Toroids as Communicators · Paper II — The Crossing

No wire between
the two sides — only
the shared flux

Wind a second coil on the same ring and you have a transformer. The first side and the second side never touch — no copper joins them — yet a message crosses, carried only by the flux they share. Paper I said a toroid passes edges, not levels. Now watch those edges leap the gap, and a PWM message arrive on a side that's electrically a world away.

v = (N/N) · v
the turns ratio scales the voltage · but the DC never crosses — only the change does
§5 · instrument three

Two windings — isolation by shared flux

A transformer is just Paper I's core with a second winding added. Drive a changing current in the primary; it makes a changing flux in the shared ring; that changing flux induces a voltage in the secondary — Faraday again, vₛ = Nₛ·dΦ/dt. The ratio of turns sets the ratio of voltages: more secondary turns, more secondary volts (and proportionally less current — power is conserved, not created).

The quiet miracle is what isn't there: no electrical connection between the two sides. The only thing they share is flux in iron. That's galvanic isolation — the secondary can float at a thousand volts above the primary and the signal still crosses, because the signal was never a current in a wire; it was a change in a field. And because it rides on change, a steady primary current induces nothing across the gap.

TWO WINDINGS · primary → shared flux → secondary · DC blocked
PWM (edges)
sine
steady DC
primary drives shared flux; the secondary answers the change, scaled by turns ratio. switch to steady DC and the secondary goes dead — no change, no crossing. the two coils never touch.
honest flagIdealised again: perfect coupling (all flux links both coils), no leakage inductance, no winding resistance, no saturation. Real transformers leak some flux, drop a little voltage, and a true square-edge rings on the leakage/parasitics. The turns ratio and isolation shown here are exact; the parasitics that limit real bandwidth and edge fidelity are what the engineering adds on top.

§6 · instrument four · the signature

The isolated PWM link — send the edges, rebuild the width

Here is the whole point of the series, working end to end. You want to send a value — a duty cycle — from one side to a side it can't touch. You can't just push the PWM across: the transformer refuses the DC, so the secondary comes out AC-coupled — the same edges, but the levels shifted to swing about zero (precisely the volt-second balance from Paper I; the average across the core must be zero). The duty cycle looks lost.

It isn't. The duty cycle was never stored in a level — it lives in the timing of the edges, and the edges crossed perfectly. So the far side reconstructs: a comparator catches each edge and rebuilds a clean full-height pulse — high on the rising edge, low on the falling — recreating the original PWM, duty intact. A low-pass filter then averages it back to the value. The message made it across a gap with no wire, riding entirely on change.

ISOLATED PWM LINK · primary PWM → AC-coupled secondary → reconstructed → recovered
row 1 primary PWM (your message) · row 2 secondary — DC stripped, swinging about zero, edges intact · row 3 reconstructed PWM on the far side · row 4 recovered value. the duty you sent arrives, across the isolation barrier.

The DC could not cross — but the DC was never the message. The message was the edges, and the edges always cross. That is why a toroid speaks PWM.

honest flagThis is the real principle behind pulse-transformer gate drive and edge-based digital isolators, simplified to its spine. Real links handle the long-pulse problem (a very high or very low duty starves the transformer of edges, so practical designs use a high-frequency carrier, edge-encoding, or periodic refresh), add dead-time and noise margins, and reconstruct with real comparators/latches that have delay. The core idea here — edge timing survives, levels don't, so rebuild from edges — is exactly how it's done.

§7 · in the wild

Where this is already running

This isn't a toy demonstration of a principle — it's the working guts of equipment you use constantly. Every one of these sends information through a tiny toroid (or a transformer very like one), across an isolation barrier, on the edges:

Ethernet magnetics
Every wired network port hides little transformers. Your data crosses them isolated — which is why a lightning surge on the cable doesn't fry the chip.
Gate-drive transformers
Power electronics switch high-voltage transistors using PWM pulses sent through a pulse transformer — the control side floats safely below the lethal side.
Digital isolators
Chips that pass logic signals across an on-die transformer, edge by edge, so two circuits at wildly different voltages can still talk.

In each, the rule from Paper I holds without exception: a steady level would die in the core, so the data is sent as edges and rebuilt on the far side. The toroid is the wall and the messenger at once.

One instrument left to close the series: the common-mode choke — the toroid wound so it ignores the noise both wires share and passes only the difference that carries the message. Selective listening, magnetic edition.

TOROIDS AS COMMUNICATORS · PAPER II — THE CROSSING
vₛ = (Nₛ/Nₚ)·vₚ · turns ratio scales voltage · the DC never crosses, only the change
secondary is AC-coupled (volt-second balance) · duty rebuilt from edge timing · checked in node
no wire between the sides — the edges leap the gap, and the message arrives