Line Voltage Into an Ethernet Switch—What Happens?
Power into Ethernet?
In the entertainment technology industry, we often run cables for Ethernet along with high power runs (like in the picture), or use hardwired Ethernet to send data directly to high power devices like dimmers, moving lights, speakers, motor drives, etc.
I’ve always wondered, what if a forklift operator ran over a power cable and an Ethernet line and crushed them together, shorting the high voltage into the low-voltage Ethernet switch? I’ve seen interfaces blown out on switches before, but I was never around when it happened. But I had a couple extra switches and some other parts laying around, so I decided to do a test; the results really surprised me.
Obviously, DON’T TRY THIS AT HOME!
Note: This post has a companion YouTube video, which you can view here:
Electrical Isolation
Being old and having worked on control systems for 40 years now, I’ve experienced all the reasons that we want isolation in electrical systems. A milestone story I remember hearing from the early days of the DMX-512 entertainment lighting control protocol was on an attraction in Vegas where they had a DMX lighting control system connecting a (low voltage) computer control console and a number of (high power) dimmer racks. There was a fire in one of the dimmer rooms. Was the damage confined to that one room with the fire? No, because after something melted and collapsed during the fire, the 120/208V line voltage got shorted to the low voltage DMX line, and that line voltage travelled through the control cables throughout the attraction, and took out the (expensive) control console and other electronics.
The solution to this? Electrical isolation, where a control or other signal is able to be transferred between two electrically operated devices without a direct electrical connection. This way, if there’s a problem on one end, it shouldn’t affect the other end, at least up to the rated limit of the isolation (with something like a lightning strike you’re in a different world). There are two primary kinds of electrical isolation: optical and galvanic (magnetic).
Traditional optical isolators electrically activate a small light source like an LED, and optically couple that with a light-activated electrical switch. This allows data to be transmitted from one device to another with no electrical connection at all. Optical isolation is used a lot in industrial control systems, and also is widely used in lighting and other entertainment systems. A fiber optic connection between Ethernet switches is also a kind of optical isolation.
The second kind of isolation is galvanic, where the desired signal is passed through magnetism, again with no direct electrical connection. One kind of galvanic isolation is a 1:1 transformer, where an AC electrical signal from a sender is used to induce a magnetic field; that magnetic field is then used to induce an electrical signal for the receiver. Again, in this way, the information or signal can be transmitted without a direct electrical connection. More on electrical isolation here.
Ethernet Isolation
Luckily for us in the entertainment industry, the designers of Ethernet, the IEEE, have mandated the inclusion of electrical isolation since the early days of the standard. “Thicknet” in 1985 included 250V of isolation; that was increased to the present day 1500V level with the roll out of 10BASE-T (10Mbps) in 1990. Ethernet was designed for the office world, which isn’t as electrically hostile as a factory or backstage at a large show. But even in the office, network cables could connect computers hundreds of feet apart, and it’s quite possible for the electrical services for those connected computers to be very different, and ground loops or significant potential voltage differences could exist. The isolation in Ethernet was intended to protect the sensitive electronics inside the connected devices (modern Gigabit Ethernet runs at only a couple volts), and 1500V of isolation is also pretty good isolation for us in the world of entertainment technology.
Inside Ethernet switches, this protection is usually packaged into “magnetics module” or “network interface transformer” chips; the protection can also be combined with the physical jack itself into a “MagJack”. You can see some of these tiny little surface mount protection devices from one of my switches in the photo (searching on some of these part numbers can lead you down some very obscure rabbit holes). The RJ-45 connection jacks are inside the metal box at the top of the picture and, if you look closely, you can see the circuit board traces leading from the chip to the jacks.
The insides of these modules are kind of wild. Here’s a couple photos from LINK-PP:
Those little circular things wrapped with wires are transformers and chokes, which provide the isolation and conditioning for each of the incoming signal circuit pairs. To see how this looks in schematic form, here’s a diagram showing one of the four wire pairs in an Ethernet interface from Würth Elektronik Group (who has an interesting Ethernet reference design detailed here).
The two vertical lines on the device in the middle of the diagram (a 1:1 transformer with a center tap that is used for Power over Ethernet amongst other things) form the magnetic dividing line between the world outside the switch on the left, with the RJ45 connector, and the internal electronics and processor of the switch, on the right.
So, what happens when these devices are pushed beyond their limits?
Note: If you’re interested, Semtech has a more detailed analysis of potential Ethernet potential faults here.
Worst Case Scenarios
I asked my friend Jim Janninck, an electrical engineer who worked on electrical interfaces in theme parks for 30+ years, what the worst-case scenarios would be for getting power into an Ethernet switch.
He outlined three approaches, from least to most potential damage:
1) Line (hot) on any one Ethernet pin, Neutral on chassis
2) Line on one wire of one pair and Neutral on a pin from another pair.
3) Line and Neutral across one pair.
I added a fourth:
4) Line and Neutral across all four pairs.
I came up with this wiring based on the four tests above:
And I wired this up into some connectors on a board:
And then I set out to test it.
The Switches
From my networking workshops, I own a couple older Netgear GS308E $25 managed switches that I don’t need any more, so I thought I’d test one of those out (paid affiliate link).
I also figured I’d buy the cheapest switches I could find, assuming they would not be properly isolated and would fail spectacularly. I bought two switches made by “Manhattan” on Walmart for $5 each (it looks like the price went up a couple dollars since I bought them).
Note: I didn’t read the fine print on this one and didn’t realize this was only a 100Mbps switch; more on that later.
I also found a gigabit switch from Cudy on Amazon for $10 and bought two:
Note: I’m not sure why you’d want one of these, but here’s is an affiliate link.
Assuming some of these would fail catastrophically, I had on standby to test a Cisco small business SG-300-10 switch that is old enough that it’s been retired from support by Cisco. But it turns out I didn’t need it!
Preliminary Test
I really wasn’t sure what was going to happen. I thought that the little sub-$10 devices might just blow up or catch fire, take out connected devices, or maybe even fail back up through the power supply and take out something else. Google Gemini agreed; when I asked, “what happens if I send 120vac into the rj 45 jack of an ethernet switch” I got:
Given all this, to be safe, I did my first tests in the backyard, fire extinguisher and hose at the ready and two different electrical circuits:
But after quick tests on a couple of the cheap switches, there were no flames or anything catastrophic, and neither the GFCI I had inline (mostly to protect me) nor the outlet strip circuit breaker tripped in any of the tests. I guess the little transformers vaporized so fast that it didn’t generate enough of a surge of current to trip the breaker, and the direct short presented by the winding (with no other path to ground) didn’t present enough leakage to trip the GFCI. In any case, since this seemed pretty manageable, I decided to move the setup inside where I could have better control over conditions, and be away from neighborhood lawn mowers and barking dogs for the videos.
Test #1 Results
This test, as outlined above, just put one line/hot lead into the switch. Jim had suggested connecting a neutral line to the chassis, but given that these switches either have plastic cases or don’t have an earth ground (they all are just powered with a two wire AC-DC wall wart), there wasn’t really any easy way to do that, and so there really wasn’t a return circuit path for the line voltage (I could have connected to the internal switch circuit ground, but that path would not be likely from an external cable problem, which is what I’m testing). So, while one line on a switch interface was sitting at 120V above ground—and this is obviously a safety concern—no damage to any of the switches occurred. You can see the results in the video.
Note: In these videos you’ll see the switches under test connected to other switches; I did that to see if--in case of a catastrophic failure--the line voltage might actually take out other connected devices.
Test #2 Results
This test put one Line/hot lead into one switch interface circuit, and the Neutral return path into another. If you look at the schematic above, if things are truly isolated, there should be no circuit path, and nothing should happen. But these are tiny little wires with tiny insulation, and I thought maybe—especially on the sub-$10 switches—they would detonate. Nope! They all flew through this with flying colors (it turns out this test wasn’t valid for the Manhattan switch, more on that later).
Test #3 Results
This is where things got interesting. This test put the hot Line on one side of a single input, and Neutral on the other, presenting a dead short to the incoming 120VAC circuit on pins 7 and 8, wire pair 4. When I fired up the line voltage into the Netgear and Cudy switches, I heard an audible pop and saw a bit of smoke.
Testing the Netgear and Cudy switches after having one circuit on an interface literally smoked, I was shocked to discover that all the other interfaces worked perfectly and passed data! This was a real surprise to me; they not only showed indicator lights when connected to other switches, but tested out with my NetAlly LinkRunner AT 1500 tester (paid affiliate link) through to my home internet (via the Cisco switch on the upper left), and successfully negotiated DHCP addresses with no problem. Pretty impressive for inexpensive switches!
Notice that I didn’t mention the Manhattan switch; that’s because it continued to work just fine through tests 1-3. Why was that—was the $5 switch super robust? No, it turns out that because—as I mentioned—I didn’t read the fine print when I bought the Manhattan switches, and they are 10/100Mbps switches only. Why would that matter? 100Mbps Ethernet only uses two out of the four pairs of wire in a Cat5e cable—and not pair 4 (pins 7 and 8) where I put the Neutral wire for test #2 or the Hot and Neutral for test #3. If you look carefully at the input jacks, you can even see only four pins connected: 1, 2, 3 and 6 (from right to left):
So this means that my line voltage wiring in Tests 1-3 didn’t even reach beyond the cable into the Manhattan switch at all.
Test #4 Results
This test put 120V AC directly across all four input circuit pairs; there’s not much way anything designed for +/- 2V could survive that, and this test gave the loudest pop and sometimes visible sparks which you can see in the video.
This test, not surprisingly, did take out an interface on the Manhattan switch. But as above after Test #3, all the other (non damaged) interfaces on all the switches worked fine and continued to pass data!
Digging Further into Test #3 and 100Mbps
Texting with Jim about the results, he said, “After Test 3, I wonder if the 1Gbps switches will still negotiate down to 100Mbps even with damage?” For this test, I had picked pins 7 and 8 just because these were physically away from the previous pins used and also are used for DC– in PoE. Coincidentally, this also meant I had picked two of the four pins not used in 100Mbps ethernet. So that means that while the input circuit pair I blew out in Test #3 was needed for Gbps Ethernet, it is not used in 100Mbps, and if the isolation worked on the wire pair level, the switch might be able to fall back to 100Mbps on the remaining, undamaged two wire pairs.
After the sparks of Test #3, I had quickly connected interfaces from each switch into other switches, and they didn’t light up right away, so I assumed those interfaces were completely blown out.
But now, I tried again with my NetAlly tester:
Sure enough, after about 10 seconds (it has to try and fail at gigabit first so it takes some time) it came up good on 100Mbps!
You can see that the switch advertised up to 1Gbps, but the actual negotiated rate was only 100Mbps. This was amazing to me, after I had just electrically abused the same interface I was now testing.
The little Netgear switch is actually a managed switch, and even its management interface worked fine with blown out interface circuits, and it dutifully reported the speed of its partially blown out input at 100Mbps (interface 4 in the screen shot).
This was all really surprising to me; here’s a video showing data passing through all three switches after damage had been inflicted on all, and two of the ports here had auto-negotiated down to 100Mbps.
Inside
Seeing a few sparks from the outside during tests #3 and #4, I wanted to get in a bit deeper and see what exactly was failing inside the switch, so I took them out of their cases. All three of these switches had been sent 120V into their Ethernet interfaces, but there’s hardly any evidence of that visually:
The isolation chips for the Manhattan switch had a little cover that could be easily pried off without removing the chip from the board (I don’t have surface mount soldering equipment to remove the others) and even in there—after all that this thing had been through (this switch had every port damaged in one way or another as you can see from my sharpie marks)—the carnage was confined to the isolation circuitry, as designed.
Here’s the most spectacular results from Tests 4 and 5 in slow motion:
If you pause the video, you can see the sparks and smoke emanating directly from those surface mount isolation chips. So the spark might cause some damage to the board, but the voltage was effectively contained.
Metering things out, it seemed that I burned out the traces on the surface mount jack for test #3.
To be thorough, I did come up with a new wiring scheme to try and see if a Manhattan 100Mbps switch would survive a legitimate test 3 with line voltage into one connected input pair, but on my first try after rewiring it burned out a different pair of traces on the connector circuit board, and at this point I figured it wasn’t worth going to further effort for a $5 switch.
Conclusion
I was really surprised by the results. Going in, I figured the little sub-$10 switches would just blow up and/or cause all kinds of catastrophic damage, possibly failing upstream and taking out connected devices or even their power supplies. I figured the Netgear switch would fare a little better, but wouldn’t be fully functional after having line voltage up one of its interfaces. But the Ethernet electrical isolation was implemented well enough on all three of these little switches that they all successfully isolated the switching electronics inside from a catastrophic 120V overvoltage situation on an input circuit. And they even isolated the other interfaces, even though they could shared wiring inside a little tiny surface mount module (I guess you’d have to try 1499V to truly test it). Pretty amazing!
I can’t say I would recommend continuing to use a switch with a fully or partially blown out port, and I generally wouldn’t use any of these cheap switches for any real show purposes (a lot more on that here); if anyone wants to send me a more expensive switch to test I’d be happy to try it (I didn’t want to risk the expensive Netgear AV switches I own). But I have to say I’m impressed with the design of these little switches!
By the way, I kept Popsicle kitteh away from the line voltage experiments, but she did help me clean up: