Very good questions - especially the one from Josh!
Josh...
When considering the effects of a line array and drivers I think the easiest/best way to consider this is in terms of the motor and the cone. With respect to the motor, Josh is right on the $. The major advantage of a good motor system with shorting rings (i.e. Scanspeak or SEAS Excel) is the very linear X-Max region. Motors that don't have such rings in their motors will have a shorter linear region within their X-Max. This is because the flux field is strongest when the voice coil is centered on the pole piece and decreases as the voice coil moves toward the limits of X-Max. John Paulsen (SEAS) explained this to me at CES. Soooo, when applying all of this goop to the concept of a line array there isn't much need for a good motor because that cone isn't moving very far. The voice coil will remain almost dead centered in the gap between top plate and pole piece.
The cone is another issue. If the cone flexes under stress at 3000hz then it will still flex under stress at 3000hz. 3000hz remains 3000hz no matter what.
There are two other issues that are fair to address here. They are the specific design of a cone driver for a line source, and the relationship between a smaller cone and cone flex. It is theoretically possible to get 30hz from a 4" cone. The reason this is not done is that the level of air displacement necessary to move 30hz at an acceptable level is substantial. With many of them it might be possible to get enough air displacement, but then there is is the issue of driver suspension. Driver dampening is normally setup for a certain frequency range. It just isn't possible for the suspension for a driver to perform ideally at all frequencies.
Next, just because a cone is smaller doesn't mean that it will be stiffer at 3000hz. I corresponded with Sigfried Linkwitz on this issue and he conveyed this truth. It was quite simple, but escaped me prior to that correspondence. I though that a smaller cone would always be stiffer because its outer edge is closer to the voice coil. This isn't always true. It is possible for a larger cone to have a stiffer cone than a smaller cone. I have a hunch that this might be true between the SEAS W18 and W15 cones, but am not sure. I certainly though the Joseph Audio RM40 (now called the Pearle) sounded nicer than the Joseph Audio RM33 in the midrange. There could be other factors with these specific speakers, but examination of the SEAS page indicates a strange rising distortion in the W15 at 2khz that is not present in the W18. I am guessing (again, not 100% sure) this could be due to cone flex. It is certainly true that the W18 and C95 cones are darn stiff at 3000hz. This isn't true with many/most smaller drivers.
Tweeters are another story. I don't believe that a cheap tweeter could ever sound good. I don't know as much about tweeters, but believe a few things. First, a cheap tweeter sounds mediocre no matter what volume level it is played at. This is a SUBJECTIVE opinion. I have listened to many some $25 tweeters. The all sound, well, like $25 tweeters. The difference between a $25 tweeter and a $75,$200, $300 seems quite nebulous, but there are differences. This leads me to the next point.
Second, I can speak with some degree of authority on the motor and dome of the Hiquphon tweeters. I will cite Osakars words as accurately as I remember them. When discussing the matter of "why the Hiquphon tweeters sound good" I really expected that he would tell me something really special about the motor structure. I figured there were shorting rings or something of the sort - nope. Oskar did spend about 10 minutes of telephone time addressing his attention to voice coil concentricity, and the tolerance between the voice coil and the pole piece. He maintains .07mm gap between the voice coil former and pole piece. He maintains .1mm gap between the voice coil and the top plate. He further explained that getting a very good magnetic linkage required a very tight tolerance and that such tolerance is not possible when using stamped driver materials. Opening up this tolerance allows the use of cheaper stamped materials, but also reduces the solid magnetic link between the voice coil and the top plate. Opening the gap increases distortion. This is obviously undesirable.
The second issue Oskar addressed was the dome. He said that most fabric domes are sprayed/doped, then pressed into shape. This creates an undesirable dispersion of the doping material on the dome. Then ferrofluid is added to damp the travel/overtravel of the dome. Oskar's domes are pressed first, then triple coated (OW1) with different coatings at differnt places on the dome for proper strength and dampening of the dome. Oskar conveyed that the SS Revelator tweeter and the Dynaudio Esotar also use this method of dampening. Oskar said that this method of doping a dome is very effective, but also very time consuming and therefore expensive.
Given what I know about tweeters (relatively little), I don't think (?) that it is possible for 20 low quality tweeters to have a sound quality equal to 1 high quality tweeter. I dunno though. This is obvously all theory. I didn't have the opportunity to a/b the pipedreams line source and the 1801, but recall that the pipedream speaker sounded very nice.
The problem of parallel/series isn't very difficult. You just hook up the drivers in a mixed series/parallel configuration and aim for about 8 ohms nominal. Given 4 8 ohm drivers the solution is simple. Hook two pairs in series, then hook those pairs in parallel. The result will be 8 ohms nominal. given 6 drivers, I'd probably hook 2 triplets in series, then hook those in parallel. This would result in 12 ohms nominal.
There are probably other issues to.
Hopefully this all makes sense. There is considerable rambling contained herein.
