A MONITORING APPROACH TO LOUDSPEAKER DESIGN (part 1)
A MONITORING APPROACH TO LOUDSPEAKER DESIGN
Design Criteria and Performance Parameters-Description of recent Anglo American Research
BY JOHN WRIGHT
It is ironic that whilst everybody knows what a loudspeaker is, nobody seems clear what it is expected to do. Thus there are speakers advocated for use in stance applications and others only suited to mono. There are efficient speakers and inefficient ones, direct radiators and omni-directional speakers and there is the elusive; 'monitoring loudspeaker'. The technical journals continually advise readers to experiment with different speakers in their homes before purchasing so as to ensure compatability to the listening environment and ancillary equipment, and at least one reviewer has used the `one man's meat is another man's poison' analogy when referring to a choice between very high quality speakers. With such divergence of opinion over performance parameters it is therefore essential to determine just what the loudspeaker is intended to do before any approach can be made as to its design.
Broadly speaking, apart from sound reinforcement systems, loudspeakers can be divided into two basic categories. There are those that are intended to give a pleasant, even if strictly incorrect, rendering of whatever signals are fed to them thus confroming with local current taste and become an extension of the artistic aspect of the aural chain. For our purposes here we will label these `commercial' loudspeakers. At the other extra there are those loudspeakers which attempt to provide an exact reproduction of the programme material fed into them, whether or not that is at all times acceptable, and this we shall call the `monitoring' approach. A wide number of `other' loudspeakers try to encompass various aspects of both approaches and thereby reorder themselves most vunerable to controversy.
Commercial loudspeakers are in themselves honest to their purpose. The individual must decide for himself whether more pleasure is to be gained from listening to a speaker which is less discriminative of quality of the programme material but provides an adequate if coloured reproduction of all signals, or whether he will suffer the frustrations of the imperfect rendition from a lower that seeks perfection. To adopt a monitoring approach is to set the design task of producing a loudspeaker that will a accurately translate whatever signal le applied to it without undue addition or subtraction, presenting the listener with a consistent aural image irrespective of the listening environment. There is little doubt as to the choice that readers here would make where honesty is allied to fidelity, but on being subjected to some monitoring attempts many members of the general public may, justifiably, disagree. Removing the main coloration and boxiness usually reveals all the distortions! Thus a monitoring approach is a hard taskmaster, near-perfect demanding signals to demonstrate its superiority over the commercial approach to which all have become conditioned. It is therefore understandable why so many `hi-fi' speakers have been designed with split personalities employing varying proportions of both approaches in an attempt to assure (financial!) success.
Whilst it is possible to derive an approach to monitor speakers in general from these conjectural observations, the field of application is still too wide to provide an explicit design brief. Since at the present state of the art no one monitoring loudspeaker can satisfy all requirements, it is necessary to establish a series of priorities for performance parameters allied to the specific application. It is relatively easy to trade bass performance for efficiency with a given enclosure, indeed it can be shown that for any given dimensions the efficiency of the system is inherently dependent upon the loudspeaker's low end extension. The required efficiency is by the listening environment, the available amplifier power and the subjective level at which it is necessary to monitor. Loudspeakers for replaying 'pop' material at levels around the threshold of pain can hardly be expected to conform to the monitoring standards defined later on if they are to be of any reasonable size and efficiency so must fall somewhere between the commercial and sound reinforcement categories. Thus we will concern ourselves solely with monitoring loudspeakers for high fidelity applications, where it is understood that the audiophile may demand a superior performance to that of the professional to whom consistency and reliability are of prime importance.
From the foregoing, two basic design criteria for a high fidelity monitoring loudspeaker can be suggested:
(1) The enclosure shall be of reasonable size and the operating principle shall be such that the frequency range may extend from about 20 Hz with the system having sufficient efficiency to reproduce the full dynamic range of the programme material fed into it, at a level sufficient to overcome typical ambient noise levels when driven by available amplifiers.
(2) The monitor must form a plane source propagator projecting the total integrated information directly to the listener and this characteristic shall be maintained under a wide range of acoustic environments, the speaker conveying very little of itself (coloration) with minimal effects due to the room and the maximum of the programme fed into it.
These basic requirements seem to elude the majority of speakers currently available and the main reason would appear to be that taste tends to favour small enclosure sizes. The monitoring approach precludes the use of onmi-directional or reflecting systems since, whilst perfectly valid in commercial contexts, results are generally influenced by the environmental location in a comparatively unpredictable manner while the phase and ambient relationships of the original programme material are modified. The possible virtues of electrostatic drivers are also unsuitable for the monitor due to problems of producing high sound levels, the doublet radiation pattern if unbaffled, and their peculiar impedance which makes than difficult to drive. Unglamorous as it might seem, the direct radiator moving coil loudspeaker still provides the most valid and practical approach to monitor.
This is not, however, to suggest that design work can commence by throwing a few paper cone drivers into a box! Indeed, in this age of plastic technology to choose conventional paper cones with their lack of uniformity, hygroscopic nature and tendency to change with age, would be unthinkable, A number of chemically manufactured diaphragm-drivers are currently available in various forms and even the small amount of research undertaken with these demonstrates their inherent superiority to most units available a few years ago. The advent of such drivers has revealed the fact that instant research had been undertaken into enclosure design, probably because such work would not have previously realised a significant subjective improvement.
The first question that arises concerns how many drivers should be employed to cover the range. Unquestionably a single loudspeaker has considerable attractions with its absence of electrical crossover problems, but to date no unit has been devised that is capable (without full-sized horn loading!) of meeting the stringent requirements of this application particularly in terms of efficiency, extended range and very low Doppler intermodulation distortion. The problem is akin to the design of a motorcar engine so flexible in its performance that it does not require a gear-box. Two units can approach the requirements if crossover is undertaken around the mid-band, but extreme difficulty is encountered in obtaining a truly homogeneous transition and the design of a bass unit capable of smooth extended response to above mid-band frequencies presents formidable problems, particularly in production. Furthermore, no high-frequency unit has been encountered that will both reproduce from mid-to supersonic frequencies accurately and simultaneously handle the power that may have to be delivered to it. Thus by far a more practical approach is to introduce a separate mid-range unit which furthermore places the crossover points away from the sensitive maid-band where the ear is most critical. Whilst removing one set of problems this does, however, simultaneously place enormous demands upon the mid-range unit employed.
Even with the use of three drivers, problems are encountered in providing a tweeter unit cable of really extended range and adequate power handling. Adding a second identical unit to incise power handling presents phasing problems due to the physical spacing between the units, and since there is considerable evidence that extended response well beyond audibility is necessary to obtain entirely convincing high-frequency reproduction, it is advantageous to introduce a fourth unit specifically designed for the super-high frequencies and capable of large power handling. These greatly improve transient performance and dispersion, being of ultra low mass chemical dome design, and the addition is most worthwhile if balanced by genuine low-frequency extension.
The possibility of employing active electronic crossover networks using separate amplifiers for each driver must be discounted on the practical grounds that it does not comply with the brief in that the monitor should be with amplifiers readily available. Experiments carried out for interest with this approach did, however, indicate that one set of problems was replaced with another for it was more difficult to achieve the required phase angle changes to compensate for the physical spacing of the units. Subjective evaluation of similarly measured results of both electronic and more conventional L-C crossovers indicated that steep rates of slope were less manageable than gradual ones if overall homegeneity was to-be maintained, although the latter do demand the use of drivers substantially smooth well beyond the nominal crossover frequencies, where a steep rate of slope was dictated it was less noticeable on the high-pass, side of the filter than on the low. It proved surprisingly difficult to predict the resultant phase relationships during crossover as component values, exact dividing frequencies and the final juxtaposition of drivers on the baffle all have a significant effect. The effect of mis-phased relationships is shown in fig. 1, where it will be noticed that it becomes more noticeable at the off axis positions causing a `suckout' and, restricted dispersion.
It was therefore necessary to take time and care to determine the optimum crossover configuration and positioning of the units relative to each other on the surface of the baffle. Furthermore, it was found necessary to mount the units in a recessed manner, so that the diaphragms we flush with the panel, if proper phase relationships were to be maintained with a minimum of diffraction. Indeed, due to the size of bass unite, optimum in-phase relationships could only be entirely maintained in one plane of rotation from the on-axis position and this observation, tether with an attempt to minimise diffraction 'steps', and employ the effectively reduced frontal radiating area to increase inward dispersion, led to the in line configuration, with mid- and high-frequency units being placed on opposite sides of the bass unit for stereo pairs.
The 'mirror image' configuration allows the speakers to be situated with the mid- and high. frequency units positioned towards the inside and away from roots boundaries so that the maximum of direct and the minimum of reflected information reaches the listener. For reasons that are not yet fully explained this strategy greatly enhances stereo imagery, simultaneously rendering additional freedom of seating arrangements, suggesting the ear makes considerable use of phase-relationships rather than mere intensity differences. The phenomenon is most dramatically illustrated by reversing the speakers so that the mid- and high-banks are towards the outside, when even with the systems well away from reflective surfaces image location suffers. The degree to which this effect is apparent is dependent upon the recording technique employed for the programme material being reproduced-so these remarks are perhaps particularly appropriate to the monitoring approach.
At various times criticism has been levelled at the use of reversible electrolytic capacitors and ferrous-core inductors in crossover networks. Electrolytic capacitors have been said to became conductive during tunes of operation causing 'transient overload', whilst coned inductors have been said to 'ring'. Despite the fact that at the lower impedances and crossover currently frequencies employed, paper capacitors and air-cored coils would be completely impractical in terms of coat and space, it was decided to conduct some research into the matter to determine what, if any, subjective benefits were to be gained. Using a multitude of ex-Government block paper capacitors and chunky, air-cored hand-wound coils a switching arrangement was made up whereby a direct comparison, between the test sample and a crossover entirely comprised of electrolytics and ferrous-core inductors, could be made. The same drive units and enclosure were employed with adjustment for differing insertion losses but, that apart, the networks were as identical as possible and, not surprisingly, the overall balance remained unaltered. There was, however, a marked difference in the 'attack' and 'cleanliness' of the sound particularly at very high and very low listening levels where transient peaks were encountered, and in every case the more 'inconvenient' network was preferred. This led to examination of the toneburst response and distortion components at various levels and frequencies. Here the differences observed could have been explained by errors in the test procedure or tolerance differences of the components used, or even a subtly different load presented to the amplifier. Again, we appear to have encountered a situation where the car is a more reliable tool than the equipment available or knowledge of how to effectively employ it!
As the intention behind the project was to develop a loudspeaker where performance, rather than price or an operating principle, was the primary criterion of merit, the choice presented itself between adopting the superior network and increasing enclosure size to accommodate it, crossing over at a higher frequency to make it more manageable, or continuing research to discover what factors precisely were contributing to the difference, and then eliminating their effects. Since a higher crossover had previously been discounted for reasons mentioned, the latter course was adopted. From the work it was found that if the tuned circuits employing ferrous-core inductors were of comparatively low Q, their substitution for the air-cored brought a negligible change which could not be detected when employed with electrolytic capacitors. Secondly it was found that by far the most critical capacitors were those feeding the tweeter and super-tweeter, which did require effective 'blocking off' on transient pulses. Comparison between the air cored coil and paper capacitor network, and a crossover comprising both types of capacitor, according to their location, and lower Q ferrous core inductor circuits was entirely favourable, the differences being rather less than could be expected from sets of components subject to normal manufacturing tolerances. Furthermore, when differences were noted, opinion varied as to which network was preferable, and this was partly (for reasons too involved to discuss here) because the components adopted presented certain other advantages if used in the fairly conventional configuration shown in fig. 2. The resistors bridging the mid- and high-frequency banks have the dual function of maintaining proper impedance relationships and also function as damping pads reducing minor resonances in the system.
It is such minor resonances that produce 'coloration' in a loudspeaker and to bring down coloration to acceptable limits seems to be the one most difficult aspect of design. In conventional systems the 'something added' can result not only from unwanted resonances in the drivers, but also from that of the enclosure. Most electrostatics at least have the advantage that they have no boxes to cause 'boxiness' and it will be found that if contemporary high quality dynamic drivers are mounted on a large open baffle they, too, are surprisingly free from coloration. Unfortunately this type of mounting is unsatisfactory in most other respects, but as an experimental tool it does provide a useful reference by which to judge added coloration from the enclosure under development. Such research reveals that the major cause of so-called cabinet coloration does not comes from the enclosure itself but is derived from internal reflections returning through the cone. Thus it can be initially stated that if coloration is to be held within reasonable limits, small or shallow enclosures should be avoided. Wherever possible the rear wave should be completely absorbed within the enclosure reducing any chance of direct internal reflection-even from damping materials. Conventional frequency response runs are of little use in assessing the success of experiments of this type as the microphone cannot distinguish the wanted from the unwanted signal and merely reads total energy output. Low level tone bursts are of use in the initial stages if studied with great care but even then the ear is able to perceive changes too small to be accurately displayed. A useful factor in evaluation here is that removing such unwanted return waves lowers the subjective efficiency of the system.
All this is not intended to underestimate the importance of reducing enclosure resonance itself to a minimum, although this sort of coloration is most pernicious at a few hundred Hertz; sealed boxes develop very high internal pressures and are most difficult to control in this respect. Whilst it has been found that natural woods resonate in a more 'pure' manner, than synthetic materials and therefore the coloration can be less objectionable, the advent of very high density building boards can, when well braced, allow less overall coloration in this respect. Stress laminating such an enclosure with Formica was found to further brace and damp the system to an extent that it was found to be sufficiently inert.
It has already been stated that the monitor should give a consistent account of itself irrespective of three listening environment and a method of avoiding undue reflections from room boundaries has been described. However, it is commonly believed that the listening room has a profound effect upon the apparent response of a speaker and there is probably much truth in this when dispersed radiation patterns are involved. Much work has been undertaken in the U.S.A. with representatives of the Frazier Corporation into the evaluation of loudspeaker responses in room situations. Frazier market a semi-professional equaliser which is inserted is the amplification chain for the purpose of 'flattening' the general response of any loudspeaker within it's intended location. Before the equaliser can be set accurately over each third octave it is necessary to take an energy reading of the loudspeakers’ response in the room and this is achieved by injecting high-level pink noise and plotting the one-tenth octave response, via filters, from the calibrated microphone. From this work there is a strong indicates that the room has far less effect upon response than might have been anticipated in respect to direct radiator systems for the same basic shape of the response curve occurs time and Again even with the loudspeaker in quite diverse surroundings. Fig. 3 shows an example of this.
Furthermore, the position of the speaker in the room has little effect except upon the response in the low hundreds of Hertz region which confirms the observations of Harwood of the B.B.C. Research Department where he shows (Wireless World April 1970) that the corner may be the worst position to place a direct radiator loudspeaker. When such experiments were carried out with the Frazier team using loudspeakers with known qualities, the mono plots showed a general similarity to those taken anechoically, the effect of rooms being rather to more or less exaggerate the known uneven of the speaker's response. A most significant observation throughout this work was the inability of most loudspeakers to produce any significant bass energy much below 100 Hz and many speakers, praised by some for their bass, showed no more than an `upper bass' resonance. Fig. 4 shows typical room responses of a number of known speakers in differing types of location, where the typical bass fall-off we are accustomed to seeing under anechoic conditions is exhibited. It would appear that loudspeakers that do not have a good response in suitable anechoic chambers do not propagate much bass in rooms either!
The problem of generating bass smoothly down to around 20 Hz without tying on the nebulous `gain' of a Corner position, or the introduction of excessive Doppler intermodulation, will be discussed in Part. Two next month where further details of a completed design will be described together with the methods employed for measuring and evaluating its performance.