Want to put "smart" lighting in your house? This smart dimmer combines a dual-microphone array and speaker, voice-command processing, and Bluetooth and Wi-Fi connectivity.
Want to put “smart” lighting in your house? Well, you could accomplish this objective by pairing a voice-activated Amazon Echo or Google/Nest Home smart appliance with a Belkin WeMo or equivalent switch. But then you’d be paying for two different widgets, and you wouldn’t be able to locally control any lights out of earshot of the appliance’s microphone, or have dimming control, only brute-force on/off facilities. If dimming is most important to you, you could go with a “smart” light bulb; only then you wouldn’t have voice control, and dimming capabilities would be restricted to a proprietary mobile app: not exactly the most user-friendly approach.
The DWVAA-1BW, rated for dimmable CFL, incandescent, or LED illumination loads up to 300 W (neutral wire required), includes both a multi-microphone array and a speaker, and natively supports Amazon’s Alexa protocol (note: no separate Echo device required). Integrated Bluetooth and Wi-Fi facilities enable communication with Amazon’s servers for voice-input decode and response purposes, as well as support via an aforementioned Leviton-branded mobile app along with Google Assistant and other protocols via IFTTT (the “If This Then That” service).
Before proceeding with the dissection, I’ll begin with a series of pre-unboxing shots:
Open up the outer box and you’ll find two things: a paper “sleeve” containing three pieces of documentation, and an inner box containing (among other things) our patient:
Extras first: there’s a wall plate, along with a set of three twist-on wire connectors:
And now for the star of the show, with a United States penny (0.75 inches/19.05 mm in diameter) for size-comparison purposes:
A few overview thoughts before proceeding; first off, the illumination “up” and “down” buttons are the largest and likely already intuitively obvious (that scratch on the “upper” one, by the way, came from the factory that way; a surprising quality-control faux pas by Leviton). Above them, at top center, is a self-illuminated “action” button that serves several purposes; initially activating Alexa for subsequent hands-free operation, as well as muting (and unmuting) the microphones.
Speaking of microphones, you’ll notice four holes, one in each corner of the face plate. Hold that thought for a moment. The speaker location is again likely already intuitively obvious per its “grill” at the bottom. And underneath it (not intuitively obvious) is another illumination source, which illuminates when ambient lights are off so you can find the dimmer in the dark, as well as providing connectivity and other device status feedback.
The two sides are relatively unmemorable; note, however, the multiple ventilation “fins” visible on both of them:
And here’s the back, which becomes much more interesting after we move the three wires out of the way:
Last but not least, here’s a peek at the top and bottom:
Time to dive in. First off, remove those four screws whose heads were visible in the prior back-view photo, which ended up being surprisingly (at least to me) long:
Release the tabs shown in the two earlier side views, wriggle the two halves apart, and voila:
Initial thoughts: note the plastic transparent light guide jutting out from beneath the speaker in the assemblage on the left. You might guess that it corresponds to an LED on the PCB “sandwich” on the right, and you’d be right (again, hold that thought). Also note the two interconnect harnesses between the PCBs on both sides: the speaker at bottom on the left is driven by circuitry on the right via red (positive) and black (negative, i.e., “ground”) wires, and a multi-contact male/female connector pair (labeled J1 on the PCB at right). That white plastic “tube” in the top left quadrant of the right-side PCB, by the way, serves no purpose that I can ascertain save for as a brace between the two PCBs to improve overall assemblage rigidity.
Next, a somewhat boring (unless you’re into passives, I suppose) closeup of the left side:
And the more interesting (at least to me) right side:
Remember the earlier mentioned light guide? At the bottom of the lowest PCB in that right-side “sandwich” are two LEDs, labeled LD1 and LD2, to which it corresponds. And speaking of which, here it is all by itself (sitting on top of the worn leather case for my 30+ year old, still working HP-15C scientific calculator):
Back to the left side. The speaker and PCB lift right out:
Specifically note the rubber gaskets around the two top “holes” leading to the outside (once again, hold that thought).
Here’s what the back of the speaker looks like in free space, clearly revealing the notch for the light guide to pass through:
And here’s the more interesting transducer front side:
Why interesting? Admittedly, it’s not terribly obvious from the photograph, so allow me to explain. The majority of the assembly around the speaker is hard plastic. But the leftmost portion (in the photograph) is rubberized material; two pieces of it, in fact, with an intermediary notch that the PCB fits through. It acts as a gasket (there’s that “hold that thought” again); also note the hole in the lowermost piece.
And now for the PCB that fits into that gasket:
Most of this we’ve already seen before, although the PCB extension that fits into the speaker is new. I’m guessing, by the way, that the Bluetooth and Wi-Fi antennas are embedded within this particular PCB (and communicate with the rest of the system via the aforementioned inter-PCB connector), both to put them upfront where transmission and reception capabilities would be most amenable and to isolate them from the noisy AC/DC and dimmer circuitry deeper in the switch as much as possible. Flip the PCB over and things get much more interesting:
First off there’s that rectangular silver IC labeled U2 in the center. Google has no idea what it is. The logo is reminiscent of that used by Micron Technology but not identical, so I don’t think that’s the supplier (and anyway, I can’t rationalize why there’d be a memory chip in this particular location). Conversely, the identities of the four front-panel switches (the two at top corresponding to the “action” button, with the associated LED in-between them, and the other two for “bright” and “dim” functions) are obvious.
And now we can finally talk about the previously mentioned four holes in the DWVAA-1BW front panel, along with their associated gaskets. There are two obvious (albeit unknown-supplier) MEMS microphones here, at position U7 in the upper right and U8 at lower left. The hole in the lower right of the DWVAA-1BW conversely seemingly serves no microphone-input function; I’m guessing it might instead act as a bass reflex port for the speaker. But what about the hole in the upper left of the DWVAA-1BW? Its location corresponds to the IC footprint labeled U6, but there’s not actually an IC there.
Here’s my theory: This is a classic two-microphone directional array configuration, intended to algorithmically cancel out common background noise, and a concept I first wrote about in EDN nearly two decades ago (I’m now feeling really old). I’m guessing that the extension to this PCB for U8 originally didn’t even exist; the initial design leveraged MEMS microphones at both U6 and U7 (recall that there’s a gasket present for both, too). Late in the design, I bet, the developers decided that there wasn’t sufficient separation between them for adequate “beam-forming” purposes, and moved the MEMS microphone at U6 to newly-created U8 instead.
Now back to the PCB assembly on the right in the earlier overview shot.
Whenever I see a sticker on top of an IC, I suspect it’s a firmware revision label and that underneath is a flash memory code storage IC. But in this case, I guessed wrong: peeling off the sticker, in fact, revealed a Kingston-branded D2516ECMDXGME (PDF) 4 Gbit x16 DDR3L 2133 (PC3L 17000) SDRAM in a 96-ball FBGA package.
Above it is a large IC with an obvious Freescale Semiconductor (now NXP Semiconductor) vendor mark, but the product markings eluded my Google-assisted identification efforts. However, I stumbled across reference to a Kontron system-on-module containing a similar-marked SoC, suggesting to me that what we have here is a member of NXP’s i.MX 8M application processor family, containing (among other things) both Arm Cortex-A53 and Cortex-M4 processor cores.
Next to both of these ICs is another chip whose identity eludes me (and whose shiny surface made even the markings hard to see in certain lighting conditions and viewing angles). Suggestions, readers?
Continuing up the PCB, in the upper left quadrant is the nonvolatile memory equivalent to the earlier-mentioned Kingston SDRAM, specifically a Sandisk SD1NBDG4-8G (PDF) 8 GByte iNAND embedded MMC flash memory drive:
And in the upper right corner is a Rohm BD71847, which the manufacturer identifies as a “System PMIC [editor note: power management IC] for i.MX 8M Mini Family.” A-ha, I was right with my earlier system processor surmise!
In attempting to explore further, I pulled the metal assembly housing the PCB sandwich (and other analog and power stuff) away from its black plastic sarcophagus:
I quickly realized that there were some seriously thick-and-numerous pogo pins interconnecting various layers, which I’d need to laboriously unsolder or snip for a complete inspection:
Check out that massive power transistor on the other side, by the way, with the overall metal assembly acting as its heat sink:
Regular readers of my teardowns already realize that, whenever possible, I try to leave equipment in sufficiently intact condition so it can later be reassembled and donated. So, in realizing that the bulk of what was left to see would be reminiscent of that in a conventional dimmer switch, I decided to content myself with peeks at interesting already-visible other bits:
However, in perusing the underside of the topmost PCB we’ve already looked at, I did discover one other critical system component:
DSP Group is a company with whom I actually have a long history. Back in the mid-90s, while I was still working in Intel’s flash memory group, we partnered with them on a reference design for what may have been the world’s first fully nonvolatile digital voice recorder (digital answering machines, for example, already existed, but they leveraged battery-backed RAM). Not only did the reference design incorporate a flash file system that comprehended the media’s inherent bit-programmable but block-erasable (not to mention non-infinitely erase cyclable) characteristics, it also enabled the use of low-cost “bad-bit” and “bad-block” flash memory components that normally would have been discarded by Intel during testing.
All of this is commonplace nowadays, of course, but again, we’re talking about development work done a quarter-century ago (once again I’m feeling really old right now)! Obviously, DSP Group has maintained its expertise and focus on various audio applications (while expanding into other markets as well), as exemplified by the DBMD5VT181A4ANI HDClear dual-core DSP in this particular design.
And with that, I turn the microphone over to you, dear readers, for your thoughts in the comments!
—Brian Dipert is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.