I started with the crossover.

This is the schematic for the KEF 104aB crossover.  The "aB" suffix indicates the impoved KEF designed "acoustic Butterworth" filter from 1979.  The original 104 speaker (no aB suffix) used the same cabinet and drivers with a simpler crossover, and was released in 1973.

Not shown on this schematic is a fuse in the minus tweeter leg, and the fact that the 2.0/2.7/3.5mH choke is connected via a switch to the rest of the circuit.

The original crossover consists of 3 inductors, which were kept since their DCR (resistance) is part of the crossover design, and a couple resistors, which are seen in this photo after removal of all the caps.  The original crossover also used 5 Elcap brand electrolytic capacitors, which should definitely be replaced at this point.  After removing the caps,  half of them measured at least twice their specified value (off by over 100%!), so not really performing as intended.  All caps were replaced with matched film caps.

The stock crossover at the top with the out of spec Elcap capacitors, and below with the new caps.  The 0.6uf cap is made up of a 0.56 and 0.047 in parallel.  The 5uf, 3.3uf, and 0.56 caps are polypropylene and the two 10uf caps are polycarbonate bypassed with 0.047 polystyrene.

The crossover also included a small panel (shown below) with a tweeter fuse and switch assembly to control midrange level.  I replaced all the wiring from the PCB to that panel with 18 gauge wire.

The bottom of the modified crossover with the new wiring.

The inside of the fuse/switch panel.  Tweeter fuse at the top, switch below.

The cabinets for the tweeter & crossovers are simple boxes with a divider to mount the crossovers.  The front panel, holding the tweeter, will be screwed in to be removable to gain access to the crossover.

A simple rectangular riser/base (on the top in this photo) will first be glued and screwed to the woofer cabinet, then the tweeter cabinet will be glued to that base to form a permanent assembly.  The ply edges of that base that you can see here were trimmed with oak and painted black as a riser for the tweeter cabinet. 

The basic construction of the tweeter cabinets completed, with the faceplates to mount the tweeters in front.  The rear has a hole for mounting the fuse/switch panel.

Checking the size of the tweeter cabinets atop the woofer.  

Covering edges/joints with Bondo so the veneer doesn't telegraph (show thru) any joints.

Veneering in progress.  Santos rosewood like the woofer cabinets.

Attaching the solid rosewood front edge trim.

Veneering completed, ready for stain.

When I built the woofer cabinets 5 years earlier, I custom mixed a stain to match the color of the rosewood LS3/5a cabinets from Stirling Broadcast (see that build here).  Fortunately I still had the custom mixed stain, this shows that stain along with several coats of satin polyurethane (same as the woofer cabinets).

At this point I removed the crossover that was built into the base of the woofer cabinet (see details of that crossover in the Part One section below, scroll way down), and I ran a single pair of wires from the input binding posts (at the bottom of this photo) up thru the woofer cabinet to feed the new 104aB crossover.

Next, the mounting base/riser for the tweeter cabinet was glued and screwed to the top of the woofer cabinet.  The black & red wires to the right are from the input binding posts in the photo above.  The green & yellow pair will go from the crossover to the B200 mid-woofer driver in the woofer cabinet, sealed with silicone glue.

The drivers were removed from the woofer cabinet to prevent any damage, then the tweeter section was glued to the base/riser as shown in this photo.

Next the crossovers were mounted inside the tweeter cabinets.  The wires on the T27 tweeters were only a few inches long so I added a terminal block at the front to connect them just behind where the faceplate will mount.

The backside of the tweeter cabinet showing the original KEF fuse/switch panel.

Wiring up the tweeter in its faceplate.  Notice the perforated metal cover over the tweeter, a hallmark of the LS3/5a design.  Used for the same purpose here - to prevent damage to the tweeter's mylar dome.

Completed assembly.  I had just enough spare Tygan grille material from the woofer project to make grilles for the tweeter cabinets, as shown on the left here.  The speaker on the right shows the T27 tweeter at the top, the B200 SP1039 mid-woofer below and the the BD139 SP1042 passive radiator at the bottom.  More detail on those drivers in the Part One section below.

The completed pair.

Using a Dayton OmniMic, this shows the frequency response at 1 meter.  Not too shabby for a 45 year old design.

The KEF 104 uses an active 8" woofer (on the left), and an 8" by 10" oval passive radiator (on the left).  Note the unusual shape of both frames, so a special jig was needed to route the recess for each of the drivers.

The active (driven by the amplifier)  8" woofer used in the KEF 104 is a  KEF model B200 SP1039.  This was a special variant to the KEF 8" B200 series with a larger magnet, designed for vented or passive radiator systems.  As you can see from the right hand sticker in the photo, these drivers were made in 1977.

This gives you an idea of the relative size of the B110 driver used in the LS3/5a (on the left) and the B200 driver used in the KEF 104.  That extra size will be important because the woofer will be required to move a lot more air at low frequencies (not that an 8" woofer is all that large).

In addition to the 8" woofer, the KEF 104 used a passive radiator, the KEF model BD139 SP1042 (a passive, no magnet version of the famous B139 oval woofer, with a styrofoam cone).  Passive radiators are sometimes used in lieu of ports/vents to reduce port noises, but the function very much like a vented speaker design.  The date code on this BD139 indicates 1980.

The first step in construction was to route the irregularly shaped recesses for the two drivers.  This requires making templates for each driver.  In fact, it requires making a (supersized) template to make the actual template for the driver.  The several step technique I used was as follows, and can be used for any irregularly shaped driver: 

1. Wrap the driver to keep dust out.  I used a couple layers of plastic wrap.

2. Screw the driver to a 1/4" oversized piece of masonite (2' by 2' in this case).

3. Using a 1/4" bit, route around the perimeter of the driver, making sure to keep the routers base against the driver at all times (thats the hard part).  That will give you a supersized cutout in the shape of the driver.  That template will next be used to make the actual driver template.

Now the driver can be removed and put away, and the cutout center portion can be tossed or reused elsewhere.

4. Attach a second piece of 1/4" masonite under the template created above, and use the first (large cutout) template to route a 1/4" cutout in the lower masonite that will be closer to the driver size.

5. Use that second template with a 1" router guide and a 1/2" router bit to make the actual recess for the driver at the specified depth.

Here you can see the 10mm deep cutout I've made in the speakers baltic birch face for the frame of the BD139.

6. I used a Jasper Circle cutting jig to cut the thru holes, which required a bit of creativity for the oval BD139 cutout, seen underway here.

The two baltic birch baffles with driver recesses.  

My cabinet construction technique follows the recommendations of George Short, formerly of North Creek Music Systems, in his Cabinet Handbook.  That  technique requires that the front and rear baffles be 3/4" baltic birch and 3/4" MDF, laminated together with Aleene's Tacky Glue (which makes a soft glue joint) for both strength and to dampen vibrations.  

Short also recommends using MDF for the cabinet top, bottom and sides, and baltic birch for extensive internal bracing, as baltic birch doesn't compress under pressure as MDF does (making MDF a poor choice for internal bracing).

Laminating the front and rear panels of Baltic birch and MDF.

The front panels, front and rear sides, after final machining showing routing recesses on the Baltic Birch front (on the bottom  in the photo) and round-over for airflow on the MDF rear side of the panels (top in the photo).

All panels and internal braces machined and ready for assembly.  Note that almost all of the panels have dado cuts to position braces and each panel to the others, making for an easy (guaranteed-square) assembly. 

Partially assembled cabinet showing some of the side braces in place.  I also installed a channel in the rear of the speaker to run a cable from the base up to the top for the LS3/5a atop this cabinet.

Fully assembled cabinets (no veneer or trim yet) with the baltic birch internal (window-frame) braces in place.  Note that although the cabinet sides and top/bottom are 3/4" MDF, the front edges are trimmed down to 1/2" wide to better match with the look of the LS3/5a cabinets.  The cabinets are also as thin as possible given the driver dimensions (much thinner than the original KEF 104s), also to look better with the LS3/5a's atop them.

Once the cabinets were assembled, I selected the veneer - I happened to have some Santos Rosewood veneer in stock that would fit the bill nicely.  Here you see one top panel veneered and I'm about to start the side panels.

Once the top and sides were veneered, I cut thin (1/8" thick) strips of solid rosewood for the 1/2" wide face trim.  Solid wood on the front edge makes it less susceptible to damage from the occasional impact than the thin veneer used on the sides.

This shows the cabinets veneered with the solid front trim in place and trimmed, ready for staining.

I did a good bit of experimenting with stain color to try to match the Stirling LS3/5a cabinets.  In the end, I got the color pretty close, but the Stirling cabinets are a bit darker than my woofer cabinets. 

The cabinets have been stained here, but have not yet had a coat of clear polyurethane to protect the finish.  That process took weeks as rosewood is an oily wood, and oil based polyurethane takes a very long time to dry between coats.

Next up: the bases.  Here are the parts, machined and ready for assembly. I use solid oak for the perimeter of the base as its intended to take the brunt of impacts from vacuum cleaners, etc.

The smaller blocks with the off-center holes are to mount Tee-nuts that for the feet.

The Tee-nuts being assembled.  The column of blocks on the left shows how the pieces will be assembled into the base.  The Tee-nut itself is glued between two layers of plywood block so it can never move or come loose.

Bases being assembled.  Lastly, I cut a bevel on the outside so they look a bit more attractive, and the base is wider than the veneered cabinet to protect the veneer.

Finished bases, upside down,  getting a coat of black paint.  Once completed, they'll screw into the bottom of the speaker cabinets so they'll be removable to repair damage and/or repaint if necessary.

This shows the completed bases, with the Soundcare Superspike feet and binding posts - Superior BP-30.  Those footers use an internal spike but with an integral base to protect the floor.

Next: the easy step of mounting the drivers, using closed cell foam for gaskets to seal the cabinet.

Once the drivers were installed, I was able to do some crossover prototyping.  As you can see here, I use an OmniMic test setup and lots of clip leads!

I originally tried variations of the crossover used for the Rogers AB1 bass module, but that couldn't get that to work well for this setup.  I eventually settled on a crossover designed by Jeff Bagby, extensively documented here, for his LS3/5A-based Continuum speaker, and that seemed to work quite well.