The audio world has some interesting problems. A particularly vexing one—that we generally don’t think about much because it makes our heads hurt—involves what we call “audibility.
This problem arises a lot as we introduce higher resolution, greater bandwidth, and improved other audio qualities to our audio systems. We want to know: can we hear the difference? If we can, then we say such an improvement is “audible,” and if we can’t, we say it isn’t.
Pretty simple, right? Well, actually, maybe not so simple! That’s the problem!
The question of “audibility” is important. Particularly during times when the technology is improving rapidly, we need to know: how much technology is enough for us? Can we “hear” 24-bit audio as opposed to 16-bit or 20-bit? Can we hear 192 kHz sampling rate as opposed to 32, 48, or 96 kHz? Why spend the money on gear that yields resolution and bandwidth that may far exceed what we really need?
There is a segment of our industry inhabited by a group of individuals called “audiophiles.” They are for the most part individuals with a real passion for their playback systems. They really get into this “audibility” business.
Fascinated by little modifications called “tweaks,” they do all sorts of apparently loopy things, including (over the years): painting CDs green, getting “special” high-performance cables at $500/foot, worrying about which way the audio “passes” through the cable, hell, even worrying about whether or not the cable is lying on the floor or being held 3" off the floor by the use of little standoffs!
All of these steps are taken for the express purpose of “improving sound quality.”
Before you start smirking about how loony these amateur guy/gals are, understand that the recording world is just as loopy and obsessive. We are hung up on 24-bit/192 kHz digital audio, pure tube-amplified analog audio, vintage hardware that sounds better, jitter, early reflections, impulse response and time smear, to mention a few things. We casually note that MDM X sounds terrible, while MDM Y has “really good converters” (which is code for “sounds really good!”).
So what’s the truth? Can we hear the effect of these things? If so, how much?
This is where the Cult of Golden Ears comes in. (This cult is not to be confused with my Golden Ears Audio Ear Training—I actually adopted the title because (a) it was a goof, and (b) I live with golden retrievers!) There are a bunch of people who either appear to or claim to have exceptional hearing (“hyperacute,” a friend calls it).
We’ve all heard golden ears stories about guy/gals who have hearing so sharp that they can hear the effect on their audio system of going from Bakelite to titanium knobs! Golden ears types are ‘sposed to be able to hear really small differences, and make comments like “Sounds a little fluffy—I think the build-out resistor on Channel 2 is a couple of ohms low,” “Gee I hate the sound of tantalum capacitors,” or “Uh-oh, the dither on the 24th bit just dropped out. They must’ve used the digital fader.”
So, by report and legend, it appears that some of us may hear these things. What does that mean? If some of us can hear the sound of a wire with reversed direction, should we specify wire direction throughout our system? If some us can hear the sound of tantalum capacitors, should we replace them all? It’s an obsessive business. But how far is going too far?
One of the things I do for a living is subjective audio testing. Clients bring me hardware and/or software and ask me to find out things about it, such as “Is our new 27-bit A-to-D converter audible compared to our competition’s 24-bit converter?” or “Will people prefer our new Grungetone 55 to the current industry favorite Fuzzwart?”
I’m supposed to give them answers such as “There is an audible difference between the 27-bit ADC and a stock 24-bit,” and “73.2% of listeners prefer the Grungetone 55 over the Fuzzwart.” If I’ve done my work well, these answers should be reproducible, which means that in application, the client will find my assertions prove predictably true. On that basis, they can claim audible superiority for 27-bit ADCs and/or Grungetone 55s.
The actual technical process of subjective testing is interesting, and I’m planning on writing an article on it. However, that actual testing process is beyond the scope of this article. Here we first need to have to look at some of the issues that come up surrounding such testing. We need to consider what it means for something to be “audible.”
This is a tough question. Simple, common-sense ways of testing turn out to give widely varying results, and usually prove to be open to challenge for a zillion different perplexing reasons. Part of the problem is semantic and philosophical, and part of the problem is psychological. Let’s talk.
The Zen of audibility
The dictionary says that “audible” is “that which can be heard.” Sounds simple enough. But sadly it isn’t.
Some rules of thumb to keep in mind are that:
• Audibility is variable as a function of levels, listeners, program, playback system, etc.
• Audibility means different things at different points in our hearing system. What we detect at the ear drum is dramatically different from what we consciously perceive in the brain. So if we hear it but don’t perceive it, is it audible? Hmmmmmm…
• Audibility changes as a function of listener knowledge. Hearing improves when we know for sure what we’re hearing (cynics call this the “it’s easier to get a good grade on a test when you know the answers beforehand” syndrome).
• Audibility changes as a function of experience. Called the “learning effect,” repeated exposure to an effect increases our sensitivity to whatever audible threshold we’re listening for and increases our level of annoyance.
• Subjective expressions of audibility are often magnified greatly and non-linearly, as in when we say “the difference in clarity is amazing” when the actual difference is so small that we can barely distinguish it from chance. What’s wild is that we still really believe it’s amazing, even when we know it’s a tiny difference. Go figure!
• Subjective recognitions of small variations in audibility are often expressed in unpredictable ways, so that a sound 0.1 dB louder may be described as “fatter” rather than “louder.”
So as you can see, audibility has a big range that is both quite variable and not very easily pinned down.
One approach has been to test for the Worst Case, using the most highly experienced and trained listeners with the best possible speakers, etc., in the best possible room with the best possible equipment. When this has been done, we can find out something about what one end of the audibility range for a given effect is, the end where all possible advantage lies with the listeners.
The problem is, this approach often doesn’t give us relevant data. It doesn’t tell us much, if anything, about real-world applications, because it violates the Range Rule, which states: The Range of the Theory must fit the Range of the Facts. The data, in this case, only fit the Worst Case, which may be inappropriate.
The Best Case approach is even worse, because now we have to establish the least trained, least experienced listeners, listening to the worst possible speakers, in the worst possible room, etc. Hardly seems worth it. Such coarse data will prove pretty useless in predicting useful standards for success in the marketplace.
What’s a threshold?
Meanwhile, we have to consider the meaning of the term “threshold.” When we talk about audibility, we usually mean a threshold of some physical or psychological magnitude below which we don’t hear a given effect or artifact and above which we do hear the effect. This point, in human perception, is called the “difference limen” or the JND (“Just-Noticeable-Difference). The normal way to establish this threashold is to measure the array of listeners’ thresholds to the Effect Under Test and then to find the average threshold, which is the threshold that approximately 50% of the listeners can hear.
So if 50% of people begin to perceive the presence of a sound at 0.00002 Pascals of pressure, then we say that that magnitude is the “threshold of hearing.” And in fact we’ve done exactly that, so that 0 dB SPL is defined as 0.00002 Pascals, and as the threshold of hearing.
What is often forgotten is the obvious corollary that if only 50% of the people can’t hear below the threshold, then obviously 50% can! So it is really foolish to say “nobody can hear it, because it’s below the threshold of audibility.” Half the population may be able to hear it!
Therefore we might wish to reconsider that threshold. If we’re really critical, we might like to set the threshold at the level where 99% of the population can’t hear the difference and only 1% can. Or if we’re feeling generous we could go the other way, finding the threshold where 99% can hear it and only 1% can’t.
If you think carefully about this, you can see that it has big implications for the design of loudspeakers, recorders, consoles, etc. We can design to pretty easy and low standards if we define the threshold of audibility as the level at which only 1% of the population can’t hear the defects. It really gets the price down!
Or we can be rigorous and design to a standard where only 1% of the population can hear the defects. But then our products will be quite expensive, and may not be competitive in the market, while only providing a marginal benefit for the golden-eared types. It’s a quandary.
Audibility in The Real World?
In the real world the question of whether or not something is “audible” is really clouded in a kind of psychological haze, and the real honest-to-God true answer is, “It depends!”
Some days you’ll hear things, some days you won’t. What you hear may be different from what someone else hears. Which loudspeaker, amp, console, etc. you use may have a big bearing on what you hear.
As a result you need to be very cautious with claims of sound quality, particularly in advertising. When we went to 20-bit audio from 16-bit, for example, we didn’t change the maximum level of the recording (0 dBFS), we changed the minimum level to a lower level.
Now, it’s worth asking: was the 16-bit level above or below the threshold of audibility? Under what conditions? How about the 20-bit level? 24-bit, anyone? Arguments abound. We hear reports ranging from “the change is totally awesome” to “it all sounds the same to me.”
While this is perplexing, it does in fact reflect the range of possible perceptions. No one is necessarily lying or mis-reporting.
So how do you cope? How do you assess the importance of “audibility” for any particular item of gear, or acoustical treatment? What’s right for you?
My own strategy has been to put on my producer’s hat and ask the question: “Does this difference or audible artifact matter to my production vision for the project?”
I first used this criterion back in 1989, when I was on a listening panel to evaluate the CBS Copycode for the National Bureau of Standards. I heard the copycode work very obviously and clearly in a couple of spots, but I wasn’t sure how big a deal it was. So I asked myself, “if I were producing this project, and I came across this artifact during production or post, would I try to fix it, or just let it go?” In that particular case, the answer was pretty definitely “I should fix it.”
So I could reasonably report to the National Bureau of Standards that I could hear a difference, and that difference was annoying enough that as a producer I wouldn’t want to release the project with the effect present.
Now this is an interesting and useful standard. If we are mixing and we move our heads by six inches, the sound will audibly change. Does it matter? Not to me.
But if it does to you, then that magnitude of difference is important, and should be treated, whatever practical difficulties that may present (I told you this business is obsessive!). If it doesn’t then we can safely ignore it, saving a heap of time and money in the bargain. Normally, we hear a whole bunch of small differences during production that don’t really matter to us, and we let them go because (a) life is too short as it is, (b) we’re over budget as it is, (c) they’ll never hear it as it is, and/or (d) it won’t bother them even if they do hear it as it is.
When we start looking at the world in this existentialist sort of manner, where things either “matter” or “don’t matter,” it becomes a little easier to orient ourselves reliably. We can zero in quickly and intuitively on which magnitudes aren’t really very relevant to us, abandon them, and move on to more important things.
When other people tell us that such a magnitude does “matter” we can take their words with a grain of salt, asking ourselves existential questions such as “Why does it matter to him/her?” and “Is he/she really hearing what he/she thinks he/she is hearing?” and even, occasionally, “I wonder what he/she is smoking these days.”
Even if said person is a world-famous engineer/producer/artist, we don’t automatically have to be concerned if we don’t hear the same differences. We need to just keep observing for ourselves, patiently, carefully and critically.
Which is to say: avoid worrying about the Emperor’s New Stereo System. Work in response to what you hear, not to what others say they hear.
Happy audibles, as we say in the NFL!
Dave Moulton owns and operates Moulton Labs, which does subjective testing for a variety of clients. He’s also the author of the book/CD package ‘Total Recording,’ as well as the Golden Ears Audio Eartraining Program.