2

The standard for ethernet says you should use a transformer for isolation.

It seems like a cap-based solution would be smaller, more versatile, and a better fit for my application.

Is there any reason why a (capacitor-based) highpass filter couldn't be used instead?

EDIT: This question is different from others because it references a specific alternative, rather than asking the generic purpose of ethernet magnetics.

Tustique
  • 559
  • 4
  • 17
  • Would the filter you are proposing eliminate the ground reference the way galvanic isolation does? – K H Dec 06 '18 at 05:55
  • It is often used in large computer backplanes for inter-board communications. This app note may help you: http://ww1.microchip.com/downloads/en/AppNotes/Capacitive%20Coupling%20Ethernet%20Transceivers%20without%20Using%20Transformers.pdf – Jack Creasey Dec 06 '18 at 05:56
  • @KH I expect it to. – Tustique Dec 06 '18 at 06:09
  • @JackCreasey that app note (and others I've come across) seems to specifically reference 10/100 ethernet. Do you know of any app notes or docs about using this with 10/100/1000 (gigabit) networking? – Tustique Dec 06 '18 at 06:29
  • Cap based isolation works, though Ti app notes state that DC bias issues occur. You can probably account for it if you control both sides of the link. http://www.ti.com/lit/an/snla088a/snla088a.pdf – Zekhariah Dec 06 '18 at 07:25
  • @Tustique Cap coupling does not work for Gigabit Ethernet. – Jack Creasey Dec 06 '18 at 18:09

1 Answers1

1

The Ethernet transformer, whether a separate component or integrated within a "mag jack" allows for complete isolation between the power and ground domain of the Ethernet controller and the cabling domain. Same thing on the other end of the cable.

Using capacitor coupling will typically mean that there is some GND sharing between subsystems or ends of the Ethernet cabling. Unfortunately the pinning of an Ethernet RJ45 jack does not have provision for a common GND connection across the Ethernet cable.

Michael Karas
  • 56,889
  • 3
  • 70
  • 138
  • Why couldn't two devices be completely isolated capacitively instead? (e.g. by placing series capacitors inline with the signal lanes) – Tustique Dec 06 '18 at 06:16
  • 1
    @Tustique Although capacitors have no direct connection between the plates and DC is blocked, **AC is not**. If you would hold one lead of a 100 nF capacitor in your hand and with the other lead, poke in a mains socket, you are very likely to receive an electric shock. With a transformer that doesn't happen as it has full isolation as it converts the energy into a magnetic field and back. The electric paths are fully separated. With a capacitor one plate can still make current flow in the other plate. – Bimpelrekkie Dec 06 '18 at 07:24
  • @Bimpelrekkie that could happen with a transformer too, though. Presumably ethernet transformers are designed to block the low frequencies of 50/60 Hz, and a filter could be too, correct? – Tustique Dec 06 '18 at 07:32
  • 1
    @Tustique That could only happen if the transformer is **badly designed**, a proper transformer will have very little capacitance between the windings. Indeed it is there but "good enough". For a capacitor only very small values will be "good enough" and those small values might block your signal too much. The ethernet transformers also play in a role in **commonmode suppression**. An RLC filter cannot do that as effectively. – Bimpelrekkie Dec 06 '18 at 07:38
  • Basically, with magnetics, the voltage between machine A and machine B look like common mode voltage at the isolator. – Toybuilder Dec 02 '19 at 01:05