The idea behind the mini was to build a small, inexpensive 2 way loudspeaker based on Martin King's transmission line models to be used in my work office.  They are also designed to be very easy to build and finish.

They use a Vifa TC11WG49-08 4.5" midwoofer and an Audax TM025F1 tweeter. The cabinet is made from 3/4" birch ply. A 2' by 4' sheet of standard grade birch ply, available from your local home improvement store, is enough for all the panels. The exterior cabinet dimensions are 5.5"W x 7.5"D x 23.5"H.  The speaker is ported in the back on the bottom using a 2" I.D. x 3.5" L black ABS pipe at the bottom. The drivers are on the centerline. The tweeter is 5.625" from the top and the woofer is 9.625" from the top. The tweeter cutout is 2" and the woofer cutout is 4." The top 2/3 of the interior is stuffed with 0.75lb/ft3 of polyfill.

The panels are all 45 degree miters.  Why all miters? Because then, there are no edges to deal with. It's not really hard, and even small irregularities can be ignored. Here's an in-progress picture.  After mitering and gluing all the panels, carefully take a palm sander and put a small bevel on all the edges. It hides any irregularities and looks reasonably good. Look here. The drivers are surface mount. The Vifa is designed this way, and the flange on the Audax is so thin, well, you can get away with it. You can even cut the holes with a half decent jig saw.

The finish is yellow shellac. Birch blotches easily. It looks good, but clear shellac probably would have been better.

The total cost, including the drivers, crossover components AND the wood, is about $150-well, maybe a couple bucks more. The first version which didn't have the RLC trap on the woofer came in under $150.

OK, so let's do the graphs and numbers.

This graph shows the MLS frequency response in red and response with the tweeter reversed polarity in black. The black curve is offset 10dB for clarity.  Note also this was designed and measured using a 15 degree off axis measurements.

Note that I forgot to set the initial time marker for the reverse null. It doesn't affect the frequency response, only the phase, which isn't displayed. But it does lead to a slight printed error at the bottom of the graph--keep in mind the actual meaningful impulse interval is 14.19-2.8ms=11.39ms, regardless of what clio spit out on the bottom of the graph.

Overall, a very good response plot. It's +/- 1.5dB from 100hz to 10k, then the gradual roll off at the expected rate for a 15 degree off-axis plot. It has a measured crossover point of 2800hz and almost full bafflestep compensation.

Now it's difficult to compare this outdoor, long interval graph with the typical in-room, measurements that most people use. You might be tempted to say there is alot of irregularity in the curve. This isn't really the case. The next graph shows the data as it would look with  typical in-room measurement parameters.

Suddenly, I'm much smoother in the midrange and the bafflestep compensation is exaggerated. Really, though, with these short intervals the resolution gradually worsens and it's hard to make much of any data, at least in this graph, below around 1k.

The next series of graphs shows the low frequency response. First,  sinusoidal port and nearfield curves. Note that the relative SPL levels have NOT been corrected. The two levels can't be compared without a correction factor. However, you can still appreciate the port tuning frequency of just over 50 Hz and the relative smoothness of the woofer nearfield until you approach the port self resonance at around 1.5k. Below the sinusoidal curve is an MLS nearfield curve. And finally, Martin's TL worksheet simulations are below this.

You can also appreciate something else from the nearfield plots. You can see the effects of baffle step compensation built into the crossover. That is, the drop from 122 dB to 116dB as you go from 100 to 1k is effect of the baffle step compensation in the low pass filter circuit, more or less. You can see the dip in the woofer response just over 50 Hz from the port resonance. It also looks over damped, which is by design.

The LspCAD frequency response prediction is shown below, and right under it the true frequency response. For purposes of modeling only, the data is smoothed with 1/6 octave smoothing. Basically the second graph is the same graph as at the very top of the page, except with 1/6 octave smoothing applied so it can be compared directly with the LspCAD model.

The actual crossover circuit is shown below.

It uses 4th order acoustic LR targets. The RC on the tweeter can't be omitted as it provides tapering of the inherent rising response of the tweeter, as well as impedance compensation. Besides, one resistor and a 1uF cap is cheap.  The RLC notch on the woofer is a bit more expensive. But it's necessary to get the right response. I used a madisound surplus 2mH coil and sale GE caps, but you could just as well use some of the new Jantzen coils and Dayton caps from PE for about the same price. No point in using an expensive coil here. Get the cheapest coil you can and subtract the resistance from R1061. Note that R1021 and R2031 are the resistances associated with the coils.

The impedance curve is shown below. Very benign. The minimum impedance is over 5 ohms. Not that you'd drive it with a tube amp. The overall sensitivity is about 82 dB.

So, how does it sound? Well, every mother loves her child-keep that in mind. For it's purpose, it's quite good. It has wide soundstage and is quite smooth in the midrange, in part due to the ML TL design. In fact, it's smooth throughout it's range. In a small room, 10x10 or less, there is enough room gain to give the illusion of low end response at modest levels. This is, of course, its Achilles' heel. At any level past modest, it is fairly easy to push the woofer past xmax. Look at Martin Kings simulation graphs above again.  Anything below 40Hz is going to cause 2-3mm of excursion on the woofer. This equates to 82 dB, when you consider the baffle step effects. So, any low frequency transient in the mid 80 dB level is about all this woofer can handle. Which also means that 65-75dB is really the most you can get out of this and still have some overhead for transients. There is no real way to get around this with a midwoofer of this size. Think about it, even if you had 5-6mm of xmax, it would only get you 3dB. If you want more SPL, you have to use a driver with more Sd.

Are there room for improvements? Sure. If the drivers were flush mounted and the edges were rounded, felt was added, more expensive drivers were chosen...well, that would be a different project altogether.  What might be worth pursing is a smaller port length to see if the port self resonance is worth trying to push up higher in frequency.