0
\$\begingroup\$

Many point-to-point communication protocols are designed to have an electrical interface with data/clock/whatever signals along with the ground signal for reference. For example, CAN is composed of +Hi, -Low, and ground. Minimum RS232 usage is TX/RX/Ground, its balanced counter-part RS485 A+/B-/GND. With exception of unbalanced interfaces, in which the GND is obviously needed to reference the signal, I've always seen the GND as needed to equalize the potentials between both ends, such that the potential difference between the equipment could be kept in safe regions. I wonder why such problem doesn't occur in ethernet (or maybe it does), where there's no GND. Here's the problem I see:

schematic

simulate this circuit – Schematic created using CircuitLab

The incoming signal TX_D1+/TX_D1- arrives and is connected to an ethernet transceiver through a transformer XFMR1. As there's no GND, an unknown potential difference exists, for instance, between TX_D1- and GND, as represented by Vpot. Well, how one guarantees that a large Vpot won't damage the transformer?

\$\endgroup\$
  • 1
    \$\begingroup\$ Your schematic is wrong. Ethernet mags are typically grounded at a center tap through a 75Ω resistor. \$\endgroup\$ – Janka May 21 '18 at 20:25
  • \$\begingroup\$ have a look at some of the ethernet magnetics data sheets ..... ethernet transformers have something like 1500V isolation \$\endgroup\$ – jsotola May 21 '18 at 20:33
  • 2
    \$\begingroup\$ @Janka You're correct, still this tap is also not grounded, see this implementation: e2e.ti.com/support/interface/ethernet/f/903/t/… \$\endgroup\$ – PDuarte May 21 '18 at 20:37
  • 1
    \$\begingroup\$ how one guarantees that a large Vpot won't damage the transformer? Of course a too large Vpot will damage the transformer. Question is, how would that high voltage get on the line in the first place? In normal use there should be no build up of voltage on the lines. And any noise or static picked up would leak to ground via the capacitors and resistors to ground. \$\endgroup\$ – Bimpelrekkie May 21 '18 at 20:46
  • 2
    \$\begingroup\$ i doubt that static electricity will damage a transformer (unless it is a lightning of course) ... there is insufficient energy \$\endgroup\$ – jsotola May 21 '18 at 20:47
1
\$\begingroup\$

The incoming signal TX_D1+/TX_D1- arrives and is connected to an ethernet transceiver through a transformer XFMR1. As there's no GND, an unknown potential difference exists, for instance, between TX_D1- and GND, as represented by Vpot. Well, how one guarantees that a large Vpot won't damage the transformer?

Your system design specification should include a maximum common mode voltage on the ethernet line, which you refer to as Vpot. With this in hand, the transformer can be specified and designed. Common mode voltage will dictate the insulation requirements for primary-secondary and for the cable winding to core. In your comment you mentioned greater than 1.5kV. Nail this number down before selecting the transformer. This is how one guarantees that a large Vpot will not damage the transformer.

\$\endgroup\$
  • \$\begingroup\$ Good insights! Do you see any other concern that might raise with this non-GND fashion interface? \$\endgroup\$ – PDuarte May 21 '18 at 23:10
  • \$\begingroup\$ There is typically a common mode choke between the RJ45 and the transformer winding. While on the topic, a year or two ago Microchip published an app note describing capacitor instead of transformer coupling their PHY interface chip, which might be of interest. ww1.microchip.com/downloads/en/AppNotes/00002190A.pdf \$\endgroup\$ – AlmostDone May 22 '18 at 1:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.