Crossover design:

For conventional passive speakers, by far the hardest element of the design is the crossover. In my experience, it's quite easy to get something that sounds OK, but making something that sounds special is very, very difficult without a measuring setup or at least a good reference speaker to perform a subjective comparison.

With this project, I was under time pressures - my house was on the market, and could sell at any time. So, in the interests of actually completing the project, I decided to "cheat", and use John Krutke's design. This was rigorously designed using comprehensive software and a decent measurement setup. It's quite possible that I'll revisit this in the future but this will really be with a view of improving my knowledge in this area. For for now I'm happy to trust John here.

I had some problems finding the exact components specified, so had to make a number of substitutions. The required 5u6 capacitor was made from a 2u2 and 3u3 in parallel, making 5u5 in theory. However, I measured the individual components and by selecting examples that were towards the higher end of their tolerance, I was able to get to 5u6. The 25 ohm resistor had to be 24 ohms, and the 8 ohm was 8.2 ohms.

I could only find the 1.5mH coil, so decided to wind the other two. I remember doing this many years ago using an electric drill and a slotted opto-coupler in conjunction with my Marconi timer/counter - I'm sure it was highly dangerous at the time! Using some scraps of timber, I decided to build a simple coil winding machine that could be safely hand-cranked.

As you can see, I used the same scrap of MDF that I tested the circle-cutting jig on! Everything else came from the scrap pile. The crank handle is simply some offcuts of 12mm birch ply, driving a 10mm section of dowel. The second piece of dowel has a small hole drilled through it, and this is to guide the wire onto the former. I left space in the middle for a mechanism to drive this dowel from side to side, but would need to find an adjustable way of gearing down the rotational speed of the main crank. The final section of the machine is concerned with regulating the tension of the wire from the spool - I'm not sure if or how this would be done on a proper machine, but I remembered that supply tension was a problem before.

This closer view shows how the tension is regulated - if you've ever looked into a VHS mechanism, you might recognise this. The supply spool is bolted onto an aluminium bar that I cut an M6 thread onto, which in turn is supported by a couple of recycled bearings. The circle of plywood was the waste from cutting the woofer opening! The highly technical length of string is used as a brake - the spring pulling on the MDF lever affords enough force to stop the spool turning. As the wire is pulled onto the coil former, it pushes the lever to the right against the spring pressure, releasing the brake. If the wire comes off the spool too freely, the spring pulls the lever back, engaging the brake and increasing the tension. This simple feedback system works remarkably well.

Next I had to make some coil formers. This was complicated because my simple coil winding jig didn't have a means of clamping a former to the crank, so the formers needed to have a 10mm hole through them so they could simply push onto the crank. My problem was not having a drill-press, so I was unable to drill such a hole at a proper right-angle to the ends. After a few failed attempts at doing this with some offcuts of curtain pole, I realised that this simply wouldn't work.

The only answer was to drill the hole before turning the timber into a former. I'm sure that this isn't a terribly safe working practice, so don't try this at home! A short length of dowel was used to hold the stock, with some hot-melt glue to stop it rotating. The formers were turned from 2 inch cubes of beech.

I found a useful on-line coil calculator at www.lalena.com, which enabled me to arrive at the best sizes, and determine the number of turns required. In practice, they needed a few more turns so I simply kept going until they measured correctly according the Marconi bridge.

The next stage is to test John Krutke's crossover using these inductors. As you saw above, the rear panel initially had 4 binding posts, enabling the crossover to be built externally on a prototype board. I must admit to being very impressed with John's work - the integration between the drive units is absolutely seamless.

Building the crossover onto the rear panels:

One thing that I really hate about commercial speakers - those cheap and nasty plastic connection terminal plates. You know the thing - those rectangular moulded plastic blocks that are seen on some very expensive models - shame! Why build a rigid box and spoil it with a thin resonant plastic tray? You can buy these from suppliers like CPC (e.g. part number LS-SRP8 or CN0246066) for buttons, but don't even think about it!

Many people elect to mount binding posts directly on the rear panel, which is better. But I've always preferred this approach:

The circular hole was cut using the same circle-cutting jig as the drivers. I used the same hole as the tweeter, but in conjunction with a larger bit to give a diameter of just over 50mm. The 1/4 inch round-over bit was used to highlight the birch ply. Then, a scrap of 6mm birch ply is glued to the inside of the panel:

I was tempted to install a layer of 12mm birch ply between these two panels which would have recessed the terminals further, but I was worried that this might decrease the volume significantly (roughly 0.12 litres, which is around 6%). We're already a bit close to the edge, as the tests above didn't include the volume of the crossover components.

And this picture shows the assembled crossover, built using point-to-point wiring. All the components are attached to the rear panel using hot-melt glue, meaning that I can remove components in the future if necessary. In an ideal world, the three inductors would be arranged at right-angles to each other, but I compromised by simply spacing them as far apart as practical. The wires from the drive units attach directly to the components. This simple construction method was chosen as it would occupy less volume than a PCB or tagboard.

Final details:

Having sanded the enclosure and removed all the dust, it was treated to three coats of Danish oil. I generally prefer this finish to varnish as it doesn't form a skin that can get chipped or discolour. Also, it doesn't raise the grain so you don't need to sand between coats. The finish gives a nice sheen, while allowing the natural beauty of the wood to show through. It also allows the wood to darken naturally with age and exposure to light.

To secure the drive units, I bought a variety of fixings including M4 machine bolts and T-nuts, but in the end I decided that wood screws were more than sufficient. I found some from CPC that had a hex head, which gives the illusion that real machine bolts were used. I first saw this trick many years ago on KEF Q55's.

Next step is to install the acoustic foam. This stuff was recycled from the Musical Fidelity Reference 2's mentioned previously, and is around an inch thick. It is normally used to line the walls of an enclosure, or rolled up and stuffed in, but I cut it into layers as this box was so small. This sequence of pictures tells the story:

Drivers mounted in enclosure	First piece of foam behind tweeter
Second layer, with round cut-out for magnet	Third layer, complete

The final layer was cut out to fit around the three crossover inductors. I remember reading somewhere that acoustic foam shouldn't be compressed too much as it can lose its effectiveness. But with these four layers, there is a gap amounting to around half a layer, so I might experiment with another layer in the future. I might also consider adding a small section around the bottom edge of the woofer, as that corner strikes me as a possible source of coloration.

Something that I might consider installing is some self-adhesive bitumen pads on the side walls as the walls are a bit thin (10-12mm). But the box responds very well to the "knuckle test", so perhaps this isn't strictly necessary. These are all options for the future, but it would be nice to have some sort of measurement setup to test for any changes.

On to the final section - conclusions and sound quality...