Moulton Laboratories
the art and science of sound
Yeah, I Think I Can Hear It! Wow! It’s Like, Totally Awesome!
Dave Moulton
December 1999

The discussion of blind versus subjective listening continues.

About hearing small differences

Last month, in our on-going discussion of the question of how audible things are, I went over the problems encountered in blind testing. It turns out that from a practical standpoint, blind testing is pretty much the only practical way to measure small audio effects without prejudice, but that it also introduces some errors into the process. I made the case that those errors are pretty small, and that blind testing works pretty well. However, and this is a big however, there are people in our business who are not satisfied with those errors and findings. They make the case that such testing tends to miss or understate the audibility of a number of small but important effects, such as increased bit resolution and sampling frequency.

And here, once again, professional prejudice kicks in as well. We all “know” (“Heh, heh,” Dave said, prejudicially) that 20-bit audio sounds A LOT better than 16-bit audio and that 24-bit audio yields a REALLY AWESOME improvement over 16-bit audio, but that none of them EVEN REMOTELY APPROACH the wonderful warm signal purity of the unadulterated analog signal! Am I right? Don’t we all just KNOW THIS, in our heart of hearts, and in our rich, wise and oh-so experienced Golden Ears?

Walk the floors of an AES convention and just try saying, to one of the many vendors on the floor, “Oh, I dunno. 16-bit audio is plenty good enough for me. Hell, I can’t even hear any significant difference between 12-bit audio and analog! What’s the big fuss about?” It’s a little like being in a Porsche dealership, sitting in the newest Carrera 4 and saying, “You know, I kinda like the new Saturn just as much.” They are all gonna look at you, man. They are gonna wrinkle their noses, check the bottoms of their shoes, chew down quick on some really strong breath mints, and they are gonna get outta there in a hurry, find some brochures to stack. Musta been somethin’ you ate. Right!

Let’s get serious here. Not too long ago, I made an educational CD that did some demonstrations of digital audio (Playback Platinum Vol. 4: Dave Moulton’s Audio Lectures – I am not making this up!). A lot of those demos were at 12 bits and less. Now 5 bits sounds really pretty hopelessly bad, and 8 bits is pretty grungy. 12-bit audio, by comparison, sounds pretty clean, and not all that different from 16-bit audio. You can hear the difference if you pay attention, but still, it’s not all that bad, if you know what I mean. If you want, check out the demo and see if what I say is a fair description of things.

How can this be? How can 12-bit audio (only 72 dB, or 4,000:1 resolution) sound “reasonably clean,” while 24-bit audio (which implies a whopping 144 dB, or 16,000,000:1 resolution!), doesn’t sound “nearly as good” as “pure” analog audio (which probably has a resolution of 120 dB, or 1,000,000:1)? How can comparatively gross differences be comparatively inoffensive while much smaller differences can be characterized as “totally awesome?” It’s a puzzler!

There is an explanation. Weird as it seems, there’s some real truth to this reality, and it’s in the psychological realm. It is helpful to understand that truth. As reader Dave Riddle points out, “Humans are lousy quantifiers, but they are great comparators.” And when we listen to program material on an on-going basis, such as in production work or listening for fun, mostly we are just “quantifying,” or making judgements about the general effectiveness of the signal’s quality for our purposes. On the other hand, when we get serious about studying “the quality of the signal,” we turn into high-rez comparators of the first order, deconstructing that signal in an obsessively detailed comparison with a reference version of the same signal. In that testing environment, the differences that are pretty much meaningless in everyday usage loom large. Why, they’re, like, totally, awesome!

I make fun of this, but I’ve been there. Take a seriously implemented A/B listening test sometime – you’ll see what I mean. We really get into it! We go inside the signal and start listening with a focused intensity that is simply unlike the attention we use for everyday production listening. The rule of thumb in the testing biz is that you’ve got to stop the testing every half-hour or so because it is such exhausting work that the test listeners are likely to keel over if you don’t give them a break that often. I don’t think I’ve ever listened as hard in my life as when I’ve been digging into the so-called “small impairments” of a signal under test using short excerpts of music and voice. Swerving around the notes, picking through the spectra, plunging down through the reverb trails, picking through the sonic confetti of noise, searching for tiny, fragmentary artifacts lurking in the auditory shadows. Man, it’s a trip! One day, we were listening to a watermarking scheme and we had the difference signal available so we could hear “where” the watermark artifacts actually were – we “knew” the answers! I went deep that day – took me about six hours to get back out of the ozone! Absolutely mesmerizing, compelling, draining and exhausting.

And when us test listeners get into that hyper-extended, spacey and somewhat exalted and stony state, EVERYTHING becomes a Moment of Epiphany, and then, when we do find something, that momentary flicker of an artifact of wrongness, that microscopic crack that reveals to us the existence of that whole crummy fake surface of The Audio Mirror, and it in turn causes the whole illusion to crack, to splinter and to crash, why, that little sucker really is LIKE TOTALLY AWFUL!

And that, gentle readers, is why we use such overblown words to describe such small differences. The testing modality leads us to it, and if we haven’t got the testing protocols to correct for those excesses of golden-eared perception, we tend to overstate, to magnify what it is that we perceive. We forget that we are listening through a microscope, if you know what I mean.

This is important to keep in mind. We need to look more carefully at the magnitudes of resolution we think that we require, and use some good old-fashioned engineering judgement to determine what those magnitudes should be. Blind testing, with it’s slightly blunted edge of resolution, helps us to do this, steering us away from extremes, from that oh-so-seductive ozone where we are really dancing with stuff that’s just about indistinguishable from chance!

Whew! Gets a little fragrant in here! Next month, we’ll go a little farther, and consider what we REALLY mean when we say something is “audible.” Uh-oh!

Thanks for listening.
Note: The following group of columns that I wrote for TV Technology are an attempt on my part to describe some of the issues surrounding our attempts to measure and evaluate the audibility of high-resolution formats. Together, I think they make an excellent short survey of these issues. I hope you find them useful.

Post a Comment


rss atom