unconsciously pays more attention to where these strange distracting noises are coming from, and less attention to the spatial portrayal of where various portions of the music are coming from in the re-created image - thereby further degrading the perceived spatial imaging of the loudspeaker.
Ninth, this spurious misbehavior of the driver, or parts thereof, is obviously non-pistonic, since it is disobeying the control of the input music signal and is uncontrollably vibrating (and breaking up) on its own (it "obeys" the music signal only insofar as it waits for the music signal to trigger this misbehavior). Thus, the acoustic output from this vibrating misbehavior is not music. It is garbage. The misbehavior does output acoustic energy, and it might sound somewhat like music (since it is modulated by the music, and thus is like a modulation distortion byproduct of the music that tracks the music). But it is not an accurate replica of the music signal.
A musically sensitive listener is not fooled by this misbehavior, and can easily tell that the acoustic energy output of this non-pistonic breakup misbehavior just doesn't sound right, doesn't sound like an accurate portrayal of real music. But the common sine wave frequency response measurement, used to evaluate and sell loudspeaker systems, gets totally fooled by this garbage output from the driver's misbehavior. The common frequency response measurement counts the garbage output as a good thing, not as a bad thing. For example, suppose that the pistonically accurate response of a driver's cone starts declining above 2000 Hz, but that the driver's dust cap starts spuriously misbehaving around 2300 Hz, thereby putting out uncontrolled garbage energy in the region around 2300 Hz. A frequency response measurement of this driver would be fooled into pronouncing this driver to be substantially flat to beyond 2300 Hz, and the design engineer or you the consumer looking at this frequency response graph would make the mistaken assumption that this seeming substantially flat response to beyond 2300 Hz means that the driver is accurately reproducing music to beyond 2300 Hz. But you'd be very wrong, and very misled by the fact that the traditional frequency response measurement cannot tell the difference between accurately pistonic signal reproduction and acoustic garbage energy produced by driver misbehavior.
Furthermore, the design engineer would never suspect from this common measurement that there is anything sonically wrong with this driver. It looks flat, so it must sound flat and sound accurate. But the human ear/brain can easily hear that something sounds wrong with this driver. Indeed, there are 9 sonically audible ways that this driver sounds wrong, due to its misbehavior that fails to show on a conventional frequency response graph - the 9 things just discussed above. One of the oldest puzzles in loudspeaker engineering is that two drivers or two loudspeaker systems, that measure the same by a conventional sine wave frequency response measurement, nevertheless sound different. It's no puzzle at all. The conventional frequency response measurement simply and utterly fails to detect and show the 9 sonic problems above, problems that are highly audible to our sensitive ear/brain mechanisms.
Human Hearing Abilities
The prehistoric cavemen who survived to become our ancestors were the ones who learned to hear and instantly analyze with their brain the difference between colorations of different materials emitting a transient sound. The cavemen who survived were the ones who could instantly tell that one transient represented the sound of a wood twig snapping, while another transient represented the sound of stone against stone. We all inherited this sensitive ability from the cavemen who had it and survived (obviously, those cavemen who lacked this ability did not survive to pass on their inability). That's why we can so sensitively detect, and are so irritatingly bothered by, foreign material colorations from drivers, or any non-pistonic misbehavior that doesn't sound accurately like the original music, accurately like the nylon string and wood body of a guitar, etc. Even if it is only a small portion of the driver that is misbehaving (say the dust cap in the center), our sensitively acute ear/brain is instantly distracted to, and instantly zeros in on, this misbehavior, on this foreign sound, garbage, and noise that sticks out like a sore thumb, in all 9 sonic ways discussed above.
Our ear/brain also has a special time slicing (or windowing) capability. We can easily hear and separately analyze what the beginning attack of a transient sounds like, and then what the after-resonances a split second later sound like. Our brain is able to effectively cut a music signal into fine, separate slices of time, and then evaluate what each fine slice of time sounds like. Is the transient attack hard or soft, sharp or rounded, bright or dull? Does the transient attack sound like the plucked guitar string is made of softish nylon, or aluminum, or hard plastic? Then, do the sounds immediately after the transient attack sound like the body of the guitar is made of fine wood, or toy plastic? Do the upper key piano strikes sound like the true piano sound of hammers made only of wood and felt hitting the metal piano strings, or do they sound more like metal hammers hitting metal strings? Our ear/brain instantly analyzes the time slice immediately after the piano key strike, to tell us whether the material sounds as reproduced by the loudspeaker match the materials of which real pianos are made, or whether they instead sound like a foreign material, e.g. metal, paper, plastic, or rubber (if they do, then that indicates almost surely that such a foreign material is misbehaving in the driver).
Thanks to the keen ability of our ear/brain to hear and instantly analyze sounds in these various ways, we can easily hear sonic problems in drivers, in those 9 ways discussed above. Moreover, our various auditory analysis abilities work in cooperative combination with each other. This means that the 9 above ways, in which misbehaving drivers sound bad, further combine to exacerbate each other, and make these 9 sonic problems sound even worse, making them even more easily detectable and even more obnoxiously irritating and fatiguing. Consider for example our ear/brain's ability to time slice, and our ear/brain's sensitivity to hearing and instantly analyzing foreign material colorations.
Let's apply both human abilities to an example, and see how they make a driver's sonic problems even more annoying. Suppose that a driver's plastic dust cap (the center portion in front of the voice coil) misbehaves and spuriously vibrates at 2300 Hz. This spurious misbehavior would cause and be heard as all of the various 9 sonic problems discussed above. Let's look at just problem 6 above. The spurious misbehavior of the dust cap at 2300 Hz will add spurious extra energy at 2300 Hz, so it will change the tonal balance of the driver, making it sound artificially brighter. But how much brighter will it sound to us? Just how artificial will this extra brightness make the perceived tonal balance sound? If the dust cap is merely 10% of the driver's radiating area, let's assume for simplicity that its total spurious addition at 2300 Hz is an extra 20%, as averaged over time (this assumes just modest Q for the misbehaving resonance). Would we hear the artificial extra brightness as being just 20% bright? No, we would hear it as being far worse than that.
First, because of our ear/brain's time slicing ability, we would hear the extra brightness of the spurious misbehavior lingering in time, after the strong musical transient, that had triggered this misbehavior, has already gone silent. In this relative music signal silence after the strong musical transient, the misbehavior at 2300 Hz, lingering after the strong musical transient that triggered it, might be just as strong as the quiet background music signal, or even stronger. So, during this slightly later time slice, our ear/brain, with its time slicing analysis ability, would hear that this artificial added brightness adds an artificial extra 100% or 200% brightness to the tonal balance of the music.
Second, because of our ear/brain's ability to instantly analyze foreign material colorations, we would instantly recognize, in each time slice, the plasticky sound of the misbehavior of the plastic dust cap, since we are very good at recognizing the characteristic sonic thumbprint of various materials as they vibrate, and it is plastic that is spuriously vibrating, out of control from the input music signal. And we would instantly recognize that this plasticky material sound is foreign to the materials that the real music instrument playing at that time is made of. Our ear/brain would instantly (and unconsciously) be distracted from the music and attracted to paying extra attention to this foreign sound (just as our caveman forefathers, in the midst of hearing other sounds, would instantly be distracted into paying extra attention to a foreign sound, such as a wood twig snapping behind them). Needless to say, as soon as our ear/brain unconsciously starts focusing extra attention on a foreign sound, it tunes out the background sounds from the real music, and this effectively dials up the loudness of the foreign sound even further (we use this same tool to advantage at a cocktail party or bar, tuning out the loud background noise, and dialing up the loudness of the person we're talking with, whom we are focusing our attention on). Since this foreign sound is also bright, adding energy at 2300 Hz, when we unconsciously pay extra attention to it and tune out the background real music, our ear/brain will be boosting the loudness of this extra added brightness even further. In short, the fact, that the added brightness of the spuriously misbehaving dust cap sounds like a foreign plastic material, and thereby unconsciously attracts our ear/brain's attention, winds up effectively boosting the perceived loudness of this brightness even further. Thus, we will hear the artificial brightness of the driver's tonal balance as being even worse than the 100% to 200% arrived at from our ear/brain's time slicing ability.
Diagnosing Driver Problems
As you can see from our thorough exploration of this subject, driver misbehavior problems are sonically a major issue. Surely, therefore, it is imperative that loudspeaker engineers be able to discover, measure, diagnose, and fix these driver misbehavior problems, when they are designing drivers or when they are selecting drivers to put into a complete loudspeaker system. But the tool that most of them rely on, the conventional sine wave frequency response taken at a 1 meter distance, is hopelessly inadequate to this task. What's wrong with this conventional measurement?
First, as noted above, this measurement can't tell the difference between accurate pistonic output and garbage misbehavior. It counts both as being "good" energy output. So it doesn't help the engineer to discover or diagnose driver misbehavior problems. Indeed, it almost conspires to hide these misbehavior problems from the engineer, even though as we now know the ear/brain of the customer and listener is superb at ferreting out and being obnoxiously irritated by these misbehavior problems, in 9 different ways.
Second, this measurement is an average over time, so it fails to duplicate the time slicing capability of the listener's ear/brain. It lumps together the driver's good behavior at the time of the initial transient with the driver's spurious misbehavior that the human ear/brain can easily hear sticking out like a sore thumb immediately after the strong initial transient has fallen to silence. Thus, it would read the driver's misbehavior in the above example as being merely 10%, as averaged over time. And measured misbehavior of merely 10% would not be visible to the engineer in the frequency response curve, since it would be hidden amidst the normal slight irregularities in frequency response. KEF in England pioneered a frequency response measurement technique that does do time slicing, and displays a 3D waterfall graph. This technique is an improvement, but it still has weaknesses such as spectral resolution of driver misbehavior.
Third, this common measurement is also an average over space. It is a spatial average of what the whole driver is doing, as seen from a far distance of 1 meter. Thus, it is still relatively insensitive to spurious misbehavior that is probably occurring in just one small part of the driver (perhaps just the dust cap or just the surround). By averaging and lumping together all spatially distinct radiating portions of the driver, this common measurement may well emphasize only the good performance of the larger radiating portion of the driver (say the main cone), and thereby hide the spurious misbehavior of a relatively small radiating portion of the driver.
The problem with this is that our ear/brain is very sensitive at detecting, emphasizing, and being irritated by even a small portion of the total radiated output, when that small portion has a foreign material coloration, as it always does (since this misbehaving portion, spuriously vibrating on its own and not under the control of the music signal, has to be made of some physical material -- and thus that spurious vibrating-on-its-own will inevitably bear the characteristic sonic thumbprint of that foreign material, which audibly obviously differs from the materials of the musical instruments that made the original sound). In short, a measurement which only looks at a spatial average of the whole driver output is irrelevant to the way that the listener's ear/brain actually hears the driver's behavior and misbehavior.
Small wonder that common frequency response measurements fail to tell us how loudspeakers really sound! Thus, it is imperative for the engineer to be able to separately measure each different portion of the driver, focusing in on each and looking for spurious misbehavior in any of the several distinct radiating parts of the driver. Celestion in England developed a laser scanning technique that could look at the behavior of each individual portion of a driver, but it was expensive to implement, so it is not in wide usage.
In previous research, we had developed a quick, easy, and inexpensive technique for identifying, isolating, and analyzing these spurious driver problems, and in a previously published article we showed the industry how to use this method. Basically, this method employs FFT analysis to do a very near field microphone probing, of the response to a step input, from individual parts of a driver's radiating area (including main diaphragm, dust cap, and surround), using time windowing to further isolate spurious artifacts in the decay portion of the driver's response to this sudden transient.
As a brief illustration of our measurement technique, here's a quick example. We were evaluating a loudspeaker system, and we heard a tonal emphasis around 2000 Hz that also had a lingering sound like a quack, and which also had a material coloration that sounded rubbery. But the conventional frequency response measurement showed nothing wrong. So we did a near field microphone probing of different portions of the suspect driver. Still nothing major showed up as measuring wrong. Were we imagining things? But then, how could this coloration have such a consistent sound on all music, with three different specifically identifiable sonic characteristics (tonal hump at 2000 Hz, lingering quack, and rubbery material coloration), if it didn't really exist? Since we were hearing this sound as a lingering quack, we followed this clue on our detective hunt, and decided to add time windowing to our near field probing, looking only at the time slice lingering after the main driving transient had already effectively fallen silent. Bingo! Suddenly, there was a huge hump at 2000 Hz in our special frequency response measurement, where the conventional frequency response measurements had shown only flat response (and by implication only accurately good driver behavior).
This hump was coming only from a small portion of the driver's radiating area, its surround, and not from its cone or dust cap, so it would not have been visible in the conventional 1 meter distance measurement that spatially averages all portions of the driver's radiating area. But it sure as heck was obvious and obnoxiously irritating to the hearing abilities of the human ear/brain, even though it was only coming from a small portion of the driver's radiating area. And the fact that this misbehavior was lingering long after the initial transient made it even more bothersome to a human listener, even though that lingering energy would have been hidden in a conventional time averaged frequency response measurement. Also, the fact that this misbehavior had that rubbery material coloration made it stick out even more like a sore thumb to a human listener, even though a conventional measurement can't hear or tell the difference between spurious rubbery noise energy and pistonically accurate sound from the main cone. If an acoustic guitar or piano or violin is reproduced with a rubbery sound, you hear it instantly, since that's a foreign material coloration, but a measuring microphone and conventional frequency response test can't tell the difference between rubbery sound and musical sound, so they irrelevantly lump all of the driver's sound output together.
Which reminds me. In this example, our special measuring technique had successfully isolated the misbehavior we heard, showing that there was indeed a tonal hump at 2000 Hz, and that it was indeed lingering after the initial driving transient, and that it was coming from just the surround of the driver (not from its main cone or dust cap). But what about that rubbery quality we heard? Well, the driver's cone was made of plastic, the dust cap was made of cloth, but what do think the surround was made of? You guessed it, rubber!
You can perform a simplified version of this test yourself, utilizing just your finger and your ear, on woofer and midrange cone drivers (don't try this on fragile tweeter domes). Gently tap each different portion of a driver, and listen closely to the sound with your ear up close. If it sounds just like a quick tick, that portion of the driver is probably performing pretty well. But if the sound has any identifiable pitch, then that portion of the driver has a ringing overhang and tonal coloration at a certain frequency or frequencies. And, if the sound also sounds rubbery, plasticky, metallic, or papery, then that portion of the driver is also imposing foreign material colorations upon all the music it handles.
M6 Driver Problems
A loudspeaker driver relies on motion and vibration to do its job, so all the physical materials of the driver's radiating area are subject to the stresses of motion and vibration, and at some point of stress all physical materials begin to spuriously misbehave. Thus, this is a pervasive problem that confronts all driver design engineers. And it very difficult to damp, control, or quell these misbehaviors to which all physical materials fall victim. A good driver design engineer expends a great deal of effort to identify, analyze, and cure these spurious misbehaviors, hopefully without too severely compromising other desirable aspects of driver performance. But this R&D effort costs money, and the clever solutions developed (exotic laminated cones, etc.) cost yet more money to implement and manufacture.
There are a number of very good drivers available, in which the spurious misbehavior
(Continued on page 108)