Capacitive sensing
A Socratic walk-through of capacitive sensing — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #How does a sheet of glass know where your finger is?
The glass on a phone does not move when you touch it. It does not click, it does not flex enough to matter, and it is an insulator, so nothing electrical passes from your skin into the machine. And yet the machine locates your fingertip to a fraction of a millimetre. Since nothing crossed the glass, whatever the phone measured must have been happening on its own side of the barrier — so what changed over there when your finger arrived?
Reasoning it through
REASONING #Start with the older answer, because it makes the newer one legible by contrast. A resistive touchscreen is two transparent conductive layers held a hair apart by tiny spacers. Press anywhere and they meet at that point; the controller drives a voltage across one layer, reads what the other picks up, and the reading is a position. Notice what that mechanism cares about: pressure, from anything at all — a fingernail, a plastic stylus, a gloved hand. Which is exactly why devices of that era shipped with a stylus, and why signature pads still use them.
Now ask what could be measured without contact, through a solid insulator. The available answer is capacitance. Put two conductors near each other with an insulator between, and they store charge when you apply a voltage — an amount per volt set by their size, their separation, and what surrounds them. Nothing has to touch for that number to exist, and nothing has to touch for it to change. So lay a grid of very fine transparent electrodes on the glass, rows one way and columns the other, and you have a field of capacitances the controller can interrogate.
What does your finger do to them? Your body is largely salt water: a conductor, and a fairly large one, loosely coupled to the surroundings. Bring it near an electrode and you offer the field somewhere new to go. Two arrangements exploit that. In self-capacitance, the controller measures how much charge each electrode holds relative to ground, and your finger adds to it. In mutual capacitance, the controller drives one row and measures how much charge couples across to each column; your finger intercepts part of the fringing field that used to arc between them, so the measured coupling falls.
The second arrangement earns its complexity. Every row-column crossing is measured independently, so two fingers produce two clearly located dips. Self-capacitance measures whole rows and whole columns, so two fingers light two rows and two columns — and the controller cannot tell the true pair of positions from the phantom pair on the other diagonal. Multi-touch is essentially why mutual sensing won.
What comes back from a scan is a small grid of numbers, spaced far more coarsely than the pointing accuracy you experience. The gap is closed arithmetically: a fingertip disturbs several neighbouring nodes at once, in proportion to how close it is to each, so the controller interpolates a centre of mass between them. Resolution comes from the maths, not from the electrode pitch.
And now the everyday failures explain themselves rather than needing separate stories. A glove is an insulator that holds the conductive part of you a millimetre or two further away, and the coupling falls off sharply with distance, so the change never reaches the threshold; touchscreen gloves work by weaving conductive thread through to your skin. A film of water is itself conductive and does not stay in one place — it couples many nodes at once, producing a broad smear with no honest centre, which is why a wet screen registers phantom taps or refuses real ones. A wooden or plastic stylus does nothing at all; a capacitive stylus is a conductive tip electrically continuous with your hand.
One honest caveat: the exact stack varies. Some sensors sit on a separate layer, some are integrated into the display itself, and every manufacturer layers its own filtering and rejection rules on top — which is why palm rejection and wet-finger behaviour differ so much between devices that share the same underlying physics.
The analogy
THE ANALOGY #Picture two people whispering across a room, one talking and one listening, with the listener reporting how loudly the whisper arrives. Someone walking between them absorbs part of it, and the reported loudness drops. The listener never touches the walker and cannot see them, but the dip says a body is there — and with many such pairs criss-crossing the room, comparing which ones went quiet, and by how much, places the body between them.
Sound is muffled by anything solid, whereas the screen only notices what is electrically conductive — a wooden ruler laid on the glass blocks nothing and is simply invisible; and a spilled film of water is less like one walker than a crowd, damping many pairs at once with no single body to locate.
Clarifying the model
THE MODEL #Two refinements are worth making explicit. First, the screen does not detect touch. It detects proximity, and touch is merely the case where proximity is greatest — which is why a hovering finger can be sensed by some hardware, and why pressing harder does nothing at all beyond flattening a little more fingertip against the glass. Second, the phone never measures a position directly. It measures a set of capacitances and computes a position that would explain them, which is what leaves room for the whole business to be fooled: any conductive object of roughly the right size and closeness is, to the controller, indistinguishable from a fingertip.
A picture of it
THE PICTURE #How to readTime runs downward, and the four columns are the controller, one row electrode, one column electrode, and your finger. The first exchange is the untouched case: the controller pulses a row, some charge couples across to the column through the field arcing above the glass, and the reading matches what was stored as normal. The second exchange is the same crossing with a finger nearby — the dashed arrow from the fingertip is not a signal being sent but a share of the field being diverted, so less arrives and the reading dips. The final step is the controller talking to itself, comparing dips at several neighbouring crossings to compute a position finer than the spacing of the grid.
What became clearer
WHAT CLEARED #Nothing crosses the glass, and nothing needs to. The screen watches a grid of tiny capacitances that exist whether or not you are there, and your finger — a chunk of conductive salt water at roughly ground potential — changes them simply by being close. Everything else follows from that one fact: the position is inferred rather than measured, gloves fail because distance kills coupling, water misbehaves because it couples everywhere at once, and a plastic stylus is invisible because the screen was never sensing touch in the first place.
Where to go next
ONWARD #- How proximity sensors, capacitive buttons and liquid-level gauges reuse the same measurement with a different electrode geometry.
- What palm rejection has to decide, and why it is a classification problem rather than a physics one.
Key terms
TERMS #| Term | What it means |
|---|---|
| Capacitance | the charge two conductors store per volt applied, set by their size, their separation, and the material around them. |
| Mutual capacitance | sensing the coupling between a driven row and a sensed column, giving one independent measurement per crossing. |
| Self-capacitance | sensing each electrode's capacitance to ground; simpler, but ambiguous with more than one finger. |
| Fringing field | the part of the field that bulges beyond the electrodes, up through the glass, where a finger can reach it. |
| Resistive touchscreen | the older design in which pressure closes a contact between two conductive layers, so any object that presses works. |
Every term the collection defines is gathered in the glossary.