I've been doing some studying of the currently favored horns and TQWT,s and such; and come to a suspicion that there's a problem that afflicts most folded horns and tubes; that they're really nowhere near their theoretical optimum shape, but a very loose polygonal approximation. I feel it very worthwhile, if we're going to construct horn and pipe loads for drivers, to trouble ourselves to make them smooth, nicely folded (if they're folded), and if they're horns, dimensionally rigorous. The problem with the advanced loadings is that it seems like everybody's trying to see how poorly they can be constructed and still work. Why? Might as well make a well-designed VB and get it over with.
        One of the things that initially attracted me to the Diatone (in addition to its rep for natural sound) was that its Thiele-Small params seem to make it suitable for an untapered TL, a much easier load to construct optimally than a folded horn; which is a very difficult load to design and construct accurately.  There are a few drivers appearing now that use this approach: the Fostex FE164 and FE168Σ, the Supravox drivers, the Moth Cicada, and two newer Fostex drivers, the FX200 and FX120.   The line drawn here is long enough not to need stuffing unto death, nice and smooth in contour, and has a novel feature in the horn mouth, which I hope will help its room coupling, and thus bass production.  Its line cross sectional area is correct for any of the 160mm / 6½" drivers, and all the ones I have take a 5¾" cutout, so this line will be plug'n'play for the Diatone, FE164, FE168Σ, and a neat looking RatShack fullrange I got on closeout, the 40-1285D.

The process / technique that makes this nice curved line possible is called kerf bending.   A kerf is the slot left by a saw blade; for a power circular saw, this is typically 1/16" to 1/8".   Kerf-bending is a way to make relatively thick pieces of wood take fairly tight bends, or radii.  Let's take a piece of, say, 3/4" plywood, say, 7 3/4" wide; taken from the long dimension of the panel, grain of the outside ply on the long run.   If we make saw cuts across the piece, almost all the way through, at regular intervals of about 1/2" to 1", we can bend the piece in a pretty tight curve.  If you look at the stand under the speaker in the picture, you'll see that we can make quite a tight bend this way.  I've; a bit of skill in this technique, having made some few constructions in this way in my other life as a cabinet maker.  If we make the kerfs (kerves?) on the inside of the piece (that is, the piece is bent so that the kerves/kerfs close and touch), and make them at regular intervals; the piece will take an even radius, determined by the spacing of the, ah, saw cuts.  The 'trick' to doing this efficiently is to design the construction to use as few radii as neccessary (which the boxes above do), and use scrap stock to determine the spacing of the kerves that will produce these radii.   There is a problem with the places where the curves feather into each other and into the straight pieces, as the little slots are only glued together at their tips (the inside surface), and the piece will come apart catastrophically in the saw when cut on the long angle.   How do I know this; hm-m?   In those sections, I fill the slot with a little triangular piece of wood; as wide as the slot at its base, and as long as the slot is deep. 3 or 4 slots at the feathered end of each of these pieces will need this treatment.   The 'profile' of the line is built up on one of the side panels (which has the profile drawn on it), using clamps and spreader sticks and tape and whatever sneaky tricks will get the job done; then the wires and any stuffing or damping is put in; then the remaining side is put on.   These are long and smooth enough that they may not require stuffing except right behind the driver, to curb reflections back through the cone; so I'll just line the area around the driver and down the curve with half inch felt, set the wiring, and close 'em up.

 
 
The left side panels marked off for the profile pieces.  I mark a little big, so I can see both pencil lines when I set the profile piece.  The full-sized profile allows me to determine the approximate length to which the profile pieces are cut before kerfing and bending, the spacing of kerf needed to give the radius I need, and the length to which the straight sections will be trimmed afterward.  These measurements are ridiculously difficult to calculate or take indirectly; but very straightforward to take directly from the pattern, simply by holding the profile piece in place.  It may be observed that, at least according to population density, the eraser is my tool of choice.

The stick with the screw driven in one end and 'save' written on it is a dedicated compass.  It has a notch at each radius increment of the drawing above, so I may draw the radii directly, without adjustment, and with repeatability.

 
 
The side panel cut to the radius of the first turn.  If you look closely, you may see that the marks on the top and rear perimeter of the panel are 5/8" from the edge, rather than 3/4", the thickness of the other pieces.  This is allowance for the edge banding that dresses the plywood edge; which is applied in the next step.

It is important after jigsawing the radius to verify that the cut edge is square.  The outer (show) edge may be just a bit high, but if it is low, it will be impossible to remove the gap between the panel and the edge banding that will occur when the banding is bent around the radius.  A little time spent with the block plane trueing the edge up pays big dividends in small gluelines.

 
 
Ultra-tech equipment being used to bend the edgebanding to approximate shape, to ease the assembly process.

There's a knack to this; you can't force it.  The section of the banding that will be bent is kept wet (I wrap it in a piece of terry soaked with hot water and put it in the sun) for about an hour.  It's held right at the spout of the ultra-tech equipment, and about five pounds of bending pressure is applied.  After about ten seconds, you'll feel it start to relax into the bend.  At this point, do not increase pressure.  Let the piece bend into your steady pressure.  Now, move on a couple of inches.  Repeat.  Patience, bruddah, patience.

If you have increased pressure, if you have not exercised patience, you will know.  You will now be holding a piece of edgebanding in each hand.  I have been doing this since 1968, and I always cut one extra piece for about each four I need to bend.

This sort of assembly is an on-the-moment thing.  You just have to observe the situation and apply pressure as needed.  There's no telling where you'll have to clamp from, or to.

Assembly under the banana tree adds to the, um, sweetness of the delivery of the, ahum, dakine.  Mana`o e ka mele.  Yeah, that's it.

I mentioned above that I have made some few curved constructions using kerf bending, but none that require the precision of curve and number of various radii of curvature of this project.  Vanity notwithstanding (and this alone is a pretty good trick, for me), I am going to walk through this project as it ambushes me and I (hopefully) solve it.

The arithmetic of the kerf spacing (for closure of the kerfs to produce a uniform curve, a radius) may be simply stated as follows: The inside diameter of the curve will be two thicknesses of the material smaller than the outside diameter.  For 0.75" plywood, this will be 1.5".  Therefore, the difference in length (circumference) between the inside and outside of a complete circle of any diameter, given this thickness, will be 1.5 π", or 4.7124".  In order to reduce 'faceting' on the outside of show pieces, I will use my nice little 7¼" Tenryu, with its one sixteenth (0.0625") kerf.  So, then, the number of kerfs to go around a complete circumference will be 4.7124" / .0625", or 4.7124 x 16 = 75.4 kerfs.  Now we have the data we need to determine the kerf spacing:  This box uses turns of the line at either 90° or 180°, or very slight variations thereof.  A 180° turn of the line will require 75.4 / 2 = 37 kerfs, and a 90° turn 75.4 / 4 = 18 kerfs.  There are no fractional kerfs; it's all or nothing for the saw. We'll see a couple paragraphs down why we chop rather than round the number of kerfs to cut.  

Using a scrap as a test piece to verify kerf spacing, two things became immediately apparent; number of kerfs to complete a 180° turn with my particular saw blade is between 33 and 34, rather than 37 kerfs.  Blade a little over a sixteenth, at least in its kerf width.  So, if the number of kerfs (hereafter N) / 90° = 16, and a quarter turn of the upper rear wall of the box is 7.875" (the outside radius of this turn) x π", or 12.37", the spacing of the kerfs will be just a little over ¾".  Trying this on the test scrap showed significant (ugly, that is) faceting on the outside of the piece.  My solution was to use the closer spacing for the narrowest turn (the 180° 'doublebacks'), just over three eighths inch, and fill the slots with tapered slips that only let the slots close about halfway.  This also gives my tiny brain one less kerf spacing to keep track of.

 
Kerfing the top-rear piece.  The setup I use is to put a long support base out to a horse (the part piece becomes pretty floppy once it's kerfed), and a supplementary back fence which is then cut through.  The kerf spacing is then marked on the fence from the cut-through slot.  This saves a lot of time in marking the pieces; I just calculate and mark off the beginning and end of the curve, and kerf away.  You can see the 'end' mark about an inch ahead of where I'm cutting, and the previous kerf registered to the spacing mark on the rear fence.

Give the piece an extra eighth or so on the measured-from end, and make all the kerfs inside the register marks.  Start about one half a kerf spacing inside the 'start' mark and continue kerfing to the last one that will fit inside the 'end' mark.

The way I make the slips is as follows:  I rip a 3 - 4 foot long, 3" wide board of soft wood, like pine or cedar, down to thickness the same as the depth of the kerfs, about 5/8".  I then set the table saw blade to an angle of 1° - 2° and the rip fence to a piece width the same as the kerf width, about a sixteenth.  The rip fence will be right next to the blade.  I then run the piece through the saw, making a first slip which is tapered on only one side, and will be used as a 'maybe' piece, in case I need a thicker one. Now I flop the board, end-for-end, and rip again, with the edge touching the rip fence only at the top point; thus making the first piece that's tapered on both faces.  I have a look at the thickness and taper; cutting a few to length, inserting in the kerfs and checking the curvature, and adjust angle of blade and position of fence as neccessary.  Now, rip, flop, rip, flop, across the board until I have the pieces I need plus a few extra.  The slips can be adjusted with the block plane afterward if neccessary, but this is laborious to do to each one, so I want to be over rather than under, but as close as possible.  Last thing, I hold them in a bundle and cut to a length about an eighth shorter than the width of the profile piece.  At this time, I also rip some slip lengths that taper to nothing at the nose, to be used in the kerfs at the ends of pieces that feather out into the wall of the box or other pieces.  Without the last 3 - 6 kerfs filled and glued with these, it's impossible to cut the feather end without the piece coming apart.  On my saw, the half-kerf tapers took a blade angle setting of 1°, and the taper-to-nothing pieces about 2°.

Now I put a slip in each kerf of the piece, put a 4' pipe clamp across the ends of the piece, and close it up until tight.  I position the piece on the side panel and see how it's going; then open up and plane the slips on the narrower edge where the curve is too open.  This is a trial-and-error process, and it took me a couple hours and some exercise of my construction vocabulary to get the first piece right.  The second piece went a lot faster.  The process can be shortened by taking advantage of the natural variance in the slips; some will be thinner, some thicker.  I fit registering the top/front edge flush on the side panel, and once it fit on the curve, I trimmed the bottom end flush to the bottom of the side panel.

 
 
Once the piece fits, it's time for assembly.  Assembly is the performance art of this sort of work, especially when using an adhesive like PVA, with its fast kick .  If you don't get butterflies when it's time to assemble, stop.  Step back and reassess, because you are not fully appreciating the predicament you are in.  I love assembly.  Review your procedure, and have everything you need convenient to grab, including the wet rag and a roll of 2" masking tape - the real sticky kind.

First the kerfs / slips have to be glued up.  I use the glue bottle held tight and perpendicular to the surface so the glue goes mostly into the kerfs.  I put a little in each kerf, then put the slips in, then another application, this one heavier.

 
 
Now, I squeegee the glue into the kerfs with a 2½" flexible knife.  Then, I close up the curve by hand, which forces the glue both into and out of the joints.  I open up the curve, and knife in again, and so on.  I do this two or three times, until the amount of squeezeout doesn't decrease; then put the long clamp across the ends, position on the side panel, and adjust the clamp for fit.

 
 
This is the only piece that I glued up and mounted to the side panel in one operation; after this it gets impossible to position the kerfed pieces on the side panel with clamps in place if the rest of the pieces are there, since they interlock.

Once the piece is glued up, I put a nice fat glue bead on the edge of the side panel and reposition the piece.  Clamping proceeds from one end, so that the piece can be urged in as I go along.  In this case I start at the top-front, which is the register end, and work on down the piece to the bottom.  The pipe clamp is left on to help the piece stay square to the side panel.  Excess glue is wiped, and the piece checked for square at the top and bottom ends to the side; adjusting by moving the bottom end of the clamp in or out.

 
 
The next piece is the front of the box, which curves under and becomes the top of the 'horn' mouth and the front wall of the back section of the line.  It is kerfed as above, and since it takes just under 34 kerfs to bend 180°, I cut 33 times, and then fold it up and try on the assembled side/rear of the box.

The reason to not cut the fractional kerf that makes up the last bit of the turn is that it's far easier and quicker to adjust like this.  I make a sanding stick in the form of a very shallow vee, narrow edge almost sharp.  A piece of 80 grit is folded across this edge, and the openings of the kerfs can be widened a bit with just one or two swipes of the stick.  No slips to mill, plane, or cuss at when they fall out of the kerf during fitting and I can't remember which kerf they were in.  Doing it this way, I can go right back and forth from the saw table to the profile on the box, and get the piece fitting correctly in just a few minutes.

Since the profiles are exactly the same, at this time I also kerf and glue up the piece that arises upward from the back of the box, turns down 180°, and terminates at the axis of turn of this piece.  Since that piece feathers into the back wall, it's much easier to fit it on this front profile which is out and accessible.  By the way; when kerfing and clamping a piece like the rear piece, which ends right at the bend terminus, you need to leave two inches or so extra on that end of the piece when kerfing; otherwise it's ridiculously difficult to clamp those last two kerfs closed.  Ask me how I know this.

Because this piece (and most of the rest) will not be assembled to the side panel when it is glued up, it will be clamped to a spreader block, which will both insure proper shape and keep the piece from twisting in the bend.  It's done as you see; the spreader is cut square and dimensional, and clamp and spreader are positioned and the clamp closed just tightly enough that the spreader can be moved.  The clamp is pushed down until it rides on the edges of the piece on both sides, and the spreader then pushed up until it lays against the bottom of the clamp.  If you look, you'll see that the far side of the piece is angling away from the spreader at the bottom; the piece has a little twist in the bend.  When the clamp is tightened, this will come out.

Before this piece is glued up, the ends are trimmed, the speaker cutout is sawn (so I can chamfer the back edge), and the 'nose' piece is glued on the inner end, just so it can be drying at the same time. This construction is taking place in the biggest rainstorm of the last five years, and drying time is at a premium.

 
 
Front glued up and sitting in position on the back / side.  Durn, this is starting to look like a speaker box!

At this point, I clamped the front to the side and stood the box up, just to get a feel for it.  Jeez, for a two meter line of 1.2 SD, it's little!  I could have made it a bit bigger.

 
 
Back under the banana tree; glueing the front on the box.  The plywood bridges are used to transfer the pressure from my short(ish) clamps to the inner run of the piece.

 
 
The feathering cut is marked off on the piece by positioning it on a duplicate drawing of the pipe profile drawn on paper (I just draw the parts I need to use for this), and cut as you will see.  I'm not going to say any more about this starting cut on the table saw; if you don't already know how to make this cut, don't do it.  This is a pretty unconventional cut, and the liklihood of kickback if you haven't made a lot of them is very high.  If you don't have a well adjusted small saw like mine, you will also be standing behind the piece when it kicks out, not leaning over the off side of the table like I am.

For those who do know this cut; use a smaller blade (this is a 7¼" Tenryu) with its thinner kerf, and remember not to cut far enough through for the offcut to be floppy on the piece, as the piece is resting on the outside corner of the offcut as its register on the saw table.  I was gratified to find that with the little slip pieces in the last four kerfs, the feathered end stays together nicely.

 
 
Finishing up the feathering cut.  If you don't have a lot of experience making weird cabinet cuts on the table saw, this is the way to make the whole cut.

The saw doesn't have to be a Japanese type, but it does have to be very sharp. The technique to making odd long cuts like this is to let the saw cut at its own rate, with only minimal downward pressure into the cut.  If the saw is at all dull, or you push the cut, the kerf will narrow and the cut will first become impossible to 'steer', then difficult to even finish, the saw binding hard in the narrowing kerf.

 
 
Trueing up the feathered surface with the block plane.  Again, the tool needs to be very sharp, as you're planing a mixed bag of crossgrain pieces.  Don't plane all the way across, plywood blows out badly on the exit side.  Turn the piece over to finish up the far edge.  Check the fit of the piece in the box until the feather is correct.

 
Assembly now proceeds repeatedly as above.  Each piece is end cut or feathered as neccessary, fit in place, and glued in.  Any pieces which have unattached ends when glued in (depends on your order of assembly which ones they will be) must be checked for square (to the side panel) when assembling.  Do not put yourself in the situation of having to compensate for incorrectly installed previous assembly;  it just gets worse and worse.

 
 
The last piece to go on is the other side panel.  The ½" felt around the driver, the fluff, wires, and the sand in the void spaces are all in, everything ready.  I put a little blob of alex caulk around the wires where they exit the void they pass through, so the sand doesn't leak.  I don't like to use rope caulk, modeling clay, or plumber's putty (all essentially the same stuff) for this, because I've found it dielectrically very funny stuff.  Once had it fall across a signal wire for several inches inside a piece of gear, and it really fouled up the sound.

I'm stuffing the first half or so of the line to a fairly light density.  I tease the polyester out a bit, and don't pack it when I put it in;  I estimate about .3 lb./cubic foot or so.  Make sure the fluff is out of the way of the joint before assembly, and drop the side panel straight down on the box.

 
 
Just like the drawing!  The right box ready to receive stuffing in the very top of the line and the speaker; the left box ready for sand, fluff, wires, and closing up.  For the very first time, I'm not going to give these the cabinet prep/finish (about a half week's work) until I listen to them for a while with all four drivers installed.  This also means they won't have the drivers inletted, though I may put some thick felt around there and round the upper corners of the box roughly;  these procedures singly, so I can check if I actually can hear any difference.

Two impressions;  yes, it's a smaller box than I thought it would be, and, rasserfratz, the thing weighs a bloody ton!

First driver in is the Diatone P610MB, the speaker for which I drew this load in the first place. I'm going to play them heavily for a few days and report back.

 
February 7, 2004.  Ah yes, the fabled Diatone.  I listened to these for about ten hours a day for a week.  The Diatone is not the giant killer cult fullrange I had heard about/hoped for. It is probably the best midrange I have heard, and super useful in that role since it only needs to be filled in above 10-12kHz and below 70-80Hz, so all the handoffs will occur well outside the psychoacoustic red zone.  Throughout its range it is super articulate and timbrally true, but it does not get that half octave at either end that would take it over the top as an actual fullrange.  Just as well, they're dead and gone.  I have two pair of them, though, and they're definitely not for sale; I have future schemes for them as extended midranges.

 
The Fostex FE164 is a considerable improvement.  It goes about 10-15 Hz lower, and about 2-3K higher than the Diatone, although its midrange is a little more bland.  It still doesn't give me that 45Hz bass that the Audax HM170G0 does; I'm not sure why I was thinking that a driver with an Fs of 60Hz and an XMAX of 1mm magically would, but there we are.  With the two drivers tried so far, this ends up being a pretty big box for the amount of bass that it makes, but this is primarily a matter of the drivers mounted, and transmission lines are more about clarity and natural sound than pounding bottom anyway.

The FE168Σ shows me a lot of promise; it has a little higher QT, a little higher XMAX, a bigger magnet, a nicer cone, and better overall construction (cast frame, damped magnet).  It's a no-turning-back deal; you can see the four additional mounting holes marked off on the baffle, and it's gonna be ugly if they don't sound good, but hey, it's experimentation.

Another fantasy I'm starting to entertain is putting the Audax system into this box, tweeter mounted in a pod on top.  This might be very instructive, since I'm already familiar with that setup in an optimum (and extremely well constructed) vented box.