A New Loudspeaker Technology
Background Loudspeaker enclosures have been the subject of continuous development since the 1920s. The types that have been identified include open baffle dipoles, horns, and closed box types sealed, vented, transmission line, and bandpass. All of these use what electrical engineers call lumped parameter models to make design decisions. The loudspeaker driver is given parameters such as cone mass, magnet strength, and so forth. Neville Thiele and Richard Small, two Australian electrical engineers, identified in the 1970s the methods to turn the driver parameters into equivalent electrical component parts, such as resistors, capacitors, and inductors. The enormous advantage of this transformation was the ability of engineers to study loudspeakers just as they had studied electrical circuits for over a hundred years. The impact of the Thiele-Small parameters was so great that virtually all quality drivers sold today are specified, designed, and built with these parameters in mind. The economic development is enormous, because having such parameters gave a way for loudspeaker designers to talk to driver designers, and driver designers to cone designers, and so forth. Before use of the TS parameters, such conversations had to occur with cut-and-try models. After the TS parameters came into widespread use, computer models could design the loudspeaker/box interface so thoroughly that a great deal of the prototyping process could be done without ever building a single box. New Technology Recently, another two Australian electrical engineers, Graeme Huon and Greg Cambrell, thought that there could be a further step in the basic method by which drivers and boxes were designed. By using the electrical engineering principles called ""distributed parameter models"" that are widely used in advanced electrical engineering, but not used in loudspeaker design to date, they proposed the idea that advantages in design could be possible using such techniques. The fundamental problems of drivers in boxes are to produce flat response over the specified frequency range, including the effects of room ""loading"" to the fullest extent possible, low group delay, low distortion, and high output capability, all combined with high efficiency for making power amplifier requirements practical. Using newly developed mathematics called Parametric Acoustic Modeling (PAM), Huon and Cambrell sought to produce improvements over the traditional methods. This paper presents results of their work. Embodied in three models of loudspeakers already, PAM provides a new way to design loudspeakers in boxes that although more complicated than other designs, offer improvements in range, response including group delay, distortion, and headroom. For instance, frequency response can be precision adjusted for the field of application, to tailor the characteristics of loudspeakers to the needs of designers. Frequency Response The world of audio is indebted to Lewis Fielder and Eric Benjamin of Dolby Laboratories for their AES paper of some years ago. In the work, the authors took a subwoofer into multiple rooms, and compared the measurements they made in the rooms with those in free space. The results was a ""room gain"" curve, showing the increase in low-frequency response of a loudspeaker caused by ""enclosing"" the output by the room walls. Since loudspeaker design programs, and even the correct use of the Thiele-Small parameters themselves, rely on a known environment for the loudspeaker such as half space, the conditions of use of the loudspeakers, in a room, differ from those given by ""cookbook"" design. What Fielder and Benjamin did was to provide a way to translate designs from the theoretical full-space or half-space model, into real rooms. PAM technology employs the Fielder-Benjamin design criteria, and produces a frequency response complementary to that of an average room, thus producing low bass extension and flatter response. Including this ""room gain"" the response and the listening location is typically flat at frequencies below those where standing waves begin to intrude on the response, within +2 dB to the lower frequency limit, which varies by model. Curve 1 shows a typical response in the design environment of half-space. The Fielder-Benjamin correction to this response will produce flat response to between 20 and 24 Hz, depending on the product chosen. Group Delay Many of the newer loudspeaker designs, especially the bandpass design with multiple chambers loading the front and back of the driver cone, produce high output, albeit over a fairly limited frequency range. This is achieved with a structure of multi-resonators, rather like a bank of organ pipes. The problem with these designs is that they produce large amounts of ""hangover,"" or resonant behavior, which results in a short input producing an output extended in the time domain. While the audibility of group delay at low frequencies was not often considered a problem, the excessive group delay of some designs made it questionable some years ago. Laurie Fincham, then at KEF, studied this problem resulting in an AES preprint on the subjective effect of low-frequency group delay in loudspeakers. Fincham found that the typical amount of delay found in even conventional loudspeakers was audible under conditions of carefully made recordings with extended bass response. PAM technology permits designing for low group delay, unlike other designs where the group delay is a function of the type and tuning. In fact, the group delay of the current PAM designs are far more controlled than that of the sealed box designs that Fincham studied. Up until now, these designs have been known as having the lowest delay available. The group delay of the PAM design is flat across the operating range, rising only at the lower corner frequency, whereas even a sealed box design has greater in-band group delay. Headroom A widely under-appreciated fact is that in a system employing a subwoofer with bass management, the subwoofer must handle all the energy supplied by the program from each of the 5.1 channels summed together. At standardized playback levels for movies, with an 83 dBC ""slow"" level calibration for ñ20 dBFSrms level pink noise, the maximum undistorted level of one channel is 103 dB SPL. The Low Frequency Enchantment channel, the 0.1, has an additional 10 dB of headroom, through the procedure of reducing the level on the medium by 10 dB, and increasing the level on playback by 10 dB. The idea behind this was to follow the equal loudness contours of human hearing down to lower frequencies in headroom capability. This was done due to a problem of the Compact Disc, that overload audibly occurs at low frequencies before middle frequencies, not through any fault in the medium, but because human hearing is less sensitive at low frequencies. Here is how the requirement is calculated for all of the channels together. Bass management extracts the low bass from all five channels and sums them together along with the LFE (0.1) channel. SPL = 20log[10L1/20 + 10L2/20 + 10L3/20 + 10L4/20 + 10L5/20 + 10L6/20], where L1 through L6 are the maximum required levels in each of the channels. Adding together 5 channels at 103 dB and one at 113 dB results in a requirement for 121 dB SPL for the subwoofer(s)! This is a worst case condition of the same signal applied to all of the channels simultaneously, and played at reference level. The reason that some systems apparently produce enough low frequency output today is statistical only: people do not play back movies at the standard level but instead average some 6-8 dB lower at home than the way a film was made, and signals rarely peak in all channels simultaneously. Testers who have played the Dolby test DVD know that practically all systems have a serious limitation in this area. The Whise 624 is designed to produce 121 dB SPL at 25 Hz and 1 m in half-space. Combined with the room gain from Fielder-Benjamin, the result in most practical home rooms is, for the first time, to achieve full headroom without excursion or power overload of the drivers down to 24 Hz. TMH Corporation provided a new test means, its ""boinker"" test signals, available from The Hollywood Edge on a series of four test CDs, that help to quantify the headroom capability of subwoofers vs. frequency. The boinker is a worst case frequency-time-level test that is easy to evaluate by ear. The test discs include level calibration sections of band-limited pink noise at ñ20 dBFSrms and boinker test signals from 10 Hz to 16 kHz on one-third octave bandcenters that just reach full scale (for one sample). The boinker signal is constrained to 1/3-octave, yet is a high transient level signal. This is achieved by the equivalent of taking a zero-crossing gated tone burst and running it through a 1/3-octave filter of the same center frequency, thus causing the worst case frequency (narrow band), level (full scale peak), and transient (only as long as it needs to be to meet the narrow band criteria) requirement simultaneously. Efficiency The question of headroom of a subwoofer is tied into that of efficiency. For if the subwoofer could handle high levels, but with only low efficiency, the power amplifier requirements could easily become impractical. For instance, achieving 121 dB SPL with a loudspeaker having a sensitivity of 89 dB SPL at 1 m and 2.83 Vrms (1W in 8 ohms) requires a 2000 watt power amplifier, which would surely break any practical loudspeaker driver. The Whise 624 has a sensitivity of 97 dB/m at 2.83 Vrms, which equals 1 W in 8 ohms. Thus 121 dB SPL requires a 250 Wpc stereo amplifier, each half of the amplifier driving one of the pair of 624 boxes. Distortion Fielder and Benjamin also studied distortion in subwoofers and found that subwoofers of ten years ago were wholly inadequate to reproduce program material with low audible distortion. This was because Fielder, in other work, had produced masking curves, useful for designing low-bit-rate coders, but also useful for weighting and studying the effects of harmonic distortion of components in the chain, including loudspeakers. He then compared the distortion of the subwoofers under test to his masking curves, and found the harmonic distortion components were quite audible. When Fielderís masking curves are plotted, harmonics and other distortion product that lie below the curves are inaudible, and those above are audible. One criteria to use is that a subwoofer is considered effectively distortionless if all of the harmonics lie below masking. Although this is a simplification because it does not consider intermodulation distortion, in general harmonic distortion is found to be more audible that IM distortion, when present in similar amounts. And designs with low harmonic distortion generally have low intermodulation distortion. Curve 2 shows a spectrum analysis of the Whise 624 measured at the listening location of the Immersive Sound Laboratory of the University of Southern California Integrated Media Systems Center. The 110 dB SPL 31.5 Hz sine wave is accompanied by harmonics that measure 43 dB down for each of the second and third harmonics, thus being around 0.7% in each harmonic, and with much lower levels of higher harmonics. Thus the THD is about 1%óremember, this is for a 100 dB SPL sine wave at 31.5 Hz! Curve 3 shows a spectrum analysis of a product that is among the highest quality on the market having equivalent cone area and excursion capability as the Whise 624. It is driven by a 400W power amplifier. The room and microphone position are the same, only this loudspeaker is installed in the corner, giving it the maximum advantage. While these curves are subject to variations due to frequency and room position of the subwoofers, they are similar at other frequencies. That is because nulls caused by standing waves can physically reduce or increase distortion at a measurement point in space. Many curves at a variety of frequencies show the overall low distortion performance of the Whise design. Compared to the Fielder masking curves, the Whise 624 at the frequencies and levels examined so far show the subwoofer to be audibly distortion free, that is, all of the harmonics lie significantly under the masking curves. Conclusion Parametric Acoustic Modeling permits the design of practical products that have: Flat frequency response in their pass band Low group delay over the pass band, rising only around the corner frequency, unlike any other design High sensitivity thus minimizing power amplifier requirements High headroom capability, 121 dB SPL in the case of the Whips 624 Low distortion, audibly distortions Practical load impedance with minimal not less than 4 ohms Reasonable size and cost given the performance For more information contact Fitz Koenig, 213 742 0030, (fkoenig@tmhlabs.com)