Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: What is ABX'ing?! (Read 2779 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

What is ABX'ing?!

Reply #1 –
Theres a FAQ on it somewhere here, but basically its a blind test.  You feed two samples, it mixes them up and asks you which is the origonal.  You take as much time as you need and run 10 or so tests (more if needed).  At the end it tells you if you could "ABX" (distinquish from orginal) the sample.  The idea being to isolate you from the quality of the audio.

Edit:  Or rather isolate the audio from you.

What is ABX'ing?!

Reply #2 –
A=sound file A
B=sound file B
X=either A or B, chosen at random, unknown to listener

The test is to listen to A and B, then to X and identify whether X=A or X=B.  If you can do this reliably (>90 to 95% statistical confidence) than you have proven that you can tell the difference between the two sound files.

In practice you might make A=original WAV file, B=compressed file.  Using a program like WinABX, Foobar2000, etc. perform the test.  You need 5 or more trials to start becoming statistically significant.  Make sure gain levels are matched (use Replaygain, etc.) because even relatively small level differences can skew results.

The test is used to verify transparency, or the lack thereof, by removing any listener bias.

[span style='font-size:8pt;line-height:100%']>>Edit: added replaygain comment[/span].<<
Was that a 1 or a 0?

What is ABX'ing?!

Reply #3 –
Quote
Make sure gain levels are matched (use Replaygain, etc.) because even relatively small level differences can skew results.

If I may pipe in for a second to emphasize the importance of this part...

Using ReplayGain is a very important aspect of ABX test preparation.  When I started using WinABX, I thought I was keeping everything equal by not using any gain adjustments at all.  While it was true that my files (at least theoretically) were at the same level, I failed to take into account clipping.

A WAV file may not have any clipping, but it might be introduced by a lossy encoding process (MP3, MPC, Ogg, AAC, etc.) at the time of compression.  Therefore, without first adjusting gain down to eliminate clipping, you may end up with two sound files at the same level, but one with clipping and one without.  In this case, you'd possibly be able to ABX them apart by identifying the clipping alone rather than by finding any other artifacts, which fouls the integrity of your ABX test (unless you're specifically testing for your ability to detect clipping).

What I do now (and there may be other/better ways of doing it) is to use WaveGain to adjust the level of the original WAV file, then encode as desired without adjusting the gain of encoded files, then test with ABX.

WaveGain can be found at RareWares.