Saramsha

No. 064  ·  19 November 2026

The Screen Sees Two, Guesses Four

A touchscreen doesn't feel your finger at one point. The wires underneath are spaced wider than a fingertip, so each touch blurs across several of them at once, and the screen has to work out the point in between. Drag the filled dot below, then switch on a second finger and watch the grid lose track of which crossing is which.

actual contact what the screen computes

About this piece

The grid never feels a point. It feels a blur

A capacitive touchscreen is two layers of transparent wires, one running in rows and one in columns, separated by a thin insulator. Your finger is conductive and grounded, so holding it near a wire drains a little of the tiny electric charge that wire is carrying. The controller chip measures that drain on every row and every column, dozens of times a second. Nothing is pressed; nothing moves. That's why a gloved hand, or the blunt plastic tip of an ordinary pen, does nothing — there's no charge path to steal.

The wires are spaced close enough — a few millimetres — that a fingertip, which is wider than that, always sits near more than one of them at once. That's deliberate. A single wire can only tell you "something touched near me," which is only as precise as the gap between wires. But if three or four neighbouring wires all register a bit of drain, weighted roughly by how close each one is, the controller can work out a point between them far more precisely than the spacing itself — the same trick an audio engineer uses to find a sound's pitch between two piano keys from how loudly each one rings. Push the wires apart, in the slider above, until only one lights up at a time, and that precision collapses straight back down to "which wire."

One finger is easy. Two is where it breaks

The simplest version of this sensor, called self-capacitance, scans each row and each column separately and only reports a total per line — it never asks "is row 3 touching column 7 right now," only "is row 3 touching something, and separately, is column 7." With one finger that's plenty: one row lights up, one column lights up, and they cross at exactly one point. With two fingers at once, two rows and two columns light up — but self-capacitance has no way to know whether the real pair was (row A, column A) and (row B, column B), or the other cross-pairing, (row A, column B) and (row B, column A). All four crossings look identical from the totals alone. This is exactly the ambiguity shown above with the toggle switched off: four equally plausible points for two real touches, a classic failure called ghosting, and it's the reason early capacitive touchscreens (and the trackpads before them) could only reliably track one finger.

Mutual capacitance fixes it by measuring every row-column crossing as its own separate sensor, rather than summing whole lines. Each row wire is driven with a tiny signal in turn, and the controller reads how much of it couples through to every column at once — so it directly measures the charge at all 100-or-so crossings on a typical phone screen, not just 20-or-so row-and-column totals. A crossing with no finger near it shows essentially nothing, even if its row and its column are each individually busy elsewhere. That's the toggle in this piece: switching it on doesn't move the dots, it reveals a value that was always true at each crossing — real touches glow, ghosts fade to nothing. Apple's multitouch trackpads and the first iPhone, in 2007, were among the first mainstream products built around scanning every crossing this way, which is why pinch-to-zoom became possible at all.

The grid, the Gaussian "finger blur," and the row/column totals here are a simplified teaching model of self- and mutual-capacitance sensing, not a literal circuit simulation; real controllers also contend with electrical noise, a finger's size and pressure changing the blur's width, and panels with far more than two touch points. Resistive touchscreens — the pre-smartphone kind, which register a physical pinch between two flexible layers — work on an entirely different principle again, by voltage division rather than capacitance, and are left out of this piece.