Can you really tell the difference between encoding methods? Two different kick drum emulations? 16-bit versus 24-bit? These aren’t subjective questions: you can actually test them–with science. (Science!) Here’s a site that lets you create your own ABX tests, or try some pre-built ones. Argue about other stuff instead!
The idea is simple. Listen to A. Listen to B. Is X … A, or B? Or which is higher, louder, whatever. Jakub Charkiewicz of theaudiostuff.com writes CDM to put us all to the test:
Two of your posts land on the same question. The 2015 Tidal hi-fi piece tells readers to run their own blind test, and the 2021 Apple Music lossless post says a double blind test can settle the bit depth argument. The blind test that older post points to does not load anymore.
So I built one on my site. Six built in pairs, including 24 bit versus 8 bit decay and a 0.5dB level mismatch, or drag in two of your own files. It reports a p-value, so 12 right out of 16 gets a verdict instead of a shrug.
Well, it’s good the Internet remembered what I wrote, as I didn’t! But indeed, you can give it a try. Jakub has some great, pre-built tests to explore, plus more tests on topics like EQ. And you can make your own. Plus, you can try this out on some friends, as well.
https://theaudiostuff.com/tools/abx-test
I was curious if Jakub was surprised by any results. He tells us:
Two main things stood out:
1. The 0.5dB volume trap where almost everyone misinterprets a tiny .5dB level difference as clearer highs or a wider soundstage rather than pure loudness. It really proves how easily human bias distorts listening tests without strict gain matching.
2. Bit-depths are something most listeners can’t reliably pick out – 24-bit vs 16-bit on full musical mixes. It only becomes distinct during extreme reverb decay or quiet synth tails down at the noise floor – which is why I included the decay pair in the tool.Applied-wise, I’ve been using it to sanity check streaming codec claims (like Tidal/Apple Lossless) and test whether high-sample-rate DAW exports actually make an audible difference on final masters vs pure placebo.
That’s a terrific idea, and like Jakub, I’m also interested in whether y’all have other stuff you’d like to see in comparison testing. Let us know in comments.
For some pre-loaded music, I randomly came across this abxtest.com site, too, which runs on the same principle. I’m linking mainly because I laughed out loud at the “prove it. or shut up about flac.” heading.
