The LAN Transformer: Signal Path and Isolation

A few weeks ago, I was looking for a clear, beginner-friendly explainer on how Power over Ethernet actually works — not a glossy marketing brochure, not a 100-page IEEE spec, just a straightforward technical walkthrough that starts from the components and builds up.

I couldn't find one.

So I decided to write it myself. And along the way, I figured I'd bring you with me.

This is the first entry in a series where we'll start from the very basics — the discrete components that make up an Ethernet port — and work our way up to complete systems. I'm not here to pretend I know everything. I'm learning too. But I've been designing hardware long enough to know where the gaps are, and I want to fill them — for you, and for myself.

We'll start with the LAN transformer — that little black box sitting between your PHY and the RJ45. You've probably seen it a hundred times. Let's finally figure out what it actually does.
If you've ever drawn an Ethernet port schematic, you know the part I'm talking about.

You have the PHY chip's differential outputs — TD+, TD−, RD+, RD− — and they go through a transformer before hitting the RJ45 connector. On the schematic, it's usually just a box labeled "T1" or "LAN Transformer," with a few windings and a center tap drawn next to it.

That little black box is the LAN transformer.

Most hardware engineers think of it as "for isolation." But pop the hood and you'll see it does a lot more than that — how the signal gets across, how DC is blocked, and how PoE power gets injected — all happen inside that same magnetic core.

This first part covers the fundamental role of the transformer in the signal path. In Part 2, we'll dive into PoE, speed, and selection.

1. Where does it sit?

First, its place in the circuit. It sits right between the PHY and the RJ45.

Signals come out of the PHY's differential pair, go into the transformer's primary winding, couple magnetically to the secondary, then through the RJ45 and out to the twisted pair.

PHY diff pair → primary winding → magnetic coupling → secondary winding → RJ45 → twisted pair

The primary side connects to the PHY ground. The secondary side connects to the chassis ground. There's no wire connecting them — the signal travels through magnetic coupling alone.

The grounds are separate.

That's the first reason it exists: the signal gets through, but the ground doesn't.

2. What's inside?

Open up a typical LAN transformer and you'll find:

A ferrite core – the physical medium for signal coupling and energy transfer
Primary winding – connected to the PHY side, usually with a center tap
Secondary winding – connected to the cable side, may also have a center tap
Insulation between windings – this handles the isolation voltage
In essence, it's an isolation transformer with a center tap.

The primary and secondary windings share the same core, but they're physically insulated from each other. That insulation has to withstand a 1500Vrms hipot test without breaking down. The signal from the primary is "replicated" to the secondary through changing magnetic flux. There's no DC path — only magnetic coupling.

The center tap is a pin brought out from the middle of the winding. It's critical for PoE, so remember where it is.

3. How does the signal get across?

The transformer carries a differential-mode signal.

The PHY's differential output stage drives TD+ and TD− with equal and opposite voltage swings. That differential voltage appears across the primary winding, generating changing flux in the core. The secondary winding picks up the corresponding differential voltage and sends it to the RJ45 and the cable.

Two key points here:

First, only changing signals pass. Magnetic coupling relies on "changing flux induces EMF." DC components don't produce changing flux, so they're blocked at the transformer. That's why it's also called a "DC-blocking transformer."

Second, common-mode signals are naturally rejected. If common-mode noise appears on the cable — both wires moving together in the same direction — the fluxes in the windings add up in the same direction and cancel each other out, so nothing transfers to the other side. In other words, the transformer itself has built-in common-mode rejection. It naturally "recognises" differential signals and "ignores" common-mode ones. That's why in some designs, the transformer can do part of the work that a common-mode choke would do.