Everyone,
Stepping back and reviewing all of the recent developments regarding the crossovers in our loudspeakers has given me cause to reflect a little. If one takes it all in, it has become obvious (at least to me) that the “big picture” is there for all to see. I will explain.
As a preface, I will say that there are other areas that apply specifically to loudspeaker design that do not come into play in the other components of the signal chain. Specifically, they are matters that pertain to the acoustic domain. Those are things like dispersion and diffraction, which, if they do have electrical analogs, are very small in comparison to their effects in the acoustical performance of loudspeakers. As such, the following does not address those issues.
In the world of engineering we have two major subdivisions regarding the transmission or reproduction of information (signals/waveforms). These divisions identify the dynamic extremes our equipment must be able to handle without injecting errors (distortion). At one end we have the weakest signal levels, which in engineering terms we refer to as the “small signal condition.” Logically, the other extreme is referred to as the “large signal condition.”
These terms are most often referred to when discussing amplifier performance, but they actually apply to any device – including loudspeakers. In the loudspeaker design world though, most designers perceive the most difficult issue to address as that of the large signal condition – and rightly so as it is the most obvious and difficult to contend with. In fact, this is so true that many designers – including myself in the past – have a tendency in their efforts to overlook the much finer details that encompass the small signal condition.
There is one other good reason for doing so as well…the designer often “believes” that he has little influence over these parameters, if they really even exist. The issues of the small signal condition are very difficult to measure – if they are measurable at all – and are essentially “built into” the parts we are forced to use. It appears to me now that once the large signal issue has been sufficiently dealt with, the “last frontier” is that of the small signal world.
As one steps back and analyses the ongoing debate and struggle that the dedicated audiophile often finds himself in regarding cables, connectors and such, we see that the small signal realm is the source of great controversy. Anybody can hear when a loudspeaker or amplifier is driven into gross distortion. In fact, the resulting distortion that occurs when any device – say even passive ones such as non air-core inductors – is over driven, is clearly audible and easily measured. We call such distortion as being the result of “saturation effects” and they must be avoided at all costs. Why? Because when the dynamic headroom of a device is extended, all errors and artifacts produced at lower levels are typically reduced as well. A device that can cleanly reproduce large signals will almost always sound better at any other level below that point.
Therefore, due to the extreme dynamic peaks modern digital sources can now achieve, virtually every designer finds himself forced to focus on this one parameter far more than that of the other extreme. Rightly so…as it makes little difference what the small signal performance is when the system produces significant distortion at, or even slightly above, average volume levels. This issue is so dominantly important that once a designer has truly achieved success in overcoming it, he can easily be deceived into thinking that there’s not much left that he can do.
In regard to amplifiers, the designer is only left with the “noise floor” at the other end of the dynamic scale, and if that is sufficiently addressed his work is essentially “done.” This assumes of course, that it is a good design such that “static” distortion and the rest are acceptably low as well. The amplifier designer has one other weapon in his arsenal to correct small signal errors though – negative feedback. The poor loudspeaker designer is not so fortunate. We run “open loop.”
In order for us to begin the effort to reduce or eliminate these small signal errors in loudspeakers, we must first address their source. The most obvious of these is “resonance.” Resonance occurs most commonly in transducer diaphragms. We call these “break-up modes” wherein the cone or membrane flexes in a manner that produces “standing waves” that continue to travel back and forth or “ring” long after any signal has passed that has acted as a stimulus to begin the process. As loudspeaker designers, we have a universe of drivers to select from that exhibit varying degrees of these effects. Ideally, we select those devices that exhibit these effects to the smallest degree possible while achieving our other large signal objectives.
Having done so, we typically find ourselves with a “completed” design and at a loss to effect much further change. But there are other sources of distortion and improvement if we choose to look more deeply. “Hysteresis effects” are always lurking in the background. Hysteresis is simply described as a sort of “memory” in a device that causes it to behave differently when signal polarities change direction. A device can “hang onto” a state that it has most recently been in before an instruction (signal) demands that it change vectors (direction of motion or current flow). Distortion is the result if this occurs.
Every real-world device manifests some level of hysteresis and loudspeaker drivers are no exception. Hysteresis effects, if they are severe enough, will generate measurable levels of distortion and thus, drivers are selected that exhibit it at a minimum in a good design. But…the type of distortion products that hysteresis causes in drivers usually manifests itself under large signal conditions, therefore, lending themselves to easy observation and correction in the driver design. While the effort to reduce hysteresis in a driver may not be easy in itself, since it can be easily measured, at least we have a place to start. Inductors in the crossover network can be obvious sources too and behave much the same under similar conditions as that of drivers. Once all these have been sufficiently dealt with, it appears we are done.
