PSS

This design is an iteration in a series of experiments in 3-way designs with side-firing woofers. The goal, which should be somewhat obvious is to build a full-range speaker with a narrow front profile, and therefore a higher-than-usual SAF.

The initial design used an 8" Seas aluminum cone woofer, a 5.5" Seas aluminum cone midrange, and a Morel MDT-20 tweeter. The Seas woofer was abandoned in favor of the more versatile and higher efficiency Peerless 10" XLS woofer, which was joined by the Seas CA15 paper cone midrange, which allowed the use of the Scan Speak 8513 3/4" tweeter. Experimenting with various combinations of woofers and midranges helped me zero in on workable combinations of driver sensitivities.

The basic difficulty in designing a side-firing 3-way is that the woofer must be crossed low, generally at less than 175Hz. This means that not only must the impedance humps of both the woofer and the midrange be compensated for, but that the woofer cannot be used to compensate for the spreading loss (a.k.a baffle step), as is traditionally done in 3-way designs. This means that the midrange must have a higher sensitivity than the woofer, which is a rare combination. Most 10" and larger woofers have a sensitivity in the high 80dB's, most mids hover around 85dB. This severely restricts the selection of drivers available.

So, if those are the rules, how did I wind up getting away with a woofer which is more sensitive than the midrange I'm using?

The answer can be found in examining the spec sheet for the Peerless 10" XLS. It has a relatively low Vas for a 10" subwoofer, but at the same time, and extremely low Qts. When modeled as a subwoofer in the box used, it has a high F3 of around 80Hz and a Qts of around 0.4. On the surface, this looks like a bad choice for a woofer, but looking a little closer shows that it begins rolling off at around 200Hz. Carefully choosing a crossover point allows the designer to also control the ultimate sensitivity of the woofer to blend it with the midrange. The low Vas and Qts also means it will do all this in a relatively small box, which plays directly into the goal of raising the SAF. Even so, I needed to add a few Ohms of series resistance to the woofer to match levels to the mid. Ideally, this design would work best with a midrange that has a sensitivity of around 89-90dB/2.83V, which is why I'm looking forward to the newly announced Seas MCA15 midrange to become available. The net results is that this design has a measured F3 of around 40Hz with a shallow roll-off and with output capabilities of about 100dB/1m per speaker.

As mentioned above, the Peerless woofer, in combination with a carefully designed crossover, can generate deep bass in a relatively small box. The woofer cavity has a net volume of around 25L. The midrange is allowed to fire into a larger volume than is normally required, which helps reduce coloration from internal reflections.

The box shown is 40"x7 3/4"x12 1/4" (HxWxD), which is typical for a floor-stander. There is a 1" bevel around the baffle to help with edge diffraction. A bevel was chosen over the typical round-over because it is easy to do using a table saw, where a round-over large enough to have the same effect would require an expensive and potentially dangerously large router bit.

The baffle is 1 1/2" thick, and all other panels are 3/4". There are 3 braces, two vented by cutting holes through them, and the third solid to separate the mid and woofer cavities. The third brace is installed at a 45-degree slope.

The baffle is made by laminating two 3/4" panels of MDF the proper size together and then cutting the bevel on the table saw. It is easier (and safer) to bevel the baffle before mounting on the box. It may take a bit of experimentation to get the saw set properly to cut the bevel. I made a mark 1" deep on the edge of the baffle, and kept moving the saw's fence until the blade just nicked the mark.

The diagonal brace is best made by cutting the brace a few inches too long, and then cutting one end at 45-degrees. The brace is then laid against the (already cut) side panel of the box and marked for length. Again, the best way to cut it to the proper length is to repeatedly trim off a bit and test fit to see how much is left to go. Don't worry, you don't need to be real accurate.

The midrange cavity should be stuffed with about half of a bag of polyfil ("bag" being the smallest bag at your local craft store, I forget the size). Cram the stuffing into the back of the wedge tightly, and left the stuffing in the front remain loose and fluffy. The woofer cavity is unstuffed.

Here is a drawing of the box.

The woofer crosses to the midrange at about 100Hz with 2nd-order LR response. The mid crosses to the tweeter at about 3.7kHz with 4th order LR response. There is an LRC network on both the mid and the woof to provide impedance compensation. These are matched to the box volume, and must be changed if the volume of either cavity is changed.

The woofer low-pass (Peerless 10" XLS (830452)...



If desired, bass level can be tuned by altering R1091. R1061 carries a lot of current, and will probably get hot. I recommend using two 14-Ohm resistors in parallel. All coils on the woofer are Erse 16ga iron-core.

The midrange band-pass (Seas CA15RLY)...



L2091 is an Erse 16ga iron-core coil. All other coils are 16ga air-core. If desired, the air-core coils can be either 14ga or 18ga.



All coils are 18ga air core. Tweeter level can be adjusted by altering R3091.

Where iron-core coils are specified, do not substitute air-core coils. Air-core coils in these values have very high DCR, which will cause problems. The small resistors in series with the coils indicate the DCR of the coil (i.e., the resistance of the coil itself, not actually a resistor). It is not important that you exactly match the DCR of the coils.

For those who want to try a second order XO, it can be done with very little change. All notes regarding coils and tuning from above still apply.

The woofer circuit is unchanged...



The midrange band-pass adds a resistor and changes C2051...



For the tweeter, a cap is removed, L3021 is increased, as is the series resistor. A ladder delay is added compensate for the fact that the woofer's AC is about 8mm behind the tweeter's AC. The all-pass coils are 18ga. Note that the tweeter's polarity is also reversed. (Note I have not listened to this the tweeter circuit, but it should sound almost identical to the one below.)



If you would rather not use the delay, the same effect can be achieved by tilting the speaker back slightly. I have tested this version and found it to sound similar to the 4th order, only a little smoother and more coherent. I found that raising the front edge of the speaker 7/8" higher than the rear produced an excellent null. It could also be done by building the cabinet with the tilt already built-in, but that makes the table-saw cuts much more complex. I think the tilt works out to about 4 degrees.

This is a fairly nice speaker, although the midrange is not up to the quality of Ella. There is a slight twinge of harshness to the Seas CA15 woofer which is sometimes audible (less so in the 2nd-order design). This speaker has adequate low-end that most people would be satisfied using it in home theater without the need for a subwoofer. Bass-heads, however, would probably want to cross over to an appropriate sub at about 40Hz to avoid over-excursion on the woofer.

From the simulations, the speaker would be classified by most commercial manufacturers as an 8-Ohm load, although a more honest interpretation would put it at about 6-Ohms. It's lowest impedance is 5.5 Ohms at about 35Hz. Phase is reasonably well-behaved, staying within +/-30 degrees, with one small excursion to about -40 degrees at 3.5kHz. Because it's efficiency is fairly low, I would say it's a moderately hard load to drive. I wouldn't want to try to power it with a typical consumer receiver without bi-amping the woofer. I would recommend that you bring at least 100W of high-quality power to the table.

Bandpass output capabilities compute out to about 105dB @ 1m per speaker from about 40Hz on up before Xmax is hit. For a pair of speakers, this equates to about 98dB at a typical listening distance of 10 feet, more than enough for most people. It would however, take 500WPC to do this. With a 200W amp, you could hit maybe 93dB at the chair, and 90dB on 100W. Your spouse will still be telling you to turn it down, although the neighbors probably won't call the cops. It should have the thermal capability to handle 200W with normal program material.

The more ambitious builder might choose to apply a Linkwitz transform and a 200WPC amp, which would satisfy all but the most demanding bass requirements.

If you don't go nuts on crossover components, you should easily be able to build a pair of these for under $700.

The next steps for this design are to find a more suitable, higher-sensitivity midrange, which would allow me to run the woofer without a series resistor. I also intend to adapt the design to use a top-quality midrange and a Hiquphon tweeter with a 2nd-order LR crossover.

These measurements are of the 4th-order design. The 2nd order would look pretty much the same. They were taken at about 0.5m with a 4ms gate. Nearfield midrange was then spliced in, and nearfield woofer was added. Response is just over +/-1dB through the passband, which extends out to 20kHz. Woofer extension measures to an F3 of 40Hz. The F10 appears to be below the point where I trust the measurement system, but I would guess it to be about 20Hz. Please note that there are only 2dB between vertical division.