Ethernet connection using capacitive coupling

Wanting to save space and weight on his project build [Florin] set out to find a way to add Ethernet connectivity without the magnetics. His ill-advised first try involved directly coupling two switches, frying both in the process. After some research he found that Ethernet hardware manufacturers have considered the need for devices without the magnetics and there are several application notes available on the subject. [Florin] followed the information that Realtek has for their devices and learned that they can be couple capacitively. After depopulating the magnetics from a second pair of switches he wired up some resistor-capacitor networks on a breadboard and got the connecting to work.

26 thoughts on “Ethernet connection using capacitive coupling

  1. On a lot of ethernet resources on the web it does seem to refer to isolating/decoupling transformers used in the transmission line.
    Why they had to call them “magnetics”, I haven’t worked out.

  2. Uh, what? Most transformers/power isolation units are basically magnets or some form of electromagnet to protect your equipment. That is why this guy fried his units the first time. The idea is, if you don’t have the protection on a certain line, if say you get a power problem (lets say, power surge) on one device, that tiny amount can affect other devices or ethernet ports itself. I had an issue where the power in my house went out. My computer was on a battery backup, but xbox 360 wasn’t. Xbox 360 ethernet adapter has never worked, but the rest of my devices still do.

  3. @Sheldon, “magnetics” is a pretty common blanket term for all components that make use of magnetic fields and magnetic materials. It’s often used because magnetics tend to be expensive, somewhat large, and require more considerations than, say, just popping down a .1F cap would.

    In a project before I’ve used an ethernet jack that had integrated magnetics. I wonder if it was actually capacitively coupled or if it just had really tiny isolation transformers.

  4. Optoisolators cost more than bits of metal at the time, and still do. I must say that I’m looking forward to 100GB ethernet devices. The future is basically awesome. Too bad politics can’t keep up.

    On a nostalgic note, old gear (like switches and hubs) 10MB AUI interfaces can be replaced with LEDs and photo-diodes [well, almost] and, when given even the most half-assed enclosures and optics, can work pretty reliably for about 300 meters.

  5. @Matt
    Despite being an electronic engineer for a good many years, I’ve never heard anyone refer to components that make use of electromagnetism as “magnetics” which is why I was questioning its origin. It’s incredibly broad and not at all useful in explaining what it actually is.
    Inductors, just like capacitors, come in many sizes and costs.

  6. @smoker_dave the standard is IEEE 802.3

    Its a “collection of IEEE standards defining the Physical Layer and Data Link Layer’s media access control (MAC) sublayer of wired Ethernet”

    The standard doesn’t refer to the transformer itself but for a network to meet that standard (and all vendors want that) you have to consider that the transformer has a large impact on the transmit template and the cable length performance which in turn have to meet the standard.

    You’ll find most transformers manufacturers have this paragraph in their datasheets: “Compliant With IEEE 802.3 and ANSI X3.236 Standard Including Baseline Wander Compensation Specification Of 350 H OCL When Biased at 8mA From OC to 70C”

    An RJ-45 jack with integrated magnetics will follow the same pattern and have those transformers in there to meet the IEEE standard.

  7. Having designed several Ethernet interfaces in my career, I’ll chime in. I’ll assume we’re talking about BASE-T here, and not about classic coax Ethernet

    The good ones all use magnetics (sorry, Sheldon, that’s what they’re called in the network biz) and transient suppressors. The cheap ones don’t. You can tell the good interfaces from the cheap ones by how they behave after a near miss during a lightning storm.

    The magnetics are there to break ground loops and couple from an unbalanced source (like a 5V to gnd signal) to an unbalanced load (like a 120-ohm CAT5 twisted pair) and back again to the receiver. The fact that you’re putting a carefully engineered, balanced drive signal on the twisted pair goes a long way (literally) to keeping the error rate down and passing FCC emissions tests.

    You can’t use opto isolators for this because they’re not fast enough. Check the specs. If you go optical, you use lasers.

  8. The reason this isn’t done for more than on-board connections is because capacitive coupling has a nasty habit of coupling all sorts of noise onto the line. In audio lines with a shared ground, there’s an equalization “pop” on connection as the capacitor draws current to reach the equilibrium state, but that’s not the worst of it.

    When you break the shared ground line and go differential, unlike a transformer, capacitors don’t protect you from common-mode noise. A shift in the ground potentials between the two ends will spike both ends of the differential pair – and while that won’t effect the differential receiver for small spikes, it’ll happily couple spikes large enough to cause damage.

  9. Transformers’ natural resonance is a convenient bandpass filter and it also conveniently decouples the (variable) capacitance of the actual cabling. I.e. the circuit above is generally fine in the lab but will never work in the real world. Rest assured that a whole lot of really smart network engineers have been trying to get magnetics out of the equation for many years as they are the largest and most expensive component in more ways than one.

  10. You don’t need transformers to drive a line with a balanced signal. For years I maintained a bunch of line driver boxes that converted RS-232 to balanced line, and IIRC used a 75113 driver and NE529 receiver chip. I say ‘maintained’ because very time we had a power glitch or lightning storm, I had to replace a dozen of those chips. Other than that, they worked okay.

    The obvious solution to protecting these chips is to use a transformer, AKA magnetics.

  11. This would’ve been better hack if it’s published like 5 years ago. Now you can even get 48port unmanaged switches under 50$ so who cares.
    But before that I would be happy hacking “uplink” between those old 24 port planet switches which required an expensive extension model.

    Realtek suck anyways. That brand always did. Hack Cisco/3Com.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s