Originally Posted by alan0354
I disagree from MY OWN finding that matching to 0.1dB is necessary, maybe to 1dB ( about 10%) is good enough because anything more, it will be obvious one is louder than the other.
Level matching to a mere 1dB precision is woefully
inadequate. Here's my own finding, done moments ago, in a double blind test showing that a level difference of less than half that, .424 dB, is easily
discernible to me in 15 out of 16 trials, hence with excellent statistical significance [p-value: 0.0003 (0.03%)]:
foo_abx 2.0.6d report
File A: 1 second of pink noise amplified .424 dB by MZ.wav
File B: 1 second of pink noise by MZ.wav
DS : OUT 1-2 (2- BEHRINGER UMC 204HD 192k)
11:04:09 : Test started.
11:04:29 : 01/01
11:04:45 : 01/02
11:04:59 : 02/03
11:05:25 : 03/04
11:05:44 : 04/05
11:06:11 : 05/06
11:06:31 : 06/07
11:06:54 : 07/08
11:07:37 : 08/09
11:07:52 : 09/10
11:08:16 : 10/11
11:08:36 : 11/12
11:08:49 : 12/13
11:09:11 : 13/14
11:09:57 : 14/15
11:11:16 : 15/16
11:11:16 : Test finished.
p-value: 0.0003 (0.03%)
-- signature --
So to summarize:
.424 dB I can clearly hear, as seen in today's off the cuff, impromptu test I did over coffee (with my annoying refrigerator compressor making tons of noise in the background, I might add)
.3 dB starts to become sketchy but is still way too close to detectable, so not good enough in my book
.2 dB is usually good enough for me if
betting small amounts of money. . . but heck, if you have the means to do .2dB why not do .1dB?
.1dB is what I need to see if I have a lot of money on the line.
UPDATE: I'm attempting to attach the test tracks should anyone else wish to ABX them.
Please stand by.
Here they are in a folder which needs to be de-zipped to use: