The phrase "they don't make 'em like they used to" sure seems to apply to this dissection candidate, which delivered eight years of solid internet service to its (at least) second life.
Way back in November 2012, I told you of my travails trying (and ultimately prevailing) to get a Linksys DPC3008 cable modem activated on Comcast/Xfinity’s broadband service. The DPC3008 was already used (factory refurbished, to be precise) when it came into my possession, which makes its robustness all the more impressive; it ran hiccup-free (aside from Comcast/Xfinity’s various periodic service issues, which weren’t its fault) for nearly eight years, including a residence migration. My sole motivation to move beyond it was that I required higher upstream bandwidth due to COVID-19-induced transitions to at-home work for everyone else (I’d been work-from-home for nearly a quarter-century at that point) and consequent (and coincident) frequent use of videoconferencing by both myself and my also-work-from-home wife. Therefore, the necessity for DOCSIS 3.1 service. Therefore the NETGEAR CM1100 successor.
The DPC3008, along with its companion power supply and never-used software CD, have sat in a plastic grocery bag ever since, awaiting charity-donation relocation (since the cable modem is no longer partitioned with Comcast/Xfinity, its new owner won’t have any problem getting it activated, at least theoretically). In advance, I thought I’d take a hopefully nondestructive peek inside and share my reconnaissance results with you. Without further ado…
The DPC3008 is a now-entry-level DOCSIS 3.0 (backward-compatible with DOCSIS 1.0 and 2.0, of course) cable modem, supporting up to 8 bonded downstream channels (340 Mbps peak) and 4 bonded upstream channels (120 Mbps peak). Its dimensions are 4.8 x 1.1 x 4.8″ (12.2 x 2.8 x 12.2 cm) and its weight is 0.7 lb. (0.3kg). Here to begin are some overview shots, as usual accompanied by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes:
As you can already see, the DPC3008 is abundantly outfitted with passive ventilation slots, a likely notable contributor to its longevity:
The yellow color of that Ethernet LAN port correctly identifies it as supporting GbE speeds:
Even more passive airflow (exhaust, in this case) vents along the top:
and the bottom (inflow this time…heat rises, don’ cha know). Note, too, the screws in each corner, which I’m already guessing are the gateway to the cable modem’s interior:
Before tackling those screws, let’s snap a closeup of the bottom side label:
along with some shots of the accompanying power supply and software disc:
And now, hopefully, inside we go:
Not much notable inside the enclosure’s top half, aside from multiple light pipes that transport the PCB LEDs’ illumination to the outside world:
And speaking of LEDs, there they are, along the PCB’s bottom edge. Note the additional unpopulated LED site toward the right, alluding to the possibility that this same PCB layout may support multiple cable modem product proliferations:
Also note the two additional screws in the right top and bottom corners of the PCB, holding it in place (the left-side PCB holes are “poke-throughs” for the already empty screw holes underneath, which helped hold the two enclosure halves together). Let’s get them out too:
This closeup of one of the two now-empty PCB screw holes also provides a zoomed-in view of one of the notable ICs on the board. Google searches for “Unihan T1210” and “TGSA-2501-R” were unenlightening, but given its physical characteristics (package, lead count, color, size and shape) and its proximity to the Ethernet connector, I’m guessing it’s the LAN transformer:
Also note the tape spanning the Ethernet connector and the Faraday cage to its left. It’s seemingly metallic in fibrous construction, leading me to suspect it acts as a grounding strap.
Similarly, a closeup of the other now-empty screw hole also revealed in detail the system DRAM IC, a Hynix H5PS5162GFR 512 Mbit DDR2 x16 SDRAM:
While we’re at it, here’s a closeup of the other memory (I’m pretty sure) IC on the board, to the right of the heat sink:
Portions of the package topside marking are obscured by the number sequence “44” (I suspect referencing the firmware version programmed in the device) hand-scribbled on top with what looks like a black magic marker. While I can’t ascertain the density, it is from Winbond. My guess is that it’s one of the company’s SPI NOR flash memories, intended for code storage.
And here’s another PCB closeup, of what seems to be the cable modem’s power subsystem, containing (among other things) three IT7612DC/DC converters (i.e., voltage regulators) from ITE Tech and three large, conveniently value-labeled inductors:
With the final screws removed, the PCB lifts right out of the enclosure’s remaining bottom half:
As previously mentioned, my as-always aspiration is to leave the torn-down patient in a sufficiently intact state that it can be reassembled in fully functional form afterwards for subsequent donation. That said, in looking at the Faraday cage in the upper left corner, I suspected I might be able to pop the lid off with a flat head screwdriver (acting as a lever arm in partnership with the heat sink) and then securely push it back on top afterwards. Let’s see:
Lots of passives inside, as expected:
The one notable IC along the bottom edge has the following four-line marking (to the best of my old, tired eyes’ abilities):
Google suggests that the “KMLG10” may be a low-noise RF amplifier from KEC Holdings, which would make sense given its location; here’s a DeviWiki entry for another cable modem with the same IC listed in the “additional chips” section.
Speaking of “leaving devices sufficiently intact that they can be reassembled and reused”, I’d hoped to also be able to get that heat sink off. But a peek at the PCB backside left me skeptical:
Note the manually twisted mounting “prongs” where they exit the PCB. I suspect that even if I got them untwisted, they’d subsequently snap upon attempted reassembly…even assuming I could get the heat sink off and then meaningfully back atop the IC in combination with some thermal tape or paste. But speaking of DeviWiki, it turns out we don’t need to bother with the heatsink, because we already know what’s underneath: Broadcom’s BCM3382 system SoC.
And with that, I’ll wrap up this teardown analysis, which I’ve already supplemented with a successful reassembly prior to writing these words. Reader comments are as-always welcome!
This article was originally published on EDN.
Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.