MARRIAGE SOUNDS GOOD

What’s the right wedding gift with 30 days of lead time when your friends are stranded across the Canadian border because of a global pandemic but they’re willing to risk it all for love and get married in a DMZ? I went with an Ikea cutting board. Well—to start.

It turns out that in places where border boundaries are blurred the acoustic offerings are slim. Without loud music (and strong drinks) no party is bompin, and without a bompin party, it’s not a wedding, so there was really only one thing to do: make a matching & linkable set of portable, hi-fidelity bluetooth speakers:

DESIGN SPECS

Cost of parts: $150 (ea.)
Loudness: 96dBSPL, 1m, @ 10% THD, A-weighted
Frequency Response: 50Hz to 20kHz ±5dB (but look at the curves down below)
Connectivity: Bluetooth 5.0
Battery: 3S Lithium-Ion, 37Wh
Runtime: 10 hours at “half volume” input (92dBA output)
Amplifier: 2x50W TPA3116D2 running @ 24V
Difficulty of Build: Dummy high—approx 120hrs from start to finish, requiring 2 CNCs and a 3D printer

In a lot of ways, this was a 2020 capstone project for me: to make something that’s loud, compact, and full of deep bass, with a 30-day conception to finish timeline, I had to pull out at least half of the dirty tricks I’ve learned over the last six years. Here’s how it went down.

ACOUSTIC DESIGN

When it comes to compact loudness with a lot of bass, excursion and efficiency are the belles of the ball, and although I simulated almost every 2-4″ driver I could find, the Dayton ND91-4 drivers (descended from long-gone AuraSound’s Neo-Radial IP) are nearly unbeatable when you factor in magnet strength, Fs, Xmax, price and weight. E.g. Peerless SLS-85S25CP04-04’s (catchy name huh) are potentially 1dB louder for a similar box size but weigh 285% more, while the Fountek FR89EX win for Xmax but need too much back volume and are 2dB less efficient…etc and so on. Just trust me on this one. In a 1.5-2.5 litre box: ND91-4, tuned low.

Tweeters are a fair sight more efficient, so down selection should be mostly driven by crossover frequency, dispersion and ease of integration. The ND91s break up right after 3kHz:

While the ND16/ND20 tweeters are truly amazing, they have to be crossed higher, and they come with a bunch of extra plastic, which clashes with the ultra compact layout I pushed.  LaVoce’s TN100.70 did the trick and can be crossed over at 1.5kHz, which was perfect—the lower a tweeter can be crossed (disclaimer: within its volume displacement limit), the better. Finally, the TN100.70 dispersion is on par with the ND20FA tweeter @ 20kHz (-15dB):

As for the port, in order to maintain compactness and b-b-bass, I had to fit 250mm of port into a 2.5L box while keeping a holdable 4-inch width so I folded it around the ND91 and then crushed the port geometry until it fit in between the driver and borders of the speaker. Tweeter in green, port/body in pink, and woofer in yellow below:

Driving the woofer and tweeter is a 3S 3500mAh battery pack (I use LG 18650s that I order B2B from the factory) paired to a 2x50W Class D TPA3116 D2 amplifier through DC-DC step up converter for maximum power delivery. WONDOM makes a wonderful TPA3116 board with the DSP integrated, which merges with their 3S MPPT Battery Management Board, although to my late-stage chagrin neither of the boards have a step up to power the TPA chipset at an adequate 24V.

BUILD PROCESS

With the acoustic design tucked away 15 days from the deadline, it was time to build. The octagonal outer shell is just a set of 22.5° mitres, tape-clamped, with the patent-pending dual-bevel 8th wall precision cut to match:

The front face was a 2 sided CNC operation, which required calibrating features for aligning the Shaper Origin I used.

Merging the two pieces with the speakers and the front mounted the port was rather easy except for some minor mishaps with a few missing microns; the t-nut I planned to use to rear-mount the woofers were exactly 300 microns short of the planned front face thickness, so after sanding I had two t-nut holes showing on the front face. The port itself had to be printed in 3 pieces because of the complexity of the geometry to fit it both on the border and between the woofer and the back panel:

With space at a premium, but also for aesthetics, I used an LED array for status lights and integrated the on switch into the potentiometer. With that in mind, I also fabbed an ultra slim 6mm bracing/sealing ring for the rear panel mounting, as a butt joint would’ve been ugly but the shell was too thin/weak for threaded inserts. Those loose microns got me again and the flange on the port interfered with the built dimensions of the rear panel, so I slotted that out, but after some truly painstaking finagling of circuit boards, 5 days before the wedding ship date, I was ready for sound test.

That’s when I realized neither the BMS nor the Amp was using a boosted rail which was causing very noticeable voltage clipping, so I had to rip everything open and shove not only a DC-DC buck converter but a giant LC ripple filter (1.3mH L and 100µF C) onto the voltage rail. The only DC-DC buck converters I had in house were straight outta Hua Qiang Bei which means the were both cheap and poorly designed. Buck converters are in general awesome, and about as efficient as one could hope (for 12 to 24V boost, I saw ~85% efficiency depending on load), but the switching causes a lot of load-dependent ripple, which adds both noise and intermodulation into the signal chain. But with that bullet bitten, and with 1 day until ship, it was time to tune. And boy does this design sound good. Sparing the details of the tuning, here’s the final frequency response with a -3dBFS sine sweep @ half input “volume.”

The 2nd harmonic distortion looks pretty high @ 50Hz but this is mostly due to the aggressive non-linear processing I added in for extra kick; a more reasonable measure of THD in this scenario are the 3rd order harmonics, which I kept below 8%. The dips in the mid band (400Hz, 800Hz) are regrettable from a data standpoint (probably due to product baffle dimensions) but overall, these speakers deliver supple bass, smooth vocals, crisp treble, and excellent definition from 50Hz all the way to 20kHz. I added a little bit of level-dependent EQ, so at maximum volume these speakers are loud enough to kick off a backyard party, and at reasonable volumes they deliver a little extra extension for a very full, deep, frequency response. In my book, a thermos-sized speaker that can fill a room down to 50Hz ticks the “bigger than it looks” box; even from another room I found my self saying “damn, these sound good.”

RETURN II

There are far more than five senses available to you in this awful wonderful human sensorium and one of them is the sense that You Could’ve Done Better. But this was not one of those times.

You ever see something and think “I bet I could make that, but better, and more cost-effective”? You ever think “I could make a portable, hi-fi, PA speaker with shit-your-pants bass, noise-complaint SPL, and art-gallery looks”? You ever sit at home and wonder “what if I went all out? What if it was way too big and way too loud and way too pretty?” Yeah, me too. This time I did it.

SCOPE

This was probably a time that I should’ve done less, but didn’t. In scoping out a project on commission I usually discuss

  • Portability
  • Loudness
  • Bass/quality
  • AssAesthetics

The discussion should and usually does occur late at night over libations which contributes to some amount of scope creep—in a good way—and in this case we started at:

  • Portable enough
  • Loud enough the neighbors want to come to the party too
  • Yass bass
  • Museum worthy

I was thinking of a very reasonable design—1 cu ft, 36V battery, maybe 2 W6-1138 (but with Neo woofers for weight). But then something terrible happened. I saw a targeted Facebook ad for the Soundboks 2; it was was full of shitty marketing claims and absurd dBSPL/battery life statements and poorly mixed dubstep (like, dubstep is fine, just don’t mix it badly or use it to tout sound quality). Here are some reference claims:

I’m not an acoustic engineer, but—wait, no, no I am. These are bullshit metrics. What kind of half-assed sound company specifies a “dB” value but no reference for the units (SPL? Re?), distance, or weighting (A? K?). I could fart at 122 “dB” for a battery life of 40 hours if I’d put the mic by my arse.

For the un-initiated this is the equivalent of saying “Oh yeah my car is really fast, it’ll do 120.” 120…what? MPH? KPH? Like when you drive it? Or when you throw it off a cliff?

Anyway fair to say this bothered me slightly and the new goal was to make a speaker that was better than the Soundboks 2. A portable party in a box. My specific objective goals were:

  • 122dBSPL (Re) @ 1m in the passband
  • Passband 40Hz to 20 kHz
  • f3 @ 38 Hz
  • Directivity controlled ± 4 dB up to 15 kHz

*For the sake of clarity, if not otherwise specified, all dB numbers in this document will be dBSPL @ 1m relative to 20 µPa.

ACOUSTIC DESIGN

Speakers assembled to front face

BASS

On the spectrum of “large/efficient” and “small/inefficient” for a constant bandwidth target, there are three main real-life ways to achieve this in the range of “reasonably portable.”

“Pro” speakers in a large box, i.e. lots of magnet, low moving mass, stiff surround. Think FaitalPro12XL

  • + efficient as hell
  • – generally 8 ohms
  • – $$
  • – Fs is often quite high

“Tang band” style in small box, i.e. lots of magnet, lots of coil, tons of moving mass, allowing for really low free-air resonance and massive linear excursion space.

  • + compact 
  • + always impressive for size
  • – $$
  • – low efficiency

“Dayton Audio” style in a medium box: Medium BL, medium mms, heavy magnet

  • + cheaper
  • + pro-sumer design means well controlled directivity, well designed in-band response
  • + reasonably efficient
  • – heavy
  • – has potential to be “worst of both worlds”

Here’s a quick comparison of the three designs plotted at a very reasonable 100Wrms:

Simulated FR @ 1Wrms

The Iron Law clearly demonstrated here: The Dayton design is in the middle for sensitivity but sacrifices on size to get extra bass. The Tang Band, which will never have the sensitivity of the DA or the FP design, loses a little bass to be small, but has overall good LF extension. The Pro design is huge and efficient but loses on LF extension. But wait! This is battery powered! We’re voltage limited! How do 2 4 ohm drivers shake out against 1 8 ohm driver?

@ Battery Nominal Voltage

The Dayton Audio design clearly wins out (dotted lines are theoretical response vs Pmax/Xmax limited response). The final question: can we kill the Soundboks? If we disregard all concerns for safety, in theory—nearly:

At 1100 Watts of input power (rms), 2 DCS205s are capable of outputting 121.2 dBSPL @ 1m @ 100Hz. The Xmax limitation cuts heavily into the bass output below that. But this design will sound better, look better, and be smaller, so DCS205 it is!

The final choice for bass—sealed, ported or passive? A simple one; sealed sucks for battery-powered. The port would’ve had to have been huge to handle the requisite volume velocity with grace, so after modeling every single DSA, I opted for 2x DS315 12″ PRs + 100g to tune to 38 Hz. These are quite wonderful passives—huge xmax, Rms for days, low enough Fs, and their ID matches the DS205s.

Full send:

MIDRANGE

Knowing that we’re in the neighborhood of 115 dBSPL@1m @ 50V input makes things a bit tricky from 300Hz to 20kHz. Pro sound options are mainly focused on output efficiency, with the sacrifice being directivity and flatness of response (DA PK165-8 below, which had neither the efficiency nor the response I was looking for):

What a terrifying directivity curve

which would essentially demand that you cross it over at 2kHz—untenable for a tweeter capable of 115 dBSPL.

Luckily, there weren’t that many options, and when one can’t sacrifice loudness, quality, or size (i.e. directivity), you must pay a lot of money. I landed on the beautiful and beautifully expensive FaitalPro M5N8 which measures like:

Their claimed efficiency of 99dB @ 1W/1m is a little short of the truth (95dB@1W/1m) but they manage an easy 117 dBSPL with xmax and plim constraints:

Throw it in (a 3D printed back volume to separate it comes in later):

HF

There was really only one option: the Peerless by Tymphany BCS25SC08, a silk dome tweeter + a (small) horn for efficiency. 98 dB, 100W of power handling (they get ragged before that, though), and the horn rather small so the directivity actually rather pleasant (this is in 30 deg intervals compared, the midrange plot was at a 45 deg intervals).

Throw that in the bag too:

And then for a back volume, the passives:

Next up is the electrical design. Stay tuned!

HELENS

Did you know
you can squeeze
oil from stone?

With enough heat / pressure
+ a dash of wanton capitalism
And you can coerce shale to cry

Teardrops are always surprisingly hot
Though less valued than shale’s
The process feels similarly igneous

I scream at the wheel
through tears taken from me:
I imagine you in the passenger seat

next to me
holding me steady
a drink

“Well, sure, it hurts—”
you start gently,
full of kindness,

and then you look at me
with eyes like we could take on
an errant volcano, with aplomb

And you say

I’m gone now
You can’t change that.
But you can do anything

So what are you going to do now?