Implementation incompatibilities between Power over Ethernet (PoE) sources and remote endpoints can doom a network topology aspiration.
Back in October 2018, when I disassembled a TRENDnet TPL-331EP powerline adapter containing a Power over Ethernet (PoE)-supportive output (see “Teardown: Powerline networking plus PoE”), I prefaced the “surgery” with a brief tutorial on PoE technology for those who (like me, before I began the project) were unfamiliar with it. The tutorial concluded as follows:
As the Wikipedia definition of PoE notes, the IEEE specification set has gone through a multi-generation evolution, with power delivery enhancements accompanying each iteration. The means by which power is carried within a 10/100 Mbps Ethernet cable also varies: with so-called “Mode A,” the power delivery takes place over the same 1-2 and 3-6 pairs used for data, whereas “Mode B” uses “spare” pairs 4-5 and 7-8. With Gigabit Ethernet, which employs all four pairs of wires for data, merging data and DC power over the same wires is the only option. Perhaps obviously, making sure your PoE power sourcing equipment (PSE) and powered devices (PDs) are mode-compliant is a critical requirement.
The wrap-up comment in the previous paragraph is, I suspect, at least part of why PoE hasn’t seemingly been widely adopted to date. If your PoE injector (or PoE support-inclusive router or switch source device, for that matter) implements one mode and your PoE splitter (or PoE support-inclusive remote device) implements another, your only recourse is a frustrating return of one or both devices to your retailer. Voltage and current incompatibilities between source and destination, which I realize in retrospect that I didn’t touch on at all in my 2018 coverage, can also result in a product return (not to mention, potentially “fried” gear).
And of course, as Wikipedia also notes, plenty of non-standard implementations also exist. While they might work fine if your network implementation is 100% single vendor, such an approach runs counter to the multi-vendor interoperability aspiration that leads to lower prices. The result of this mishmash is clear when you look at comparative prices: while a standard eight-port GbE switch from TP-Link sells for less than $20 on Amazon, for example, its PoE-supportive counterpart (with PoE on only four of the ports) will cost you $59.99. And expand PoE support to all eight ports, as a D-Link GbE switch also found on Amazon does, and you’ll pay almost $100.
To wit, this time, we’ll be disassembling a PoE injector—specifically Ubiquiti Networks’ 24-24W-G-US:
I’ll begin with a brief story of how I originally acquired the unit. Back in July of last year, I came across it in the midst of a multi-product Newegg promotion listing. The product was titled “Open Box: Ubiquiti Networks POE-24-24W-G-US PoE Adapter & Injector,” and as you’ll see below, the primary photo that accompanied it implied that I’d be getting both a PoE injector and a splitter:
When it arrived, however, only the injector was in the shipping package, and although Newegg customer “support” tried to hide behind the company’s Open Box Product disclaimer, I refused to relent, pointing out that a) I was a long-time and lucrative-to-Newegg customer and b) the title implied, and the photo blatantly showed, that I’d be getting two products, not one. I got my refund, and I got to keep the injector. And since I don’t own any Ubiquiti PoE-supportive access points, I decided to take it apart and see what’s inside, instead.
Next, here’s the “decoder ring” for the product name: The first “24” indicates that it outputs 24 V over the Ethernet connection; the following “24W” means what it says—24 W, alternatively indicating that the unit outputs 1 A max current; “G” means that it supports GbE connections; and “US” means that the power cord has a US-compatible NEMA 5-15 wall outlet connection.
One more comment before diving in. A 24-V output might seem odd to you—it was to me, considering that most network client devices run at voltages like 5 V, 9 V and 12 V. Keep in mind that transmitting DC enables more straightforward demuxing of power-vs-data at the receiving end, versus if AC is transmitted as with the powerline networking also supported by the TRENDnet TPL-331EP I mentioned earlier. However, the DC power also experiences more rapid “loss” (versus AC) as it passes from source to destination along the thin-wire Ethernet cable (right, Thomas Edison?). DC-to-DC voltage down conversion at the PoE destination device (or the splitter) is therefore common.
Let’s now dive in, starting with some box (and unboxing) shots:
Nothing inside the box save for the unit and its associated power cable; not even a slip of quick-start literature:
And here’s our victim, accompanied as usual by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes (the 24-24W-G-US specs include dimensions of 88 x 57 x 33 mm (3.47 x 2.24 x 1.30″) and a weight of 158.5 g (5.59 oz).
Bottom, with optional mounting plate still attached:
And input and power-included output end:
Step one: let’s get that mounting plate slid off the bottom of the unit:
Here’s a close-up of the label with all of the nitty gritty spec details. Note from the verbiage that the DC power is ONLY present on pins 4 and 5 (+ voltage) and 7 and 8 (- voltage). Translation: this is not a higher power 4PPoE (or if you prefer, Ultra PoE) which uses all four twisted-pairs in the Ethernet cable to transmit not only data but also DC.
Time to dive in. There are no screws underneath the label; truth be told, I couldn’t even get the label to peel off! Turns out that the two halves of the case are ultrasonic-welded together, so I broke out my hacksaw:
The underside of the top half (right, above) is pretty staid; the only thing of note is a translucent piece of plastic for the LED underneath to shine through:
The bottom half is much more interesting, although as a long-admitted “digital” guy, many of the particulars remain somewhat nebulous to me:
A couple of lumps of white paste (one on either side) holding the PCB in place were easily severed with a box cutter:
Looking at the PCB’s underside first, there’s not much to note aside from a bunch of passives, visible traces, solder points and, near the top and between the two Ethernet connections, a reset button (whose function the dearth of documentation doesn’t elucidate):
Flip the PCB over and things get more interesting:
Again as the “digital”guy, the IC in the upper right corner drew my immediate attention; it’s a 100/1000 BASE-T transformer module, the G24102HKG from M-Tek Electronics. Aside from that, it’s all analog—power and passives as far as I can see. A transformer and big coiled inductor in the center, two equally massive capacitors (check out these photos of what happens when one of them blows its top!), and several beefy power transistors thermal paste-mated to metal heatsinks are highlights. Here are some side-angled shots to close out the story:
And with that, I’ll wrap up and turn the pen over to you power afficionados to fill in the gaps in the comments!
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.