Introduction
The term evolutionary has been used with deliberation, for much of what follows, is in the individual cases not revolutionary or even new. However, in the existing loudspeaker industry, it is not unusual to find such a great commitment to a certain method of design, or production, (in part excusable by a concomitant to commitment to production tooling) that it is quite difficult to make any major breakthroughs that are practical. While solving one problem in a driver often reveals yet other problems, the solution often is important in that it increases the diversity of ways in which a re-designed driver may be employed. - Even if this is not realized by the design engineer involved.
Evolution is based on the appreciation of the possible in existing designs. Successful evolution is based on the appreciation of a very large number of optional possibilities, together with the knowledge needed to combine the best options available. And there is no denying that the desire to combine these elements is often thwarted by technical problems that they need to be solved.
While the problems and their solution are a science; the selection and combination of design elements is often an art; and if there is such a thing as craftsmanship in design, the combination of science and art together with a integrity in execution must be that quality.
What follows are a
brief summation of the qualities and elements used in the design of the
Watson Laboratories Loudspeaker systems. It is not so much an attempt
to convince the reader of the validity of our approach, as an "a posteriori"
presentation of information which we hope will enable the people who have
heard and who own these systems to understand how it was that the self-evident
results were achieved.
Loudspeaker baffling and coloration - the design of an integrated speaker system
Remarks have and will be made on the number of different techniques which have been used in the design of the Watson Laboratories speaker systems, and that considering these techniques, that it could have been assumed that difficulties would have been encountered in obtaining a cohesive, unified sound.
This is what makes the assembly of a series of drivers, as much of an art as a science. In the critical mid-range, for example, the speakers are operated as pure dipoles not only to help "reconcile" the speaker to the listening room and the demands of stereo imaging, but because this method of baffling produced the least system-intrusive coloration. It was possible to have very fast movement of the low mid-range speaker cone as necessary for piano and drum transients without problems with enclosure reflections; room reflections fortunately are delayed enough to avoid masking effects.
The woofer posed a major problem. While it has been demonstrated that a large dipole can generate excellent low-frequency signals, it is rather dependent upon both orientation of the "piston" surface to the rooms' major axis, and the size of the room. For this reason a pressure coupled system was chosen; and this done, the design problem became one of confining or controlling the back wave of the speaker in such a way as to produce the minimal compliance distortion, obtain the greatest frequency linearity, and to obtain satisfactory efficiency. It was felt that based on a series of in-house tests, that the frequency response had to be extended well down below 20 Hz with a smooth and moderate-rate roll off, free from any peaks which could lead to acoustic feedback problems in turntables. For this reason a first order enclosure was chosen, together with a refinement of the "Gas-linearized Compliance" system. The use of 10" speakers allowed the cone motion to be controlled so as to cause no crossover hiatus in coloration between the woofer system and the low mid-range. The cross-over frequency was chosen to minimize the need for extended range response on the woofer; at 600 Hz we wanted its output to be at least 18 dB down on the crossover, and about 10-15 dB down as a "radiator".
The baffle dimensions
on both the mid-range and lower mid-range units were chosen to cause a
smooth roll off of the units effective coupling at low frequencies, the
use of a 18 dB/octave network both assisted in obtaining the maximal pulse
response and stopped any low frequency components from requiring excursions
of the "free cone" drivers such that they would be operating in a non-linear
range of the voice-coil assembly.
By employing a soft dome tweeter good dispersion could be combined
with very low distortion characteristics. Special constant-amplitude
phase correction networks were used to smooth the crossover region.
The crossovers are 3rd order Bessel Function networks for best pulse response; air cone inductors are used except in the woofer where the advantages of a large-core-section alloy core in providing minimal inductor resistance far outweighed to very slight inductor non-linearity experienced at or near the absolute limit of the power rating of the speaker system. The antennuator networks used were placed at the input of the respective filter sections as this proved to provide a more stable crossover performance.
The cabinet design
is unusual, but it places the greatest mass of the enclosure at the lowest
and least obtrusive point, while insuring that the back of the upper section
cannot be placed too close to a rear wall. The interchangeability
of trim finishes will ensure that should any cabinet woodwork be damaged,
it can be replaced or repaired at minimum cost either in speaker down time
or in dollars. In addition it allows the dealer to offer more options
on trim without staggering stocking costs.
We feel that the result is a system which will combine reasonable size
with long life and reliability, in the process achieving a realism and
transparency which up to now has been more the subject of speculation than
actual fact.
The gas-linearized-compliance woofer system
Developed by William Wright in 1969, this system is a modification of the "Acoustic Suspension" system used in small cabinet volume low frequency speaker systems, eliminating or modifying some of the restrictions imposed by conventional "acoustic suspension" designs.
Air when used as a spring suffers from the disadvantage of being non-linear. When the air is compressed, its temperature rises, this heat is radiated and in part lost to the system; such that when the air expands again, and cools, it no longer occupies its former volume until it warms up to the original temperature. This is termed adiabatic. While the hystresis effect is small, it is significant in that it contributes a pronounced non-linearity to the system. The normal method of controlling this non-linearity is to employ cubes of fiberglass inside the enclosure. These are not there as most people think, to absorb the rear wave of the speaker, but to act as a large-surface-area-heat sink such that the heat given off by the compressed air is held in the glass fibers and is readily available to re-heat the air during the expansion cycle. This technique somewhat linearises the use of air as a spring. This is an attempt to make the “spring” in the air act as if it was isothermal – that is, the surface of the fiberglass act as a “heat sink” and does not require the air itself to store or release heat!
However a much better alternative is the employment of a gas or gas mixture, within the enclosure, where the gas has a high specific heat. This allows the gas to store a larger amount of heat, less is lost during the compression cycle, and thus during expansion, there is less hystresis error. By the employment of such a gas, non-linerities can be substantially reduced over those present when air is used.
An added bonus is the major reduction in the self-resonant frequency of the speaker enclosure. Because the speed of sound in the gas is much lower than the speed of sound in air, the virtual dimension of the enclosure, expressed in wavelengths, is much larger.
There are several different configurations, which can be employed to contain the gas used. The whole enclosure, including the speaker cone and suspension means may be made gas tight, with attendant problems caused by variations both in atmospheric pressure on the surface, and during air shipment. These can be reduced somewhat by use of a small plastic bag (of perhaps 15 to 30% of the enclosure volume) which is air filled and connected to the outside air. When the gas pressure rises, this bag will collapse. However it must be remembered that during the un-pressurized air shipment the atmospheric pressure may be reduced by over 50%, thus the bag should occupy at least 50% of the enclosure volume if it is to provide full shipment protection.
A somewhat easier system to employ in manufacture is to confine the "ideal gas" to smaller individual plastic bags which occupy only about 50 to 70% of the available enclosure volume. The balance of the enclosure volume is filled with either fiberglass or long fiber wool; and a high density packing which can be compressed only by the expansion of the gas bags during air shipment. As several foam materials are available with very high-density skins, a very effective compromise can be reached between shipment protection, and distortion reduction.
A further advantage of the use of individual plastic bags (which themselves are never filled to occupy more than 50% of their expanded capacity), is that they randomize the acoustic path lengths within the enclosure such that the rectangular enclosure is no longer an acoustic parallelepiped. This is very effective in reducing the "Q" of the enclosure resonance’s.
In the recent past, several companies have resorted to the use of motional feedback in order to reduce the distortion present in that area of the low-frequency speaker's operation where its cone movement is primarily under control of the suspension and enclosure gas compliance. The use of a "Gas-Linearized Compliance" system obviates the need for such an expensive solution. The coupling efficiency of a piston can be increased through the use of a larger piston area; as the use of a dense gas increased the apparent area of a piston by from 9 to 16 times it is not difficult to obtain excellent coupling between a cone speaker and a dense gas lying in front of it; and if this gas in turn is bounded by a sealing diaphragm having a surface area some 9 to 16 times the area of the cone speaker employed as a driver, the increase in effective coupling will be carried out through the whole system. When properly employed this allows the simulation of acoustic transformers such as horns without the non-linearity’s of these devices.
Development work is
underway at Wright Electroacoustics and Watson Laboratories on employment
of the latter techniques for other than low frequency speaker systems as
well.
Reducing distortion in the critical mid-range
At one time, much of the distortion present in mid-range speakers was caused by mass-movement-anomalies in the cone structure. The cone was said to "cymbal" with different areas moving contrary to the motion of the voice coil. There were two schools of thought on these phenomena. Some designers thought it inevitable, and provided concentric rows of corrugations around the cone's surface. This, so that ideally the area of the moving part of the cone could be progressively reduced with rising frequency. Others attempted to make the cone more rigid; there was even a compromise school which divided the cone's surface into two or more well defined concentric areas with viscous damped compliance folds between each area.
With the emergence
of materials such as Bexedrene and Kevlar* it became possible to mould
very light cone structures which with the application of a minimum of damping
material, could be safely said to move as a piston even at their highest
rated frequency. But during this time, very little attention was
paid to the non-linearity’s imposed on the cone by the compliance.
Conventional thinking was that the resonant frequency of a mid-range
driver simply imposed a lower limit on its useful range. If you wanted
to use a driver from 400 Hz upward, then the resonance could be at or slightly
lower than 400 Hz. But what was overlooked was the coloration imposed
by the resonance at frequencies at even several times the resonant frequency.
Because Compliance-effect-degradation starts to occur well above resonance
in a driver it is ideally advisable to have the lower operating range of
the driver stop octaves above its resonance; especially where 6 dB/octave
networks are used.
In tests we found that a three to four octave separation between the resonant frequency-together with 18 dB/octave roll-off in the crossover resulted in very low coloration’s and in an increase in the elusive "transparency" effect in the speaker system.
As a result, the driver's
employed in the Model 10 and Model 7 speaker's mid-range and lower mid-range
operate well above their resonant frequencies (4 octaves and 3 octaves
respectively).
Phase coherency in a loudspeaker system
It has become popular to emphasize the phase coherency or the "time-alignment" of a multiple driver speaker system at a time when most psychoacoustical experts are divided about the ability of the listener to hear small phase differences. Part of the confusion is, we suspect, because of the difference between steady state tests and transient tests; which is analogous to the situation only now resolving itself in the design of amplifiers to purely steady state standards rather than to both steady state and transient standards.
"Phase Coherency" or
"Time-Alignment" of a speaker system isn't simply the vertical alignment
of either the different driver's voice coils or their radiating centers;
although it has an element of both. Once the crossover network has
been designed such that time-alignment is theoretically possible, better
results in terms of listening tests can often be obtained by aligning the
different drivers to obtain the best replica of unit step or pulse, rise
time and maintenance of the step being the criteria; rather than alignment
to produce a simple steady-state phase coherency. Because the characteristics
of most drivers are not identical under transient and steady state operation
it is not unusual to find major apparent positional discrepancies between
the two types of operation. In this condition we have chosen to use
the transient criteria as our basis of design.
Harmonic distortion and speaker acceptability
It comes as a considerable surprise to most audiophiles to find out just how much harmonic distortion is present in even the best loudspeaker designs. They understandably find this difficult to reconcile with the emphasis placed on low distortion in a preamplifier or power amplifier.
However too little has been said in the literature relating high-order harmonics distortion to unacceptability of sound and it is these "dissonant" harmonics which present in very minute quantities cause the most distress to the listener. Because of the signal bandwidth and the presence of discontinuities in the transfer functions of most power amplifiers, they produce large amounts of high order distortion components. When combined with the Transient Intermodulation Distortion that is excaberated by improper feedback design, which produces even larger amounts of high order distortion products, the result is not pleasant. Electronic equipment is more prone to High Order Harmonic Distortion product masking and coloration than loudspeakers.
One noted manufacturer marketed a line of equipment for several years, in which the system bandwidth was deliberately reduced in order to reduce the problem. In effect the "mushy" overload characteristics of tubes (which does reduce the High Order products produced on overload) is the thing that has lead many listeners to prefer tube equipment.
In designing the Watson
Laboratories speaker systems, every effort was made to reduce the harmonic
distortion present to 2nd and 3rd harmonics, any high order products were
considered to be unacceptable.
*Dupont's Kevlar ™