Sixty-four tiny ferrite rings on a grid of wires, the way computers stored their bits from the 1950s into the 1970s. Tap a ring to flip its bit. Each wire carries only half the current needed to switch a ring, so only the ring where the two energised wires cross gets enough to flip.
For kids: Bit Blocks tells the same story with blocks and water, in English and German →
A ferrite core has a nearly square hysteresis loop. Below a sharp threshold the field barely disturbs it; above it, the whole ring snaps to the other direction of magnetisation, and it stays there with no power.
The controller sends half the switching current down one X wire and half down one Y wire. Fourteen cores feel one half and stay put. Only the core where the two wires cross gets the full current and flips. This is called coincident-current selection.
A core cannot be asked which way it is magnetised. To read it, the controller tries to write a 0. If the core held a 1, it flips and induces a short voltage pulse in the sense wire, which runs through every core. If it held a 0, nothing happens and the sense wire stays quiet.
The bit is known now, but the core holds 0. So every read is followed by a write that puts the bit back. Untick write back after read and use Read all to see what happens without it.
Every write starts by clearing the core to 0. Then the controller drives X and Y towards 1. To store a 0 instead, the inhibit wire carries half a current in the opposite direction, so the chosen core only gets half and stays 0.
In a real machine there was one plane per bit of the word. A whole word was read and written at once, all its bits sharing the same X and Y wires. Here one plane is shown, and each row holds one byte (most significant bit on the left).