Dual-diaphragm compression drivers have been a fixture in a specific subset of professional sound reinforcement applications for a number of years. Instead of being built with a single membrane – like a conventional compression driver – these devices feature two differently sized diaphragms.
They’re an attractive proposition because they can deliver a wider bandwidth output than conventional compression drivers and can also be fixed to a single horn, thus providing constant directivity over the entire frequency range.
These benefits come with a number of acoustical compromises, however, which manifest themselves as noise and distortion in the reproduced signal. They also demand the use of an external crossover which itself introduces phasing and coherence issues.
I’ve spent much of my career refining and improving compression driver designs, and together with Dr. Jack Oclee-Brown (the former research engineer and now head of R&D for Celestion’s sister company Kent Engineering Foundry, or KEF for short), we took on the task of developing a compression driver that could produce a very wide bandwidth signal, without the compromises inherent in the industry standard dual diaphragm design.
Several years of development later, the result was the Axi 2050: wideband compression driver that is able to drive a high SPL, low distortion signal – with a single diaphragm – from a stated 300 Hz to 20 kHz (horn permitting).
The compression driver itself began life almost a century ago as an ingenious way of producing a more efficient loudspeaker output, by exploiting the radiation resistance between the device and a horn that it was designed to be coupled to in order to function effectively.
By the early 1930s, the key operational features of compression drivers were established. Engineers were able to develop mathematical models to aid design – calculations that enabled the prediction of the response of a highly simplified driver from an equivalent electrical circuit.
These models were successful in describing the basic performance of the driver but weren’t capable of demonstrating more complex behaviors. They showed the overall response trend, but major response variations which limited the bandwidth of compression drivers, especially at high frequencies, such as diaphragm breakup and cavity modes, had not been accounted for.
Although the engineers of the day were aware that complex effects limited driver performance, they often could not decide which effects were responsible for which shortcoming. One such example was the question of what caused the large mid and high frequency dips in the response curve. Various theories were proposed; interference due to path length variations and cavity resonances were prime suspects (Figure 1).
Phase plugs were developed as a means to mitigate these problems, but it was only in 1953 that acoustical researcher Bob Smith went a step further, publishing a paper that analyzed the resonances in the cavity between an idealized flat, resonance-free, diaphragm and a phase plug.
Smith’s realization that the phase plug channels excite “cavity resonances” was not intuitively obvious but this discovery was the first step towards a deeper understanding of the acoustics, giving birth to the idea of avoiding excitation of resonance by positioning that excitation in a nodal position; a technique that came to be known as “modal balancing” (Figure 2).
Perhaps because the paper was published in an academic journal it was not applied to a real driver with a dome diaphragm until 1970 and some designers persisted with the simpler concept of path length differences until that time.
Refining Smith’s Approach
By the turn of the millennium the use of FEM (Finite Element Modeling) with high-power computing was an established tool at Celestion. In 2003 at the AES, I presented a paper illustrating a “virtual prototype” in which computer models of magnetic, electrical and vibro-acoustic behavior were combined to predict a driver’s complete response, including diaphragm and cavity resonances.
This was a great step forward for compression driver development since it allowed the results of the mathematical simulation to be compared with the performance of a real device, enabling the virtual model to be used as part of the design iteration. This was not just a time saver for the engineer; it also allowed many more iterations to be performed and the chance to visualize the pressure and motion inside the driver, greatly enhancing the engineer’s ability to understand any performance issues, and to resolve them.
During the development of a new generation of compression driver for Celestion, I realized that the results I was obtaining did not entirely conform to Smith’s method. In response, I introduced Oclee-Brown – a member of the research team he was leading – to the problem in the hope of improving on Smith’s approach. After many months of work, an improved theoretical model of the compression driver diaphragm and phase-plug cavity was developed.
The new approach modeled a curved diaphragm as opposed to the flat one which Smith had done, resulting in a further reduction in the excitation of cavity resonances, thereby lowering output distortion. It also piqued the curiosity of the two of us about how the diaphragm and cavity resonances interact. This work was subsequently the subject of a patent and was presented at the AES in 2007 and went on to be the subject of Oclee-Brown’s PhD thesis.
As the PhD work progressed, it became apparent that the design “ideal” of avoiding cavity resonances in compression drivers was precluded by the extremely strong excitation from the typical diaphragm modes. It eventually became clear that the problem originated in trying, and failing, to achieve the goal so treasured by loudspeaker engineers: the “rigid piston.”
So a new approach was tried in which the motion of the diaphragm was adjusted to avoid exciting its resonances, applying a novel approach that Oclee-Brown referred to as “coupling maps,” which were used to evaluate the extent of resonant coupling in the frequency range of interest. The equations revealed that if the diaphragm resonances were exactly the right type of profile, the negative and positive components of the excitation could be made equal, so they don’t excite the cavity modes.
Figure 3 and Figure 4 show the second mode for air and structure respectively. The pressure resulting from movement is combined across its surface and for every combination of structural and acoustic mode to produce the coupling map. In effect, diaphragm resonances may occur within the operating band and not radiate.
All Roads Lead To Axi
While Oclee-Brown was completing his PhD, I was working on a wideband compression driver array, exploring the impact of this more complex design by combining the outputs of midrange and high frequency drivers – a dual diaphragm approach. Two diaphragms meant additional cavities, which resulted in acoustic resonances that were extremely hard to avoid. Solving this seemed like an ideal road-test for Oclee-Brown’s coupling maps technique.
Further analysis led to the idea that a single driver using the new approach might just be able to cover the wider bandwidth without the compromises necessitated by the use of more than one diaphragm. It was a very risky path to take, especially since the amount of work involved was very substantial.
It came down to a very practical problem: finding a diaphragm geometry that would allow the profile of the resonant motion to be tailored to avoid exciting the cavity resonances, as well as provide sufficient air movement to give good output at relatively low frequencies. A viable diaphragm geometry was necessary to enable the rest of the development work to take place.
My inspiration for the diaphragm actually came from a very unusual place – the roof of the original KEF factory. The roof of the factory was made from corrugated material in the form of a Nissen hut (a prefabricated hut used by the military during World War II). Here a thin sheet of material is made relatively rigid and strong by bending it in a repeating pattern. Mimicking this tried-and-tested architectural principle, I produced a number of diaphragm pattern designs, which I then analyzed and compared using the “coupling maps” technique.
Several geometries were tried, but one emerged as the most suitable and the process of optimization began. This was not just a matter of clicking a red button in the software labeled “optimize,” but rather required fundamental changes to the design approach, including improvements to the coupling maps technique, and a crude attempt to ensure formability, making sure that at the end of the process it might be possible to manufacture the diaphragm successfully.
After producing a good attempt, the performance was reviewed and the “goal posts” moved to the use of a larger 5-inch voice coil to further increase low-frequency output. During this process, literally thousands of finite element models were solved, emphasizing the critical nature of the computerized modeling process; it would have been completely impossible to generate that many iterations using the time honored “cut-and-try” methodology.
Following several months work, an optimized diaphragm design yielded gratifying results. The first radiating mode was at 9 kHz, much higher than would be expected for a dome radiator with the same low-frequency output (5.25-inch diameter)! This was a really exciting result since the bandwidth could be extended beyond 9 kHz and the team decided to push ahead with this design and develop it to prototype stage.
Making It A Reality
Because the driver was completely new and original, every constituent part needed designing and optimizing. Routing the air channels, magnet design, voice coil design and rear loading all were designed and then refined using FEM techniques.
The first prototype driver was built using individually machined parts; but the diaphragm needed to be fully tooled. Waiting for the diaphragm tool to be completed was a stressful time since all of the past work was dependent on whether the diaphragm could be actually be formed as a single piece of titanium.
Fortunately, the approximation used as a criterion for formability turned out to be right, and after a few heart-stopping weeks of tooling adjustments, a serviceable diaphragm was produced. Meanwhile, a suitable manufacturing process for the phase plug was found and a voice coil tooled. After the many years of intense research and development work a new driver was born.
The resultant product is the Axi2050, which employs the thin, annular, gently curved 175 mm-diameter titanium diaphragm as designed by me and Oclee-Brown, together with a single 125 mm voice coil coupled to a multiple-aperture phase plug.
The precise nature of the diaphragm’s three-dimensional corrugation and its rotational “axiperiodic” symmetry (hence the product’s name) reduces the modal resonances in the driver to an absolute minimum, while the slight overall curvature of the diaphragm ensures that the initial breakup mode is very high for a diaphragm of this size.
The titanium diaphragm is thin and very low in mass, which makes it is sensitive enough to accurately reproduce frequencies up to 20 kHz, but at the same time sufficiently rigid and large in diameter to reproduce frequencies down to 300 Hz and below (horn dependent), at high SPLs. The large diameter also confers another benefit – the compliance of the driver changes little with large excursions at low frequencies, which keeps low-frequency distortion (to which human hearing is particularly sensitive) at a minimum.
Finally, the wideband response means one driver can cover the entire mid/high-frequency range, doing away with the need for a MF-to-HF crossover, making this an exceptionally low-distortion device across the frequency band.
Many companies claim that their new products are the result of many years of R&D; in the case of the Axi2050, that’s certainly true. The Celestion team worked for several years on mathematical tools and modeling software before any hardware was designed at all. But as a result, and thanks to the power of modern computing, we’ve been able to bring the performance of this wideband MF/HF driver to its optimum while gaining better insight into the physics at work in the products. That’s a win-win.
The Celestion Axi2050 is the first product in what may well become a new range of AxiPeriodic compression drivers, and provides a good example of what computers can do for product design when coupled with engineering innovation.