Bumping this thread and using a recent file discussed on ASR in another matter.
I used the original file and used two 2nd order allpass filter in Audacity (allpass2 s 100 1) and (allpass2 s 200 0.7). Both were tested in ABX against the original with the following results (the 100 Hz Q=1 was easy, the 200 Hz Q=0.7 was harder).
Questions: Are the files correctly made with respect to filter and is the result expected?
Shared with Dropbox
www.dropbox.com
Shared with Dropbox
www.dropbox.com
Shared with Dropbox
www.dropbox.com
I compared the frequency magnitude and phase spectra for the 100 Hz allpass. I applied a bit of tukey shading on the ends of the waveforms, and performed a single FFT zero padded to 2^20. Results are attached. I believe the magnitudes are well matched, with the differences appearing in the nulls, which are always higly variable and don't impact our perception at all. The power spectra has 2 plots, with the orig in blue and the allpass in red. You can just notice a bit of blue extending lower than the red nulls at 35 Hz and 165 Hz for example.
The phase plot shows 360 degrees of phase rotation, centered at 100 Hz.
I would say this is a valid test.
Regarding the result, I believe that's expected as well. It's been established that phase differences can be heard in this frequency band. Do the same thing at 3000 Hz instead of 100, and you'll likely fail. However, while differences are audible, I am not sure it's been established that there's a clear preference, or that listeners have a clear sense of one being "right" and the other "wrong" or "distorted". At least not at the scale of phase distortions caused by filters people typically employ.
My concern is that people point to results like this and exaggerate the impact of phase, and give undeserved credit to manufacturers who address inconsequential phase distortions in the kHz range because that's easy, while doing nothing for lower frequencies, because that's more difficult. There's plenty of marketing material claiming that linear phase speaker systems are a cut above the others, because that one attribute is the secret to audio bliss. Sometimes these are great systems that have the other more important aspects right, and the reason for their greatness is misattributed. Sometimes it's systems that have more fundamental flaws, and don't approach greatness.