Capacitive Touch Is a Finger Changing an Electric Field

The capacitive touch buttons CERN built in the 1970s measured about 200 picofarads each, and a nearby finger raised that by only about 10%, according to Bent Stumpe and Christine Sutton writing in CERN Courier. That small nudge is all a capacitive screen has to go on. Capacitive touch means the screen finds your finger by how it changes the electric field between conductive lines hidden under the glass, with no pressure involved. Anything that changes that field the way a finger does counts as a touch, which is why gloves, water and styluses behave the way they do.

A bare finger and a cathode-ray tube, 1965

Eric Johnson of the Royal Radar Establishment in Malvern, England, published a short letter on a touch display for computers in 1965, and US patent 3,482,241, granted in 1969, lays out the design. Wires sit embedded in the screen in front of a cathode-ray tube, one per site, and the computer relabels them as the task changes. The patent says the circuit uses the operator’s natural capacitance to earth, and it lists a signal from a bare finger as a requirement. Bill Buxton’s history of multi-touch records that touch screens were in use for air traffic control in Britain by the late 1960s, and that Johnson’s was still in service in the late 1990s.

Copper lines too fine to see, 1972

Stumpe proposed CERN’s version in a handwritten note dated 11 March 1972, to control the Super Proton Synchrotron then under construction. His screen was a set of capacitors etched in copper on glass, with lines 80 micrometers wide and 80 micrometers apart so the pattern vanished to the eye, and a lacquer coat kept fingers off the metal. The sensing circuit tracked a phase shift in an oscillator locked to a 120 kilohertz reference, which kept it steady against noise and drift. The control room opened in 1976 with these screens, and some ran for about 30 years. The authors hedge that this was apparently the first application of a capacitive touch screen anywhere, though Johnson’s air traffic system came earlier.

Put CERN’s numbers next to a modern figure and the scale shifts. Atmel’s John Carey wrote in a 2009 Embedded.com article that a finger’s effect is on the order of one picofarad, against background capacitances of tens of nanofarads. One picofarad over ten nanofarads is one part in 10,000, so a projected capacitive panel of the kind in touchscreen phones hunts for a change at least 1,000 times smaller, as a fraction, than CERN’s 10%. These are different designs, a 16-button panel and a screen full of electrodes, so read the comparison as two data points, not a trend.

From one touch to ten

Johnson’s and Stumpe’s screens read one touch at a time. Buxton credits Bob Boie at Bell Labs with a transparent capacitive array over a cathode-ray tube in 1984, and a University of Toronto capacitive tablet in 1985 that reported many touches at once. FingerWorks, which grew out of Wayne Westerman’s doctoral research, shipped capacitive multi-touch pads in 2002. Apple bought the company in early 2005, and Buxton links that work to the 2007 iPhone.

The grid designs are called projected capacitive, because the sensor film only projects an electric field that a controller chip drives, Carey explains. The word “capacitive” also hides a quirk. According to Microchip’s design note AN2934, a self-capacitance measurement sees a finger as added capacitance, while a mutual-capacitance measurement, which watches the coupling between a row and a column, usually sees a reduction. Mutual capacitance reports each touch as its own coordinate pair, while self capacitance can confuse two fingers into two possible pairs, which Carey calls ghosting. The same Microchip note uses capacitive touch for buttons, sliders and wheels as well as screens.

A stylus has to look like a finger

A passive stylus is a piece of conductive material that works as an extension of your hand, write two Cypress Semiconductor engineers in a 2013 eeNews Europe article, and the screen has no easy way to tell it from a finger. Size is the problem. They approximate signal as proportional to contact area, so a 2 millimeter tip gives about 25 times less signal than a typical 10 millimeter touch, since (10 ÷ 2)² is 25. A tip that small can also fit inside one 5 millimeter sensor cell, and its position then reads as stuck at the cell’s center until it crosses into the next. An active stylus carries electronics and a power source and transmits to the host, which allows hover, pressure sensing, buttons and an eraser.

What a glove does to a picofarad

A glove adds distance, and distance costs signal. Microchip’s design note models a fingertip over a sensor as one plate of a capacitor, the electrode as the other, and the cover as the gap, with the cover thickness in the denominator. Its worked example, an 8 millimeter fingertip over a 12 millimeter electrode under a 1 millimeter plastic cover with a relative permittivity of 2, comes to about 0.89 picofarads. Changing only the gap gives this:

I computed the rows from Microchip’s formula, keeping the same fingertip and permittivity, so a real glove material will differ. The note itself warns that the approximation loses accuracy when the plate is not much larger than the gap, so read the lower rows as a rough shape. Even so, going from a 1 millimeter gap to 5 leaves one fifth of the signal.

Controllers answer with extra measurements. Microchip’s 2019 datasheet for one automotive controller says self capacitance can detect single touches through thick gloves that mutual capacitance alone may miss, and it lists single-finger glove touches up to 5 millimeters thick, subject to the sensor stack-up. That is a vendor specification for one chip, and a phone’s tuning will differ.

When water counts as a touch

In Microchip’s model of a single sensor node, a water droplet raises the coupling between the two electrodes. That reads as a negative “anti-touch,” because the droplet gives the field no return path to ground. If the droplet also reaches another conductor, such as a neighboring circuit element, the same note says the result can tip toward a false touch. The datasheet lists no false touch with condensation or a water drop up to 22 millimeters across, and says the controller keeps tracking one finger through moisture by switching to self capacitance once it detects water. Both are vendor claims. Samsung’s support page for touchscreen problems says a wet screen can make touch misbehave and that even waterproof models can have limited touch in water.

Samsung’s page also advises clean, dry hands and avoiding non-conductive objects such as a ballpoint pen. Dry the glass and your fingertip, because water that bridges to a grounded conductor can read as a touch, and for gloves choose the thinnest pair you can, or any glove or tool that gives a conductive path from your skin to a fingertip-sized patch, which is the same idea as a passive stylus.