SPEAKER DESIGN THEORY
By: Albert Von Schweikert, Chief Design Engineer
The following is a multi-part article that highlights the design philosophies of Von Schweikert Audio. All material herein is copywritten by AVS in 1981.
“My dissatisfaction with the lack of realism in contemporary speaker design has led me on a long quest. More than thirty years ago, my experiments conducted at California Institute of Technology enabled me to discover several important psychoacoustic principles.”
The first Discovery was that the ear/brain hearing mechanism can sense differences between certain types of sound wave patterns and uses this recognition for identification and spatial localization of sound sources. For instance, an omnidirectional wave pattern consisting of spherical sound waves can be differentiated from highly directional beam waves. A computer in the brain compares data arriving at each ear and computes directional (stereophonic) localization from arrival times, frequency, phase, and amplitude responses, among other things. This data is stored for later processing, and over a sufficient learning process, becomes an acoustic reference bank. Differences in the data arriving at each ear conveys stereo information, for instance, including spatial localization and timbre recognition of previously heard tones or other sonic sources.
An omnidirectional source radiating a spherical sound pressure wave is comparable to an acoustic musical instrument such as a guitar, piano, or drum. A directional source (read: conventional forward-firing speaker system), however, does not sound precisely the same, nor does it load an average listening room in the same manner, due to non-linear frequency response combined with time and phase delays in the off-axis response. These aberrations contribute to warped sound waves that are neither coherent nor accurate to the original spherical waves and can be easily heard as such, no matter how accurate the system appears to measure on axis. These aberrations are highlighted due to reflected energy from boundaries such as the floor, ceiling, and walls. Although previously documented, these effects were not considered to be of prime importance prior to my research, but had tremendously important psychoacoustic implications, as I discovered. Several researchers “rediscovered” this effect and published their findings in the Journal Of Audio Engineering many years later.
Desired Axial Response
In 1976, I developed a small two-way speaker system that exhibited perfect measurements by the existing standards of the day. The design was a Time Aligned two-way speaker using a 6.5″ woofer, with first order phase coherent crossovers. The impulse response was pretty good, considering the drivers being used, and the frequency response was exceptionally flat on the axis directly in front of the speaker’s tweeter. Yet side-by-side comparisons with an acoustic guitar as a sound source revealed that the prototype lacked an essential realism. One evening, while I was listening to my creation with a piano recording in that magic sweet spot where the music seemed to come together, my wife was washing dishes. I was complaining about my disappointment with the sound quality, so my wife Linda, far off-axis in the kitchen, remarked that the sound was muffled and did not float in the air like the sound of our Hardmann (circa 1899) upright grand piano. This somewhat startled me, since in the narrow “sweet spot” where I sat, there did not appear to be problems with muffling nor image recreation. Moving around the room, I quickly realized that the narrow listening area, typical of most speaker systems, was very detrimental to the illusion of realism.
Further research in the lab let me to believe that our ear/brain hearing mechanism somehow compares subtle cues such as radiation patterns, among others, to recognize and identify sonic information. I had recognized, of course, that the sound changed dramatically when I stood or moved around the room, but was not concerned with this behavior since all other speakers I had heard exhibited the same problem! I decided that the brain must somehow compare these subtle cues (like sound wave recognition patterns) to stored information from past experience. Thus the brain knew that the sound from the speaker could not be radiated from a live piano, since the sound waves from the speaker did not match the radiation pattern of sound waves coming from the instrument. Obviously the piano, being an omnidirectional radiator, involved the entire room with its radiation pattern, while my highly directional prototype speaker, did not. Amazingly, listening to one speaker up close did sound highly realistic, much as a very good pair of headphones. It was the directional pattern of the system that was flawed!
I hurried to the lab to conduct a series of off-axis response measurements on a 180-degree horizontal and vertical axis. The results, although dismal as expected, excited me, since the off-axis radiation pattern was clearly non-linear and was perhaps related to the lack of realism I was experiencing! Several years of advanced experiments regarding directivity patterns and driver behavior later proved my theory to have merit. To the layman not schooled in conventional theory leading to a status quo in engineering design, this is not perhaps a surprising discovery, since it would seem intuitive to design a speaker to project sound in the same manner as live instruments.
Acoustic Inverse Replication
Additional research led to my further discovery that recording microphones encode the musical signal with their overlaying pickup response patterns. After making a series of recordings, using several different microphones, it was obvious during playback that the mics not only had tonal differences related to frequency response errors, but also created different types of imaging patterns. The perception of depth and space was not only dependent on the recording environment and mic placement, but also on the mic’s off-axis polar response. For this reason, I decided to engineer an adjustable ambience retrieval system radiating from the rear of the VR (virtual reality) speakers, in able to recreate the space and depth heard in the concert hall when the spaced omni method of recording is used.
Thus, a correctly designed speaker system should project the inverse of the mic signal, acting as a decoder to translate the original sound field. I have termed my design for this decoding as Inverse Acoustic Replication. TM The Virtual Reality series of designs was developed from several important concepts related to microphone pick-up patterns. These concepts are based on the consistent phase/frequency relationships in the polar response pattern of the mics, which was later reverse engineered into the VR speaker systems.
First Versus Fourth Order
Experiments validated the concept of consistent (not the same as coherent) phase vs. frequency linearity in a 180-degree arc around the speaker system, and appeared to work far better than Time Coherency limited to the axial tweeter response. As is commonly known, first order crossovers have severe problems with driver overlap, which lead to an effect called lobing. This problem is related to the fact that the drivers can sum perfectly only on one very narrow axis, since the path length from the drivers to all other axes cannot sum to unity, in either frequency, phase, or transient response! This not-surprising effect is due to the mathematics governing wave transmission and is easily verified by simple experiments or “doing the math.”
Thus the measured polar vertical off-axis response, for instance +/- 180 degrees, of speakers using first order crossovers will typically exhibit amplitude dips and peaks of up to 18dB as a result of the lobing effects caused by uneven path lengths and will have severe phase distortion as well. The ear/brain hearing mechanism can easily hear this effect, due to reflected response from the room boundaries even though the listener may be seated on the perfect axis. Not amazingly, the ear is far more critical than any type of test equipment yet devised, so these effects cannot be ignored on a psychoacoustic level, especially in a normally reverberant living rooms where the off-axis response dominates the perceived frequency and phase response.
Off-Axis Phase Vs. Amplitude
I have termed my method of enabling consistent phase vs. frequency behavior in the off axis response as the Global Axis Integration Network TM (GAIN), since my design constructs a consistent polar response both in the amplitude and time domains, both horizontally and vertically. Not only does this radiation pattern enable the listener to perceive well-balanced frequency and harmonic integration from almost anywhere in the listening side of the room, but also enhances sound-stage imaging over a 180 degree axis horizontally and 70 degrees vertically. This is especially important psychoacoustically, since the ear/brain hearing mechanism responds favorably to this reconstructed sound wave pattern.
This Global Axis Integration method consists of a carefully engineered radiation pattern created by front and rear driver arrays. Proprietary circuits form steep 24dB acoustic crossover slopes at specially selected frequencies without the penalties of induced ringing and excessive phase delay. These slopes are necessary to limit lobing effects and non-linear off-axis response, and actually enable the consistent phase behavior necessary between drivers. The architecture of the circuitry resembles first and second order filters combined with Zobel conjugate compensators in parallel. By using a minimum of high quality parts in series with the drivers, the sound remains transparent, yet the control over phase and amplitude can be corrected with the paralleled Zobel circuits. To my knowledge, I was the first person to have engineered this type of circuit, which was finally finished during the summer of 1984 when I sold the first VR-4 design to a recording studio. That first design was called the Vortex Screen, as it resembled a large, thin panel speaker system (but was in fact, a cone system). VR-4 was my code name for virtual reality in four dimensions: time, amplitude, phase, and “space” so I later changed the name from Vortex to VR-4.
HISTORICAL FOOTNOTE: I find it quite amusing that a young engineer (who was still in diapers when I brought the VR-4 to market) has recently claimed in magazine advertisements to “have invented the first crossover which closely correlates phase versus amplitude.“ I suppose that this young fellow is not well read in this field!
Listening Vs. Measurement
Long-term listening sessions have shown very good correlation between the engineering target-response patterns and perceived musicality. Although initially I listened to the designs under high quality phonograph and tape sources, I later designed a test using live music sources to compare to the speaker output. In a large room, with sonically absorbent panels in the middle of the room, we compared the sound of an acoustic guitar at one end of the room being replicated by the speaker at the other end. We used several expensive mics and a tube preamp for these tests and alternated between different kinds of portable instruments, such as brass, chimes, snare drum, trumpet, saxophone, harmonica, and of course the human voice. In the beginning, circa late 1970s, we were humbled by the lack of realism in our original designs. Over the years, by conducting research into the distortions and colorations caused by the drive units, circuits, and enclosures, we were able to eliminate or greatly reduce the factors that contributed to the sonic colorations.
Critical evaluation of these new engineering principles by several magazines has resulted in highly favorable reviews and comments. Although not an exact inverse of the mic signal, the AIR and GAIN designs use psychoacoustic principles to work with the listening room. Ambience retrieval, imaging clues, and soundstage transparency are combined with wide band frequency response, low distortion, and ultra-low levels of coloration. This combination of engineering goals has resulted in unprecedented levels of realism not achieved in competing speaker designs, regardless of cost.
New Driver Technology
My latest research has concerned the cone material and motor. Driver manufacturers continuously develop new cone materials in their search for the holy grail of perfect measurements. For instance, the new metal cones have fantastic stiffness and “punch” for bass frequencies, but unfortunately do not fare as well when used in the midrange and treble ranges due to the high Q peaks. Research I have been working on lately indicates that oxide coatings and anodization of the metal can eliminate or greatly reduce the perceived effects of the “metallic” signature. However, these coatings are not utilized on “stock” versions of the drivers we have been specifying, making custom versions mandatory.
On the other hand, old materials such as paper have been upgraded with composites such as carbon fiber powder, Kevlar threads, or plastic compounds injected into the paper when the cone is being formed. Our new VR-44, VR-5 Anniversary Mk2, VR-9SE Mk2, and VR-11SE Mk2 use a new cone material that consists of five different elements, some mixed together in a slurry, with other elements layered on top to provide a composite cone that eliminates audible artifacts known as “colorations.” As a reference driver, the Quad electrostatic has been equaled by our new midrange drivers, thus vindicating our new technology. Our European engineering partners have been instrumental in providing the necessary R&D so I could focus on system design which follows my general theories regarding transducer behavior.
Although many materials have been tested for treble reproduction, including metals, ceramics, paper, plastic film, and others, the plain-Jane fabric dome has seemed to remain at the top of the audiophile’s list for smoother response. However, these are not the “old school” fabrics from the 1970’s, but are rather specialized composites utilizing carbon fiber, banana fiber, cotton, and plastic resin impregnation to provide the necessary stiffness and internal damping.
New motor technology has been discovered and implemented in many of our new products. The distortion caused by the voice coil moving in a non-symmetrical magnetic field has been greatly reduced by new mechanical designs of the pole piece, shorted-turn voice coils, and new magnetic materials. We have reduced 90% of the previous distortions, leading to greatly improved transparency and smoothness up to the 40kHz range. Some of our speaker systems employ ribbon super-tweeters with response up to 50kHz-100kHz!
Triple Wall Cabinet Technology
Lately, several high profile speaker companies have been conducting advertising campaigns using their cabinet construction methods as the topic of discussion. One manufacturer of ultra-expensive speakers claims that solid aluminum cabinets are ‘The Best In The World’ when compared to any wood-based cabinet. A second manufacturer, using resin impregnated MDF, claims that their ‘proprietary ABC mystery material’ is the best choice. Since all of these claims seem to have scientific facts behind them, and since it is well known that inexpensive thin-wall speaker cabinets can contribute a high degree of coloration due to panel resonance, VSA Corp. has undertaken a two year scientific study of panel resonances, their audibility, and methods of reduction. As we wanted to conduct the highest level of analysis, we used a famous university’s lab, which is well stocked with laser interferometer, Finite Element Analysis software, and Fast Fourier Transform-equipped computer programs. This paper seeks to inform the reader of a new Pat. Pend. design now utilized on all Von Schweikert Audio Mk2 speaker systems, using a triple layer of constrained damping materials with opposing Q factors that behave as “active” noise cancellation. In essence, our cabinet design utilizes a multiple laminate of three different materials, since there is no perfect single material with ‘magic’ properties. In addition, our design goes one step further – decoupling of the drive units from the baffle, and in the UniField Series, the baffle is further decoupled from the cabinet proper. This is accomplished by our proprietary use of a visco-elastic gasket that provides a mechanical barrier to vibration transmission.
The Audibility of Cabinet Wall Resonances
Readers will instinctively realize that in order to reproduce a musical signal faithfully, an accurately designed loudspeaker should not add or subtract anything from the musical signal. Speaker systems claimed to be “tuned to the orchestra” by the use of resonant cabinet designs are nonsense. An accurate speaker should have no sound of its own, including cabinet panel resonances. After all, speaker systems are not ‘generators’ of music, they are ‘reproducers’ of music, which is self explanatory. Below is a graph showing a 7” bass-midrange driver mounted on a VSA speaker cabinet with our “active” noise cancelling Triple-Wall design compared to a solid cabinet made from the industry-standard MDF (medium density fiberboard). The difference is dramatic! We all know that an accurate loudspeaker must trace the signal as quickly as possible, and any ‘time smear’ emanating from a ringing baffle or resonant cabinet wall will be highly audible if the amplitude of the vibration is sufficient. In addition, if the cabinet wall resonance falls into a frequency range that is excited constantly by music or the human voice, i.e., 100-400Hz, the resonances will be even more audible. Finally, due to the nature of high Q versus low Q effects, the type of damping utilized will have a large impact on the success of the method. A narrow, high Q resonance found at 1kHz may not be as audible as a low Q resonance found at 150Hz, so the type of damping and its effect on the resonance is highly frequency dependent. For that reason, a wide bandwidth resonance reduction method necessitates the use of several different types of materials, and their individual Q factors need to interact with each other in order to be effective at all offending frequencies. Successful implementation will result in greater clarity.
#1. Comparison of Damped Vs. Undamped Baffles
NOTES TO ILLUSTRATION #1, ABOVE
In illustration #1 above, the Red line is the accelerometer-measured response of a 7” mid-woofer mounted on a baffle using our Triple-Wall Noise Cancelling technique. Note how quickly the impulse excitation has been damped. The 7” mid-woofer frame has been mounted onto the baffle using a synthetic clay absorptive gasket, effectively isolating its vibration from the front baffle. In turn, the baffle is decoupled from the cabinet itself, using the same compliant gasket method, thereby reducing themechanical vibration transfer from the mid-woofer to the rest of the cabinet. The baffle is a 25mm MDF resin-soaked panel that has a specific Q-design. The Blue line is the response of the same driver, hard-mounted on a standard cabinet baffle made from MDF “wood” which is used in 99.9” of the speakers on the market, but without our proprietary compliant decoupling or our “Opposing-Q” active technology. Note how the driver and baffle are still ringing up to 50 milliseconds after the impulse. Since the mid-woofer is mechanically attached to the baffle, and the baffle is mechanically attached to the cabinet, there is a high degree of unwanted transfer of vibration into the cabinet, effectively adding “smear” (a lack of clarity due to the constant ringing of the cabinet walls). In addition, the baffle and cabinet walls are ringing since they are not damped. Ringing results in very poor sound due to smearing of the music. Since the surface area of a speaker cabinet can be up to 10 to 40 times the area of the cones, very little motion of the cabinet walls is necessary to become audible. The implication is that the cabinet vibration might have as much energy as the output of the driver itself, and this unwanted energy must be cancelled if true clarity is desired.
Below is a photo of VSA’s Pat. Pend. wall construction, as used in every Von Schweikert Audio speaker system offered today.