A flat response
Aug 1, 1999 12:00 PM,
The loudspeaker, as we know it, has been around for some 70 years. In thattime, it has undergone a myriad of refinements. Today, however, we havepretty much the best that can be achieved. That is not to say thatimprovements are still not being made, particularly as new computer-basedmeasurement and mathematical modeling tools also develop. These techniquesare allowing new insights into loudspeaker behavior, both in terms of theacoustic radiation itself and the mechanical/air load interactions at theradiating surface using, for example, finite element analysis. Thistechnique also allows other underlying aspects of loudspeaker design to beinvestigated, such as the complex magnetic field right inside the polepieces and motor assembly, parameters that cannot be measured, but can nowbe visualized with powerful computer analysis. Advanced FFT techniques,correlation measures, time windowing, acoustic intensity, interferencespectrometry and laser scanning, to name but a few of the newer measurementtechniques, are enabling corresponding and unprecedented advances to bemade in the complexity and innovation of the measured data. Despite this,there have been few new major breakthroughs or radical changes to the waywe that think about or use loudspeakers for a considerable number of years.This is about to change.
NXT (New Transducers) was established just three years ago to exploit anovel discovery made in noise control materials technology by a UKgovernment research agency, and the distributed-mode loudspeaker was born.As we shall see, this new class of device has acoustic properties radicallydifferent from conventional loudspeakers and are such as to set manypreconceptions and approaches to loudspeaker system design on their head.NXT scientists are involved in a huge research program not only into thebasic mechanics and acoustics of distributed-mode technology, but also intothe corresponding psychoacoustic and measurement aspects. In the past twoyears, more than 20 papers have been presented at Audio Engineering SocietyConventions and US/UK acoustic conferences on these topics, indicating thepotential significance of this new technology (see listing in referencesection). NXT itself, however, does not make loudspeakers; it operatespurely as a research, development and licensing company. Distributed-modeloudspeaker products or loudspeaker modules will instead be made by thegrowing worldwide network of 100 plus companies who have already taken outmanufacturing and user licenses.
The fundamental difference between distributed-mode loudspeakers andconventional cones or pistonic devices relates to the way in which sound isgenerated and radiated. The distributed-mode loudspeaker essentiallyconsists of a thin, stiff panel that is set into vibration by means of aspecial electro-acoustic exciter. The exciter is normally a moving coildevice, but piezo-electric and other forms of excitation can be usedequally. The panel is excited by an exciter (or multiple exciters)carefully positioned and designed to excite the natural resonant modalstructure of the panel optimally. Although it would seem counter-intuitiveand in direct opposition to the cumulative wisdom of 70 years ofloudspeaker design to design and excite a resonant structure, whenappropriately effected in a controlled manner, it can lead to someinteresting effects.
The point is that unlike air, the panel material is dispersive, and a densemodal structure soon builds up. Studying the panel vibration by means of ascanning laser and time windowing the measurement shows the modal vibrationto become rapidly complex. The panel can be considered as a pseudo-randomlyvibrating object, for at any given instant, different areas are excitedwith different amplitudes and phases. Figure 1 illustrates the point. Thepanel can be thought of as a whole series of individual radiators, eachradiating sound effectively independently of its neighbor but summing inthe far field to give the desired response. The greater the modal density,the greater the number of these radiators and the greater the random natureof the vibration and the greater lack of correlation between them. It isthis lack of correlation of the radiation over the surface that is one ofthe primary keys to the successful operation of a distributed-modeloudspeaker. In a conventional cone or pistonic device, the resonant modeswould lead to coloration and other detrimental effects, the completereverse is the case with the distributed-mode loudspeaker. This is becauseunlike a pistonic cone loudspeaker, where the objective is to move oraccelerate the complete radiating surface as a whole, leading to coherent,phase-related radiation across the entire surface, the distributed-modeloudspeaker is the complete reverse of this. Different parts of adistributed-mode loudspeaker panel radiate at different times and are notdirectly correlated with each other, thus creating a diffuse radiationcharacteristic. Because the resultant wavefront is not phase coherent, itwill not produce the strong coloration effects associated with resonancesin conventional loudspeakers, nor the local boundary specular reflection effects. Furthermore, the surface motional vibration of the distributed-mode loudspeaker is orders of magnitude lower (micron scale) than a pistonic conedriver because it does not act like a piston in order to move the air andthus, radiate sound.
This matching to the air load impedance is also different. In an ideal DML,the radiation resistance is insignificant and is constant with frequency.As a result, the diaphragm dimensions no longer control directivity. Thismeans that you can make the radiating area as large as you like without thehigh-frequency output becoming confined to a narrow solid angle about theforward axis, as is the case with conventional drivers. Figure 2illustrates the point and presents the polar diagrams for a conventional 4inch to 6 inch (102 mm to 152 mm) cone loudspeaker and an NXTdistributed-mode loudspeaker.
In contrast, by acting as a piston, the diaphragm of a conventional drivermoves as a rigid whole, or, at least, that is the designer’s aim. Inacoustic terms, such a loudspeaker is mass-controlled over most of itspassband. For a given input voltage, the motor generates a force that isconstant with frequency, and the diaphragm resists with a mass (its ownmoving mass plus that of the air load). By Newton’s second law of motion (F= ma), the acceleration of the diaphragm is constant with frequency. As aconsequence, its displacement decreases as the signal frequency rises at arate of 12 dB/octave.
At low frequencies, where the wavelength in air is large compared to thedimensions of the diaphragm, this is as desired. The real part of thediaphragm’s radiation resistance, into which the driver dissipates acousticpower, increases with frequency at exactly the same rate as the diaphragm’sdisplacement decreases with the result that the acoustic power output isconstant. As the frequency rises and the wavelength in air decreases to thepoint where it becomes comparable in dimension to the size of thediaphragm, a significant change occurs. Instead of continuing to rise, thereal part of the radiation resistance reaches a limiting value andessentially becomes constant for all higher frequencies. Consequently, thediaphragm’s acoustic power output begins to fall at 12 dB/octave. This doesnot necessarily mean that the on-axis pressure response drops; in practice,the diaphragm’s acoustic output becomes restricted to progressivelynarrower solid angles. The radiation becomes directional, and theloudspeaker begins to beam.
Variation of directivity with frequency is one of the major problems ofloudspeaker and sound system design. Whereas the on-axis response of agiven loudspeaker may well be flat, the frequency dependent directivity andsubsequent off-axis response will not be, and so the direct, earlyreflected and reverberant sound fields in a room or space will all havedifferent tonal balances. In hi-fi and multimedia systems, this can affectnot only the overall sound quality and perceived coloration, but also thestereo imaging. In larger spaces employing sound reinforcement or publicaddress systems, this inherent characteristic can also affect perceivedspeech clarity and intelligibility. Consequently, most loudspeaker systemsuse multiple drive units (two- and three-way systems) of progressivelydecreasing diaphragm size in order to maintain a wide coverage angle. This,however, still generally means that the loudspeaker’s directivity variessignificantly with frequency.
Although the acoustic behavior of an NXT panel appears random, the designprocess is, in fact, completely deterministic. The key parameters thataffect a panel’s performance are the size and shape of the panel (which maybe curved), the position of the exciters, the bending stiffness, thesurface density and the internal damping. Provided that these are known, itis then possible to predict the acoustic performance of a panel with a highdegree of accuracy.
Distributed-mode loudspeaker panels using moving coil exciters offerremarkably benign amp loads being essentially resistive at low and midfrequencies as shown in Figure 3. As the frequency rises, the inductance ofthe voice coil becomes significant and the impedance increases as the loadbecomes reactive. Nowhere is there a low impedance and large phase angle incombination.
Unlike conventional diaphragms, where the moving mass determines the upperlimit of the frequency response, with distributed-mode loudspeaker panelsthere is no equivalent restriction. As a result, the technology is trulyscalable. You can make a large panel without impairing dispersion orhigh-frequency response. In fact, the performance actually improves as thepanel size increases because the frequency of the fundamental bending waveis lowered, which not only extends the bass response, but also increasesthe modal density at mid and high frequencies.
That the loudspeaker is operating entirely in resonance seems entirely atodds with conventional loudspeaker design practices in which it is wellknown that resonances are to be avoided at all costs. Provided that theyare correctly designed, distributed-mode loudspeakers do not sound coloredas might be expected. This is due to their complex radiation and itsdecorrelated nature. Distributed-mode panels also exhibit an unusualimpulse response that displays a fast rise time or initial transient, butincorporates a decaying resonant tail (see Figure 4). This impulseresponse, however, can lead to a flat frequency response and, as alreadymentioned, a flat acoustic power output.
An interesting outcome of the unique sound-radiating properties of thedistributed-mode loudspeaker is that it can be used unbaffled, for example,as a free-standing loudspeaker or as a hanging, talking signboard becausethe acoustic power output radiated from the back sums nondestructively withthe sound from the front instead of canceling. This is attributable to thecomplexity of the distributed modal radiation and uncorrelated phase of theindividual radiating elements as seen from the far field.
The uncorrelated nature of the radiation has some other interestingeffects. First, the panels interact far less strongly with local boundariesand thereby exhibit significantly less comb filtering and coloration ascompared to conventional loudspeakers. This is illustrated in Figure 5,which is a spatial interference spectrograph measured by the author for anNXT panel and 4 inch (102 mm) cone device located in close proximity to aroom boundary. The distinct striation pattern of the inherent combfiltering of the cone device can be seen to be radically reduced for thedistributed-mode loudspeaker, both measurably and audibly.
This reduction in comb filtering and reduced interaction with reflectingsurfaces has a number of implications. For example, it means thatdistributed-mode loudspeaker can be used in close proximity to roomboundaries or reflecting surfaces with reduced effect on stereo imaging andtimbre. Furthermore, recent research has shown that the off axis and, inparticular, the rear radiation from a distributed-mode loudspeaker arehighly decorrelated with respect to the forward radiation. This suggeststhat the majority of early reflections within a typical listening room willalso be decorrelated and therefore effectively act as diffused reflections- a highly desirable goal but without the need for acoustically diffusingwall treatments. This property, coupled with the wide dispersion, resultsin a very uniform sound field within a listening room. This is partiallyillustrated in Figure 6, which shows the fall off with distance in atypical, well-damped listening room for an NXT panel and a two-wayloudspeaker. Examining the in-room impulse response, it shows thedistributed-mode panel to generate a significantly greater number of earlyreflections than even a wide-dispersion conventional two-way comparableunit. The listener, therefore, becomes far more immersed in the soundfield, which should potentially offer a greater sense of spaciousness andis often reported as an increase in perceived loudness.
Although exhibiting wide dispersion, distributed-mode loudspeakers stillappear to offer precise, stable imaging. In typical domestic conditions,the stereo sweet spot often extends further than that experienced withconventional designs primarily because the wide dispersion and thereduction of destructive boundary/room interaction effects. Thesecharacteristics, together with the slower rate of in-room sound level, falloff with distance and greater diffuse reflection density, suggest that DMpanels should be particularly well suited to multi-channel home theatresystems. In addition to the wide dispersion and low visual impact, thediffuse nature of their sound radiation ensures the required surroundchannel diffusion, so that listeners are not conscious of the surroundloudspeakers as distinct entities. Furthermore, in video projectionsystems, the screen itself can be a large-panel loudspeaker, therebyensuring perfect synchronization of sound with picture.
DM loudspeakers, although inherently bipolar, can be enclosed. Theenclosure can be shallow (approximately 2 inches or 51 mm), and so a lowprofile loudspeaker can still result. Enclosing a panel changes the bipolardirectivity pattern to wide dispersion forward radiation, but the diffusenature of the radiation is maintained. Also, by enclosing the panel in aknown way, the response can be optimized, and the unit can then be wall orboundary mounted with only minimal acoustic effect.
A further aspect of the wide dispersion of the NXT panels can be seen inFigure 7 and utilized in ceiling-based distributed sound systems. In Figure7, the fall off with lateral distance is shown and compared to a 4 inch(102 mm) cone device. The measurements were made in a test ceiling underanechoic conditions and show how the high-frequency (4 kHz) coverage ismaintained. The off-axis, fall off with distance for the panel can be seento be minimal up to the limit of the 16.4 feet (5 m) measurement set up.The measurements suggest that an even-coverage sound system can beachieved, with fewer units required even than when using relatively smalland highly dispersive 4 inch (102 mm) cones. The wide dispersion, however,does mean that for high intelligibility to be maintained, the reverberationtime of the space needs to be reasonably well controlled, as is the case inmost office and similar commercial developments.
Initial tests by the author in spaces of 1.5 to 2 seconds, also show tomatch popular ceiling tiles, projection screens, wood paneling, andpictures also open up a whole new raft of audio possibilities, even withoutthe added acoustic benefits. For further information, consult the NXT Website at www.NXTsound.com.