Quote:
|
Originally Posted by Sycraft
Ok, great, however what we empiricists do NOT know is if we have discovered everything that is important to human hearing.
|
Of course measurements do not reflect everything that may affect perception. After all, distortion figures are but summary statistics, and don't tell you the detailed profile of the distortion, which also matters to perception besides the magnitude of the distortion. Again, this is not at issue. A properly conducted blind test covers all known and unknown matters of perception. Denying this is denying rationality and science as a way to understand the world -- it turns into a religion.
I've said this several times at the various audio forums: by far the greatest difference between live sound (and I listen to live classical and jazz concerts extensively) and electronically reproduced one is in creating a proper soundstage, which is to some extent a geometric problem, and includes the recording arrangement's ability to record the proper directional distribution of sound, and the reproduction at the other end of such directionality in a way that parasitics are eliminated. By parasitics I mean things such as room reflections and more importantly, sound intended for one ear reaching the other ear, which is not a problem with headphones but it is a problem for speakers. Generally, stereo separation is crudely increased in the recording in order to accomodate for this crosstalk, but the results are not life-like (and worst for headphones). The proper way to deal with this is signal processing and speaker arrangements which can eliminate most of the crosstalk (there are numerous examples of related research that can be found on the web). This way binaural recordings can be listened to properly with speakers. But those have limitations too, such as different HRTFs (head related transfer functions) on the recording head or dummy head, and the listeners head/ears, which makes especially up/down and front/back positioning less effective than if the same shape head/ears were used (again, perhaps signal processing can accomodate for this if the listener's HRTF is known; imagine everyone getting it measured and feeding it into the signal processor as a custom profile
); etc. etc.