...are worse than 3-way?

Yes.. and no. The thing is a certain combination of drivers might require a crossover double the cost of the drivers. It'll sound good, but elegant it's not. And there's a cost in power loss, too. Not to mention that more drivers are naturally less efficient. So you can use a hugely powerful amplifier to drive a speaker that behaves like a load of bricks- sort of like an american 'sports' car. Or you can design a speaker that is much more responsive; efficient drivers coupled through a crossover with the least components; sort of like an english sports car. It's elegant, and should be cheaper, too. Such elegance is easier to attain with 2 driver than 3. In fact, past 2 way you often see a leap to 3.5 or other exotic multidriver solutions with a crossover that could heat a room.

Full Range Drivers then?

So if less is more, even less should be even more, amirite? Full range drivers cover the entire spectrum, and so you need only one speaker, in theory. However, they have weak bass that requires an enclosure that can invert the phase. So a bass-reflex, horn, transmission line.. All of which bring their own issues to the table. The biggest issue however is that they are a point source. Beautiful for that warm classical music where the whole room comes into play, acting as a concert hall. But I'm looking to build a home theatre system. This means as less dispersion as possible: rear speakers will act as the 'room'. Ideally you're looking for a 'sweet spot' the size of your couch.

Well, what about horns?

A horn speaker has a beautifully directed dispersion pattern, just what the doctor ordered. However, a horn only really works well on a specific frequency... you'd need lots of different horns to take advantage of one driver. There are, in fact, designs that do just that. Volvotreter's site gives you a good idea of how it would look- and the WAF and placement issues of such a system.


Another point against horns is the 'beaming' on axis where higher frequncies are bundled more than mid ones. There's solutions for this though, a simple one being a cut-off and seperate tweeter- but then we're back to the complexity of crossovers.

So you can't win?

Now we're getting somewhere... Don't forget the all-important WAF, neither. And try and remember that counter to that is a gadget freak trying to express himself. The more obscure your speaker design, the happier you might think you'll be. But is it worth the hassle?

The plan.

A two-way design to keep costs down. Now for a home theatre system the most important sound placement is that of the human voice. So there's a good reason to keep that crossover as far away from the human voice as possible. I've found a few references placing the highest frequency at 3k or 3400hz and the lowest at 85 hz. THX puts the subwoofer crossover at 80hz. Given the slope this means the speaker need not go lower than 50 hz and no higher then 3400hz. A tweeter will pick up the high end, the low end will be handed to a subwoofer.

Caveat

Now there's a caveat. I'm assuming such a driver will give a good flat response in the range of the human voice. It might however be that a design with a lower crossover has better results. It will depend on the crossover design, driver performance, baffle step, etc. As the Dutch say: 'meten is weten'.

Checklist

I Started by making made a list of DIY 2way speakers that have been built to good results. I used many sites, but in the end only the speakers found on lautsprechershop.de (strassacker)made it as their information was consistent- after setting a standard I could easily add other designs. I wanted to spend less than 300 euro a pair for pieces, excluding woodwork, glue, possible replacement binding posts and those spiky things which my SO never need know the cost of.
First step in triage was to check the crossover frequency. The average crossover turned out to be 2650hz. But there were many crossovers on or over 3000hz, wich suited my 'human voice' theory.
At the bottom end I needed to make sure they could follow the THX cutoff down from 80hz to 50hz. Not many passed that test.
The next thing I wanted to check for was the simplicity of the crossover. Itt might be a simplistic ide, but I do believe that 'less is more'. However there is more to it; i also believe that issues should be dealt with in the enclosure design as much as possible. Well, mostly so I can show off what I picked up on speaker design, which is not on the electronics side.. I never did do my electronics homework (sorry mr. Van Haelst).
The average amount of parts was 7.5, although the Visaton 5750 set with 16 parts (guessed from picture) took the crown.
The last thing I wanted to look for was efficiency. Here I had to rely on the values from strassacker.com. All of these criteria narrowed down my list to a handful of contenders

Flip a coin..

..and you know immediately what you wish for. Not that the selected speakers were bad, but they just didn't get my heart pumping.
The speaker I was hoping for was the CerAl 4.2 And this mainly because of the look of it's aluminum Tang Band drivers.
So why did it fail to make the selection? First of al it had a cut-off frequency of 2500hz. Which meant depending on the slope it would cut over in the high end of human speech. Again, this need not necessarily be an issue; I needed to take a look at a response graph.
At the bottom end though, we were in trouble... 69hz at -8db? It's not a problem as far as the human hearing is concerned- we stop hearing direction at about 150hz- but would THX's crossover slope work? If not I'd neet to set a higher crosover in the receiver, wich might not be the optimal way to treat the material.
If the design had one glaring problem it was being new. No-one had built them yet, no-one had tested them yet, outside of K+T magzine.

From the ground up

Either way, I had a functional stero set, so as I came into a limited budget I prioritized bought the subwoofer set. I looked at the subwoofer enclosure that came with the set, and while I could certainly go smaller (bassexciter anyone?) I liked the size. In fact, it had about the same volume as the 'chinese' cabinets I was looking to build and put under my components. After some research on downfiring subwoofers I settled on a temporary re-design with the same volume, and made some sketches for the missus. My brownie points appeared sufficient; permission was granted, more or less. Acting fearlessly I orderered the parts- and that is where the saga ends, for now.

To the subwoofer!

In strange aeons..

..one learns much, much more than a mere mortal should. About voices for example, and listening. I had come across the whole 'voice' thing a lot while researching for my speaker choice. Most of it was off-handed remarks by designers and hobbyists.
'The human voice is placed mostly between 100hz and 2000hz' or other thumb rules of this kind. At one point, I dug deeper, and found actual voice experts treat the subject amtter slightly different. Here's a link that shows you how different.

One nugget of gold on the same site was a test where you could check your own 'frequency response'. This was my response, with crappy headphones, during daytime across from a construction site:


(I'll try to measure my response again, at home and in the evening- there's another construction site right around the corner- with sennheiser open back headphones) Beware: this shows relative sensitivity rather than absolute unless you calibrated your system. But the site gave me a good idea of what is important: not the voice, which covers the whole spectrum, rather than the bit always mentioned, but the range we can hear, and in that range the harmonics and timing. Put an orchestra in a cathedral and its acoustics become more important than the instruments'. Feed a recording trough your ears and- exactamundo.