A multi-day weather-induced, utility-instigated electricity cutoff thankfully left this engineer’s residence and its contents largely unscathed…with one geriatric smart speaker exception.

Last December’s high-wind-induced extended power outage thankfully didn’t cause notable damage to our home or its contents. But to say we escaped completely unscathed would still be a (slight) overstatement. When we returned home after Xcel Energy restored power, I noticed that several of our Amazon Echo devices—two second-generation Echo Dots:

and a first-generation Echo:

exhibited perpetually rotating topside-light patterns characteristic of an imperfect bootup:

I’ve occasionally encountered this misbehavior before with various Echo-family devices, and as in the past, power-cycling the two Echo Dots got them going again. But despite multiple attempts, both power- and reset switch-based, I couldn’t convince the Echo to resurrect itself:

Oddly enough, albeit presumably indicative of an underlying distributed-processing system architecture, the top-panel microphone mute button still seemed to operate as expected, at least from an LED-illumination response standpoint:

But the Echo never came online or, more broadly, was responsive to voice-command attempts. Not to mention the perpetually rotating topside light whose video you’ve already seen. That said, I’d been using it for a long time, and they’re apparently prone to such misbehavior sooner or later. And that said, it’s two same-generation siblings in the residence were still working fine, and the first-generation Echo doesn’t support Amazon’s latest Alexa+ enhancements anyway (the second-generation Echo I replaced it with, conversely, is Alexa+-cognizant).

At this point, I’ll reiterate something you’ve read from me in variously worded ways plenty of times before: when a device dies, it frequently then turns into a teardown candidate. I’d disassembled the first-gen Echo before, for publication more than a decade ago, to be exact:

But as with my Tile Mate teardown a few months back, trying to figure out why a device has died is often reason enough for me to entertain a dissection revisit. Plus, in re-reading my earlier coverage, I was reminded that my teardown presentations have become more verbose (whether that word choice translates into “comprehensive” or “long-winded” is up to you to decide) in the last decade. And although I’m still primarily snapping photos using a smartphone, the integrated camera has gotten a significant upgrade in the intervening years. So…here goes!

Foot first

You’ve already seen the reset switch accessible via a hole in the device’s rubberized “foot”:

Here’s another bottom-side closeup, this time of the various product markings, including the always-informative FCC ID (ZWJ-0823):

As was also the case last time, the “foot” (both here and in subsequent photos accompanied by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes) peeled right off:

exposing to view the circuit testing (and firmware programming, I’m also guessing) conductive pad array, along with four screw heads:

I suspect you already know what comes next:

Voilà. Our first, but definitely not our last, PCB is now exposed for visual perusal:

Its functions include internal power generation and management, along with analog-to-digital audio conversion and subsequent amplification for the main (combined midrange and tweeter) and subwoofer speakers. Not to mention the aforementioned hardware reset switch. Let’s next disconnect the associated cabling:

thereby enabling completion of the bottom assembly’s separation from the remainder of the device. To wit, keep in mind in viewing and analyzing both this and subsequent images that the Amazon Echo is currently oriented upside-down, i.e., it’s resting on its top edge:

I’ve got the power

A few more initial images of the PCB from various perspectives and proximities follow:

A bottom assembly flip-over first-time reveals the external power input connector:

And before continuing, I’ll supply a few additional power-related images and comments (power…supply…get it? Ahem). First off, here’s the “wall wart”:

also including closeups of its specs and “barrel” connector:

I’ve had a few second-generation “Dot” devices’ external power supplies fail in the past; the end result is either a flat-out refusal to start at all or a perpetual repetition of partial boots followed by abrupt restarts. In those cases, the consistent “fix” was straightforward and non-wasteful. Since the AC/DC converter with USB-A output was distinct from the USB-A to microUSB cable that fed the device, I could just swap in a replacement for the former and be up and running again in no time. Every time I did this, by the way, I wondered how many Echo Dots prematurely ended up in the landfill due to typical-consumer ignorance of both the exhibited issue’s root cause and simple resolution solution.

The Amazon Echo, as with subsequent-generation Echo Dots, is different: the AC/DC converter and power cable are one integrated unit. And I’d also yet to encounter an Echo-family power supply failure that presented itself as a partial boot followed by a “hang”. Nevertheless, my obvious first step was still to tether a multimeter to the power connector and confirm that I was measuring the expected voltage. And then, since this preliminary outcome wasn’t in and of itself definitive (the power supply could still be peak current-compromised, to the earlier partial-boot-and-restart scenario, and something that the light power supply loading by the multimeter wouldn’t expose), I borrowed a power supply from one of the other first-generation Echoes in the house and sadly confirmed that I still saw the same “hang” behavior as before.

IC details

Back to the bottom of the device. Next, let’s remove the PCB from the assembly, an easily accomplished task:

Here, for comparisons sake, is the comparable PCB (and vantage point view) from my initial February 2016 teardown of the Amazon Echo:

Quoting from that earlier writeup:

At top is the DC power input jack. At bottom is the ribbon cable connector. On the right are the two speaker connectors, one white and the other black. Near the center is a Texas Instruments TLV320DAC3203 stereo audio codec (PDF). On the far left is Texas Instruments’ TPS53312 step-down converter. And the PCB is otherwise covered with assorted large inductors, “can” capacitors, and other passives.

The only thing I might add on revisit of my earlier prose is mention of the multi-LED cluster at center of the PCB, which works in conjunction with the clear plastic lightpipe you probably noticed in the earlier device interior shot, taken post-bottom assembly removal:

to route the device’s power and connectivity status to a backside indicator located above the indent for the power cord. In retrospect, I wish I would have noted not only the perpetually rotating LED pattern up top but also the information communicated (or not) by this secondary LED set, as it might have assisted in diagnosing whatever had gone awry with the Echo.

Next are some side views:

And now for the PCB’s other side:

once again accompanied by that of its decade-plus ago predecessor:

along with a requote of my prior prose:

Flip the PCB over and the comparatively sparse underside contains several notable elements, beyond even more passives. At top is, again, the dual screw-reinforced backside of the DC power input jack. At bottom is the system reset button. In between them are the previously mentioned test points. And in the middle of the left half of the PCB is the audio amplifier, again from Texas Instruments (the TPA3110D2 Class D Stereo device, to be precise).

You more recently saw mention of the TPA3110’s higher power, PFFB-supportive successor, Texas Instruments’ TPA3255, within a class D-based audio amplifier teardown I did last fall.

Symmetrical sound redirection

Onward. With the lightpipe out of the way, the conical black plastic piece underneath it (as currently oriented; above it in normal operation) slides right out:

Below it is the main full-range speaker, handling all but the lowest audio frequencies, which instead route to the subwoofer we’ll see shortly:

The plastic piece’s contours, with the cone end pointed toward the speaker (which normally points downward), uniformly redirect generated sound out the mesh sides of the device:

Locating the brains

At this point, speaking of “mesh”, let’s press “pause” on the speaker and redirect our attention to sliding that metal mostly-mesh outer chassis off instead:

Take the inner assembly:

Rotate it horizontally by 180°, and another PCB appears:

The connector at bottom in this upside-down (versus norm) orientation mates to a flex cable that also routes to the top assembly:

And the one at the other end mates to the earlier-seen flex cable that also ends up at the bottom assembly:

Let’s cut away the foam surrounding much of the insides, so we can see what’s underneath it:

That’s more like it:

There’s that RFID tag again, which I’d first showcased a decade-plus ago:

Here’s our first glimpse of the subwoofer which, like the full-range speaker you saw earlier, points downward in the device’s normal operating orientation. The full-range speaker’s rear housing, which you’ll see shortly, is rounded, and akin to the cone-shaped piece ahead of the full-range speaker, similarly redirects the subwoofer’s primary output out the sides:

And here once again is what I’m calling the “digital PCB”, now free of any foam obscurant:

See those four screw heads? Buh-bye:

In removing one of them, which promptly and firmly re-attached itself to the side of the internal assembly, I was reminded that there’s a sizeable subwoofer-inclusive magnet inside:

As before, I’ll start with the PCB’s outside:

Next is its decade-plus ago, still attached, 90°-rotated-in-comparison counterpart’s image:

And a reprint of the prior associated prose (perhaps obviously referencing locations in the earlier-oriented photo):

In the middle, toward the top is Texas Instruments’ DM3725CUS100 “digital media processor” SoC. It’s fairly diminutive in processing chops, compared to the application processors in modern smartphones and tablets, containing only a single-core 1 GHz ARM Cortex-A8 CPU. My best guess, therefore, is that it primarily handles the Echo’s speech recognition features, with “heavier lifting” redirected to Amazon’s servers via the device’s Wi-Fi connection. Speaking of which, the shiny-packaged IC below the DM3725CUS100 is a Qualcomm Atheros QCA6234X-AM2D Wi-Fi and Bluetooth module, also found in Amazon’s Fire TV and Fire HD tablet. The corresponding antennas are etched into the PCB, on either side.

Volatile and nonvolatile memory are a necessity, of course, and in this case they respectively take the form of a Samsung K4X2G323PD-8GD8 1 Gbit 200 MHz x32 mobile DDR SRAM (in the upper left corner) and Sandisk SDIN7DP2-4G 4 GByte iNAND embedded flash memory drive (below it). A standalone power management IC is also pretty much a guarantee in a product like this, and the Echo doesn’t disappoint; on the right edge of the PCB is a Texas Instruments TPS65910A1.

Next is another set of PCB side shots; note that as with its bottom-located predecessor, this particular board is impressively “meaty” from a thickness standpoint:

Finally, what’s underneath? A decade back, I wrote, “I didn’t bother showing you the underside of this PCB, by the way, because there’s nothing really to see … unless you’re into a bunch of additional passives, that is.” As you’ll see, “a bunch of” was arguably even overstating reality:

To the summit

One more PCB to go; the seemingly still-functional one up top, starting with the removal of a side screw:

The top assembly’s now gone:

More accurately, I’d just momentarily set it aside:

Four more screws to remove (how many times have I already said that in this piece?):

And now a pictorial sequence of the steps necessary to fully-as-possible expose the PCB to view:

Last time, I wrote: “I didn’t bother with a shot of the underside of the PCB; the only contents of note are switches corresponding to the top-side microphone-mute and device-setup buttons.” This time, I’m instead going the extra mile for you, dear readers:

See, like I said before; just switches:

The other side of the PCB, of which you’ve already caught several glimpses, is more interesting:

Here’s the image of that same PCB (and side of it) from last time, once again albeit notably rotated 90 degrees as compared to the new version:

Regarding the gear structure in one corner, I previously said that “Echo contains a rotating upper “cuff” which, among other things, acts as a manually operated alternative to voice command-driven volume up/down operations.” And the gear? It “provides cuff position and speed-of-rotation information.”

And what about the various visible ICs? Again, I requote (again, with location references to the original version of the photo, not the newer one):

Toward the top is a humble Texas Instruments SN74LVC74A dual positive-edge-triggered D-Type flip-flop (ironically the largest IC on this particular PCB). Toward the center are four Texas Instruments TLV320ADC3101 stereo ADCs. They surround one of the seven microphones, at center in gold. And they are surrounded by four Texas Instruments LP55231 9-output LED drivers. The other six microphones are symmetrically located along the rim of the PCB; one of them isn’t visible in the photo, obscured by the ribbon cable. And on either side of each of those edge-located microphones is an LED, twelve total in the design.

Transducers redux

Now let’s return to those two speakers—main and subwoofer—that you initially saw earlier. A decade-plus ago and regarding the foam seen surrounding the internal assembly after removing the mostly-mesh metal outer chassis, I wrote:

Underneath the thin black fabric layer surrounding the chassis is the woofer, along with its corresponding bass reflex port. To see them in detail, check out iFixit’s website; the electronics aspects of the design were my primary focus. The fabric’s purpose may be at least in part to diffuse the speakers’ outputs, thereby delivering the 360º sound that Amazon promotes. It may also dampen vibration at high volume.

This time, curiosity got the better of me, and I decided to peruse them for myself, sharing the images with you in the process. The first step, however, was to finish removing the main speaker. Hey, look. Four more screws!

The rounded rear of the main speaker’s acoustic suspension enclosure, as mentioned earlier, uniformly redirects the subwoofer’s primary sound output around the mesh sides of the device.

Speaking of the subwoofer…

And, last but not least, the curiously shaped (as you’ll soon see) bass reflex structure, intended to boost the subwoofer’s low frequency efficiency:

You’ve actually seen its associated port already, in one of the post-foam-removal internal assembly side shots. Here’s a closeup:

Remove two screws, also from the inside sides of the device:

And the bass reflex structure then slips right out the Echo’s now-speakers-less bottom end (in normal operation; top as currently upside-down oriented):

Wutdunit?

(why yes, I did just create my own word)…

Unfortunately, unless you saw an old-vs-new teardown disparity I’d overlooked in any of the PCB photos I’ve shared here (a bulging capacitor, perhaps?), we’re left with the same question I posed at the start of this writeup: what caused this Amazon Echo to fail? The power subsystem in the bottom assembly seems to still be intact, given that the top assembly’s various LEDs and microphone mute switch continue to function (likely removing them from the root-cause list as well). I doubt that a failure in either/both the bottom assembly’s audio subsystem digital-to-analog and amplification stages would suffice to bring the device completely to its knees, either.

That leaves what I previously referred to as the “digital PCB”. A corrupted firmware image, perhaps the result of power loss mid-update, is one possibility. While the Echo was still present in my list of activated (albeit in this case, not found) devices, before I removed it in an ultimately fruitless attempt to fully factory-reset and then revive it, its settings in the Alexa app indicated that a firmware update was pending. In retrospect, I wish I would have snagged a screenshot at the time; you’ll have to take my word for it. I initially discounted this firmware-corruption possibility because:

• Amazon certainly wouldn’t design a device that “bricked” so easily, absent any sort of user-friendly recovery scheme…would it?(??!!!!)
• And at the time, my other two, still-functional first-generation Echoes’ settings displayed the exact same “firmware update available” messages. I therefore assumed they were bogus remnants of the first-generation Echo’s dearth of Alexa+ support.

But in re-looking at their settings just now, those messages are now gone. There’s no user-controlled way to manually initiate a firmware update; Amazon automatically “pushes” them (presumably at a time when it senses that a given device isn’t in use, for example). So…maybe?

The other, more benign possibility is simply that some circuit on that particular (or another) board has “gone south”, taking the entire device down with it. Nearing 3,000 words in, I’m going to wrap up at this point and turn the keyboard over to you for your theories and broader thoughts on this teardown in the comments!

—Brian Dipert is the Principal at Sierra Media and a former technical editor at EDN Magazine, where he still regularly contributes as a freelancer.

Related Content

• Preemptive utilities shutdown oversight: Too much, too little, or just right?
• Amazon and Google: Can you AI-upgrade the smart home while being frugal?
• Teardown: Amazon’s Echo voice-activated virtual assistant
• A hard-life Tile Mate goes under the knife

The post A defunct Amazon Echo: Where did its acumen go? appeared first on EDN.