Like most of us, when I was a kid, I suffered from a bad case of technolust. Among my favorite reading matter were catalogs from Allied Radio, the monthly Popular Electronics, and the publications of the American Radio Relay League, the organization of ham radio operators, of which I was a member — although, at the age of nine, not a very accomplished one. One of my favorite books of theirs was a slight volume called The World Above 50 MHz, which talked in vague terms about how hams of the future might be able to take advantage of what were then considered “trash” radio frequencies. Although these high frequencies (or very short wavelengths, as we used to refer to them) might be useable for things like commercial television and FM radio broadcasts, they were not much good for any sort of long-distance communication, which is what hams lived for. Novice ham operators (like me) were allowed to operate voice transmitters way up there around 144 MHz, where they wouldn’t bother anyone. Much more valuable to experienced hams was the spectrum between 2 and 30 MHz, because signals at those “short-wave” frequencies could bounce off the ionosphere and travel around the world. Above 50 MHz was of interest only to “experimenters.”
Since those days, the VHF, UHF, SHF and EHF (“S” for Super and “E” for Extra, in case you were wondering) bands have, of course, been used for a stunning variety of purposes and are now viciously fought over by a plethora of different wireless services, both real and imagined. With the advent of the communications satellite, short-wave radio (with the exception of propaganda broadcasts into unfriendly nations) went more or less the way of the dinosaur. Congested spectra made signals that did not bounce, but instead beamed right through the ionosphere, much more valuable.
Today, we all talk nonchalantly about 700MHz digital television broadcasts, 900MHz wireless phones, 1.9GHz digital cell phones and Ku-band satellites, which use the 13 to 18GHz band. In Europe, research and field testing are going on into frequencies as high as 60 GHz — which would have a great future in point-to-point transmissions, were it not that they tend to be disrupted by things like snow.
But you don’t want to hear me talk about radio, you want me to talk (if at all) about audio. Just as the radio amateurs of yore considered 50 MHz the top of the useable spectrum, audio engineers and enthusiasts have long regarded the 20kHz upper limit of human hearing as an inviolate parameter, and signals above that simply didn’t need to be dealt with. In the days of analog, this proved to be very helpful, because the physics of audio transducers and media — the tape and tape heads, microphones and speakers — made recording ultrasonic frequencies a difficult proposition indeed: Above about 10 kHz, for every tiny increase in high-frequency response, was an enormous increase in cost.
This is not to say that the world above 20 kHz has been completely ignored. Phonograph records have long been — at least in theory — capable of playing higher frequencies; in fact, the ill-fated ’70s quadrophonic LP system known as “CD-4” (for “Compatible Discrete,” not the other kind of CD) took advantage of this fact by putting the back/front difference signal onto a 30kHz carrier, which meant that signals as high as 45 kHz were being cut into the grooves (even though they did tend to disappear after a few listens).
And there have always been laboratory instruments, both transducers and recorders, capable of handling ultrasonic frequencies. Some of these, like B&K microphones, have trickled their way down into the professional audio world.
In analog audio, the top-end response of a system generally sort of fades away gracefully the higher you go, not unlike the noise floor. In digital, however, because of the need for a fixed-clock frequency, there is an absolute limit, the Nyquist frequency, above which signals cannot be processed, period. Ever since the first digital systems — F1, DAT and CD — put a brick wall up in front of any frequencies over 20 kHz, voices have been grumbling that this really wasn’t high enough. Many of the grumblings were objections to the phase distortion that the lowpass filters engendered below 20k, but some were complaints that we were losing “detail” in the higher frequencies, which could never be recovered, and that in order to achieve true “high fidelity,” our audio systems needed to stretch further into the ultrasonic realm.
In recent years, as the cost of bandwidth and digital storage has plummeted, those grumblings have become a deafening roar. Much of the pro audio community is now regarding ordinary CDs (which were once described, lest we forget, as “pure, perfect sound forever”) with something approaching contempt, while even sub-$1,000 home recording systems are boasting about the superiority of their 96kHz sampling rates so that those of us with plain old 44.1 and 48kHz hardware are feeling left behind.
But do we really need these higher sampling rates? Or is the whole thing, as some say, just a marketing scam to shame us into junking perfectly good equipment and buying all new stuff? Or is it perhaps something that we audio pros have latched onto to make us feel superior to the great unwashed millions who are forced to listen to lo-fi audio that tops out at a paltry 20k?
Well, the question of whether we need to do this actually has to be broken down into three questions: First of all, can we hear sounds above 20 kHz? If not, do ultrasonic frequencies somehow influence sounds that we can hear? And if they don’t, is there something about higher sampling rates that makes the stuff in the audible band sound better? I’m going to deal with the first two parts this month and then try to tackle the third in another column.
There’s no question that there is plenty of sound energy far above 20 kHz in the musical and natural worlds. For some striking evidence of this, take a look at a paper (www.cco.caltech.edu/~boyk/spectra/spectra.htm) by James Boyk, a pianist and electrical engineer at Caltech, which shows, among other things, that the spectrum of a trumpet with a Harmon mute slopes down linearly and smoothly from 2 kHz all the way to 102 kHz — and probably beyond, but that’s where his spectrum analyzer quits.
But do those frequencies actually reach us? As I wrote here a couple of months ago, Dr. Chris Halpin, an audiologist at Massachusetts Eye and Ear Infirmary and an erstwhile electronic musician, says that there’s no way to measure a human’s response to sounds above 13 kHz or so. It doesn’t mean we don’t hear sounds up there, it just means that they have not figured out any objective means to quantify our sensitivity to them — or lack of it. Whatever it is, however, it’s awfully small. According to the tests Dr. Halpin gave me, for example, my sensitivity at 10 kHz is down some 60 dB and plummeting. If the curve were to continue, my sensitivity at 20 kHz would be -120 dB, which is probably comfortably below the noise floor of my nervous system.
Interestingly, the audiologists tell us that we also can’t differentiate between dissimilar frequencies in that range. When anything above about 12 kHz tickles the cilia deep in our inner ears, it registers in our brains as “high,” and that’s about all the information we get from it. It might appear, then, that as long as there is something going on in the top octave, it sounds perfectly okay to us.
But surely people have done some hard research to see if we respond to frequencies above 20k, haven’t they? Well, a search of the literature actually turns up one — count ’em, one — formal study in this area. A paper presented at the AES Convention in October 1991 by five gentlemen from the National Institute of Multimedia Education in Japan is titled “High-Frequency Sound Above the Audible Range Affects Brain Electric Activity and Sound Perception.” (It is preprint number 3207, available from the AES at www.aes.org.) It’s quite a fascinating document with some pretty weird results. The researchers stuck electrodes on the scalps of a rather small sample of people — 10 men and six women, ages 20 to 34 — and played them the recorded sounds of a Gamelan orchestra, a source that is very rich in high harmonics. The subjects heard the recordings through two systems: one with response out to 40 kHz, and the other with a lowpass filter at 26 kHz.
In A/B testing, the subjects could not hear any difference between the two systems. However, the researchers found that six of the 16 subjects showed a marked increase in brain electrical activity, which started anywhere from 20 to 80 seconds after the music started, when they were listening to the 40kHz sound. Six of the subjects showed a slight increase, while the remaining four showed a slight decrease in brain activity. There were a couple of other tests in the paper, and those results are equally curious but hardly more compelling. Among the remarkable conclusions that the researchers made from this experiment: Not only do we indeed respond to sounds above 26 kHz, but our standard method of real-time A/B evaluation of audio systems is not valid!
Interesting stuff but a little out there, I think you’d agree. And because the results haven’t been confirmed or followed up, as far as I know, in the 11 years since the experiment, it would seem that we’d have a lot of company. So maybe I’m missing something, but I think that it’s a bit much for the entire audio industry to reinvent itself based on this one light-years-from-definitive study. It reminds me of that infamous paper that equated a person’s ability to resist pressure on his outstretched arm with whether he was hearing digital or analog source material. Whatever happened to that guy? Perhaps he’s working up an experiment to prove that subjects listening to 96kHz digital audio on their Walkmans can run a marathon 73 seconds faster than those listening to 44.1kHz audio!
There are less-formal experiments that purport to show that we can hear above 20 kHz, and perhaps the best known of these is the one that Rupert Neve — whom I have a tremendous amount of admiration for, although I think he’s completely wrong on this — does. He plays his audience a 10kHz sine wave and then a 10kHz square wave, and everyone in the place agrees that the two waves sound different. Therefore, he concludes, because the lowest harmonic above the fundamental in a square wave is the third, we are hearing 30 kHz!
Of course this is, as the English say, “tosh,” and many before me have pointed this out. There are a lot of reasons why we can hear the difference between those two tones, none of which have anything to do with ultrasonic sensitivity. One is simply that the energy of a square wave is higher than a sine wave at the same nominal amplitude, so the square wave sounds louder. Another is that any transformers in the signal path, unless they are exquisitely designed and constructed for passing such high frequencies, will introduce slewing and intermodulation distortion from the square wave — not only from the third harmonic, but from all the odd harmonics above it — that will have products well inside the audible range. And, if somehow a perfectly amplified 10kHz square wave were to make it all the way to the speakers, then the speakers would create their own distortion, which would be quite different from the distortion a sine wave would make.
You can easily prove this for yourself by running a 10kHz sine and a 10kHz square through a guitar amp. You will immediately hear the difference, even though the amp probably doesn’t have much response at all above about 6 kHz. My friend Leon Janikian, a longtime audio engineer and a professor at Northeastern University, re-creates Neve’s experiment for his classes, but with an additional step: He plays the two signals from oscillators and then records those same signals onto a 44.1kHz DAT machine and plays them back. Then he asks the group if they can discern any difference between the first pair and the second. Regardless of the order in which he plays them, the students can easily differentiate between the sine wave and the square wave, but they can’t hear any difference between the pre-DAT and the post-DAT signals. Because the DAT isn’t recording anything at all above 21 kHz, it’s obviously not energy at 30 kHz that the students (whose high-frequency response is probably a lot better than Leon’s) are hearing.
So, it seems like the audio world above 20 kHz, unlike the radio world of my youth above 50 MHz, will probably not turn out to be very important. As I said, there may be other reasons why high sampling rates are helpful, and they are very much worth discussing, but not because your old technology, and mine, is missing anything.
Paul Lehrman is at long last beginning to get over his technolust. Thanks to Richard Elen and David Moulton for their contributions and suggestions.