David, A few more questions. . .
Efficiency. Clearly, the basic benefit of increased driver efficiency is an increase in SPL for a given electrical power level. Louder can be better in an absolute sense, and high efficiency can let one use a low power amp that has particular merits other than high power.
Measures taken to increase efficiency tend to have trade-offs. I can only give a few simplistic examples, here, and the existence and degree of trade-offs are not absolute in every case. Adding a horn to a driver improves the impedance match to the air load, which increases efficiency, but has the potential for adding some cavity coloration and diffraction effects. Narrowing the magnetic gap in a cone speaker's motor increases efficiency, but alters its characteristics in other ways, where balancing off the side effects can lead to increased distortion or coloration. Narrowing the electrical gap in an ESL increases its efficiency, but demands a set of compromises that ultimately increase distortion. There are lots of other examples.
JansZen One crossover count. The use of two crossovers is the inevitable result of the JansZen One being a three-way system.
It is a three way system for one of the two usual reasons: to improve dispersion uniformity across the frequency spectrum. Drivers have a dispersion angle that decreases linearly with increasing frequency. When there are drivers of different sizes, with the smaller ones carrying the higher frequencies, then for the portions of the spectrum that each one carries, the dispersions can be made to overlap advantageously. This works out nicely for all driver technologies, because smaller drivers are generally more appropriate for higher frequencies anyway.
The other usual reason for using more than two driver sizes is not important in the case of JansZen or ML or any ESL maker I am aware of, which is to avoid placing a crossover frequency in the 1 kHz to 3 kHz band where the ear is most sensitive to the slightest problems. In 2-way dynamic systems, driver capabilities usually force the crossover frequency into this band, and athough the sonic interference and phase problems can be largely addressed by the use of complex, 24 dB/octave crossovers, such crossovers have their own problems.
Crossover nefariousness. Generally, the topic is very complex, but crossovers are not inherently bad, although there are many factors that can affect the result. It is easy to get a mediocre result, and possible to get a bad result, I could say, even with all the software available for designing them. Unfortunately, as with most complicated things, for best results, preparation, thought and experimentation are required. Cost tradeoffs can also affect performance. Crossovers that operate on the source signal before amplification offer a lot more flexibility than passive crossovers operating on the amplified signal, and allow a much closer approach in general to ideal behavior, but they require multiple amplifiers, one for each driver.
Some insist that crossovers are bad and the best response is created by not having any crossover at all, while placing all the frequencies together onto one transducer, or onto a compound transducer where the sum of the parts acts as one. This no-crossover concept has intuitive appeal, but the underlying concept is fallacious.
The first problem is that there is always a crossover, except in one very specific case. In an ESL, there has to be something to counteract an ESL's naturally rising response with increasing frequency. The natural increase is 6 dB/octave for an ideal point source, and 3 dB/octave for an ideal line source. This occurs because an ESL is not mass loaded, at least not at audio frequencies, so unlike a dynamic speaker, it does not take increasing amounts of energy to move its membrane as the frequency increases. At the same time, as with any transducer, the dispersion angle decreases with frequency. Without mass loading to attenuate an ESL's output, this narrowing of the angle concentrates the same amount of sound into decreasing volumes as the frequency increases.
To attain a flat response, this characteristic must be counteracted, and the means for doing so is for all practical intents a crossover. It operates on the entire audio band at once, rather than short frequency bands, but it works in essentially the same way. The one exception is an infinite plane ESL transducer, which will not require compensation for varying dispersion, because there is no variation -- there will be zero dispersion at all frequencies. Sanders Sound Systems makes an ESL that approximates such a plane wave, at least at mid and high frequencies, and I presume this allows the elimination of any type of crossover or compensation circuitry above where the dynamic woofer crosses out. Since there is practically no dispersion from the ES transducers, they must be aimed directly at a single listener for full spectrum sound, but I must say that IMO the effect at this position is remarkable.
Then there are the two trade-offs:
1) A full spectrum ES transducer of practical size driven by way of a single circuit will have high capacitance, and thus low impedance at high frequencies. Impedances below 1 Ohm at 20 kHz are not unheard of. The capacitance will also add lag to an amplifier's feedback loop, potentially destabilizing those with modest phase margin. This is somewhat relieved by the response compensation circuitry, i.e., the crossover, but not altogether so.
2) If one is not intentionally designing a plane wave transducer, the extreme narrowing of the dispersion at high frequencies from the large area has to be managed. To get dispersion, the transducer must either be curved or broken up into facets that are arranged at a series of angles relative to one another. When the solution is a curved membrane, the overall width leads to lobing at lower frequencies than one would get from a narrow transducer, increasing the likelihood of their audibility, i.e., changes in the high frequency response and phase as one makes small changes in position might be noticeable. It certainly is with pink noise, but music is another animal. The use of facets has a similar effect, where each facet in the aggregate has wider dispersion for a given frequency than the entire width would. Since each of the individual "beams" overlap by a finite amount, this creates a similar set of diffraction and interference effects as the curved arrangement.
It is important to note, however, that the effects of both tradeoffs are not necessarily audible, or when they are, not necesarily objectionable. As usual, let your ear be your guide.
In a dynamic speaker, the mass loading supplied by the cone, coil, bobbin, etc. acts as a continuous crossover. This is in every sense the mechanical equivalent of the electrical circuit needed for ESL's.
On an interesting historical note, there was once an ESL company, Acoustat, proclaiming that crossovers are bad, but still using classical, 2-way crossovers in their speakers. Although they did apply the entire spectrum to their panels, it got there after being split by an ordinary crossover into upper and lower frequency ranges, sent to separate high and low frequency transformers, and then recombined after being stepped up. Electrical recombination is probably better than acoustical, but there is never a need in the first place with multi-way ESL's or ESL hybrids to cross over at a troublesome frequency, since ESL tweeters can produce a very wide spectrum without difficulty. Others may be doing the split transformer thing now, for all I know. Seems like a good idea, especially if one admits it's a crossover, because with a full-range-spectrum-to-all-panels type of ESL, the transformers become easier to make, and the capacitance at the binding posts should be thereby reduced, increasing the impedance at high frequencies and improving the amplifier load.
JansZen One crossover frequency. Already talked about it a bit, but yes, lower is better, down to a point, so to speak. This is because the more spectrum carried by the ESL panels, the better, until one reaches the point where dynamic woofers can do the job just as well, and this also happens to be the point at which floor effects around the speaker become inaudible compared to broader room effects.
Dual woofers. Please refer to my answer to User211's (Justin's) question, near the middle of page 3 of this thread. posted 07-10-2009, 10:31 AM
Bass accuracy. A speaker with a reasonably flat frequency response in a reasonably flat room has at least frequency response accuracy going for it, although it may still be far from accurate. Bass is a part of the spectrum, of course, and ideally should be a part of that flat frequency response. One can then boost or cut bass to suit personal preference. A flat frequency response in the bass, though, can be particularly deceptive, because the accuracy can still be quite poor. Of course, the usual problems are audible, although some will prefer various forms of inaccurate response.
Bass extension. At first, this seems like a question about how deep the bass should go, and the answer is, as deep as you like, but of course there is more to it than that. Some prefer natural (accurate) bass, and others prefer bass that is subject to various distortions or adulterations.
An all time favorite is the extra whump that one can get from a ported enclosure with a low cutoff, which is in addition to the extra extension one gets compared to a sealed enclosure. It is a side effect of poor time domain performance, where the cone motion is poorly controlled near and below resonance, and tends to overshoot.
FWIW, at this time, even when the porting and tuning are done in ways that maximize the time domain performance, we believe that a woofer using a ported enclosure can not be made to integrate convincingly with an ESL.
Another favorite is a bass response that has a broad hump in the vicinity of 100 Hz, which can substitute for deeper bass according to many people's perceptions.
2nd harmonic distortion is pleasing to many people, especially in the bass.
A woofer with 2nd and 3rd harmonic distortion in a certain ratio will create the impression that the fundamental is present at higher SPL than it actually is, i.e., using psychoacoustics to deepen the apparent bass response. There has been at least one iconic speaker that was much smaller than one would expect is necessary, which relied in part on this effect. Teenagers and young adults may be happy with it, but mature listeners who find it convincing will also find it fatiguing.
If one has decided that natural bass is preferable, where smoothness and flatness of response along with a lack of distortion or coloration is the object, then there is still the matter of how deep to go. Our Model One in its standard configuration is flat within 3 dB in-room to about 30 Hz. This is below the lowest note on a bass guitar or double bass, which is E1 at 41 Hz, and matches the lowest note on a 5-string double bass, which is B0 at 31 Hz. Bass is reproduced very convincingly and realistically, although those who are accustomed to boom boom bass may find it lacking.
To radiate the kinds of artificial bass content that some types of music specialize in, such as hip hop, deeper extension and some deep bass boost are needed, but for that, subwoofers are available, although to retain accurate reproduction of musical content, most will have to be modified to cross out at a low enough frequency and at 6 dB/octave. Accuracy is not important in this context, but it is still worth mating the sub properly to avoid humping up the rest of the lower bass spectrum.
