Assumptions

In any loudspeaker design or loudspeaker simulation there are assumptions made. Understanding these assumptions is critical to understanding the resultant product. We feel the need to outline our design philosophy (assumptions) so that users know what they can expect the simulation to do. We constantly get requests and questions about features. Some we implement others we do not. We implement those features that we feel are important and we do not implement those that we do not believe are important. Importance to us starts with the major factors which govern the subjective perception of a loudspeaker system moving down to those factors which do not have a major effect on the perception. We are interested in sound reproduction. Effects that have not been proven to make a significant audible difference in the design are not deemed important no matter how important intuition says they are.

Simulations need not be highly accurate. This is because, from experience, the production variations in driver manufacture are usually greater than the accuracy of a detailed simulation. Doing an analysis to five decimal places makes little sense in a +- 10 % world. Simulations, therefore, are more useful for indicating the general design approach and studying design alternatives than performing detailed analysis. Detailed analysis can sometimes be useful for large production- low cost components or systems where "squeezing every penny" out of the design is fruitful, but they are seldom useful when the design is not highly cost driven (i.e. pennies count) . Given this approach, the simulations need to be fast to allow for multiple trials, where the program is placed into a kind of human optimization loop.

Speak is the result of more than twenty years of work on acoustic simulations. We use it extensively for our own research into fundamental questions of design and several of our various publications. It has been used to develop and prove out numerous concepts, many of which have resulted in patents or patents pending. We make it available for others to experiment with also.

We have always believed that polar/power response is just as important, if not more important than the axial response. The reasons for this are well documented in our books. More and more designers and researchers have also been coming to this same conclusion. It is these aspects of the design that we are most interested in and Speak reflects that belief. Speak excels at modeling the polar response of loudspeakers and systems

As discussed in our book, we believe that rooms should not be too dead because then they sound dead. A live room makes the power response very important. Dead rooms need only control the systems axial response, but since these rooms don't sound good anyways it bringing us back to the importance of the power response. This belief does not apply at low frequencies (below the Schroeder frequency), but that's another lecture in itself.

Distortion is also important and Speak does an good job at this. The problem here is that second and third order non-linearity's are not the ones that cause the most audible effects. As such, all orders higher than the third should be absolutely minimized. There must be third order non-linearity since no flux field or suspension can allow for infinite travel. There is usually second since complete symmetry is difficult to achieve. Thus the lowest two terms must always be present to some degree and therefore they must be understood and controlled. Higher order nonlinearities should not be present under normal operation.

Cabinet diffraction is another interesting subject, however, since there is no "good" cabinet diffraction, a "good" design minimizes these effects. Diffraction is easy to avoid in principle but not always easy to implement in practice. Diffraction occurs at any change in a surfaces slope and an increased sharpness of the edge increases the diffraction. Mathematically speaking the amount of diffraction is proportional to the second derivative of the surface. This means, in effect, that rounded edges will minimize diffraction - the smoother the better. Taking this concept to its extreme yields a sphere - the enclosure with the lowest possible diffraction. For this reason we have implemented a spherical enclosure model because no enclosure can have less than this amount of diffraction (except for an "infinite" baffle which can never really exist in practice). We see no reason to model something which we know very well should not be in a design to begin with.

Optimization is an area that we have thought about, but alas this is not easy. It is an easy matter to implement a filter optimizer which flattens the axial response of a measured set of drivers. But, as we said, we do not feel that this is a good tradeoff if it results in a poor polar response. Optimizing the power response is an entirely different matter. In our opinion, optimization is best done by a knowledgeable designer using a rapid analysis tool like Speak. We may someday implement a form of optimization, but it is not our first priority.

We have added phase response in later releases mostly for interest. Phase has never been shown to be a principle effect. . SPEAK has always used complex internal calculations, so the phase was always there, but plotting it was not of much interest. We do not calculate phase or group delay as these delays serve no useful purpose they are not audible.

Our current interests are in the areas of improved subwoofer design (mostly using levers), the measurement of stable and accurate loudspeaker parameters (most methods in use today are not at all stable) including the measurement of parameter nonlinearity, a very hot topic as of late, and thermal effects (some of these exist in Speak today). We have one issued patent in this area and several more pending.

Last modified May 20, 2008 by Earl Geddes