and MP3 designers are trying to make the compression so good that you can't hear the difference between the lossy compressed file and the original.

What is your goal? I can't tell if you are interested in the science, or the sound.

Do you want to know what you are missing? If you can't hear the difference, you are not missing anything! If you are concerned that other listeners will hear the compression artifacts in your AAC files, go ahead and use a higher bitrate, go with a non-lossy format, or don't use any compression.

ANALOGY - It's like you are trying to use a microscope to look for scratches on your car. If you can't see the scratches with your eyes, you can consider the paint-job perfect. Maybe some people with extra-sharp vision, or with lots of inspection experience can see the scratches. (In the audio/audiophile world, critical listeners who can hear tiny details that most people miss are said to have "golden ears".) If you take a photo of the car, the scratches might not show-up... even with a microscope. But, the loss of detail in the photo is not a problem. It is "useless" data than can safely be thrown-away.

If you want to get a "feel" for what the scientific differences are, listen to the 64kbps sample. The "weaknesses" and "deterioration" are similar with higher bitrates, they just are not as bad. People with more critical ears might be able to hear these tiny defects... They will agree with you that the 128 sample sounds better than the 64, but they will tell you the WAV sounds better than the 128.

I don't understand what you mean by "binary" in this context.

yeah exactly.. am trying to look at scratches in my car with a micrsoscope because i cant see them with the naked eye!

I am not looking at the encoding science just the final sound...

but thanks ill seach for the killer files...

So you want to tell the difference between:

a. a box with a key inside.
b. a prop, an empty box.

You can't see through the box. What are you missing when you see the box on stage? Assuming the box is well-made, nothing.

Well here is a picture of the difference between lossless CD audio and LAME V2 MP3:



Not sure if that means anything to you, though.

if you cant convince them, confuse them!

I can get this with audacity...




i cant see through the box but someone with better eyes can! so i need i microscope to see what the other person saw!

Well if you can get that form Audacity, listen to the difference audio, that is your answer, but it isn't the whole picture. I can hear a lot in the difference file but a big part of lossy audio compression is masking.

Masking is when you have sound A, say a loud cymbal hit and sound B, a quiet triangle hit, you can't hear sound B because of sound A. So the lossy compressor discards sound B. On it's own you can hear sound B but with sound A over it you cannot.

So basically, the "extra bits" are things you cannot hear.

That other person must have been Superman.

Or there was a hidden hole in the box which can be noticed if you know where to look. But the shape of he hole would need to be familiar as well.

You don't need a microscope for that.

The audio world is full of Supermen.

As mentioned above, this is not a real repesentation of the difference between an encoded and the source file. Instead, I think it would be a more real comparison if you take the spectrum from the encoded and the source file (you can export it in Audition 3.0), load them into Photoshop, and subtract one from another. You will probably still have some small differences because of the phase shifts, but you would probably notice the higher frequencies to represent the great difference.

That's what the image in the lower-right is already showing.