The same guys from the article and this enlightening video released a small software (available also in the debian repos) called squishybox, aimed to perform A/B/X tests on audio files; today is a rainy day, I have some audio file from the last field recording to process but I’ve also some spare time to test the tester software, so here it is.
Note that I’m not an audiophile, or at least I don’t consider myself an audiophile and my knowledge of this is (still) too limited. All I want to do is learn how to use the software, get some memorandum and listen to music.
The software can be found at this address.
After installing and reading a bit of the man page of squishyball I’ve grab a flac version (16 bit/44.1 kHz) of this song, chosen due to my pleasure and a sort of interesting dynamics into it.
I’ve converted it to OGG Vorbis
oggenc -q 2 file.flac so at the moment I
have this two files:
Boy_Lilikoi.flac: FLAC audio bitstream data, 16 bit, stereo, 44.1 kHz, 11896416 samples Boy_Lilikoi.ogg: Ogg data, Vorbis audio, stereo, 44100 Hz, ~128000 bps, created by: Xiph.Org libVorbis I
So to perform the test just run
squishyball -a Boy_Lilikoi.flac Boy_Lilikoi.ogg -B -s 60 -e 70 -n 10 -R
-ameans that the A/B test should be performed
- Filenames are filenames…:-)
-Bmeans to beep when switching from one sample to another
-s 60 -e 70means the start and end of the sample to be listen. I’ve chosen this window for it contains both a delicate voice + guitar + effects and a more “powerful” part. Default is to play the whole track until end or until you flip to the other track.
-n 10means that 10 tests will be performed (you have to decide 10 times if the track you are listening is one or another).
-Rmeans that every time you switch from A to B it restarts the sample instead of going on from the play point.
The interface is this nice curses-based gui:
While playing you simply have to decide (in simple A/B test mode), switching
b keys wich “version” sounds better. Once you have decided
B key and the software skips to the next trial.
When you have done it prints an output with the results and some statistics:
Normalizing all inputs by -4.6dB... done A/B/X test results: Correct sample identified 4/5 trials. Probability of 4 or better correct via random chance: 18.75% Testing metadata: Total time spent testing: 85.80 Total seeks: 0 Beep flip used 10 times. Undo was not used.
At the moment I cannot distinguish really anything from anything…
Another different test from A/B is the A/B/X one where the tracks specified in
the command line are compared with the “X” one. You have to identify which one
(A or B) is identical to X. Obviously X changes from time to time.
Once you listened the three tracks you press
B if you think track X is
equal to track A or B.
The command I use for this is
squishyball -ab Boy_Lilikoi.flac Boy_Lilikoi.minus1.ogg -n 3 -B -g
-g option (
here) is used to print
immediately the result (not so professional…)
Note: a simple site that allows you to perform such tests without install anything.
Note2: an important advice from the man page:
Many modern audio playback systems (such as PulseAudio or the ALSA 'default' device) give no means of determining if the requested playback paramters are actually being used by the hardware, or if the audio system is helpfully converting everything to some other supported depth/rate. When using these systems, squishyball has no way of knowing if 16-/24-bit playback or sample rate is being honored. Automatic conversion can affect audible playback quality; be careful to verify actual system behavior.