There is one consistent misconception that non-scientists have about science, and that is the Albert Einstein problem – the idea that major scientific problems can be solved by a lone genius who sweeps away decades of dead-end sawdust and comes up with a miraculous total solution. That stuff only ever happens in the movies, and even then the genius scientist either uses his new invention to take over the world, or becomes indentured in the service of a mad megalomaniac with the same idea. I don’t recall off-hand a movie where the mad genius starts a high-end audio company.
Back in the real world, science doesn’t typically work like that. Individual scientists make progress in tiny steps, none of them particularly noticeable outside of their own niche fields. But the net result, seen from 10,000 feet, is a remarkable overall rate of progress. 20 years ago, my first home internet connection came via a dial-up modem, which was so slow that its speed was measured in baud. Today, my home internet connection has 25Mb/s download speed and 15Mb/s upload speed, and I can access it wirelessly anywhere in my home. Nobody invented that. The state of the art simply progressed from dial-up to FTTH in an uncountable sequence of individual tiny nibbles…and it continues to progress today at approximately the same rate.
Because of the success of the scientific method in virtually every aspect of our modern lives, it is natural to look at the problems we would like to resolve in our high-end audio world, and expect the scientific method to provide the most obvious path forward. We’ve all come across the biggest elephant in the audiophile skeptic’s user manual, the double blind test (DBT), which is inevitably cast as the Gold Standard for the scientific method. And although we’ve also seen countless rebuffs to the DBT argument, they suffer in comparison to the apparent simplicity of the DBT proposition. In the end, nobody ends up satisfied, whichever side of the fence you may sit on.
The core of the audiophile problem is that the objective of a high-end audio system is to provide a satisfactory audio experience to its owner. That’s the beginning and end of it. The purpose is most assuredly not to satisfy the rationalist expectations of an armchair expert somewhere in Seattle. But the manufacturer of audio equipment is in many ways in a similar position. He knows no more than the armchair expert what the owner is experiencing, yet he has his own set of rational expectations. But where the armchair expert can pontificate to his heart’s delight with no concern for the consequences, if the manufacturer’s equipment isn’t satisfying his owners’ expectations he does have a real-world problem on his hands.
What’s to stop the scientific method from addressing this core problem? The fundamental issue of science at play here is how, exactly, we can know what a listener is hearing, and – more to the point – how we can quantify it. For example, we can ask a listener whether he hears a good stereo image, and he can tell us yes he does, or no he doesn’t, or he can maybe describe in detail how he perceives the stereo image. But we can’t quantify it with a score on a scale of 1–100, such that you and I can both listen to the exact same system and report independently, reliably, and reproducibly, that we are hearing a stereo image with an imaging property of 74.3%. It’s a problem because science is only really helpful with things it can meaningfully measure.
That wasn’t a problem for the guy who designed my DSL modem. He had a set of straightforward performance specifications as long as your arm that he had to meet, and he wasn’t at all concerned by whether or not a customer in Peoria would find his product preferable in a DBT. Meeting the performance specifications was all the product had to do, and, frankly, all the customer in Peoria was expecting from it. In fact, exceeding the specs would be a problem, one which he would ultimately solve by re-designing it using cheaper and cheaper parts until it just met all the specs.
Now, the armchair expert in Seattle doesn’t care either, and just prates on about how this, that, or the other piece of audiophilia is snake oil unless proven otherwise by a DBT. The audio manufacturer, on the other hand, cares deeply. He still needs to be satisfying his customers. It’s an existential challenge for him. And in the absence of a proven scientific method with which to address his toughest design challenges, he has to rely on his experience and skills.
So why doesn’t science just step in and address these issues? The answer is that there is a huge gulf between what we know and what we need to know. We’ve pretty much come as far as we can with our understanding of what the ear can detect. Far enough to know that the things we still don’t know in that area aren’t what are holding us back. After all, it is clear that mature audiophiles with measurably degraded hearing are still critically demanding of high-end audio systems, and are still among the most valued judges of a quality system.
Front-line basic research interest today is more concerned with how the brain perceives what the ears detect. And the biggest challenges facing scientists in that field is being able to quantify what it is that the brain is perceiving. We are only taking our first baby steps in this field at the moment, and it will be a long time before major developments in the field will give rise to significant advances in high-end audio.
This was brought home to me during some listening tests with Prof. Edgar Choueiri’s BACCH system, which I have previously written about in Copper #60 and #61. In particular, I was playing with the system set up such that a pair of headphones emulates the sound of a pair of loudspeakers in the room. If you do the demo properly, even the most fastidious audiophile can be totally fooled into thinking he or she is listening to the speakers when in fact they are listening to the headphones. But what is interesting is that I found that this illusion is easily shattered. Under certain conditions my brain would stop perceiving a pair of external loudspeakers, and instead perceived the diffuse sound field typical of a set of headphones. I then found I could usually force my brain into switching back to the loudspeaker illusion by concentrating on looking hard at the speakers, but it took a surprising degree of effort. The realization that the brain can interpret the exact same high-quality sound presented to the ears as being one of two binary sound fields – and switch between them – was very interesting. The scientific challenge is to independently measure and quantify effects such as these, and we are nowhere near being able to do that.
It illustrates at least one aspect of the problem of conducting any sort of structured audiophile listening test. If an audiophile wants to judge the impact of a change to his system, he will first want to ensure that his listening experience is set up to be exactly how he is accustomed to hearing it. After all, our systems are there to be listened to, and we set them up so that our listening experiences are as well-optimized as we can make them. However, that set-up may not be conducive to a formalized test, with appropriate controls. If, for the purposes of the test, I rig your listening seat with electrodes in the cushion which periodically deliver a serious jolt to your rear end, I suggest that it is unlikely that the outcome of the test is going to be reliable. That’s an extreme example, but the core point is a very strong one. External factors influence how our brains perceive what our ears hear, and we don’t yet understand that well enough to take proper account of it in a scientifically controlled test. It is clearly a layered and complex matter.
Where does that leave us, then, if we wish to bring the scientific method to a study of audiophile matters? The bottom line is that if we wish to definitively study the audible impact of a cable (or an isolation mount, or passive preamplifier, or whatever), science does not yet give us the ability to make objective measurements of the critical outcomes. Science can’t even tell us what the listener actually hears, because the listener could be a youngster with perfect hearing or an older guy whose hearing has demonstrably degraded, or someone with a bit of an ear infection…and in any case those factors don’t seem to make a fundamental difference to a listener’s ability to perceive the key qualities of a high-end audio system. Neither is science able to tell us how the listeners’ brains are perceiving the sounds we play for them, something which we know is affected greatly by external influences. We’re not even able to observe them, let alone quantify them.
Consequently we’re reduced to asking listeners to self-report what they are hearing, with the results being generally unsatisfactory when employed in a scientific approach to high-end audiophilia, not to mention unacceptable to the armchair expert in Seattle. The alternative is to be some kind of a science fundamentalist and proffer as irrefutable truth the notion that if I can’t measure it, you can’t hear it. But like fundamentalism in all walks of life, I’m not sure how, if at all, that can ever lead to progress.
The primary take-away is that this is a situation which does not really lend itself to the scientific mind. The curious mind, yes. But the professional scientific mind, no. Because the professional scientist wants to see a way forward, and the road forward is paved with experiments yielding hard, meaningful data, which can be used to either prove or disprove an interesting theory. Invite a reputable audio research scientist to oversee a DBT involving the audibility of some piece of audiophile gear and your response will involve strings of garlic, signs of the cross, and lengthy unexpected commitments in Antarctica.
What is left is what the leading lights in our industry actually do. It is a combination of trust in the designers’ own ears (and those of a limited number of trusted colleagues and acquaintances), and proven test methodologies which have withstood the test of time. These tests set about measuring things that are known to correlate well with desirable qualitative results. Things like distortion in amplifiers are good examples…it is clear that lower distortion has got be a Good Thing. But we mustn’t forget that there was a time when we only measured THD, and in our haste to drive THD values ever lower we made amplifiers that actually sounded worse. We needed to learn about IMD, and how to measure it, in order to make real progress. We are currently in a similar place with phase response, and we are learning that linear phase response tends to correlate with better sound. But does it follow that hard-line efforts to minimize phase nonlinearity will always pay de facto sonic dividends? We will find out in time.
Science will play its role in all this. For example, linearizing phase response requires a lot of in-depth scientific knowledge, not only in understanding the implications of tinkering with phase response, but also in correctly measuring it. But the end results will continue to be driven by what listeners hear – or, being pedantic about it, what they perceive they hear.
By the way, if there’s an armchair expert out there in Seattle – which, on the balance of probabilities there surely must be – rest assured that I don’t have you, specifically, in mind.