Obligatory finished product first:


Sometimes things come together, and sometimes they come together perfectly. This was not either of those scenarios; it’s actually just a name pun. I think the journey began confidently over beers, but the tolerances involved in interpreting what someone means by “portable and loud but doesn’t have to be too loud but also make it look really cool” can allow for a lot of design doubt (by no fault of their own–it’s just hard to gauge what reference points people have for “small” and “loud”), and so by the time I packed the Tesselator out, I had built 6 separate designs, each one but the last dusted in a fine sheen of “not-quite-good-enough.” This is their story (dim the lights).


Try 1 was actually pretty awesome. Basically, I wanted to see what the hype was about with the HiVi B4N’s. Ports in small boxes often of chuff me the wrong way and the client wanted “big circles on the front,” which I interpreted to mean speakers. Plus, I go for passive radiators when I can…and so I went for a passive radiator design. I had been having luck with asymmetry, and I wanted to carry a “T” motif through the design.


The problem with the B4N’s that all the fanboys won’t admit is there’s an insanely high Q 15 dB break up mode right at 3kHz, and it likes to jump around depending on boundary conditions, air temperature, zodiac sign, etc. So I wanted to be at least 15 dB down by 3kHz which meant a tweeter that could hit 1.5 kHz, and for directivity reasons, I decided on a 500 Hz crossover, which obviously meant I was going to use the Aurasound NS1s.

Then I found a sweet spot of plywood that I could waterfall from top to front face to edge, cut with confidence, laid out some paper circles for test fit, and very poorly lock-mitered the shit out of the wood.

Lock miters as promised:IMG_2498.JPGThe separated volume is for the electronics–lesson learned from previous projects is when trying to attain a good seal, either get better at electrical engineering, or compartmentalize your bad work.

Of course, I still overestimated my abilities and placed the batteries in the acoustic chamber for space reasons. The white boxes are the enclosures for the NS1s.IMG_2522.JPG

I also had the idiotic notion that using banana plugs as pass-throughs would be simplest, but not only did I get the polarities wrong, it turns out banana plugs are super expensive and take up tons of space:IMG_2503.JPG

I didn’t manage to fuck up the miters too much and the face is perhaps lovable by more than it’s mother:IMG_2947.JPG

I cut out some purple heart and embedded some glow in the dark for the volume knob:


With all that shit sorted, it was time to make an absolute mess of the ASP. The signal chain starts with a power-source isolated bluetooth chip, which is split by an op-amp active crossover, the low frequencies going to a china-market bought TDA7492 class D amp and then to the B4N’s while the high frequencies are padded down by potentiometer and sent to a similarly procured TPA3118D2 amp. The TDA7492 is rated for 40W into 8 ohms @ 25V @10% THD, which works reasonably with the B4N’s 25W RMS rating. Typically it’s better to spec an amp with more headroom (@ less THD) over the continuous power rating of a woofer in order to match the crest factor of music, but I didn’t think of that at the time.


This is the last build I used analog signal processing on, partially because of the above mess of wires. Here’s the terrible wire management in context:


I opted for a glow in the dark,  3D printed, inset handle to preserve the form factor, and then slapped some spar varnish all over that bad boy and called it a day.



There are far more than five senses available in the bleak sensorium of human existence, and one of them is the sense that you could’ve done better. Honestly, I was pretty happy with Round 1, but it was just not quite there. It was a little too big, and the lock miter bit I used for the edging was one of those cheap amazon finds that reflect their pricing in their quality. So, I started completely anew…by taking an old project that had been called into half-hearted existence with 3 other siblings in a similarly iterative process that finally yielded the Krump Kanon and cutting it in half. In general, this approach is poor.

It sucked for multiple reasons, some of which were that it was ugly and sounded bad and was still too big. Essentially, it failed to meet any of the criteria laid forth.



I then tried a new design that was basically Round 1 but with half the stuff in half the space. It also sucked. I was convinced that it wouldn’t because of my experiential lesson on KK Round 2–“efficiency is king”–but it turns out that only works if you have a pleasing natural response or some good DSP.

It was doubly a shame because the wood that went into the box was beautiful, but for some misguided reason, I used the cheap lock miter bit from Round 1 and, completely to my surprise, it didn’t work well the second time either.



I then decided that everything I had decided was wrong, that efficiency wasn’t king, and it was all about extension. I went back to some of my “super-compact design” notes and decided to drag some micro-subwoofer Tang Bands into wretched existence. The only problem is that tuning a small box to subwoofer frequencies requires a long-ass tube (because the air spring in a small box is relatively stiff, you need a lot of acoustic mass in the resonating port to get a low resonance frequency), and long-ass ports are very inconvenient to fit into small boxes (not a problem encountered in my daily life). I had a minor stroke of brilliance stroke and decided to make a port that was both a long-ass tube AND a handle, therefore circumnavigating this issue.  Here is the relatively tiny box, which looks shitty because I had come up with the terrible play that I’d wrap the whole thing in carbon fiber once assembled:


And the incredibly sleek and not at all awkwardly protruding port/handle design. IMG_2794

I set the thing up, hit play and was, for the first time in a long time, pleasantly surprised. Here’s a casual video of it in a living room (turn ya sound up and throw on some head phones to appreciate the FIDELITY that’s SPEWING out of this BOOMBOX).

For such a tiny little thing, it was really moving air. It had real potential until I yield-stress tested the carbon fiber with a hammer.


Not really much to go on about here. It was ugly. I underestimated how weird it would look to have the speakers sticking out of the face instead of flush mounted, and the thing looks like a damn bug-eyed pug.


In a surprisingly reflective and narratively satisfying moment, I decided to combine the lessons of the last 5 iterations. I drew up a plan for a small, relatively efficient boombox with precise waterfall miters, inset speakers, DSP, and a port handle. And no fucking carbon fiber.



On to the even more boring stuff. Yes, yes, I know the stereo image is going to be ruined by placing the “tweeters”  on top of each other. But it looks cool, and there’s no point in attempting to get stereo width out of a box narrower than one’s head.

Anyway, it’s got 2x TB W3-1876 in a mono “sub” configuration, sitting in a 3.7L box stuffed with light polyfill, tuned to 48 Hz with a 12″ long by 1.2″ diameter port. This theoretically gives an f3 of 42 Hz. The port is a 3D printed 3-section design that was epoxied together for surface finish and adhesion. It’s flared on both sides equally for symmetry. The “tweeters” are 1″ W1-1070SH, which are sitting in a 0.1L box and crossed over in a 48 dB/oct LW DSP crossover at 500 Hz. The outer dimensions are approx 4.5″Hx4.5″Dx14″ and the 80Wh battery supplies 24V (nominal) to a China Black Market TDA7492 (to run the woofers) and a CBM TPA3118D2 (for the tweeters) for about 8h of quite listening and 4 hours of TURNT listening  MiniDSP 2×4 runs the tuning, and the bluetooth is run by an APT-X Bluetooth 4.0 chip. The advantages of this chip are high quality transmission with surprisingly low radio noise, but by some trick of China-blackmarket circuitry, it manages to clip it’s output stage at maximum source volumes. I suspect they added a NE5532 output buffer but didn’t manage the gain properly. The numbers on the edge display battery voltage, which is my lazy solution for a battery gauge.

The wood itself is is 1/4″ maple ply, reinforced on the interior with another 1/8″ of ultra-stiff epoxy and some bracing. I finished the wood Water-Lox high gloss finish, which I enjoyed for the simplicity of use and quality of finish. It brings out the grain and luster of the wood beautifully, and it dries quickly into a reasonably durable exterior finish.


Subjectively, the thing is awesome. It sounds far bigger than it looks, and with DSP trickery, there are little concerns of over-excursion despite a relatively low tuning for such small woofers and such a small box. The stereo image is shit for previously mentioned issues, but it manages to have pretty laid-back directivity, which is all you could hope for from a small source.

Final assessment: can fill a living room with danceably loud music, yet it is small enough to hand carry to a barbecue. Ship it.



IMG_2730.JPGMuch like worried parents will fuss over a child before sending them off into the world, I fidgeted over the details of this lil guy, attempting to delay the inevitable departure, filled with pride and worry at the rigors he’ll face out in the real world. Unlike most worried parents, I eventually said “fuck it,” and dropped this fucker off at the local Fedex, to be shipped cross country in a large cardboard box.


The details were particularly sweat-able on this build, as this was essentially the third iteration on the concept, starting with a beast of the beats that went to Keith, and then a semi-pro configuration that went nowhere. Here are the vague details of the build:



Obviously, the first detail to isolate is the Lego theme. Legos, by the way, are a fairly mediocre permanent construction material. Turns out the 10-micron precision makes them fairly expensive from a cost/volume stand point (a small enclosure requires a lot of legos). Had my little siblings help me build lego boxes to compare the looks. Turns out rainbow is a crowd favorite


Fast forward a few days, after much designing, some deliberation, and then cutting, I’ve got a wood box to match the lego box. I believe I designed for 4L internal volume for each NS3 driver, which, in retrospect, I feel was too much. However, once the wood is cut…alea acta est. I went for a seamless miter approach on this build, to avoid the ugly “end-grain” of the birch plywood. I wrapped the grain around the “depth” of the box, but the grain of the “face” does not flow into the edges. So far I haven’t figured out a solution to this that works out in our boring 3 dimensional Euclidean / Newtonian universe.


Here we can see the translucence of the 3d prints, pre assembly. Originally, I had not planned for there to be a VU lightstrip in this build, but then I realized that since I built all the electronics off of the wood box, fitting them into the lego box, which had 5/6 faces constrained already, would be extremely difficult. At this point, I also realized I miscalculated the amount of space wires take up. Medusa rears her ugly cable management head yet again.


Now you see what I mean.


Analog signal processor demonstrating that I am the particular type of person that loves neat things but does not love making things neat, and so I live in a constant torment of my own devising. I chose a scheme that allows for a bass shelf at lower volumes, but flattens the EQ towards higher volumes—it’s essentially a loudness compensation circuit, except that since I have no reference for the actual loudness of the output (due to lack of information about source gain, listener position). I call it the “party” compensation circuit, because while one might enjoy deep sonorous extension at lower listening levels, once your friends roll through, 14 beers deep each, you’re cranking that fucking volume knob. And while the NS3’s have a lot of allowable excursion before crashing, the garbage bass lines that litter the hip hop soundscape are essentially glorified sine waves that will fuck your shit up. The obvious and simple solution to this is a simple dynamic gain-tied or signal-adaptive high-pass for excursion limiting, or a multi-band compressor. I leave these endeavors as an exercise to the reader until I have the time to implement them on the next build.IMG_2409.JPG

Pre-wood finishing. After disliking the gumminess and amber tint of marine spar varnish, I chose to use tung oil to bring out the figure. Then I sealed the box+3d printed parts with epoxy (bar top) for strength/durability, and finally, for UV protection and hardness, I finished with a clear gloss polycrylic. Here’s what it looks like finished:



So after all that talk of efficiency v extension, I figured I’d actually build something. It worked out that I ended up building on both sides of the coin, as the “Extension” build ended up being pretty gigantic. But the review of both approaches is in and it turns out that in some regards efficiency is more pleasant; I used the Faital Pro 5FE100 in a ~8L enclosure tuned to about 45-50 hz with a 3 inch x 15 inch port which I 3D-printed.  It WOMPS. Sounds great, crisp, powerful. Really pushes the “effortless” bass feeling, until you get below resonance and the woofer starts whacking around. It becomes excursion limited quite fast on songs with a lot of low frequency content, and it sounds pretty alarming. This forces my hand in ASP to put in a high pass around 32Hz-20Hz, adding complexity and heartache. But all in all, not a bad design.

As for the extension, which also sounds good, a different set of issues comes into play. I went with the Tang Band W5-1138SMF ~14L tuned to around 40 hz with a PR, and for starters, the thing is fucking gigantic. Despite the apparent size, 40 hz in 14L is pretty damn good, and it definitely has a presence visually and sonically. It sounds large, powerful, and it’s exceedingly hard to push the woofer itself into distortion. However  pushing the low efficiency of that driver is it’s ridiculous excursion capability, which brings an unexpected issue into play: high excursion means a lot of acceleration. A lot of acceleration mechanically means all kinds of shit is moving around, including things coupled to the boombox through the surface it’s on; put the boombox on the counter and all of a sudden the beer bottles on the counter start wiggling around too. Additionally, in the electrical domain, a lot of acceleration means a lot of voltage, which means more batteries.

These factors push the design to seem a lot less effortless than the 5FE100, potentially because the 1138 design seems to promote higher THD. It’s not necessarily a fair comparison, because the ASP, amp, and battery management on the 5FE100 design happens to be a lot better (I built it second, learned from previous mistakes). Normalizing for those factors, though, the smaller size is pretty great as it turns out that there is usually very little content from 40 Hz and below, and because larger boxes tend to be less efficient in terms of materials.


This slideshow requires JavaScript.


So for  the KrumpKanon I’m going with the W5-1138SMF, and for Keith’s, the W6 version of the same driver. Why? Because it has the excursion, Fs, and general specs to have solid bandwidth and linearity at loud volumes. Unfortunately, if you look at the models:



The yellow line represents the optimal ported design, while the blue is the optimal 8″ Dayton Audio PR design. The vented design is more efficient, has higher bandwidth, and requires less box volume. However, for a 4″x2.5″ port, opening I’d need 36 inches of port to get the 12L box resonating at 36 Hz. That’s absurd, and what’s more absurd is that at 50W, the airspeed would be ~40 m/s. That’s 90 mph, and the air has to move 90 mph through a folded port 3ft long. I have low expectations for the laminarity of the air flow in that scenario, but an educated guess tells me it would sound like an elephant farting through a muffler.

PR’s it is. With dual 8 inch passive radiators, I could be pretty near the vented design, but a) that’s a ridiculous amount of radiator area for a 5 inch  area, b) that’s 2x the cost, machining, surface area dedicated to the radiator, and c) trying to fit two 8-inch radiators, a 5-inch woofer, and 2 full ranges onto the front of the boom box would be ridiculous. Why not throw them on the back? Well, then the front will look stupid, barren. I want the raw visual and auditory power of a fat fucking 8″ blasting you in the face. But fitting 2 8’s on to the front of a reasonably sized boom box (using the esteemed [boombox name to go here once I find my notes] as a template) would be geometrically dubious.

So at the sacrifice of about 1/3 of an octave of bass, I’m going with a single 8. I hate leaving bass on the table, especially when it’s easy, efficient bass, but given the constraints and my desire for a simplistic design…it’ll have to do. Additionally, and this is the most important part, but, if you look at an analysis of the most popular music being played these days, there’s  little content beneath 40 Hz, and therefore, little need for sound below 40. So I guess 40 Hz it is, and we’ll hope it sounds good. There’s only way to find out.








After modeling 20+ woofers, including the attractive ScanSpeak classic series, the illustrious semi-pro Morel and FaitalPRO designs, and the uncategorizable SB Acoustic woofers, I’ve decided that there are only two viable designs. For the purpose of listening to  party music in party situations (Keith’s BuffBox), I want something like 104 dB at 48-42Hz at max, solid directivity, and about 3dB head room to handle over-compressed music and power compression that comes with party scenarios. For Elise’s use (KuadKrumpKannon), I’m probably going to want something slightly smaller and with slightly less performance. We can go with 100dB 50-40Hz with 3 dB of head room.

That leaves me with two options, the classic crux, the bane of boomboxes, the designer’s dilemma,  the iron law. For a given size, you can either have more bass but lose sensitivity, or have less bass bandwidth but more loudness. The third unspoken option is get both and pay $$$.

One the one hand, I could go for a 2.1 type set up with the boombox-friendly micro-subwoofer designs like the Tang Band 1138-SMF with it’s extreme 45 Hz free-air resonance, it’s lovely 9.25mm linear x-max, 5-something liter Vas and 82 dB sensitivity. Yeah, it sacrifices sensitivity for low Fs+high xMax at a decent cost point, but with 2^6.3 watts power handling we can potentially get to 99 dB at 1m down to 32 Hz in a 15L cabinet. In my experience, having deep reach with slightly less output can be more enjoyable despite not being as overwhelmingly loud. Essentially, I’d give up loudness for an extra  bass half-octave.

Alternatively, I can go for sensitivity over bandwidth. The FaitalPRO 5FE100 is a great example of this, as it maintains decent excursion and could potentially hit low 50’s in a small box while still being 6 dB more sensitive than the 1138-SMF. In fact, it could  hit 103 dB  in a 8L box, for a 51 Hz-3dB points, or and 40 Hz -3dB in a 15L box. It actually could handle producing about 108 dB (that’s at 100W, it can handle 160W) but above and below the port resonance we’d be we’ll out of the linear excursion range of the woofer. It starts to get directive at 1.5 kHz (beam forming begins at f=c/λ which is 2.7 kHz for a 5 inch woofer), so I could easily cross it over with some equally sensitive 1-2.5 inch speakers around 300 hz and call it a day.  Lots of out put, lots of head room, good package. Only two factors to consider: 1) whether it will work with available  passive radiators and 2) What Would Heathor Trainor Think? About the missing half octave?

On the other hand, I could go for a more sensitive design dual-fullrange or dual woofer tweeter with slightly less reach, but I get 3dB off the bat because I’ll have two woofers. The NS4 would be perfect for this but alas, they are gone from the world. The second best thing is the AuraSound NS3-193-8A , and it’s copy cat brothers: Dayton ND90, ND91, ND105, and DS115 are all great candidates for this. But after modeling I found:

  • AuraSound NS3: Great response, pretty design, 50hz in a 4L box. Play nice with passive radiators. All around solid drivers at a good price. Definitely on the table. The only issue is they’re not that sensitive and they don’t hit super duper hard.
  • ND90: good for this application, but still, lots of ripple. -3dB at 40-38 in 9-7.7 liters. Good news is xMech for days, so bottoming out shouldn’t be an issue for these ND series. Doesn’t work
  • Dayton ND91: great for 1-4 liter enclosures. Looots of ripple otherwise, still a lot of ripple.
  • ND105: lots of ripple, needs at least 14L but can get down to 33 hZ  at -3dB, or in 9L, 38 Hz with a bit of extra efficiency in the pass band.
  • DS115: Kind of a middle ground. In 6 liters, could go to 50 hZ -3dB. They look pretty, and with two of them, I could get 104 dB while staying in the linear region.

The downsides to this approach are: raised complexity (I’ll have to fit 2 4th order systems in the same form factor), limited design flexibility (long ports are almost out of the question and passive radiators are harder to fit), potentially greater cost, and of course, half the space which usually ends up meaning less bass.

Here is a graphical explanation of all that jazz ^.

TF ELise session

Transfer Function of Speaker. -3dB point marked in red.

But what do the colors mean? See next photo.

20w legend

Legend of which colors match up with with speaker/box designs.

Transfer functions are great but we want to know the absolute level.

20W comparison

Absolute level. Compare the sensitivity vs the bandwidth of the various drivers. The ND105 manages to be a good combination of both.

But how loud do they really get with linear excursion constraints?

maxSPL Elise session

Maximum (linear excursion) loudness. Note the FaitalPROs (magenta, green, teal) manages massive power handling until the excursion limit kicks in. The previously impressive ND105 (orange) has little-to-no advantage over the 1138SMF (yellow) due to its lower power handling and limited excursion.

So…what to do?


Probably a lot. Keith says something that’d be good for a house party, as far as portability I was thinking something around the size/weight of the one you made for connor like 10″x20″x10″-ish and somewhere around 15 pounds (I’ve been lifting so weight isn’t too big of a concern)

i guess no more than 25 pounds? i really have no idea.”