Hi-Res ABX Tests Prove It…Not So Fast!

My email has been flooded by friends, audiophiles, and manufacturers asking if I’ve seen Michael Lavorgna’s recent post “ABX Test Prove Hi-Res Audio Is Legit“. As a major proponent of high-resolution audio/music, you might assume that I’d be cheering and forwarding the URL of his piece to everyone in the industry. I’m not impressed. And as much as it pains me to agree with anything the designer of the test states, the reality is much more complex than Michael states.

You can read all about the raging debate between Amir and Arny over at What’s Best Forum…frankly I gave up reading their rants many months ago. The essential question is whether a recording done at 96 kHz/24-bits can be perceived as different if presented to the listener at 44.1 kHz or 32 kHz. This issue looms large in the hearts and minds of audiophiles and audio enthusiasts who want to know if “high-res” is the next big thing in the world of music reproduction or whether it’s just a hoax to extract more money from uninformed consumers. It’s actually both…but I’ll save that assertion for the last part of the post.

So did Amir “prove” that at least one human being can successfully distinguish between the lower and higher sample rate? No, he didn’t.

In the same article, Michael refers to a similar test that was done collaboratively by Scott Wilkinson and myself over at the AVS Forum…although no mention was made of my contribution to this ABX comparison. You can visit their site and read about the AVS-AIX test by clicking here. The AVS Forum comparison was done using real music tracks rather than a few jangling keys.

The testing methodology is the problem. And it was a problem in the AVS Forum comparison as well…and both Scott and I knew it. It all comes down to the fact that people aren’t honest. Unless I can get you to come to my studio and submit to a rigorous procedure that compares a high-res files vs. the same recording in standard-res, I can’t trust you to do the test yourself. It’s too easy to cheat. If I was taking the test here at my desk, I would be able to tell you which file was which without even listening to them. In fact, I would be able to identify them without even opening them. The bigger files…the 96 kHz/24-bit one…would be double the size and take twice as long to download.

So Scott and I decided we had to sample rate the standard-resolution file back up to 96 kHz/24-bits to remove the temptation to simply look at the file specs. Even doing the SRC back to 96/24 doesn’t guarantee that people won’t cheat. Just open the files and look at the spectra.

And every time you process or convert a file, there are changes to the bitstream. Maybe that’s what Amir was using to tell the two files apart? It’s a fair question to ask. What if the testing methodology was changed to avoid any conversions? Put a very high quality microphone in front of an instrument (not keys) that produces lots of ultrasonic frequencies and record through identical converters at 96 kHz/24-bits and 44.1 kHz/16-bits. Then play those files and do an ABX test. People would challenge the quality of the conversion but this would be closer to determining the realities of high-res vs std-res.

Amir proved nothing.

More tomorrow.