How do we choose optimal points of crossover for 2- and 3-way constructions?

The SP38 is part of a journey into investigating the various ways of constructing a full range speaker covering a frequency range from 40-20,000 Hz. Above is depicted four different ways of achieving this goal.

Example 1 is the traditional 3-way utilising a 6-8” woofer, a 4-5” midrange driver working from around 350 Hz to 3-4 kHz and a 1” tweeter handling treble from around 3 kHz to 20 kHz.
Some people will take the 4-5” down to 150 Hz, but to my mind we are then talking satellites + subwoofer. If we want the midrange driver to handle upper bass/lower midrange we need a 6” driver (WATT/PUPPY).
The classical 3-way will be dealt with in the Acapella WB (wide baffle speaker).

Example 2 is an unusual approach – an oldtimer – using a 3-4” driver from around 800 Hz to 3-4 kHz, maybe even up to 5 kHz and a ¾”-1” tweeter up to 20 kHz. This was a quite common approach some 20-30 years ago and it definitely has its merits and it’s a gut feeling there will interesting things to learn from this set-up. See below. This is taken from a recent issue of HiFiWorld and displays the ear's sensitivity and strongly suggests that keeping away from the 1.5-4 kHz region may be an advantage. This is very much in line with the SP38 and Zahra + coming constructions of 3-way systems.
And by the way: An interesting article "Modern Speaking" by Noel Keywood, HiFi-World, October issue 2005.

This classic 3-way is going to part of my learning programme. First the TJL3W introducing the W11/W12 between the SEAS W18E001 and the ribbon tweeter and I might try the 18W/8531G00 and the ScanSpeak 13M/8640 or the 12M/4531G00, the latter if I can borrow a pair. I’ve got a pair of mint condition, second hand 13M/8640s, but the price of the new 12M is terrible.
A 3” driver can handle well down to 500 Hz and when used from only 700-900 Hz it has an ideal size for handling frequencies above up to around 4 kHz. At 4 kHz the wavelength is 8.6 cm. More than the cone diameter of a typical 3” driver. These may be classified as 4” drivers, but we’re talking 55 cm^2 cone area. Quite small - and usually fast and by "fast" i mean they can ad a speed to the upper midrange than makes you wonder why we keep building the classical 2-ways.

One reader has pointed out that taking the point of crossover between mid and tweeter up to 4 kHz does not remove this "problem" out of the range critical to the ear's highest sensitivity. 4 kHz is actually right at the peak of the ear's sensitivity. True! I have no explanation for the subjective benefits of adding a small upper midrange handling the 800-4000 Hz range compared to a 1" dome going down to e.g. 2.5 kHz or a suitable 5-6" going up to e.g. 2.5 kHz. Possibly the explanation is that by adding a small upper midrange we may have a driver with significant reduced distortion in this critical area compared to two drivers that work at the very limit of their optimal operational range.

Spendor, Peak Consult and Verity Audio have no trouble (?) taking a suitable midrange driver up to 3800 Hz - and with apparent success. And I'm sure that the choices they have made are very well considered. They simply want to have one single driver handling the entire midrange and lower treble to get rid of this problem. However, taking a 5" AudioTechnology driver up to 3800 Hz does have an impact on the overall voicing of a speaker. I've tried taking my AudioTechnology 4H52 up to 4500 Hz crossing over to a SS 9500 treble unit (4th order slopes) and I wasn't too thrilled. The dispersion from a 100 cm^2 cone at 4 kHz is not the best, but possibly a 2nd order filter will make a suitable blend of sounds to enhance the midrange driver's limited ability to perform well at such high frequencies.Comments are welcome.

Example 3 is the SP38/13 approach and to my knowledge not much used, if used at all. And it’s a pity, because few 6-7” drivers handle well above 1500 Hz. Well, some 6-7” kevlar and polypropylene drivers have a tolerable presentation above 2 kHz with a smooth roll-off with no disturbing peaks to spoil the treble quality. But large cones start beaming when having to handle higher frequencies and very few drivers get away with it successfully – or we may have to use low order filters to make a blend of sounds from the midbass and treble thus overcoming some of the shortcomings of the midbass driver.
Drivers like the ScanSpeak D3806/8200 are rare and I haven’t been able to find an exact replacement that will basically handle the frequency range from 1,000 – 13,000 Hz.
I’m sure that some will object here, shake their heads and tell me there’s no problem in taking a 1” dome down to 2 kHz, so why the agony? And this is true; the 9700 tweeter goes clean down to 800 Hz.

Above the 9700 FR and yes, down to 800 Hz, but does it sound good? I’ve tried it many times; a 6” woofer (8545) and the 9500/9700 down to 1,800-2,000 Hz and I don’t like it. I don’t like 1" tweeters fiddling with upper midrange.

What I have left out in this presentation is the true middome*, although the 3” 12M/4531G00 actually is a kind of middome, a large voice coil carrying a large paper dome and a narrow cone connecting the dome to the surround. Fortunately this driver uses a very small magnet (neodymium?), has a vented pole piece and should have proper rear ventilation. Not an easy thing to make when drivers get this small.
*: Forget the 3" Vifa middome. It's crap. The old Dynaudio D52s are good and so are the horrific expensive middomes from ATC.

Example 4 is the all-time classic. A 6-7” midbass married to a 1” dome, usually with a point of crossover around 2,500-3,000 Hz.
This is obviously the most cost-effective way of getting it all from two drivers. Simple crossover – if the drivers are performing well – simple cabinetry, no need for middriver housing, etc.

As can be seen we’re moving in octaves here: 400-800-1600-3200-12800. Well, I have never seen a point of crossover at 6,400 Hz as few tweeters have trouble handling 3 octaves from 3 kHz to the limit of audibility. A good upper-middome like the D52 may be able to handle a p.o.c. around 5.6 kHz, but I’ll leave out this option.
The points of crossover (p.o.c.) are serious choices to be made and as a general rule of thumb we should avoid p.o.cs. in the midrange where the ear is most sensitive. This has been good and sound practice for decades, but we (diy people) may now have the tools to make it work (LspCAD and similar) - or do we?

The planned TJL3W will be the first time* I make a 3-way with a p.o.c. at around 800 Hz. The initial modelling doesn't look too bad. Below simulated response by measured response from 1 x W18E001 + W11CY001 + TJL ribbon. Below is seen the response with inverted polarity of middriver.

*: 26-11-2005: Events took over here and before I'll get around to the TJL3W, I have done two systems using small upper midrange drivers. The first one reported here: Zahra. And read below here.
The next one was a 3-way from 18W/8531G00 + 12M/4631G00 + D2904/7000, the Ekta.

Up-date: 30-10-2005

Example 2:
Making a pair of loudspeakers for a friend made me experience the second example of a 3-way with a point of crossover between bass and mid at around 800 Hz. This is quite an interesting exercise and I'm quite amazed by the quality of the sound and the ease with which the 3" mid-driver handles the 800-4000 Hz frequency range.
Actually the speaker is a 3½-way with two Scan-Speak 18W/8543-00 bass drivers, the 13M/8636-00 kevlar/fabric dome for mid and the D2904/9800 for treble.

The simulation of the crossover took some evenings and resulted in a fairly simple crossover providing a good phase tracking between drivers as seen below. This was fortunately confirmed when I came to the actual set-up of the crossover. This is not always the case.

LspCAD simulation: This almost looks too good with a nice frequency response and a
beautiful phase-tracking between drivers over a wide range in the over-lapping areas.

The speaker's SPL response has an overall tilt of approx. 3 dB from 100 Hz to 10 kHz. Close to the BBC-dip target giving the best balance of sound - from this speaker. It doesn't work for all, but here it does. Raising the tweeter level by only 1 dB make this driver clearly too high.

This speaker will hopefully be reported in detail later but here at the final stage of fine-tuning the benefits of having a mid-driver handling the most difficult range where the ear has its highest sensitivity is very clear. It's probably the speaker with the highest level of transparency I have made. It can be played loud and is very revealing to the programme material, but fortunately in a non-aggressive way. The small kevlar mid-driver with the fabric dome is certainly a fine driver deserves a much wider use than seen. Actually I don't know of any commercial speaker using this excellent mid-driver.

27-11-2005: Now available: Zahra.