What do audiophiles hate the most? Well, there’s Bose. And the late Julian Hirsch. After those two, it would probably be blind testing, specifically ABX testing. Why? Because the results of ABX testing tend to conflict with much of what audiophiles believe. The topic of ABX testing may be poised for another look with the recent emergence of the Audio by Van Alstine AVA ABX, which to my knowledge is the first commercially produced ABX box released in more than a decade. In this article, I’ll discuss what ABX testing is, explain the criticisms of ABX testing, and get a little into my first experiences with the AVA ABX.
When I found out about the AVA ABX from reading the comments section of this website, I immediately contacted Audio by Van Alstine’s namesake, Frank Van Alstine, to see if I could borrow one to try out and then buy if it met my needs. I was attracted to it not for its ABX capabilities, but because it looked like a well-made and versatile switcher I could use in my reviews. I have a good switching system that I designed for this purpose; however, like most hand-built, one-off electronic products, it’s not very reliable. I could see from the interior shot of the AVA ABX that it was built the same way my switcher is–with high-quality relays, minimalist controls for level matching, and a switching system. But the AVA ABX was designed by an experienced audio engineer, Dan Kuechle, who has the knowledge and resources to build products with professional-grade reliability.
What is ABX testing? I’ve been using the AVA ABX for level-matching and switching in my reviews for a few months, but I hadn’t actually experimented with the ABX function until recently. Here’s how ABX testing works: The ABX box presents two audio signals, A and B, plus a third, X. X is either A or B; the assignment is random, and it changes (or doesn’t change) with every trial. So you listen to A, listen to B, listen to X, and then decide whether X is A or B. Then you or the test administrator activates a function on the ABX box that displays whether X was A or B for each trial.
Random guessing will, after enough trials, result in correct selections 50 percent of the time. So, to prove there’s a significant difference between A and B, you’d have to correctly identify X somewhere between 50 and 100 percent of the time. Even someone randomly guessing might get 6 or 7 out of 10 right, so the results aren’t meaningful unless you can do even better than that. For a 95 percent confidence level (a typical standard for statistical significance), you’d have to have correct identifications on 23 out of 24 trials. That’s three test sessions on the AVA ABX, which provides eight trials per test session–quite a high hurdle.
Full story Via hometheaterreview.com