My cabinet-making collaborator, Mike Spurlock, wanted a simple
system for his beach cabin. "That's not too hard, Mike, you
just build this," I said, drawing a simple picture of a transmission-line using a 5.5" Vifa midbass driver and a 1" Scan-Speak soft-dome tweeter.
As I filled in the details, I commented, "These particular
drivers are really easy to use, since they're quite flat on their
own, sound pretty good, and only need a minimal crossover. The
transmission-line ought to provide better bass and midrange than
a closed box, and it also avoids the need for an expensive stand."
Mike then asked the key question: "Why don't you also draw
me a D'Appolito configuration, Lynn? That'll be twice as efficient,
so I can use my new 300B triode amp."
"Hmmm," said I, in my best I'll-deal-with-this-problem-later
System Administrator style. "Let me draw that one on my computer,
so we can have accurate dimensions to work from. Let's talk about
this some more ..."
As I worked on the Canvas computer drawings, it dawned on me that the
new loudspeaker might meet the needs of more people than Mike.
After all, I did need something to talk about for my next article. (The Ariel made its debut in Vol. 4-4/Fall 1993 of Positive Feedback magazine, followed by the Mark II revision in Vol 5-1.). Maybe something a bit more significant was beginning to stir ... time to do a little reflecting on what this project was really all about.
Three key ideas emerged:
-
Higher-than-average efficiency combined with natural
and lifelike sound, spacious imaging, and a slightly warm tonal
balance (not a thin-sounding minimonitor balance, in other words).
-
A conceptually simple design that leaves room for the builder
to have fun with parts tweaking. The crossover would be simple
enough to allow sonic experimentation with different brands of
capacitors, inductors, and resistors. The transmission-line could
be fine-tuned for different rooms by altering the area of the
exit region and by trying different fill materials (wool, crimped
Fortrel, etc.) in the straight pipe section behind the drivers.
- Straightforward construction. I realized that most skilled
woodworkers don't like building weird shapes that have large radii.
Even though a box is pretty close to the worst possible shape
for a loudspeaker system, that's what woodworkers tell me they
want to build.
An experienced woodworker with an accurate table saw, a precise
miter box, a router, and a complete set of clamps (or better yet,
a vacuum bag) should be able to complete this system. Do not,
repeat not, attempt this system if this is your first
loudspeaker. On a scale of difficulty, it is definitely more complex
than an ordinary bookshelf speaker, but is simpler than a
Lowther horn ... it's comparable to a typical transmission-line
in complexity and requirements for very accurate (1/32" tolerance
across 4 feet) woodworking.
Since the inspiration for this system came out of the clear blue
sky, I named it after the Archangel Ariel. For me, the name Ariel
captures the way it feels to have a new creation emerge from the
great beyond, come to life, and go out into the world. Thanks
to Positive Feedback magazine and the Internet, people all over
the world are enjoying the Ariel.
Since "keep it simple" was uppermost in my mind at the
beginning of this project, I picked drivers that were up-to-date,
easy to work with, and had first-class subjective qualities without
requiring tricky equalization circuits in the crossover. (Not
many drivers qualify!)
The Vifa 5.5" P13WH-00-08 midbass is refreshingly free of
the muddled and confused sound of typical 7" and 8" polypropylene
drivers. This is almost certainly due to the small size of the
cone, which makes it more rigid, as well as careful Vifa engineering
of the vented pole-piece assembly. The frequency response of this
driver is also pretty remarkable, with a ruler-flat midband gently
descending to a smooth 12 dB/Octave rolloff beginning at 5kHz.
That means the crossover can be simple, and won't require midband
equalization.
The flatness of the 5.5" Vifa driver is more unusual than
it first appears. Since the ear is so sensitive to spectral imbalances
in the midrange, most high-fidelity speakers have traditionally
used modest equalization in the crossover to straighten out the
crucial midrange. It is only in the last few years that we have
started to see high-performance drivers that don't require any
equalization, and the Vifa 5.5" midbass is an excellent example.
(The only other one I can even think of is the Focal 6V415, and
even that one has a small bump at 4.5kHz).
I first heard the 5.5" Vifa in the Lineaum LFX, and was impressed
with the wonderfully transparent midrange. I was not so impressed
with the thin bass. I figured if this driver had any future as
a good midbass unit, a pair of them driving a 6 foot transmission-line
would bring it out. Doubling up would gain 6dB of efficiency and
headroom, and the transmission-line would extend the bass down
to the F3 of the driver - 60Hz. By contrast, a vented box with
conventional Theile/Small QB3 tuning begins a 24dB/octave rolloff
at 80Hz.
The two orthodox approaches to get deeper, more powerful bass
would be:
-
Select a larger midbass driver. 7" or 8" would be the
logical choice, and would get down to 40 or 50Hz.
OR
-
Retain the 5.5" Vifa, use it as a midrange driver, and add
a 10" or 12" paper-cone woofer in a separate vented
box. That would get down to 40Hz with no trouble.
High-end speakers in the middle price points ($1000 to $5000) usually compromise the mids in favor of getting more of that intense bass that many
audiophiles want. Increasing the size of the midbass driver reduces the midrange quality in the case of the Vifa (and many
other drivers). The 6.5" and 8" Vifa drivers need modest
midrange equalization, and after you go to all that extra trouble
and expense in the crossover, they still don't sound as
good as their little brother (EQ is not a substitute for quality).
Going to a 3-way system by adding another passive crossover at 200 to 300Hz is just asking for trouble in terms of large, expensive, and sonically intrusive capacitors and iron-core inductors. Assuming you spent a small fortune on crossover parts, there's still the problem of having the crossover fall in the range where the ear is most sensitive to phase distortion (100 to 800Hz). And for what? Most 10, 12, or 15-inch poly-cone woofers don't have such great bass anyway. It's loud and deep all right, but they can't keep up with good mids and tweeters. That's why most 3-way systems avoid using the best midranges and tweeters - although top-quality mids and tweeters don't add much to the price, they expose the poor quality of the generic poly-cone woofer. Good woofers quickly run into serious money, and it takes bi-amping to really do them justice.
The design tradeoffs in the Ariel are very high quality midrange and treble, focus and immediacy across the spectrum, and good efficiency - at the expense of deep and powerful bass. I tell most folks to go ahead and build the Ariel, and see if they want more bass after they live with it for a few months (it's a bit like acclimating to Quad ESL's or Lowthers). If you'd like a bit more extension, don't rush out and buy just any commercial subwoofer; most of them can't keep up with speed and resolution of the Ariels. (With the notable exception of the REL subwoofers, which work just fine ... I own a REL Strata II.) Take a look at the Good Bass section for more info on which woofers (not subwoofers!) are a good match for the Ariel and ME2's.
Returning to the midrange, I cannot recommend any other midbass driver as a substitute. If you want to use any other drivers you are completely on your own.
There are a handful of top-rank tweeters to choose from, unlike good mids, which are scarce. This is the result of many advances in materials sciences, acoustic holography, and computer modeling in the last decade. Soft domes in particular have made great strides, and are head and shoulders above the popular soft-domes of the Seventies. The best soft-domes (Scan-Speak D2905/9500 and D2905/9700) now surpass the best metal-domes.
The issue of subjective coloration goes a bit deeper than flat response and good MLSSA waterfall plots. The best tweeters meet all the technical criteria very well; what remains are the subjective differences. This is where I part company with the detail-at-all-costs brigade, and go for the one that sounds the most true-to-life. I am not a fan of the "analytical" cookie-cutter sound so popular in the high-end magazines. My criteria are much simpler: does it sound like real people singing in my living room? Does the clarinet sound like a real clarinet? Is it real; does it live and breathe?
The Scan-Speak D2905 family of tweeters communicate this subtle essence in the most natural way I've heard so far. Here are my personal impressions of the top three models:
-
D2905/9500: This Scan-Speak tweeter is a combination of the 9300 dome assembly, front plate, and Ferrofluid damping, combined with the 9900 aerodynamic non-resonant rear chamber. Since I have a few reservations about the sonics of the 9900, that makes this the best all-around choice for the Ariel and ME2. The reports from the Ariel Builders Club all agree: this is their favorite tweeter ... guess I better replace my old sticky-dome 9000's with the 9500's one of these days.
-
D2905/9700: A 9900 with a flat faceplate. I've heard a rumor that Siegfried Linkwitz commissioned this tweeter for the Audio Artistry line of speakers. He wanted all the features of a 9900 without the shallow horn, thus the 9700. No question about it, they sound very good in the Audio Artistry speakers - with Linkwitz as the designer, they should!
The 9700, along with the 9900, do not use any Ferrofluid in the voice-coil gap. (A very unusual feature in any tweeter these days.) That creates a large peak in the impedance curve at the tweeter resonance of 600Hz, thus requiring a precisely tuned notch filter in the crossover. More significantly, this also means the tweeter portion of the Ariel and ME2 crossover requires a complete redesign for the 9700 - if you use the "stock" crossover there is a real possibility of damaging the tweeter and the certainty of very high IM distortion as the result of uncontrolled excursion at the tweeter resonance.
-
D2905/9900: The top of the Scan-Speak driver line. Unfortunately, this tweeter isn't my cup of tea. I haven't heard any speakers that use this tweeter that are to my taste; the measured wrinkles in the response in the 12 to 18kHz region are audible (to me, at least), and I am not a fan of short horns, which is what the 9900 uses as a faceplate. (Short horn = "directivity control element".)
In any event the standard Ariel and ME2 crossovers require a complete redesign for the 9900. If you select the 9900, you are on your own. Make sure you have CLIO, MLSSA, and a calibrated 1/2" condenser microphone nearby. If you are a very advanced builder with a fully equipped lab, this could be an interesting challenge.
In purely objective terms, the Vifa P13WH-00-08 and Scan-Speak D2905/9500 are the flattest, most neutral drivers in the world; the group-delay vs. frequency and MLSSA waterfall results are truly exceptional. Not to mention they sound great as well!
Don't take my word for it - other builders of the Ariels have called me up long-distance and told me they were fooled into thinking a party was having an impromptu sing-in in their living-room. One fellow even thought one of his next-door neighbors was humming along with the music - turned out the humming was on the recording, and he had never heard it before! With a moderate-power vacuum-tube amplifier, voices, piano, and clarinet have an "in-the-room" presence that is relaxed and natural ... quite different than the "high-end" sound you hear at the shows.
Or, The Neverending Story.
The first Ariels were pretty simple; just a two-fold transmission line going down, striking a corner reflector, and going back up the rear of the cabinet, with the vent near the top of the cabinet. All three drivers shared a common chamber that fired down into the line; curious readers can order a back issue of Vol 4-4. of Positive Feedback magazine if they are curious to see the (badly-scanned) artwork. I was pleased with the quality of the mids and highs, but the bass was not much better than a minimonitor (but lacking the "squashed" or compressed sound of small-box enclosures). To keep things simple, let's call this pair the Mark 1.
A brief chat with an old friend, Ron Nelson (of Nelson-Reed loudspeakers), revealed that vents have to be close to the floor ... preferably right on it ... if you really want to energize the room. Both Ron and I, along with many other speaker designers, noticed that an interesting boundary effect occurs close to the floor. Get the driver or vent within a critical distance of zero to three inches, and the whole room lights up, almost as though the bass waves were travelling freely along the entire room periphery ... maybe they do, for all I know. If you move the driver or vent just a few inches further away from the floor, the effect disappears, and the speaker starts sounding like a minimonitor again, all strain and grunt. Ron and I surmised this is an unusual form of horn-loading, with the driver or vent loading into it's mirror-image below the floor, raising the efficiency and power-handling, and decreasing distortion.
So the vent at the top had to go. Yes, I could have simply modified the enclosure for a three-fold line, a very old TL design going back to 1971, but before I went down that road, I heard a very impressive demonstration that turned me around. Mike Spurlock (you remember him at the beginning of this story) contacted an architect friend in Tacoma, WA, named Randy Hedgebeth, and Randy brought down his new TL speaker using a single Focal 6 1/2" fiberglass midbass and a 1" Focal Kevlar tweeter.
The bass quality on Randy's TL was impressive, particularly since I'd been used to listening to a subwoofer that used a pair of push-pull 10" Peerless CC-line drivers in a Bessel-aligned 6 cubic foot cabinet. The little 6 1/2" Focal matched the "big guy" in power and reach, and was in a superior class for speed and detail. All from a pint-sized Focal driver! Fortunately, Randy wasn't interested in being coy about his design (none of this "proprietary" jive), and showed me an article he was writing for Speaker Builder magazine.
The key features of Randy's new type of transmission line are:
A) A hybrid of a classic 1/4 wave TL combined with a folded labyrinth for the last 1/3 to 1/2 of it's length, similar to the "Brother Jon" by Robert Spear and Alex Thornhill in Volume 13, Number 4 and 5 of Speaker Builder magazine.
B) Placing the driver at the 4/5 point along the line instead of the far end and filling the last 1/5 completely with foam absorbing blocks.
C) Leaving the line unfilled and using damping pads of 1/4" and 3/8" felt at appropriate locations around the drivers. (As far as I know, these last two features are unique to Randy's design).
Randy's article also provides an impressive set of equations to balance driver area, free-air resonance, and overall line length. Considering the poverty of literature on transmission-line loudspeakers, and the contempt with which mainstream engineers treat designs outside the charmed circle cast by Theil and Small, I consider Randy's article a major step forward.
After hearing the first version of the Ariels, Randy encouraged me to build a new pair that took advantage of what he had discovered about TL systems. This conversation became the genesis of the Mark 1.9 and Mark 2 Ariels. The Mark 1.9's are the ones you see in the pictures of my living room, and the Mark 2's were the first version to go on the Web back in 1995 or 1996. (Ah yes! When the Web was young, in the days of Mosaic and Netscape 2.0, when there were only 200,000 sites and your browser would crash after 15 minutes. That's part of the reason this site looks so simple ... the HTML is hand-coded and old!)
Unless you're building a Mark 2, the description is mostly of historical interest. This version pioneered the hybrid TL/labyrinth concept for the Ariel, and I've stuck to it through thick and thin. The traditional problem with TL's has been the conflict of enough filling for the 80-200Hz region while avoiding cone-loading problems from excess damping ... audible as squashed dynamics. With the labyrinth smoothing the bass in the 80-200 Hz region, the amount of filling can be greatly reduced, and the latest versions of the Ariel use very little filling compared to traditional 3-fold lines. To me, a mark of success in any TL is a having a minimum of filling while keeping response smooth ... it's not easy to do both.
Looking at the Mark 2 drawing after the passage of some time, I can see that this version went through a few changes, such as the addition of a crossover box at the back of the enclosure base. It has a TL that starts as a split line, angled to minimize reflections in the region directly behind the driver (essential in any speaker), then joins together at the labyrinth portion. There is a variable-height feature that has generated a lot of comment over the years, since I left that detail for the builder to figure out. The combination of interior plywood and exterior MDF has also been carried forward in all other Ariels, since this provides a rigid inner structure (transmitting vibration away from the drivers) while having a lossier, more damped outer shell.
It was a chorus of complaints about the 15-degree twisted interior panel that prompted the revised Mark 3 version, along with a desire to provide a much smoother interior airflow for the straight portion of the line (airflow in the labyrinth is intentionally rough). There is a second drawing that has a detail difference in the mounting of the slanted panels. As expected, people who built them reported smoother upper bass and lower midrange as a result of the less boxy shape of the interior.
The most notable differences in the Mark 4 version are the acoustically-matched twin transmission lines, with an fully-independent line for each driver. (Thanks again to the Ariel Builders Club members for the suggestion.) Each line uses a smoothly tapered reverse-horn profile from the back of the driver to the entrance of the labyrinth. In subjective terms, there are significant improvements in the upper-bass and lower-midrange region from 100 to 500Hz (piano, male voice, and acoustic bass). Although the original Y-shaped line seemed like a good idea at the time, the time-delayed interaction between the two drivers was muddling the midbass to some extent. The two fully independent and matched transmission lines removes the possibility of a time-delayed reflection from one driver to the next.
One problem that builders with rooms with low ceilings (8 feet) were encountering was excessive bass. The variable-height feature that confounded many builders was also not proving effective enough in limiting the output of the vent. Another conversation with Ron Nelson, along with experimentation by the members of the Ariel Builders Club, showed moving the vent off the floor by only a few inches made a major difference in the low-ceilinged, bass-heavy rooms.
Eddie Draaisma built a solid-sided box that was 4 inches high, raising the entire speaker 4 inches higher, while keeping the base-plate resting on the floor. For those of you who already have Ariels in a smaller room, you can try three different boxes ... 2.5", 4", and 6" high ... and see which works best for you. Changing the vent height by just a few inches makes a very large difference to the overall bass level, since it controls the efficiency of the room coupling.
Another possibility is building the much simpler ME2 which I've heard work very well in really small rooms. The best solution, though, is giving the Ariel more latitude for different rooms. This prompted the Mark 4.5 version, which is pretty much a Mark 4 with a separate rear vent, which allows the construction of an optional variable-height "chimney" to raise the effective vent opening a few inches off the floor if necessary. Note this chimney can be as simple as sectioned PVC pipe split in half, or as elaborate as a separate wooden sub-assembly.
A month or so later, Zoltan Kis modified his Mark 4 Ariels with dual side vents. (This version can also take advantage of split PVC pipe to extend the vents vertically). By this time, the naming was getting more than a little confusing, so I threw in the towel and simply re-named the whole series the Mark 5. Not just any Mark 5, but the Mark 5b, 5r, and 5s. So there. Three versions, build the one you want, you choose where you want the vent to come out. Baseplate, rear, or sides. In theory, they should all sound the same since bass wavelengths are 10 feet or more at the vent frequency, but in practice, the orientation of the vent will probably make a difference.
The actual production version is the Mark 6. This version is acoustically streamlined and optimized for automated NC-controlled woodcutting machinery. Note that the drawing omits the usual filling and wool pads shown in the other versions; this was done as a convenience for converting the drawing to a 3D DXF file. If you want to build the Mark 6 for yourself, add in the filling and wool damping pads shown in the other versions of the speaker. Although very difficult for a home constructor, this version should have the highest performance due to the gradual and progressive changes of cross-section profile in the transmission line. The goal with successive versions is a smooth approximation of a reverse-horn profile that transitions to a labyrinth at the far end.
I admit no less than six versions of the Ariel might seem intimidating at first, and probably a little annoying to those of you who are in the middle of project. In practice, though, most of this is ancient history. There are no reasons to build an early versions ... the ABC'ers report better results with each successive version. For those of you in mid-project, should you abandon what you have and switch to the latest version? I can't answer that. It's the same as whether or not you should upgrade your computer now, or wait next year. The improvements are noticeable, especially in the midbass, but the essential character of the Ariel has been there since the Mark 2. The improvements have all been in the direction of greater transparency, reduction of coloration, and optimizing the 100 to 500Hz region, which has always been the greatest challenge for any transmission-line speaker. Once you hear the TL sound, though, it's not easy to go back to closed or vented-box designs.
I placed the tweeter in an off-center D'Appolito configuration,
hoping to mitigate the upper-mid forwardness that many commercial
D'Appolito systems exhibited. By creating a lateral asymmetry
based on the golden-section ratio, it would help damp any standing-waves
that might form on the front panel. There would still be a potential
for a vertical standing-wave, but at least it wouldn't be doubled
in magnitude by a left-right symmetry as well.
Why did I depart from the well-known symmetric D'Appolito configuration?
Well, it's a long explanation ...
As I mentioned at the beginning of this article, a box is a terrible
shape for a loudspeaker. After all, a loudspeaker operates at
the speed of sound (not a figure of speech), and what was
the last time you saw an airplane that looked like a box? Just
like an aircraft, sharp edges and bumps create shockwaves at the
rough spots, and these shockwaves create drag in an aircraft and
secondary reflections in a loudspeaker. Since these reflections
have delays comparable to the delays created by the head and shoulders
of the listener, the edge reflections confuse the stereo image,
flatten the depth perspective, and create listening fatigue.
The truly correct approach is to use gently sweeping aerodynamic
shapes, resulting in speaker cabinets with large-radius curves
at least equal to the radius of the midrange cone. Gee, you might
think, nobody does this. Why? Cost and cosmetic expectations.
That's all.
In aircraft, safety and performance rule, dictating the shape
of the aircraft. After all, badly designed aircraft waste fuel
and crash; by comparison, badly designed loudspeakers only destroy
the sense of musical realism and beauty.
Having thoroughly demolished the hi-fi pretensions of the box-shaped speaker, here I go telling you to go ahead and build one. Well, I heard the reactions of woodworkers in many places, and it's pretty clear that exotic (by loudspeaker standards) shapes require exotic machinery and skills. Besides, most of us are used to the sound of box speakers, defects and all. So I decided to minimize the subjective effect of the reflections by carefully spreading out their timing.
This is where the golden section ratios come in. It is the ratio
called phi, and it is close to 1:1.618, one of the best
ratios to choose if you are trying to discourage standing waves.
All of the driver placements on the front panel are determined
by a series of golden sections, with the left-to-right tweeter
location being the most important ... fortunately, the midbass
drivers have fewer problems with diffraction, so the layout is
less critical.
So I took the standard D'Appolito configuration and offset the
tweeter just a bit, following the golden section phi ratio. This results in both the edge distance having a left-to-right ratio of 1:1.618.
In effect, the left side of the tweeter sees a cabinet
that is 1.618 times smaller than the right side. (The pattern
is reversed on the other speaker, of course.)
The intentional asymmetry dissipates energy from standing-waves
on the front panel, which provides a measurable, and audible,
improvement on the frequency and impulse response compared to
conventional D'Appolito systems. Although it takes a bit more
time to build mirror-imaged Left and Right cabinets, it's worth
the effort.
(NOTE: The imaging will be ruined if you forget to mirror-image
the left and right speaker. I found that out the hard way at the
1975 Consumer Electronics Show when the Audionics shipping department
accidentally shipped us a pair of Left speakers by mistake. We
had to ship a Right speaker by overnight air freight ... not cheap
for a 120lb. TLM-200.)
Panel dimensions:
All panels are nominally 3/4" thick, with the exception of
the double-thickness front panel, which is two pieces of 3/4"
MDF laminated together. Remember to accurately measure the thickness of all of your panels; we had to adjust our dimensions by 1/32" throughout the system
because our MDF was just a little thicker than the expected 3/4 inches. Note the rear panel thickness may be thicker than 3/4" if desired; you may choose 1", 1 1/4", or more if you want. Only the exterior dimensions change, while the interior measurements stay the same. Some builders report good results with extra-quiet cabinets with extra-thick rear panels.
All Exterior panels are premium-grade MDF (we used high-density
Canadian Rangerboard) and all Interior panels
are plywood (plywood is 3 to 4 times stronger in compression and
tension than MDF composites). You could do worse than use multi-ply
Baltic Birch plywood, which is the choice of the BBC. The amount
of wood in the Ariel is pretty modest, so you needn't be deterred
by the cost of the Baltic Birch. The Interior panels slide into grooves that are dadoed or rebated a quarter-inch into the front, side, and rear Exterior panels. I do not recommend screws,
butt joints, or biscuit joiners, since the mechanical properties
and air-seal are nowhere close to the performance of a well-made
dado (rebate) groove. Think about how the best furniture and musical instruments are made and you'll be heading in the right direction.
The mechanical integrity of the cabinet comes from the glued dado (or rebated) joints, not fasteners. Precise construction and test-fitting assures that all glued surfaces will mate accurately. If the cabinet is accurately constructed, it can be completely pre-assembled and test-fitted without gluing, which lets you check for interference and clearance problems before you commit to gluing.
A perennial question concerns the exact way to cut the dado/rebate and the edges of the labyrinth sub-panels. One way is to make a dado/rebate all the way across the cabinet interior, and simply make a simple right-angle saw cut in the sub-panel, as shown in the drawings. Another way is to stop the dado/rebate where the sub-panel ends, and round the edge of the sub-panel so it fits right into the rounded end of the routed-out dado/rebate channel. Which is right? Both are!
You see, turbulence in the labyrinth area is fine. The whole function of the labyrinth is to slow the air-flow; it is very different than a horn, where smoothness is everything. Although it's not pretty to have the dado/rebate extend beyond portions where the sub-panel stop, the extra turbulence introduced by a 1/4' deep groove really isn't all that important acoustically. But it is ugly even if you can't see it ... you know it's there, after all.
So if esthetics are important to you, and I admit if you're building an Ariel in the first place, they probably are ... then the alternative approach of rounding-off the ends of the sub-panels (sometimes called "bullheads") and stopping the routed-out dado/rebate right where the rounded panel ends is tidier and certainly a lot nicer looking. And if you have the great good fortune to make the cabinet with computer-controlled routers, this is the only way to go; everything fits exactly into place after the NC router automatically does all the hard parts. What about dimensioning? It's probably a good idea to have the rounded-off sub-panels extend just a wee bit further into the 2" wide aperture, maybe 1/8" or so, to make up for the wood that was taken off in the rounding-off operation. This keeps the cross-section of the TL fairly constant, which is a good thing to do.
The midbass flange diameter is 5.53"(5 17/32"), flange
depth is 0.188" (3/16"), and the diameter of the inside
hole is 4.56" (4 9/16"). T-nuts spaced at 45, 135, 225,
and 315 degrees are recommended for driver mounting, since it
facilitates removal and testing as well as providing a predictable
amount of torque on each bolt. I use 10/24 by 2" hex bolts
for appearance, durability, and the ease of adjusting torque evenly
on the driver frame.
(I do not recommend using wood screws to mount drivers.
Wood screws tear up the wood, making driver replacement difficult,
and cannot be torqued to uniform tension without further damaging
the wood. Wood screws are not compatible with drivers that are accurately and precisely mounted to the front panel.)
Although it's a hassle, you'll get improved driver performance
and easier mounting for the T-nuts if you make large-radius scalloped
cut-aways between the T-nut mountings, leaving about a 3/4"
lip for each T-nut mount. This makes connecting the hook-up wires
and lightly filling the interior of the enclosure much easier,
as well improving the airflow around the magnet.
The tweeter flange diameter is 4.125," flange depth is 0.156"
(5/32"), and the diameter of the inside hole is 3.25."
T-nuts spaced at 45, 135, 225, and 315 degrees are recommended
for driver mounting. As before, a scalloped opening between the
T-nuts is a good idea; not for airflow, which is zero for the
tweeter, but to be able to get the hook-up wires connected to
the tweeter without bending the rather fragile terminals on the
tweeter.
Although it's not very clear in the drawing, the tweeter chamber is acoustically isolated from the midbass TL sections. If you want to open a hole into this section, either make it barely large enough for the tweeter wire and seal it with RTV silicone, or drill a 1/2" hole in the bottom section only, then cover it with felt.
Be careful about setting up the router depth and getting a good air-seal with the gasket material. Even extremely small air leaks will degrade the quality of bass significantly. (This is true for all types of cabinets, not just the Ariel.)
Many people have expressed concern about the large radii on the left and right sides of the cabinets. I would like to comfort you and say that a 3/4" radius works just fine, but I can't. The first version of the Ariel had 3/4" radius edges, and I hate to say it, but the current version has a projected stereo image area at least 20 to 30% better ... with the same drivers and crossover as the first version. So yes, the large radius is really important, although I grant it is not easy to construct.
To get the correct radii on the cabinet requires a 3 hp router (about US$280 for a 1/2" shank 3 hp DeWalt), and expensive bits. (The 3/4" router bit is Amana #48520 and costs US$53, and the 1.25" bit is Amana #49524 and costs US$153. Ouch!) It is probably most cost-effective to go in with another builder for this kind of investment ... fortunately the large-radius bits are useful for making other kinds of high-performance speakers as well.
Important Note for Metric Woodworkers: Although all of the drivers are made in metric Scandinavia, I've converted the dimensions to the nearest 1/16" or 1/32" for the convenience of woodworkers using traditional measure. If you convert the driver dimensions back to metric again, the errors will add up, and the mounting holes could easily be the wrong size. Save yourself a lot of pain and aggravation and simply measure the drivers directly when you get them in the mail. Whatever you do, don't start building the cabinet before you get the drivers!
Since you will be building the cabinets in metric units, use the interior dimensions as the starting point for all other measurements. 18 or 20mm MDF and plywood are excellent choices for building the cabinets; just keep the interior dimensions similar and let the exterior dimensions fall where they may. In case you're curious, the left-to-right centerline of the tweeter is determined by the Golden Section ratio of 1:1.618 measured from the midpoint of the radiused edges.
In some countries MDF or even plywood is not available; if this is your situation, use the softer woods for the exterior cabinet, and the hardest woods for the interior elements. Make sure that wood that you use has been kiln-dried, and has been kept dry at all times between the kiln-drier and your cabinet shop. It doesn't do any good to use kiln-dried wood that's been sitting in a mud puddle!
Access Door and Binding Posts:
Well, they have to go somewhere
(like the old Vaudeville joke, everybody's gotta be someplace!)
If you put them here and make the panel itself removable, you
can insert the wool and/or crimped Fortrel after you've built
the enclosure. For convenience, the inner cutout can be 4.56"
(4 9/16") in diameter to match the driver cutouts.
The binding posts are then mounted on a 5.53" (5 17/32")
diameter disk of 1/4" Masonite or transparent acrylic, which
is fitted into the routed-out flange with a gasket. Don't forget
about a good air seal here also, so be sure to use a gasket and
6 or 8 screws (or better yet, bolts and T-nuts) to mount the Masonite
or acrylic disk.
Don't expect anything terribly profound, this is just what I observed
while I helped Mike build the first prototypes.
WEAR EAR PROTECTORS!!! Even though I wore them during all
of the sawing and routing work, my ears still rang with a 9 kHz
whistle for many hours afterwards. Hearing tones like this means
the nerve cells in the cochlea (inner ear) have been traumatized
and that you should keep things very QUIET for a day or
two afterwards. As for me, it made me aware that audiophile woodworkers
should go out of their way to use extra high quality ear protection
... what I was wearing obviously wasn't good enough.
Mike and I were exasperated by a persistent series of small (1/16"
to 1/32") errors in the lengths of the outside panels and
in the corresponding dado locations, which required a fair amount
of rework as we went along. We later traced this to an inaccurate
miter box combined with not cutting all 8 outside panels at the
same time. Make sure your miter box is set to exactly
90 degrees, and don't allow any domestic interruptions
to a sequence of related operations.
The various dado-cut (rebate) slots make it possible to test-fit the various portions of the enclosure as you assemble them. This lets you
check if any tolerances have gone astray or if any panels have
the wrong dimensions. This is an essential step! Make sure you
do this frequently as you build up the various assemblies that
go into the enclosure.
The final step before gluing is to pre-assemble the enclosure
and check for tight, accurate fit in all locations. You'll find
that it assembles like a big Chinese puzzle and fits quite snugly
without glue. If all the pieces fit, you can disassemble, clean
up all of the interior surfaces, glue the felt down in the appropriate
places, get out your clamps, and go ahead with the final assembly
and gluing procedure.
I've spoken to two different constructors who used a vacuum bag
to laminate the front panel and also do the final enclosure clamping.
This sounds like a really great idea; if you've got the facilities,
it neatly sidesteps all of the troublesome warping problems that
can arise from conventional clamping procedures, as well as assuring
uniformly high pressures on all glued surfaces.
The decorative faceplate of the tweeters (and mids) must be mounted flush with the front panel of the enclosure. Even tiny variations create reflections in the time domain which are measurable and audible. On the subject of reflections, if you must use a grillcloth and grill frame, mount it about 2" away from the front panel using the old Infinity-style standoffs. The major problem with grill cloth isn't the absorption, which is quite minor in sheer fabrics, but the reflection, which causes roughness in the high frequencies if it is any closer than about 1" to the dome.
Testing for panel damping is really simple; play loud orchestral or choral music through the loudspeaker and press your ear directly on the rear or side panel of the enclosure in an area close to the drivers. With many speakers, you'll hear a loud, offensive drone, which you'll then discover is still audible when you listen to the speaker in the normal seated position. One of the most important differences between mass-fi and advanced loudspeakers is the quietness of the cabinet panels; with the best speakers, the cabinet is almost totally silent.
In the Ariel and ME2, the panels have very small free areas, and full-width plywood cross-members are used to increase rigidity. Small Deflex pads may optionally absorb the remaining vibration in the panel directly behind the driver, but these should be used in moderation or not at all with the later versions of the speaker. The best application for Deflex may not be on the cabinet walls at all, but glued to the magnets of the drivers (with a small aperture removed for venting the driver).
To minimize acoustic standing-waves inside the cabinet, use 1/2" or 3/8" thickness of F13-density 85% wool felt to line the areas around the drivers (as shown on the construction diagrams). This is an industrial felt, light grey in color, and is used for noise damping in truck headliners and under heavy machinery in factories. It can be found at gasket-supply houses in the industrial part of a major city. The panel directly behind the driver is most critical, since it will reflect energy right back at the driver if not lined with a heavily absorbent material, such as industrial felt or Deflex damping pads.
My experience with foam damping pads (2" thick #2 gray polyethylene foam) has not been very successful. Using the foam at various locations within the enclosure all resulted in a substantial loss of impact and realism and the addition of a toneless "gray" coloration to the overall sound (by contrast, the measurements showed a tiny improvement). My partner, Karna, immediately objected. (Her exact words were "They sound constipated! Take it out!") She made me remove all the foam blocks from the speakers, and sure enough, they sounded much better and more alive. So watch out when you use foam damping material.
The straight portions of the line can be very lightly filled with cleaned, combed, and carded long-fiber wool, crimped Fortrel, Acousta-Stuf, or polyfill. (The labyrinth itself is intentionally left unfilled.) The wool appears to be most effective in the midband, and provides a sweet and transparent sound. By contrast, crimped Fortrel or polyfill has only a moderate damping effect in the midband and tends to selectively damp the bass. If you use wool, make double-sure it is moth-proofed; having moths in your house is no joke! (Bold experimenters might want to try tiny 2" by 6.5" pillows of feathers or goose down located directly above the entrance to the labyrinth ... these materials are very lossy and will absorb line reflections quite efficiently, although some bass may be lost.)
The removable panel that supports the binding posts at the rear of the enclosure is removable for a reason; it allows you to add or remove filling material long after the cabinet is built, so the system can be tuned for your own personal tastes and listening environment.