What if you designed your ultimate speaker?

A good question to ask is: Why nobody else is doing it (1/2 octave design)?
Do you know something they don't or is it simply not cost efficient for others to manufacture? Is it for you?
Dealer markup is often in order of 50%. Company markup on components is at least 50%. It means that company has to build it for less than 1/4 of the sale price. Each of $2k speaker pair has to cost in parts less than $250 to break even - including cost of cabinet. If you can make speaker box for $50 (don't even attempt curved walls) and crossover for another $50 we're left with $150 for all transducers. Good tweeter alone can be more than that.

Wouldn't be better to simplify design and use quality transducers instead of array of junk.

Good components are expensive. On one end you have Mylar capacitor that cost maybe a quarter (often reason for tweeter glare) while on the other Duelund caps at $500 a piece. I assume you will use at least good polypropylene cap at $5-$10 and a lot of them. Many people believe that any form of plastic introduces glare and the best caps are oil/silver or paper/copper (like Duelund).

Some people believe that instruments with complex harmonic structure like piano can only be faithfully reproduced with headphones because of speaker's crossovers. Is adding more crossovers going to help preserving the phase?

Yes, we are very far from the sound of live performance but speaker is only one element in the chain. Dynamics are already compressed in studio and even more harm is done in the media/playback. Your comparison of the tweeter's membrane to size of the cymbals is not fair. You're likely to be 300' from cymbals and 10' from the tweeter. Since power quadruples when distance doubles the same tweeter at 10' requires 1000x less power than tweeter at 300' - Cymbals don't look that big anymore in proper scale.

80Hz that you mentioned might be just harmonics of bass refleks tuned to about 40Hz. Many smaller bass refleks speakers with extended bass show strong hump around 80-100Hz.

20 way? Are you going to have 20 inclosures too? Is there anyway to predict how 20 drivers are going to react in the same box? So what if the X-over is only sending them 1/2 octave, what about back waves (passive atunatoin) from the other drivers. And what about point soucre imagining, that is probably more inportant than getting every last Hz correct, imhp. Rememeber, a speaker is a box and a driver?

I did think of having a speaker for every insturment in the band. Fed by a recording with a chanel for each insturment. It maybe could work for a trio or quartet?

Zaiksman, have you seriously thought about the size of the speakers? Say to cover 20Hz to 30Hz you use a 15" driver. To cover 30-40Hz you use a 12" driver. For 40-60Hz a 10" driver. And so on... Are you getting a picture? To design enclosures that would properly accomodate the bass and lower midrange drivers alone you would be talking about something the size of a minivan! Of course you could get around this by designing the most complex concentric driver system ever envisioned by mankind. But conceptually isn't that what the Walsh driver or the Quad 63s onward are? They are far easier to implement than what you propose and IMO more elegant solutions.

Allow me to point out that there are and have long been not only 2-way, 3-way and 4-way speakers, but also 5-way and even 6-way designs. So to my mind the multi-way concept in general is clearly a scalable one, and don't let the 17-way thing throw you -- it's not fundamentally different in kind, only in degree.

There's nothing so complicated about it in theory that it couldn't be done, and in fact my supposition is that it might actually be easier to cross-over drivers whose passbands are more restricted and whose physical designs can therefore be made more similar to their adjacent neighbors. Certainly it would seem to me the crossovers themselves needn't be anything but simple, first-order types, without extra compensating elements, since each driver could be electromechanically optimized to have its flatest reponse within its narrow passband, and out-of-band misbehavior would be far enough removed to not be an issue even with a low-order rolloff, I think.

Likewise, my point about not needing 'exotic' tweeter materials, like diamond or beryllium, doesn't imply the use of "junk" drivers. It simply means that if a tweeter isn't called upon to handle 3 octaves by itself, but rather shares a divided workload, with a more restricted passband, then each tweeter can be sized optimally for its particular passband. So the elimination of the need for a relatively large 1" diameter in the top-octave driver ("large" relative to the wavelengths in that octave, in order to go down to a circa 3KHz crossover point with the midrange, as in most conventional designs) naturally leads to lower mass and greater rigidity without resorting to exotic diaphragm materials, while pushing the breakup resonance well above the audioband as well as gaining wider dispersion in the process. (A 1" tweeter becomes significantly directional above 10KHz.)

By the same token, each driver in this system would experience less excursion, meaning lower distortion (and heat), perhaps mitigating some normally conflicting motor design demands. But of course none of this would mean the drivers in such a system couldn't or wouldn't benefit from the decades of development that's gone into making drivers capable of high performance in covering much wider passbands.

For me, the question of why is it that conventional tweeters are commonly allowed to run past the point that they begin to beam in the top octave, raised the further question of just how many -- let's call it n-ways -- 'should' an ultimate speaker design be? An idealized number (like say, one)? Or an arbitrary number? Because that's all dividing the audioband into bass, mid and treble ranges represents -- an arbitrary division, having no particular physical, musical, or acoustical basis, although 3-way design can work decently, as can 1-way and 2-way designs, and we're used to it. A two-range division might have the most basis in reality: Fundamentals producable by pitched instruments and voices, and above that their overtones. But that's clearly not precise enough for audiophiles when it comes to describing either music or especially sound, which is why we talk crazy about stuff like "upper mid-bass" and "low treble", and it isn't necessarily the best engineering solution either.

How about 1/3 octave divisions, like are found in a studio graphic equalizer? Seems like too much, and again it's arbitrary. But eventually I hit on the idea -- given that the loudspeaker is the leading source of harmonic distortion (and also nonlinear distortions) in the audio reproduction chain, and given that the prevailing distortion product is the second harmonic -- that a reasonable basis for arriving at an 'ideal' number of divisions could be to take advantage of the crossover network to help usefully surpress the second harmonic of the lowest fundamental within each driver's passband, leading to at minimum half-octave divisions (assuming the minimal -6dB/octave slopes). And in this case as in many others, the minimum required level of complexity to acheive a certain engineering goal seems best -- thus a 17-way system (assuming my math and the overall concept aren't basically in error concerning this whole notion*).

But some of you posters are getting close to something that I should bring up now, which is of course that there's nothing new or unique about the idea of using many more drivers than is typical in order to help lower distortion and increase power-handling/dynamic capability, which is a big part of what I'm proposing. Elevick above mentioned Pipedreams -- just one of the designs that have aimed to achieve a simulated or quasi-line-source radiation pattern through the use of vertical line arrays consisting of multiple, closely-spaced monopole or dipole point-source mids and tweeters, and all of which I guess are spiritual descendents of the Infinity I.R.S. design, ultimate versions of which used up to over 100 drivers.

Obviously, dividing up the workload among that many drivers, with that much total radiating area and motor volume, will bring the THD way down vs. conventional singled- or doubled-driver designs, even without employing my 2nd harmonic crossover-surpression theory. So what's the benefit of considering a very 'high-way' design such as what I'm proposing, when these existing mega-designs can achieve that low-THD end using only a standard 3-way configuration? Next post...

*Here's the frequency intervals in cycles per second (Hz), calculated by starting with 20Hz and proceeding upward in 1/2-octave steps (multiplying by 1.5), rounded to the nearest integer:

1) 20-30
2) 30-45
3) 45-68
4) 68-101
5) 101-152
6) 152-228
7) 228-342
8) 342-513
9) 513-769
10) 769-1,153
11) 1,153-1,730
12) 1,730-2,595
13) 2,595-3,892
14) 3,892-5,839
15) 5,839-8,758
16) 8,758-13,137
17) 13,137-19,705

Interval #18 in this series would be 19,705-29,558, though of course there would be no need for an upper cutoff, and the whole interval may be considered ultrasonic and therefore perhaps unecessary. Alternatively, if the series is begun at 20,000 cycles and divided by 1.5 in 17 steps, it does end very close to 20Hz. In any case, the precise numbers aren't what I'm presuming to be important here -- it's that all of the driver frequency ranges are restricted such that they permit at least a >3dB surpression of the 2nd harmonic for any fundamental tone within each particular range, thus in theory providing what could hopefully be a worthwhile reduction in aggregate THD over the entire audioband.

There is no doubt that removing lower frequencies from a driver will help it perform more smoothly in its upper range.
My Feeble mind can't grasp, Why beyond maybe a 5 way.
Bass, mid bass, mid range, mid high, high.
When you are talking the benefits of reducing cone break up, thats about it. I cannot see a 17 or 20 way working. Theoretically yes, but in practice, finding drivers that complement each other, ie radiation patterns, sensitivity matching, enclosure type and size, etc.
Don't get me wrong, you can certainly use more, I'm just asking where the benefit stops. Tim