People have been researching the audibility of various distortions of audio signals since at least the 50s, we have a pretty dang good idea of what is audible and what is not when measured in controlled conditions in terms of absolute volume, harmonic distortion, intermodulation, phase shift, crosstalk, etc. But most importantly, we know beyond reasonable doubt that human senses are very sensitive to bias, and the humble $9 Apple dongle is unlikely to fare well in a sighted a/b test.