Ethernet will probably do fine. The differential pair ends in a transformer, so unless you pump so much current in that it physically melts the transformer or wire, you'll be fine. USB should be similar, but apparently is not. (In the case of the USB killer, the amount of time where 200V is applied to the circuit is probably very tiny. I imagine it's just a charge pump that charges a tiny capacitor and discharges it.)

Ethernet is designed this way because the differential pair can apparently pick up a bit of a DC bias when laid next to cables in the ceiling/wall. While researching this post, I found that the spec requires the transformer/inductor to work correctly with a 30mA bias current. I am too lazy to go measure what the DC resistance is, though, so I don't know whether or not 200V is close to that.

When I was in high school, I wired up a cat 5 cable such that each pair was connected to one phase of the mains. I then plugged it directly into my iMac. There was no damage to the computer and the Ethernet port worked fine afterwards as well. (The integrated monitor went crazy; this was in the CRT days and presumably the Ethernet transceiver's magnetics make a good degaussing coil).

Oh dear god, why did you connect your iMac’s Ethernet port to mains? Did you expect it to survive?!

The fact that Ethernet functions the same with a voltage bias across the leads is exploited for PoE — power over Ethernet.

> Oh dear god, why did you connect your iMac’s Ethernet port to mains?

Because I was 15 and bored.

Wouldn't you be more bored if you found yourself without an iMac?

In college I stripped one end of an Ethernet cord and plugged it into household current. The other end I had hooked up to an old laptop I no longer wanted. The wire exploded, and the motherboard was fried. YMMV. Laptop was an old Sony Vaio c. 2001 vintage if it makes a difference.

(On the plus side, this made up for sleeping through that physics class where my professor did the exploding wires demo.)

Household current is AC, the transformer will pass it through.

The transformer mainly protects you from a voltage difference between a twisted pair and ground. Much less so from voltage across a twisted pair. But saturation could help limit the current spike, at least.

That makes sense. The pair shouldn't have a voltage on it, that's why it's a twisted pair.

> Is there any device that is supposed to survive this?

There used to be...

As a boy - and this is many, many years ago - I once, ehm, accidentally connected the external speaker-plug on an old vacuum tube household radio to the 220V mains.

A very loud bang, and the house went dark. The radio, however, was fine, although with an interesting new tonal slant to everything.

As a young man, I grabbed a 6146B tube before it discharged. I felt fine, until the pain started.

Because radios in those days were made properly, without planned obsolescence!

That's the filter of history at work. By and large, old radios were not built to last. The only ones that survive are the ones that, because of good design, good luck, or gentle use, managed to make it to today.

I repair & build vacuum tube equipment as a hobby.

You'd need opto-isolated data lines. I can't see any common devices having that.

For a while, macbooks were built this way: https://github.com/usbkill/community/wiki/Consumer-Laptop-Ha...

But the recent/2018 ones are vulnerable? Why? Cost cutting?

Probably related to the switch to USB-C, which started with the 2016 models.

Would that even meet the USB spec? You could isolate the entire USB module, but I don't think you can just isolate the data lines.

And you'd still have to deal with the big spikes coming in on the USB power pins.

Newer Apple devices should be protected.

Why?

For a time, macbooks had opto-isolated USB ports[0]. I think they don’t anymore though.

[0] https://github.com/usbkill/community/wiki/Consumer-Laptop-Ha...

Call me highly skeptical. Every reference I've seen to this claim seems to point back to usbkill.com making it itself.

The problem is: optoisolators are very slow. Building one which goes faster than 12Mbps would be nontrivial; I do not think such a thing existed in 2014 (which usbkill is claiming). This implies that Apple has alien technology.

I think it's something else.

  The problem is: optoisolators are very slow.

Many are but they don't have to be - an optoisolator is just a zero-length fiber optic cable. Ain't much higher bandwidth than a fiber optic cable.

Not saying apple laptops ever had optoisolated USB though.

IMHO the problem with isolating USB is the 480Mbit ("USB 2.0") bidirectional signal.

12Mbps/1.5Mbps is easy. ADI has an integrated magnetic isolator chip. For hobbyist use you can get this assembled on a board ready to use as an ebay special.

The 5Gbit of USB3 is two unidirectional links and so is also easy (monoprice and elsewhere sell optical USB3 cables).

And lest you think you'll isolate at 5Gbit and then use a hub to translate to 480, USB3 doesn't work that way - the backwards compatibility is provided by a complete USB 2 bus in parallel.

For general isolating (say an electronics workbench), it's easiest to consider the USB host floating (eg a dedicated bench computer), and rely on ethernet (T or SX) for the isolation.

Beyond that the host has to supply 5V to the USB device, couldn't you just cause some damage by sending 200V on these lines? You might only fry the USB power supply but that's enough to be a major pain.

I vaguely recall the argument (maybe buried in the USBKill site) that manufacturers could ~easily limit maximum voltage across the data lines. Maybe using a resettable fuse? I know that Raspberry Pi limit USB input voltage to ~5V, for example.

Raspberry Pi limits the input voltage with a Zener diode across the input. In a Zener used this way, the current flow through the diode varies with the voltage difference between the input and the output - so with enough voltage input you can still make it fail, as you can exceed the max power dissipation of the diode and then it will burn - and after that, your circuit is not protected anymore.

I can confirm that you can indeed destroy an RPI0 by accidentally connecting a power supply to the USB port while another power supply is connected to the power supply port.

OK, so is there any other purely electronic protection method? Or are optoisolators the only option?