I played around with Cool Edit a little bit.
I ripped a song with EAC to WAV and encode it to MPC Std, Xtr, and Ins (and MP3 APS)
Then I decode all the files back to WAV-File. Now I load the original WAV with CE and "Mix-Paste" (both channels inverted, Overlap Mix) with a decoded file. The result is a quiet wave sounds like... hmmm... can't explain in english ;o)
The higher the quality of the mpc file, the quieter was the difference. The MP3 APS wave had roughly the same loudness as the MPC Xtreme wave.
Am I right that this "difference wave" is the "lossy" part of the original wave after encoding?
You are right, what you are hearing is actually all of the data that got thrown away during the encoding process.
Keep in mind that what you hear, and how loud it is has nothing to do with the actual percieved quality. You want the encoder to throw away all of that unneccessary sound, you can't hear anyway. The more data is thrown away, without you hearing a difference, the more efficient the encoder.
Just out of curioisity, what are the exact steps after loading a decoded WAV and the original WAV that one must do in Cool Edit Pro to try this? I am not trying to decide which format and bitrate to use based on it(I'm stuck at -q 6 Ogg Vorbis for now, since it's transparent to me), yet it seems like quite an interesting experiment, and I don't know my way around Cool Edit Pro that well. I'm still experimenting.
As MiskyWhixer said, it's simply a matter of opening both files in CoolEdit, select all and copy one of them to the clipboard, switch to the other, then click Edit | Mix Paste and make sure it's coming from the clipboard, not looping, "Overlap (mix)", no crossfade, and the volume is set to 100 for both channels and the Invert checkbox is checked for both channels. Once you click OK you'll end up with the difference.
But don't base your codec decision on it! It is possible that a quieter difference will be more noticable than a loader difference when added back to the audio.
You might want to check this thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=11286&) where this issue was previously discussed.
I tend to believe the less volume the difference file has the better the encoder, assuming both encodes are already transparent and have the same bitrate, because it approximates the original wave closer.
Microsoft followed the same line of argument with WMA7. They observed the difference between WMA vs. original and MP3 vs. original and found that WMA sounded better because the difference to the original was smaller.
Unfortunately WMA7 sounds worse than MP3 in most cases.
A codec which doesn't throw away the inaudible stuff, which leads to an increase in difference, can be considered inefficient. You want a psychoaccoustic codec to throw away as much as possible while still sounding the same.
edit: Thasp, please also be aware that some codecs introduce a small delay when you encode and then decode again so you have to align the waves carefully before subtracting them from another. You can use Cool Edit Pro's multi track features to accomplish this. There is also another way of inverting one of the wavs without using the clipboard: goto effects>amplitude>channel mixer and set it up like this:
New Left Channel: L 100 / R 0
New Right Channel: L 0 / R 100
and tick "invert" on both instances.
A codec which doesn't throw away the inaudible stuff, which leads to an increase in difference, can be considered inefficient. You want a psychoaccoustic codec to throw away as much as possible while still sounding the same.
Yes but when you say "as much as possible" it doesn't mean more information is discarded when the difference file volume is high. It is much more complicated than that from an information theoretical perspective. You want to discard as much inaudible information as possible so that coding with less number of bits will be realized.