How the test is done:
First the original track is placed in a multitrack digital audio computer workstation program as shown below in red.
Next, the mp3 made from that original track is placed in a track below (yellow) and flipped upside-down (polarity inverted) so that the two tracks are mixed together "out of phase." (Note "out of phase" is the term commonly and incorrectly used to describe polarity inversion, which is equivalent to flipping the "+" and "-" or red and black wires on a speaker.)
With volume levels of the original and the mp3 tracks exactly matched, mixing the two tracks together will leave only the difference between the two red and yellow tracks -- the mp3 subtracted from the original. If the two tracks are really identical, then no sound whatever will be heard from the mix of the two tracks.
Listening to the difference green track gives us two pieces of information -- the volume of the difference, and the nature or "tone" of that difference. The difference track then, is the sound of what is done to the original signal in the process of making an mp3 out of it. One might argue that this difference is the result of subtraction—subtracting the mp3 from the WAV, but it is also ADDITIONAL sound in the form of distortion, that is added into the sonic end result, since the ear cannot distinguish between distortion that is added or subtracted.
As you will hear, the difference is a disgusting abomination, and clearly something no one in their right mind would willingly pay for with hard-earned money knowing the hidden truth in advance.
This screen shot shows how the original and the mp3 were carefully aligned--accurate to the individual digital audio sample--to remove errors due to track timing. The "FX2" box highlight reflects the use of a polarity inversion applied to the mp3 track. Polarity inversion merely flips the audio wave form upside-down so that it will cancel out totally. Two identical sound files treated this way (e.g. if we used two originals), when mixed together (to make our green result), would produce a track of total silence. The reason that the third green subtraction track is not silent is because of the differences between the original and the mp3 tracks.