I've done a careful analysis of MQA.
First off, I can prove that people can hear a difference (at least sometimes) with MQA because I myself, through my very modest system, can hear a difference with music files processed with MQA vs. identical ones which are not, but rather than just taking my word for it [which nobody should without evidence gathered under controlled conditions to preclude the possibility of expectation bias clouding the results], I can actually demonstrate this ability via the free, easy-to-use, scientifically valid, double blind, level matched audibility test, Foobar ABX
, showing strong statistical significance in my perfect 10 out of 10 score:
foo_abx 2.0.1 report
File A: Blagutten 30s MQA deblurred.flac
File B: Blagutten 30s NO deblur.flac
DS : Primary Sound Driver
12:36:20 : Test started.
12:38:46 : 01/01
12:39:09 : 02/02
12:39:22 : 03/03
12:39:37 : 04/04
12:39:54 : 05/05
12:40:08 : 06/06
12:40:35 : 07/07
12:41:01 : 08/08
12:41:14 : 09/09
12:41:31 : 10/10
12:41:31 : Test finished.
Probability that you were guessing: 0.1%
-- signature --
Bob Stuart of Meridian, one of the main MQA inventors, would have us believe the audible difference is due to "deblurring" accomplished by MQA processing however that's not the case at all. Want to know the real difference? It is a change in the music's perceived "air" and transparency due to a subtle change [Note: an increase
] in the noise floor shaping in the audible range which any CD quality "limited" music could just as easily replicate if such a goal was desired [no need for "Hi-Res" to manifest such a change] starting around 8 or 9 kHz and increasing as you move higher.
Here is a close up of the noise floor's frequency response change, originally published by Archimago, which shows this and clued me in on how and what to focus on when I did the above, double blind test, getting a perfect score. From the color legend we see blue
are the MQA processed music, with and without MQA decoding. The red and yellow have no MQA processing applied to the music:
As we can see from the measurements, MQA processing, both decoded and not decoded, actually adds considerable high frequency noise in the audible frequency range
(+12dB at 15kHz and nearly double that at 20kHz) . . . but this isn't perceived as "noise", per se
, it is perceived as "air" or "sheen".