VR-55 AKTIVE
~ ENGINEERING GOALS AND METHODS ~
As high-fidelity sources and electronics have been improving over the years, it has been acknowledged that the loudspeaker has been traditionally the weakest link in the chain. Any electro-mechanical device that will transform an electrical signal into air motion by necessity will have drawbacks and multiple trade-offs in every aspect of the device.
Distortion and coloration will occur in the drive units, crossover circuits used to blend the drive units, and cabinets that house the assembly, no matter if the device is a moving coil design in a sealed, ported, or open baffle enclosure. Even panel-type transducers like an electrostatic or other planar design such as a ribbon or Magnaplanar type of device have multiple sources of distortion and coloration. These distortions will be addressed later in this paper.
WHAT TYPE OF TRANSDUCER IS MOST ACCURATE?
To complicate matters, there is a lack of agreement on how exactly to “push” the air with the most fidelity. Does using a planar device result in greater fidelity to the source, or is the “plain jane” moving coil driver more suitable to accurately reproduce the signal faithfully?
If you were a design engineer, exactly how would you choose to design the most accurate speaker system possible, given the cost constraints of a budget and room décor?
TYPE OF DRIVE SYSTEM REQUIRED FOR HIGH RESOLUTION
Since there are several ways to “push air” to reproduce a sound wave, what type of design makes the most sense? A thorough examination of the recording process makes this an easy answer: the goal of a loudspeaker is to reproduce the microphone’s characteristics. This was my thesis at Cal Tech, so I’m quite familiar with the topic after studying the various aspects of sound reproduction for three years (1976-79).
My research partners and I used FFT equipment to measure the signals on the recordings of the day, which consisted of tape recordings and vinyl records. My goal was to design a speaker system that was as accurate as a recording mic. I had realized after three years of research that the actual goal of an accurate speaker system was to behave as a recording microphone in reverse. Dr. Richard C. Heyser, my mentor, agreed with me that a microphone/speaker "system" had to be seen as an "encode/decode" situation. I had come to that conclusion after studying the Dolby Noise Reduction System along with the recording RIAA curves used to record and play back vinyl records. The idea that a speaker had no actual goal except to "sound good" did not sit well with me, as a study of psychoacoustics led me to the premise that there was only one correct way to design a speaker that would replicate the electrical signals on a recording.
MICROPHONE CHARACTERISTICS
Since almost every recording microphone is a small 15-30mm point source and often has an omnidirectional or cardioid pickup pattern, the characteristics of a recording microphone have to be inverted by a loudspeaker system, if the original sound wave pattern on the recording is going to be reproduced. After all, there is no “music” on a recording, only the electrical voltages from the recording microphones.
Here are the characteristics of a “perfect” loudspeaker, which would closely mimic the recording microphone’s behavior:
1. Small point source. Since mic’s have a small diameter capsule, it can pick up sounds far off axis. Wide baffle speakers cannot mimic a wide pickup pattern, as the baffle will enable reflections to occur and narrow the directivity of the radiation pattern. We have designed our front baffle to be as narrow as the drive units. The edges of the baffle are highly chamfered to eliminate edge diffraction. That is why the speaker looks like a pyramid: the tweeter at the top is much narrower than the woofers, so form follows function.
2. Full range response. Most microphones pick up 20Hz to 20kHz, so an accurate speaker must also duplicate this response. We have designed cone bass drivers with response flat to 21Hz, and the -6dB point is 16Hz. At the top end, the Beryllium dome tweeter has response flat to 40kHz. According to the Hilbert Transform, if you wish to have flat frequency response with flat phase, the treble pass band must be accurate to an octave above the recording's high frequency limit, since the phase response starts to rotate towards 90-degrees when approaching cutoff. Thus, for a tweeter to have flat frequency AND phase tracking, a tweeter that is flat to 40kHz must be utilized.
3. Wide dispersion pattern that can mimic the mic’s pickup pattern. Many types of speaker designs have a narrow or limited dispersion pattern, which cannot reproduce the microphone’s pickup pattern. The “perfect” loudspeaker would have small drivers and a very narrow front baffle that tapers off at the edges (for minimum diffraction). That is why the VR-55 uses five small drivers on a minimum-width baffle (the fifth driver is the rear-firing Ambience Retrieval ribbon tweeter on the top rear of the cabinet).
4. High dynamic range. Many recording mic’s have a very high dynamic range, so the loudspeaker should also have as much dynamic range as possible, without any peaks being created by horn loading or high efficiency speakers with no deep bass. There are two reasons why we use the “booster” amplifier with our VR-55 Aktive: a. The dynamic range is increased by a factor of 6dB, and b. The VR-55’s owner can adjust the bass level to his room and recording. That is what we engineers call a “double home run.”
5. Low distortion. Since a typical recording mic will have distortion far below 1%, the speaker system should use Low Distortion technology in every aspect of the design, from the driver cones and motors to the cabinet vibration. We spent almost four years searching for the most accurate drive units available, then we had their factories modify their designs for our application. Accuton of Germany has developed technology over the past decade to reduce distortion in their moving assemblies and motors. Along with our suggested modifications, the hybrid ceramic/honeycomb woofers have approximately 0.5% distortion at 91dB. This is an amazing feat! Similarly, the Accuton ceramic/Kevlar midrange driver and Scanspeak modified Beryllium dome tweeter with our modifications have less than 0.3% and 0.1% respectively. Since the drive units do not add any distortion, they are relatively “colorless” and have almost no self-noise. In fact, although some may believe this statement to be “heresy,” our drivers have less distortion and faster transient response than the fabled Quad ESL electrostat. Audio Precision and Hewlett Packard measurement systems have verified this, along with side-by-side listening tests using the A/B/X tests with a group of listeners.
6. Fast transient response. Mic’s have very fast transient response, since their diaphragms are very light and small in size. This means that the transducers utilized must be designed with large magnets and low moving mass. More details are supplied later in the Transducer Section, below.
7. No cabinet resonance. The highest quality Neumann, Telefunken, Sony, and RCA mics have no resonant hollow body surrounding the mic capsule, so the loudspeaker system needs to be housed in a non-resonant cabinet. Special requirements are necessary to design a cabinet that cannot vibrate or resonate. Note that most cabinets have a large surface area, perhaps 20 times larger than the transducer square area. This means that the cabinet panels have only to move 1/20th of the distance of the driver excursion for the cabinet to be as audible as the sound from the drive units!
8. To eliminate any source of sound or resonance from the VR-55 Aktive panels, they are formed of three different materials. We call this design the Triple Wall Noise Reduction System.
a. HDF outer layer, good for accepting paint and hard enough to machine to tight specifications. Wilson uses resin impregnated HDF in their best cabinets and this is a very good starting point.
b. Artificial stone inner layer, made like concrete with crushed gravel, resin, and powdered mineral dust. This far denser and heavier layer has a resonant factor that is half of the HDF layer and has a different Q factor, along with weight to stiffness ratio that was also necessary.
c. A rubber adhesive damping layer is used to bond the two different panels together, providing an elastic coupling that is also sound-proofing in action.
d. The premise is that the two different layers of materials resonate at opposing frequencies and Q’s, such as the addition of +1 and -1 = 0. In a side-by-side test of our Triple Wall Noise Reduction design against 1” thick aircraft-grade aluminum, the Triple Wall design was quieter and more musical, with less “ringing” of the cabinet behind the music. The sound came from a “darker” background. The Triple Wall design complements the Low Distortion drivers, resulting in one of the most transparent speakers ever built.
9. No electrical circuit distortion or phase error. Since microphones are “one ways” with no electrical circuitry except for a preamp stage, they are by definition “phase coherent with linear phase. Any single driver system that does not use a crossover is very band-limited, with very shallow bass and dull highs. Two-way or higher systems sound better, but must be Time Aligned and Phase Coherent like our VR-55 if the goal is to sound like a single driver. We accomplished this by designing the world’s first steep slope phase coherent crossover circuit at Cal Tech in the mid 1970’s. See Page 7 for more detail.
SPECIAL CAVEATS TO NOTE
Since there are no vertical line source microphones that work well enough to record an acoustic event, a speaker should not be designed as a vertical line source. You can write off panel speakers in general due to their low dynamic range and lack of point source image focus. Most panel drivers also exhibit “clamped edge” effects which create a chaotic transient response decay, due to reflections from the panels to the clamped edges. In cone drivers, these reflections are damped by the rubber edge.
That leaves the humble cone speaker (dynamic moving coil transducer) as our best bet. However, it is well known that cone drivers have multiple problems that must be addressed if the system is to be as accurate to the signal as possible.
A. Most cones have coloration from cone flex if made from soft plastics, such as polypropylene or Bextrene/TPX, and even treated paper has distortion caused by the paper fibers rubbing against each other during rapid cone excitation. It is obvious that the cone should not flex or bend, and it also has to be extremely light and damped. Most stiff cones will ring at high frequencies, so internal damping must be used to prevent this.
B. Most cone drivers use low-tech motors first designed in the 1920’s with permanent magnets. A common voice coil driven by a ferrite magnet can achieve a distortion measurement of up to 37%, which we measured at Cal Tech. Granted, this occurs at high volume levels, but even at normal levels, distortion was almost 8%. This is an unacceptable number and would create a “fuzzy” sound quality with a lack of clarity to the depth of the sound field.
HOW WE DESIGNED THE VR-55 TRANSDUCERS
We realized that the cone had to be as light and stiff as possible, yet it had to have some control over the ringing modes often seen with metal or ceramic cones. After spending several years listening to and testing various cones, we concluded that only two cone materials were in the running for best sound and measurements: magnesium cones with a ceramic damping layer, and ceramic cones with a honeycomb layer. For our new VR-55 series, we chose the ceramic cones as the best choice for this particular application, since they are more efficient than the magnesium cones. Although we use ceramic coated magnesium drivers in our tall point source systems like our relatively new VR-100XS, the midrange drivers are low in sensitivity and need to be used in pairs. One of our VR-55 design goals was to keep the speaker height low enough to use in small rooms, so the Accuton midrange driver became the default driver to base the system around.
WOOFER DESIGN
The original Accuton ceramic woofer cone had ringing that was unacceptable and many of the test samples arrived with cracks in the cone from shipping. We found it necessary to redesign the cone composition. Using a composite layer of two cones, one being a honeycomb of Kevlar paper bonded onto the ceramic cone, prevented both of the problems of fragile cones and ringing. To drive the woofers, we developed a twin magnet system and a unique voice coil system with an 11-ohm impedance. This gave us the required low distortion measurement that we were seeking. This driver took three years to develop, as Accuton’s engineering staff was as particular as we were.
MIDRANGE DESIGN
The midrange driver presented a different set of problems. As we wished to use the midrange as a quasi-full range, we wanted a frequency response of 50Hz to 10kHz. After testing and listening to every midrange driver, including electrostatic panels, AMT (air motion transformers, which are folded ribbons), and conventional ribbons, we concentrated on cone units due to their wider bandwidth, higher output, and potential to match the recording mic’s characteristics. Most of the cones we tried were colored or peaky, so we decided to use our fallback position of using a composite. In our original VR-5 series, we had used composites since 2001, and the original Aerogel cones by Audax were quite good for their day. However, we always realized that it would be possible to use even lighter and stiffer cones, with more powerful magnets that would have less distortion. These requirements led to the development of the Accuton ceramic cone with a Kevlar damping layer bonded onto the surface of the rear of the cone.
To get down to 50Hz requires a heavy duty cone with a soft suspension, but trying to force a small woofer to vibrate accurately at 10,000 cycles per second requires a very light weight moving assembly. These two requirements are in opposition to each other, so the development of the midrange driver became a balancing act. By using new technology for the suspension, we are able to push the midrange down to 50 Hz without distortion. To achieve the goal of 10,000 cycles per second, we had to develop a cone profile shape that would not beam the high frequencies into a narrow pattern that would not blend with the wide dispersion of the 1” tweeter. In fact, focused imaging for the most accurate sound stage requires a constant dispersion pattern off-axis between the midrange driver and tweeter. If there is a lack of blending of the dispersion between the two drivers, the image will not be focused nor solid. That meant that we had to develop the midrange and tweeter as a unit and not separately. Therefore, this required cooperation between Accuton (ceramic midrange) and Scanspeak (Beryllium tweeter). Being the end users, VSA became the chief engineering leaders, with the suppliers building the drivers to our specifications rather than vice versa.
TWEETER
There are many choices of tweeter designs on the market. You could choose between a silk or fabric dome, known for smoothness, but compromised by a lack of upper frequencies and accurate reproduction of an impulse due to the bending wave nature of a soft dome. The hard domes fabricated of ceramic, titanium, and artificial diamond-coated units all held promise, as well as ribbon tweeters. After careful evaluation of their sonic and measured characteristics, we chose Beryllium for its upper frequency response, low distortion, wide dispersion, and lack of coloration except for slight sibilance in the female vocal range, which we corrected by vapor depositing a thin layer of vinyl (similar to Saran Wrap) onto the surface of the dome and rubber edge surround. This can be seen as a light gray layer on the surface of the assembly.
Beryllium is one of the lightest and stiffest membranes available, and is comparable to artificial diamond coated drivers. Although Beryllium has gotten a “bad rap” from early designs, we have implemented our version with Neodymium magnets for low distortion and have damped the surface of the dome with a very thin layer of vinyl (similar to Saran Wrap) to eliminate the “bright” or “hard” sound of the early types of Beryllium domes on the market. If there is a better tweeter than what we’re using in our VR-55 Aktive, we haven’t heard it. (A similar Beryllium design is used by Magico in their Q7 Mk2 at $225,000/pr). It is widely agreed that the Mk2 generation of Beryllium tweeters is at the 'state-of-the-art.'
REAR AMBIENCE RETRIEVAL SYSTEM
Since high quality recording microphones have pickup patterns that are omnidirectional or cardioid in behavior, they receive sound from 360-degrees. In fact, the best classical recordings that can simulate the sound of a concert hall were recorded with a pair of omnidirectional Neumann or Telefunken mics in the 1950’s up to the present day. The rear of the mics pick up the hall reverb, which is encoded into the recording. This hall ambience can be retrieved by using a decoder and another tweeter mounted on the rear of the cabinet.
Since the back of the mic is picking up sound that was reflected from the side and rear walls of the hall, the signal is out-of-phase with the front wave received by the mic from the orchestra itself. By implementing a passive QS decoder, used during the era of Quadraphonic sound, the hall reverb and depth signal can be decoded from the original signal, then fed into the rear Ambience Tweeter. The VR-55 uses a 3” tall ribbon tweeter to disperse the ambience to the rear. If the ambience was not adjustable, you would end up with problems due to either too loud of a signal when placed close to a hard reflective wall, or too low of a signal if placed out into the room and perhaps even firing the rear tweeter into an absorptive surface, such as curtains. The successful implementation of the Ambience Retrieval System requires a level control to match the rear output to the room and it’s absorption characteristics. We have worked on this system for the past 32 years, so we believe we have done this correctly. I might add that our speaker systems are the only ones available with this particular feature, which is of great benefit to listeners of live recordings, whether classical, jazz, folk, or rock. The sense of depth and hall reverberation is a hallmark of the VR-5 platform.
The benefits of the Ambience Retrieval System go beyond that of enabling a beautiful sense of depth. An additional benefit is that listeners off-axis can hear a three dimensional sound stage, since the rear-firing tweeter fills in the gaps of the front wave’s dispersion pattern. If you walk around the room while the VR-55 is playing, you can easily determine that is is one of the most “open” sounding speakers ever created. You can hear a dimensional image from anywhere in the room, not just the sweet spot. If you’re a fan of live music, you will love the VR-55. [Although the MBL Radialstrahler has omnidirectional imaging and sounds quite nice in very large rooms, it is very problematic to set them up in small to normal sized rooms, as they have no controls to direct the sound wave patterns like the VR-55 offers].
GLOBAL AXIS INTEGRATION NETWORK
Since a truly great speaker system requires a wide bandwidth with flat on axis and flat off axis response, the use of a three-way system is the best way to achieve this goal. However, “normal” crossover circuits add distortion, along with uneven driver output levels (dips and peaks) unless the driver reactances are controlled.
Driver "reactance" consists of the inherent inductance of the voice coil, the mechanical capacitance (energy storage) of the driver suspension, and varying impedance of the voice coil at different frequencies, based on the movement of the voice coil in the magnetic field. If you have ever seen the impedance curve of a moving coil (cone speaker) driver, you know that is is NOT a constant 8-ohm load that is seen by the crossover filter circuit. The roller coaster impedance curve shows the actual "reactance" if the driver. It is well known to advanced researchers in the field of acoustics that a "textbook" circuit cannot adequately compensate for these reactances without additional forms of circuits used for compensation. The design of a "complete" circuit that can keep the impedance from varying over the bandwidth of the driver is difficult or impossible to design without some form of computer aided circuit analysis program. At Cal Tech, we used a proprietary version of a SPICE circuit simulator to quantify and compensate for the sixteen different sets of reactances found in a typical moving coil woofer. Attempting to do this without advanced knowledge and a large computer was next to impossible in the era before 1999.
In addition, text book circuits above First Order (6dB/octave) create phase shift distortion. We realized in the mid 1970’s that a new type of circuit had to be invented, as “text book” circuits did not work correctly with the reactance created by the voice coil moving in a magnetic field. First order circuits have so much overlap between the drivers that you can hear cymbals being reproduced by woofers (an ugly sound) and midrange coming out of small tweeter diaphragms that can easily be driven into massive distortion at high levels. Clearly (pun intended), it was time for a better circuit design. We had rejected second order filters due to the 180-degrees of phase shift, which forced the speaker designer to run the midrange out-of-phase with the woofer and tweeter, as we wanted to use a set of drivers that all pushed and pulled together for accurate pulse replication.
As our lab project, several students and I (as team leader) decided to invent a new type of circuit that would not chop up the signal and then attempt to blend the pieces back together at the driver level, with resulting phase shift distortion or drivers wired out of phase with each other. With Sigfried Linkwist of HP (Hewlett Packard) as our consultant and Dr. Richard C. Heyser of JPL as our advisor and lab professor, we began our work. Our first project was to study the entire body of literature on circuit design, from telephone/telegraph circuitry up to present day circuitry (circa 1976). We studied the theories of Western Electric, Bell Telephone Labs, General Electric, and RCA Laboratory researchers and subsequent text books regarding filter design. It seemed that no modern research had been done regarding the effects of filter circuits and no one had even tried to invent new circuitry, most likely due to the difficult nature of passive circuit design driving an unknown reactive load.
Today, an active digital crossover can be designed to have any type of filter shape, along with other aspects of filter design. However, active filters never seemed to sound as “pure” and direct as a passive circuit, whether using analog or digital circuitry. We all felt that a completely active speaker system with built-in amplifiers for midrange and treble would not sound as good as using one high quality "audiophile-grade" amplifier.
Our New Crossover Design
Based on the cascade principle where one circuit can be fed into another following circuit, our team at Cal Tech developed what we believe to be the world’s first steep slope crossover that was Phase Consistant (wave-coherent but not time-coherent) with no audible distortion, wave form or otherwise. The time lag from the steep slopes of our G.A.I.N. circuit are below the ear/brain hearing mechanism to detect; Fastl and Zwicker (German research scientists who wrote a seminal book on the psychoacoustics of the hearing process) found that the most critical listeners could not hear a pulse train as solitary sound bites if the pulses were generated at a faster rate than 5 ms.
Our Global Axis Integration Network blends the drivers together “harmonically” instead of chopping the signal into pieces and then attempting to reassemble it at the drivers. This is due to the nature of the cascading filters, which employ negative feedback from the driver’s back- EMF signal as “read” by the input/ouput comparator employed in our G.A.I.N. design. The time delay between the original pulse and the replicated signal is 2.87 ms, for the entire speaker system, which is below the threshold of audibility.
Our G.A.I.N. circuit also contains EQ circuits to modify the driver’s response shapes and blending at the cross points. Although quite novel, the circuit is easily implemented and uses the best parts we can find. We use Mundorf gold and silver foil encapsulated in oil along with V-Cap copper foil in Teflon capacitors in a "recipe" for best transparency, Goertz and Perfect Lay inductors wound with single crystal copper, and Mundorf metal film resistors are specified due to superior sound and power handling. All internal wiring is Single Crystal Copper manufactured for us by Delphi Aerospace, as used in the Signature Series of audio cables by Master-Built. On special order is the ULTRA internal wiring, which is a special alloy of rare earth metals combined with precious metals and an outer shield designed to reject the magnetic fields inside the cabinets, created by the large magnets and inductor coils.
SONIC GOAL AND MEASUREMENT PROTOCOL
Our goal was to achieve the most transparent sound quality combined with holographic image focus and sound stage depth. Using live versus the speaker a/b testing, we compared the microphone feed from a live trumpet, saxophone, clarinet, violin, acoustic guitar, harmonica, and a wide range of percussion instruments, side by side to the VR-55. Most importantly, we used several male and female vocalists to sing through the speakers. After careful adjustment of the filter circuits along with the quantity and type of internal damping (stuffing), we found that the VR-55 could replicate the sound of the voice extremely well, with no sibilance, lower midrange “chestiness,” or other type of coloration.
These listening tests were done after six months of computer testing, using both analog and digital test equipment. We measure frequency response flatness, phase response accuracy, dispersion patterning, and distortion of the system. Both listening and instrumentation testing proved that we have met our goal with our new VR-55 Aktive design. Having won five awards for Best Sound Of The Show in the US and Asia, we believe that many customers also like what we have achieved. It is possible that the VR-55 Aktive is one of the most “live” sounding speakers on the market today. We hope that you agree!
Copyright 2015 by Albert Von Schweikert
