Mini-monitor (14K)Originally, these were going to be built into a standard MDF box and simply veneered or painted. But then I had a crazy idea that would result in a unique and beautiful enclosure assuming, of course, that it stood a chance of being practical...

Birch ply

A while ago at work, a large number of shelf units were required. Several sheets of 23mm birch ply were ripped into 8 inch wide planks and assembled as required, meanwhile the offcuts sat around unused and were eventually skipped - I've got several pieces, all perfectly machined to width, and approximately 42 inches long. Needless to say, I'm planning to build some shelf units too!

Noticing that the diameters of the two drive units add up to nearly 8 inches made me think about using this ply for the baffle. I retrieved a plank from the attic and found that the exact width was 198mm. Take off 70mm and 116mm for the for the tweeter and bass units respectively, and you're left with 12mm. Allow a spacing of 2mm between the drive units, leaving 5mm of "margin". A nice compact size...

Looking at my small stocks of "recycled" 12mm birch ply, I quickly realised that I didn't have enough to form the walls of a conventional box. While musing about the possibilities, including the rather rash option of actually buying some (!), inspiration struck...

The end-grain of birch ply seems to be in fashion at them moment. With that in mind, I was planning to cut the sides, top and bottom panels such that the end-grain of these would surround the baffle. These panels would be mitred at 45 degrees, just like the Pen Audio Rebel 2's that were reviewed in the June 2004 issue of Hi-Fi News. But thinking about exposed end-grain, and the thickness of those planks of birch, I hit on the idea of laminating several pieces of 23mm ply to form the depth of the box, giving the impression that the speaker was in fact hewn from a 6 inch thick piece of birch ply!

The challenge of this appealed as much as the finished aesthetic. It would be wasteful, as most of the birch ply would be thrown away - I guess this is why I've never seen this approach before. Accurately cutting the birch ply would be tricky with only a jigsaw, but I do have a power-sander!

So, how practical is this daft idea? The first thing to consider is how many "layers" would be required to make a suitable box:

Table 1 - Box parameters as a function of "layers"
Layers Internal Depth External Depth Volume (litres) F3 (Hz) Q
1 2.3 4.7 0.41 144 1.53
2 4.6 7.0 0.82 115 1.13
3 6.9 9.3 1.22 104.6 0.96
4 9.2 11.6 1.63 99.4 0.86
5 11.5 13.9 2.04 97.2 0.80
6 13.8 16.2 2.45 96.7 0.75
7 16.1 18.5 2.86 96.1 0.71
8 18.4 20.8 3.27 96.1 0.69
9 20.7 23.1 3.67 96.1 0.67
10 23.0 25.4 4.08 96.7 0.65

Note: External width and height are 12.6 by 19.8cm. Assume a wall thickness of 12mm, giving an internal width and depth of 10.2cm and 17.4cm respectively. Front and rear panels are 12mm birch ply. Volume taken by drivers, crossovers, etc., and the effects of stuffing are not included at this stage. Results are predicted by WinISD, using average of measured TS results.

Needless to say, one or two layers wouldn't work because the driver magnet would stop the back cover going on! But even if this wasn't an issue, the results would be fairly horrible thanks to the high F3 and Q. The theoretical optimum is given by 7 layers, resulting in a system Q of 0.71. But it's tempting to try 5 or 6 layers, as the graphs drawn by WinISD don't look very different. A depth of 16.2cm "feels" about right - I've always liked the look of boxes that are deeper than their width. However, building up the box from layers means that I can audition it at each stage.

Cutting the layers

This was going to be a repetitive task, which calls out for some sort of jig. Encouraged by the results of the jigsaw and straightedge used for the MDF prototypes, I built a simple right-angle guide from a piece of hardwood and an offcut of Conti-board:

Right-angle jig (28K)

Dead simple, but it works remarkably well. I cut seven pieces in about half an hour, and they all worked out pretty well. The top and bottom are perfectly straight, which forms a reference to help when assembling them, and the imperfections on the sides can be arranged to average out, meaning they disappear during sanding.

To hollow out the layers, I found that the best approach was to make the cuts free-hand with the jigsaw, having first drilled holes in each corner. The edges weren't perfect, but actually this isn't a big problem because the internal walls won't be seen and you could argue that the uneven edges would prevent any standing waves forming!

But that said, I was worried that there would be a significant loss of internal volume, and wondered about a more professional approach using the router. I have a home-made jig for cutting trenches or housing joints (or dados in the US) - surely this could be used somehow?

My router and dado jig (31K)

This jig is simply a piece of 6mm MDF with 3 pieces of oak that were precisely machined at work to be perfectly straight. The bottom piece forms a "fence" for the work, and is glued firmly to the MDF. The left piece is also fixed, and is set at a perfect right-angle to the bottom piece; the right-hand piece is adjustable. The two top pieces simply confine the side-to-side movement of the router, allowing you to cut slots in the work at any width required. For an example of the results, have a look at this page of the workshop conversion...

This jig had nearly all of the elements required for the job in hand. I simply added one more piece of hardwood, as shown here:

Modified dado jig (25K)

So now it's just a case of locating the layer in the jig, clamping it to the bench, and cleaning up the jigsaw cut. You have to make sure you stop at each corner, and as it's quite hard to see the corners through the base of the router and the jig, some compromises will be inevitable there. Each edge requires two passes, as my straight-cutting bit isn't long enough to do the entire thickness in one go.

Cutting the baffle openings

I wanted to rebate the drivers to make the project look somewhat more professional. Also, there is plenty of scope for sonic problems with surface-mounted drivers - have a look at this page on John "Zaphoid" Krutke's site for some more information. Having spent ages thinking about the various ways to achieve this, I arrived at a delightfully simple solution:

Circle-cutting jig (30K)

I could have done with a circle-cutting jig to cut that circle! However, that doesn't matter. As you can see, it's just a piece of 6mm MDF that replaces the sole-plate of the router. The holes give particular diameters in conjunction with a straight-cutting bit of the relevant size. The holes are exactly the right size to give a tight fit over an M3 spacer that is bolted to the work in the centre of the hole. I tested this on a scrap of MDF, and the results were just beautiful!

Here's my step by step guide to flush-mounting drivers:


The assembly order needed careful planning. The baffle was initially screwed to the first layer and was trimmed to the exact size using the router fitted with a flush-trimming bit. This meant I could measure the centres of the box and make the pilot holes for the circle-cutting operations detailed above. I knew that I'd need to separate them again to route the rear of the baffle, but having completed the baffle, it was glued to the first layer and the screws were removed as they were no longer required.

Enclosure components

For the rear panel, I used a similar procedure but as the rear panel will remain removable, there are eight screws on the outside edge. The four 4mm banana sockets are temporarily fitted to enable the crossover to be optimised - you'll see later how we fit the terminals.

Before gluing the middle layers together, I wanted to determine exactly how many were needed. We saw before that seven layers would result in an enclosure volume of 2.86 litres, giving a "perfect" Q of 0.71 and an F3 of 96.1Hz. But as adding stuffing to an enclosure increases the apparent volume of air seen by the driver, I thought that I might get away with using 6 or even 5 layers. It was time to make some measurements:

Table 2 - Measured performance as a function of "layers". Driver 2 used.
Layers Internal Depth External Depth Volume (litres) Fs (Hz)
3 6.9 9.3 1.22 148
4 9.2 11.6 1.63 130
5 11.5 13.9 2.04 116
5 (with stuffing) 11.5 13.9 2.04 105.2
6 13.8 16.2 2.45 109.9
6 (with stuffing) 13.8 16.2 2.45 104.3

Because the layers were held together by hand, the box wouldn't be completely airtight. This means the results should improve when the box has been completed. Perhaps surprisingly, it seems that there is little benefit to be gained by moving beyond a well-stuffed five layer box. Along with these measurements, I also couldn't hear a difference. So, decision made - 5 layers it is.

The layers can now be glued together. Because the glue takes at least four hours to set, I wanted to glue all the layers in one go. But I quickly discovered that this is a bad move, as you don't really get enough "open time" with PVA initially, but it also takes ages to set properly - therefore after a rush to get the glue evenly distributed evenly, there's a panic to get the clamps installed and tightened. And while tightening the clamps, the pieces invariably move slightly!

So the first box was slightly off-square from front to back. After unclamping the assembly, it was actually a bit worrying, but after planing and sanding, the error is only a mm or two. However, for the second box, I decided to play things safe, and took the time to carefully assemble the layers one by one. With hindsight, this project would have been the perfect excuse to buy an air compressor and a brad-gun!

Complete enclosure (37K)

And here's a picture of the first enclosure, after much sanding! After getting the different layers nearly the same size using a power plane and the power sander, I was able to hand-sand the box. For this, I used a scrap of 18mm MDF as a sanding block, and started with 60-grit paper. This quickly flattened all the high spots, and squared up all the edges beautifully. I was amazed at how straight and crisp the corners became after this. From there, I went to 120, 180, 240 and finally 320 - by this time, you couldn't detect any of the laminations of the box by touch, even with a sharp fingernail. In order to ensure a perfectly flat and square enclosure, I taped complete sheets of sandpaper to the bench for the final sanding operations.

It took a long time to sand these perfectly, but there's something quite therapeutic and rewarding about spending an afternoon sat on the drive sanding while enjoying the summer sun!