Dick Olsher Dahlia speaker

Do-It-Yourself (DIY) loudspeaker projects are quite common in the UK, where details about several excellent designs, including a recent one by Martin Colloms, have been published for public domain consumption. Stateside, the situation is rather grim, where only an occasional subwoofer project (always popular) makes it into the commercial magazines.

Whatever this imbalance may say about American audiophiles, it speaks well for the British attitude toward good sound; home construction offers the audiophile a ticket to kilobuck sound quality at a fraction of the cost. Generally, there's a ratio of at least 4:1 between a speaker's manufacturing cost (labor and materials) and its retail price. Thus, speakers that retail for $1000 a pair might cost only $250 to produce. (Before you write your congressman, you should realize that this is standard business practice, and in no way implies any undue greed on the part of the industry. There are usually two middlemen to be paid as well: the distributor and the dealer.) There are big bucks to be saved in home construction, but also a few "ifs and buts".

DIY projects are not for everyone. There are three things you must have: patience, patience, and more patience. After all, you'll want your speakers to be perfect after your first-ever production run, and you won't be able to work out the bugs in manufacture. You'll have no one but yourself to blame if your screwdriver slips and punctures a cone or your soldering gun chars your finger or (Heaven forbid!) the cone itself (footnote 1). And when all is said and done, there are no warranties, either expressed or implied. A nasty letter to whoever dreamed up the sorry mess you tackled is about the best you can do. If you're still game, read on.

To those who wonder why I called this the Dahlia, the answer is simple: while I was working on the project, my daughter made me promise I'd name it after her, and I did.

The Dahlias are not state-of-the-art, but they do make a strong statement in the under-$1000-a-pair price range, and I believe they'll compete favorably with anything in this range (footnote 2). Total parts cost for two Dahlias will run between $200 and $300.

The design evolved from my search for an outstanding midrange-woofer driver. Good speakers start with good drivers. There is no way you can squeeze good sound out of mediocre drivers, no matter how accurate their time alignment and how inspired their crossover or enclosure design; the inherent in-band distortions and nonlinearities of the drivers will still be there. So when Audax of France unveiled what they call an "advanced formulation plastic" cone, which they dubbed TPX, I naturally had to investigate.

TPX sounds about as involving as STP oil treatment, but, choice of names aside, Audax claims that this material is superior to polypropylene for loudspeaker diaphragm. It is slightly less dense than polyprope, therefore more efficient; a lighter cone is possible with TPX. The areas where TPX appears to trounce polyprope, however, are in internal damping and rigidity. TPX is twice as rigid and three times better internally damped than polyprope. Diaphragm rigidity is especially important in the middle octaves, where cone flexion (breakup) becomes a problem.


In the bass region, it is easy to move the whole surface of a cone in unison, even a paper one. This orderly piston action tends to fall apart in the midrange, however, with some areas of the cone pushing while others pull. With Dahlia, the amplitude of such deformations is much less than from paper or even polyprope, resulting in lower distortion, smoother response, and better transient response.

Audax's 8" cone also features a surround made of a new synthetic compound called Norsorex which has remarkably little recoil. This stuff is dead! (At the Winter 1985 CES, Audax demonstrated the characteristics of this material by dropping balls made of standard PVC rubber surround material and Norsorex onto the bare floor. The Norsorex ball didn't go anywhere. It makes a lousy handball, but an excellent surround: it effectively reduces reflection of vibrational energy back into the diaphragm. This means better transient response with less ringing.

The lab data provided by Audax show excellent toneburst test results, which is exactly what you'd expect. The decay part of the toneburst is very well behaved, with minimal ringing. I'm happy to report that the TPX driver used in the Dahlia is indeed a great midrange/woofer, the cleanest such driver I've come across, and the raison d'etre of this project.

The Audax TPX driver, introduced in its final form in late 1984, hasn't been around long enough to have been adopted by any of the major manufacturers. I understand, however, that John Bau at Spica is working with a custom version of this driver for a two-way system slated for introduction in 1986.

Design Philosophy
One of the Dahlia's major design objectives was to set a new standard—at an affordable price—for midrange clarity and transparency from moving-coil loudspeakers. To that end, design priorities were arranged so as to preserve the TPX driver's potential in the midrange. For example, the crossover point was set at 3kHz—rather high for an 8" cone—in order to preserve the cohesiveness of the middle octaves. An effort was made to keep the cabinet size moderate at 1.36 cubic feet internal volume, with the idea of minimizing panel resonances. The dictum "flat (response) is beautiful" was abandoned in favor of a more subjectively flat response in the lower octaves. There's a 2–3dB peak centered at 75Hz, which compensates for a somewhat lean balance in the lower midrange. A bass-reflex enclosure was chosen to allow a reasonable enclosure size without pushing the system resonance even higher, again to maintain minimal flexing of the cabinet walls.

Both the high- and low-pass crossovers are fourth-order Linkwitz-Riley types that produce a symmetrical acoustic radiation pattern through the crossover region. The steep 24dB/octave slopes prevent the woofer and tweeter from contributing significantly to the overall acoustic output over a widely shared frequency band, thus minimizing acoustic interactions in the radiation pattern of the speaker.

I've never cared for first-order (6dB/octave) crossovers. It's true that the output of their electrical network is a perfect replica of the input, but the acoustical output of the speaker ain't necessarily so. Because the tweeter and woofer do not radiate from the same point in space, usually being separated by at least several inches on the front baffle, there will be interference between them. The extent of this interference depends on the actual separation of the drivers as well as the crossover frequency and inherent driver phase differences.

The big problem with first-order filters is their nonselective nature. Both drivers contribute substantially to the output over a very wide, overlapping frequency band. What you've got here is two voices competing for attention, and this, in my experience, increases perceived mud and fuzz in the frequency region shared by the two drivers. I'm not implying, however, that 24dB/octave filters are perfect. The steeper the filter's slope, the more it will tend to ring on transients. This ringing will not necessarily be as bad as the problems introduced by the first-order crossover, however; I believe that this is the right tradeoff to make in the pursuit of midband clarity. Note that the low pass section has an impedance equalizing network that compensates for the voice coil inductance of the woofer, making the woofer appear an essentially constant resistive load to the crossover network.

The tweeter used is the Audax HD12x9D25 1" soft-dome unit. This tried and true driver is used in a number of highly regarded speakers, but is surely no longer state-of the art. Nevertheless, it is readily available and mates well with the TPX cone.

The cabinet plans are shown in fig.1. Actually, for my first prototype I used the Tabor House cabinet, which is roughly the correct volume and has the proper baffle cutouts. (The Tabor House is a popular English kit available from Falcon Acoustics in the UK.) I ended up using a Woodstyle customized cabinet made of ¾ chipboard, with an "H" brace, for the final version. Woodstyle is an OEM cabinet maker, but their fully assembled cabinet should be available through Just Speakers or A&S Speakers for those of you with no desire for woodworking.


Fig.1 Dick Olsher Dahlia enclosure layout.

The crossover schematic is shown in fig.2. All coils are of the air-core type, wound with wire no lighter than 18 gauge. I tried several types of capacitors, from inexpensive mylar films to the very expensive Sidereal caps, which sounded best. Next best was a French metallized polypropylene called Solen, available from Just Speakers. For the extra money, the Sidereals give cleaner and slightly less sibilant treble transient attacks; these would be the best choice for the perfectionist. (The pair of Dahlias reviewed by JGH have Sidereals in them.)


Fig.2 Dick Olsher Dahlia crossover network (one channel).

I used long-fiber wool, in the form of a quilt batten (also known as a carded batt), for cabinet stuffing. Available locally (for me) from the Rio Grande Wool Mill, this batten's random weave is far preferable to the yarn variety because it does not clump up in the bottom of the cabinet. (It's great stuff for transmission-line enclosures.) About half a pound of wool per enclosure is optimum. Be sure to drape the wool over the internal brace so that the wool actually occupies the central volume of the enclosure. This is where standing waves develop, and where the wool will do the most good in attenuating them. Incidentally, long-fiber wool is technically 2.5–4.5" long and, naturally, is sheared only from those sheep that qualify. I recommend wool over foam or (even worse) fiberglass, because it attenuates midbass frequencies without significant absorption of the deep bass around 30Hz.

When mounting the woofer, be sure to isolate it from the front baffle with either Mortite rope caulk (or equivalent) or foam insulation—partly to prevent air leaks around the driver basket, partly to prevent rattles. It will also diminish, to some extent, the amount of vibrational energy transmitted to the enclosure walls from the driver basket, thus helping to control panel resonances.

To reduce diffraction effects, the tweeter should be surrounded with strips of ¼-thick felt or soft plastic foam on the front baffle, 1–2" on all sides. This will reduce HF frequency radiation toward the sharp edges of the enclosure, and will reduce diffraction effects as well.

The Dahlia speaker must be tilted back 10–20° in order to obtain the proper tonal balance, and to approach proper time alignment of the drivers. Rather than complicate the woodworking with a sloping front baffle, I decided to rely on a tiltback stand. The Chicago Speaker Stand (model RJ100) works very nicely with these speakers, and are highly recommended. Otherwise, experiment with the tilt—there is a noticeable difference between the tilt and no-tilt positions, the former yielding a gently rolling off high end that is more natural.

The bass-reflex port, 3" in diameter, is loaded with a 5" duct cut from a cardboard mailing tube. (A PVC pipe of the proper diameter would be even better because of its greater rigidity.)

The Sound
What follows is a synopsis of my feelings about the sound of the Dahlias. In all of recorded history, there hasn't been a speaker designer who disliked his own creation, so I asked JGH for a second, more objective opinion.

As I see it, the real glory of these speakers is their midband clarity, transparency, and resolution of low level detail. Electrostatics they ain't, but they come close. In fact, they come closer than most other cone loudspeakers. Their excellence extends right through the presence region, brasses are reproduced cleanly and with remarkable conviction, and the overall sound is best described as spacious, light, and airy.

There is a bit of leanness about the warmth region, which contributes to the perception of lightness. Imaging and focus within the soundstage are very good, with excellent depth. But I keep coming back to the midrange—timbres are right on, and there is an uncanny ability to resolve fine musical textures.

Unfortunately, there are debits. There is very little deep bass; don't look for much below 50Hz. The midbass is pretty good, but a tad on the lumpy side. Also, the upper treble lacks the ultimate in delicacy and speed. Finally, as is typical of European speaker drivers, power handling and SPL levels are limited.

The speaker will not handle more than about 50 watts of pink noise, so don't use it for PA applications or satisfying SPLs in large rooms. I see the Dahlia as a music-lover's speaker. It is capable of dynamic bloom on even full-scale orchestral music—provided it's not pushed too hard. Here is the purity of a classic minimonitor, but in a larger, better balanced package.—Dick Olsher

Footnote 1: If this sounds a bit heartless, remember that your finger will heal with time; the cone won't.—Dick Olsher

Footnote 2: As the designer of the Dahlia, DO is clearly unbiased.—J. Gordon Holt