Teardown: Dissecting Snap’s third-gen AR glasses

Article By : Brian Dipert

Hear the words “smart glasses” and, if you’re like me, you automatically think of the full-blown implementation of the concept: something you wear on your head and through which you can view the real world, containing a camera that also captures and interprets real-world images, coupled with a display (standalone or leveraging the glasses’ lenses) that feeds you information to augment the real-world data you’re seeing. Snap’s Spectacles are a notable subset of that full-blown implementation, intentionally so. Simply stated, they’re (at least in their current form) essentially cameras that you wear on your head, capturing still and video images that eventually make it to your Snap account via a wirelessly tethered smartphone or other mobile device (they support both Bluetooth and Wi-Fi Direct connections) to share with your followers.

The 3^rd-gen iteration of the concept is targeted at “fans of high fashion and artists who relish new creative tools,” quoting from a review in The Verge; the accompanying video (see “Snap Spectacles 3 review: here we go again”) says it slightly differently: “…intended for a more fashion-forward and creative audience”—one with a bigger wallet and more disposable income, apparently. This time around they cost $380.

For that price adder, you get two cameras, one at each temple, each capable of capturing video at 1216 x 1216 pixels (still only translating to 720p widescreen, however) and 60 fps, or 1728 x 1729 pixel still images. Spectacles 3 even embeds a GPS receiver, so you can “geotag” your captured images. You get four microphones, with each pair still in an array setup, for “stereo” audio capture. The frames are (mostly) metal, albeit no longer water resistant. And the carrying case is “premium” (whatever that means).

Let’s get to tearing down! I’ll as usual begin with some external packaging photos:

Slip off the outer cardboard sleeve…

and the contents begin to come into view:

Setting aside the (not-so) mysterious tissue paper-wrapped piece for a bit, underneath you’ll find two cardboard boxes. The first one contains two scant pieces of literature, along with a cleaning cloth:

while the second houses the USB-to-USB-C charging cable:

Underneath them is a Google Cardboard design-based 3D viewer:

Putting the peripheral component opening acts aside, let’s get to the headliner. Surprise; it’s the glasses! I’ll begin with some overviews of that “premium” case:

Guess what’s inside? You guessed right!

Let’s focus first on the case itself. It also contains a battery pack, which Snap claims can recharge the glasses’ batteries up to four times before it’s depleted and requires recharge itself. A four-LED line inside indicates charge progress:

You may have already noticed that as shipped from the factory, there’s a piece of yellow-and-transparent plastic between the contacts on the case and the glasses themselves (see photo above). Remove it and, even with the case batteries fully charged, the LED-status sequence begins all over again, until the glasses themselves are fully charged:

One other nifty thing about the case: the charge-status LEDs automatically light up for a few seconds when you open the flap; there’s no button that you need to press to get status:

This suggests to me that there’s a magnetic switch in the front half of the case, which interacts with the magnet strip at the edge of the case “flap.” Speaking of which, what’s in that case? Diving in:

Fearful of acrid smoke and heat, and resisted by stout adhesive, I admittedly didn’t get too far. What I suspect is that there’s a rechargeable battery in each half of the case, along with charging circuitry, cabling interconnecting the two halves and the aforementioned LEDs and magnetic switch. Sorry, folks, I’m not up for setting either myself or my house on fire, so I’m not going to press on further. Instead, let’s switch gears and focus on the glasses themselves.

Here are some overview shots, as-usual accompanied by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes. The glasses themselves have a frame size of 47 x 25 x 153 mm and a weight of 56.5 g:

Admittedly, the style’s not my particular “cup of tea,” which perhaps just means that I’m not “fashion-forward and creative?” I’ll let you be the judge. Anyhoo, unfold the arms, make like you’re going to put them on your head, and here’s what you’ll see:

Are you thinking that those oval-shaped plastic “shells” on both arms, near the hinges, would be perfect for putting batteries underneath? Me too; hold that thought. Snap claims, by the way, that you can capture and sync up to 70 videos per full charge, and that the glasses fully recharge in their carrying case in about 75 minutes.

Here’s what the arms look like from the outside, also showcasing the switch on top of each, which finds use in initiating still and video image capture, etc.

And here’s a closeup of both arms from the inside:

Flip the glasses upside down and you’ll now be able to see two additional things on either side: a multicolor LED (which I initially thought might be one of the microphones, although the location for such a function was odd) and a screw, which I initially assumed might be a pathway to the insides but didn’t pan out (again, hold that thought):

Last but not least, overview-wise at least, here’s a peek at the charging contacts built into the glasses’ top bar:

And here is a close-up of one of the pair of temple camera-plus-LED structures from the front (shown here is the right side structure):

This is what they look like (in one of their multiple modes) when the glasses are operational:

Speaking of which, here’s what one of the inner LEDs looks like while the glasses are capturing images:

Now let’s take a look at the (roughly) oval plastic pieces on the insides of both glasses arms, underneath which (as I mentioned earlier) I suspect are rechargeable Li-ion battery packs. If you look closely at the earlier photos of one if the arms with the glasses upside-down, you’ll see tiny gaps (at one end, and underneath) in the otherwise sturdily glued seam between plastic and metal around the entire circumference. My suspicion that these are “vent holes” supports my conjecture that batteries are inside, but there’s only one way to find out. No amount of effort attempting to crack the glue using a thin flat head screwdriver inserted in the gaps as a lever arm was successful. So I instead turned to my heat gun (right glasses arm first):

I was right! (But you already knew that, right?) Now for the left side:

Another battery! In both cases, you can also see the “shutter” switch atop the arm, from the inside. One other thing I’ll draw to your attention is the two hex head screws near the hinge on both sides. I’m equipped to remove them, courtesy of the hex drivers in my iFixit 64-bit driver kit, but I couldn’t gain purchase on them. Turns out they’re covered by some sort of a transparent but sturdy coating that seemed immune to my application of heat, which again, I didn’t overdo for fear of an acrid smoke and heat result.

Next let’s separate the plastic-vs-metal portions of the top bar, once again assisted by the heat gun. I began at the middle (in the vicinity of the earlier shown charging contacts), moving outward toward each temple. Here’s what the resultant middle portion of the plastic (back) half of the top bar looks like, showing the wiring that runs between both temples, whose power-and-other functions will become clear shortly:

And here’s the backside of the corresponding front (metal) half of the top bar, also showing the bridge and lens rims:

As you can see, one of the lenses has already popped out in the process of separating the two halves of the top bar, revealing the cutout in it for the associated camera:

Also revealed in the earlier image is the fundamental function of the earlier noted screws in the rear of each temple. Snap has partnered with Lensabl to offer (at an incremental price) Spectacles 3 complete with prescription lenses. Think about it for a second and you’ll realize that the prescription-or-not decision is one likely made late in the glasses manufacturing process. Pop in the desired lenses, tighten the screws, and they’re good to go to the retailer or (in the prescription case) directly to the customer.

Back to the plastic half of the top bar (and rewinding in time a bit): here’s our first glimpse of the camera-and-other cluster in the left temple (again, a reminder that in this write-up, I refer to “right” and “left” from the wearer’s perspective; said another way, the right temple is in the upper left corner of the glasses when you’re looking at them from the front):

The shield in front of the camera assembly promptly fell off, of its own accord revealing the LED ring below it, and around the camera lens:

Here’s its right-module twin, with the shield still attached:

And here’s an overview shot of the entire assemblage:

Let’s switch back to the metal front half for a bit. Here are closeups of both temples:

After removing those remaining plastic-and-metal standalone pieces there’s nothing substantive left afterwards, at least that I can discern, so it’s time to get those temple modules out for a closer look. I’m going to share this next set of shots in chronological order, and editorial-free. The extraction was done solely using a thin, small flat head screwdriver, and, as you’ll see, the resulting right-temple module is a three-PCB sandwich, with flex PCB cable interconnecting each layer:

One task left, at least for this module: get that Faraday cage top off to see what’s underneath:

Here’s what’s left after the right module was removed, by the way:

And for grins, I manually reinserted it in the metal frame to approximate its normal orientation:

Now for the left-temple module, which I initially thought about skipping over, assuming it was identical to its right-side sibling. I’m glad I went ahead, because it’s quite different (for one thing, it’s a dual-PCB sandwich this time):

And another Faraday cage whose top is begging for a rip-off:

For these last few shots, I switched from my smartphone to a standalone digital point-and-shoot camera. Its macro mode let me get closer, but the results are arguably better-or-worse. Complicating matters, of course, is the fact that many of the components on the PCBs are extraordinarily tiny and/or their package markings are faint-to-nonexistent. That said, I was still able to discern some interesting bits, which you’ll hopefully also be able to discern between these and the prior photos (otherwise just take my word for it!). Right-side module first:

The most obvious IC is the Samsung 4 GByte NAND flash memory used for local image storage, on the PCB in the upper left quadrant. While writing this piece, I relooked at the photo and noticed the foam pad on the middle PCB; pulling it away with a tweezers revealed only a flex PCB (between it and the camera PCB) connector underneath. I’m guessing that one of the flex PCBs handles signals associated with the camera itself, while the other deals with the LCD array around the camera lens. Flipping it over:

The IC of note here is on the PCB to the right: it’s obviously from Intersil (now a subsidiary of Renesas Electronics) but the second and third lines of its markings are a question mark:

BR5732
D937AR

They don’t correspond to any IC I can find via Google. Reader insights are welcomed. Intersil’s core expertise is in power management and precision analog, so I’m guessing that it (along with other ICs and passives in this module) deal with digital power maintenance and/or battery charging.

Now for the left module:

The large chip on the right-side PCB, underneath the Faraday cage, is (unsurprisingly) Qualcomm’s QCA9377 single-chip dual-band Wi-Fi and Bluetooth (both standard and low energy) controller. But that curiously shaped piece of metal nearby is not the Wi-Fi/Bluetooth antenna. Thanks to FCC documentation (ID: 2AIRN-003), we know that it’s the metal glasses frame itself—specifically the portion around the left lens (check out the “TempConfidential_Internal picture rev.1” link once you log in to the FCC site). Clever!

So what’s that “curiously shaped piece of metal” for? Flip the PCB over for the answer:

Note that big IC on the right. Its markings are faint but discernable, thanks to my inexpensive but effective illuminated magnifier:

©010D
ZOEM8B
C02010
14 CC3

It’s a GNSS (global navigation satellite system) receiver, from u-blox. And that “curiously shaped piece of metal” on the other side? It’s the associated GPS antenna!

Back to this side of the PCB. Next to the GPS receiver is another design oddity, Nordic Semiconductor’s N52832 Bluetooth 5.2 SoC. This is a parts portfolio that I’ve encountered before and it continues to confuse me: why are there two different chips (this and the aforementioned one from Qualcomm) that both profess to handle Bluetooth duties?

That’s all I can discern; everything else is too tiny, too faint or absent any markings whatsoever. Leaving me with a couple of nagging questions:

• Where’s the system processor? At minimum, there’s got to be some intelligence somewhere for responding to user switch presses, controlling LEDs, managing the microphones and cameras (and the data generated by them) and the like.
• Is the audio and image data getting compressed and otherwise processed in any way prior to storage, or is it just getting dumped “raw” on the flash memory with all subsequent processing done in the mobile device post-transfer? The first-generation Spectacles included an Ambarella SoC for glasses-native processing tasks, but unless I’ve overlooked something obvious, I’m not seeing an equivalent here. Were the Spectacles 3 voice-controllable, for example, there’d need to be an integrated audio processor from a company such as DSP Group or Syntiant, but in this simpler scenario…
• And speaking of the microphones (MEMS, presumably), where the heck are they? Supposedly there are four total, but I can’t find any of them. My guess is that they’re around the cameras (two per module), with sound leaking in from the outside through narrow gaps between camera assembly and frame (therefore the lack of water-tolerance this time around) but I’m completely befuddled. Then again, what else is new?

I’ve just passed through 3,600 words, which must be some kind of record, and my poor editors also have over 120 images to deal with (let’s see how many of them end up on the cutting room floor). So I’m going to stop, with the hopes that you’ve also made it to the end, and turn it over to you for your comments. Were a Snap engineer, for example, to fill in the missing pieces along with adding more knowledge nuggets…anonymously, of course…just sayin’…

This article was originally published on EDN.

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.