Jump to page content

Capacitive sensing

Contents

Overview

Capacitive keyboards use electrical capacitance instead of conduction to detect keystrokes. Instead of the switch passing current when actuated, each switch is a variable capacitor, with (depending on the design) an increase or decrease in capacitance indicating switch operation.

With conventional matrix scan capacitive keyboards the sensing takes place directly under the plunger. This is where the return spring would be in many other switch types. As a consequence, capacitive keyboards tend to have the return spring or buckling rubber sleeve on the outside, underneath the keycap. Topre’s conical spring method is unusual in allowing the rubber dome to be placed safely inside each key.

The majority of capacitive keyboards are matrix scanned, with centralised electronics that query each key in sequence. However, there were also designs where each key had its own dedicated capacitance sensor. The various implementations are listed where possible in the approximate order in which they were invented.

Reliability

Regarding the IBM Model F, inventor Richard Hunter Harris stated:

The capacitive sensing signal is generally proportional to 1/d where d is the thickness of the gap between the pivot plate (flipper) and the copper surface. With very small gaps signal differences can vary significantly with gap changes. This makes the design very sensitive to dust, particulate contamination and part tolerances that effect this gap. This was likely the main exposure that lead to Keyboard M.

Clear details are difficult to obtain about switch reliability, and why keyboards seemingly only detected keystrokes when the movable element was in contact with the PCB. There is no such limitation with Topre keyboards.

Selective shielding

Around 1967, IKOR introduced an AC coupling arrangement between a bank of emitter pads and corresponding PCB tracks (or so the patent appears to indicate). Selective shielding between the emitter and receiver lines allowed for each output bit to be defined separately, making this a self-encoding technique. Unlike other capacitive sensing techniques, there is no variable capacitor involved: the capacitance emitter and receiver are permanently fixed in position and operate at a fixed capacitance. Every output bit has its own separate AC emitter and receiver, with each receiver coupled capacitively to its emitter. A metal shield within the key bears apertures that allow the electromagnetic field to pass for each output bit that should be enabled. In its inactive state, the apertures in the shield do not line up with emitter–receiver pairs, and no output bits are active. As the key is pressed, the shield is lowered, allowing the AC signal to pass through as the apertures in the shield align with the emitter–receiver pairs. In IKOR’s design, there are two facing sensing arrays to provide the required number of output bits.

This implementation was vertically bulky, even with the apertures and holes placed into two columns, and the design did not seem to last for many years. It may be possible to achieve a non-encoding version where each key has a small shield that only enables a single bit. This would reduce the keyboard’s profile significantly.

Capacitance-controlled transistor

Around 1970, Control Devices invented a method for detecting capacitance by connecting the switch’s variable capacitor to the base of a transistor. Unlike the movable shield method, this approach uses each key as a variable capacitor just as with the matrix-scanned types. Each key has its own dedicated capacitance sensor, comprising a transistor, a capacitor and two resistors. Control Devices arranged for each key to be connected to a diode matrix for output encoding. The per-key portion of this circuit is shown below, taken from the Control Devices patent:

The fixed capacitor is held charged, and this keeps the diode matrix columns high. When the key is pressed, an AC signal passes from the oscillator into the key’s transistor, and this drains the key’s capacitor, grounding the diode matrix columns. There is no matrix and no scanner, and thus no way to determine the co-ordinates of an individual key. Each key must therefore be connected to a means of providing its identity. One option is two-of-N coding. Another option is a diode matrix, as shown above.

In reality, the circuit was more complex than the diagram shown above. The diode matrix output required amplification, and to deal with the individual columns triggering at different times, synchronisation was required to withhold the output code until all the columns had activated and settled. This method was considered low-cost at the time, with a per-keyboard price of under $100, or under $600 in February 2021 prices.

A similar arrangement can be found in the operator’s console for the Monarch 120 Call Collect System; this panel was manufactured by Pye Electro-Devices. Here, only a single resistor is used per key. The technical description handbook for the Monarch 120B notes:

The operator’s keypad is driven by a 100 kHz oscillator. When the operator’s finger is inserted in the key depression, a capacitive coupling to earth of about 1 pF is applied (about 1 pF at the point of application of the finger in series with, say, 100 pF to 150 pF body capacity to earth).

This is sufficient to cause a shift in DC output level from a transistor circuit driven by the oscillator and apply a DC marking condition on a lead corresponding to the key. The circuit elements and the way in which the keypad is connected into the system are shown in Figure 55.

The per-key circuit per the Monarch 120B Compact Call Collect System Technical Description Part 2 (page 106) is as follows:

The encoding process is again diode-based, but instead of producing output codes directly, it produces a form of two-of-N code that is fed to a separate circuit. (There is some disagreement within the documentation as to whether the oscillator ran at 100 or 110 kHz.)

The indication is that PED only produced touch-contact panels and keyboards with this technology, i.e. they were not full-travel switches: each key station responds purely to the operator’s fingers, with no moving parts. Creation of full-travel switches using this circuit appears to be possible however.

Metal dome

Metal dome capacitive keyboards were introduced by Colorado Instruments in 1970. These keyboards used a snap-action metal dome as the movable portion of the variable capacitor, providing a click sound and feel in direct conjunction with actuation. Example keyboards are depicted in magazine articles, but these are yet to be seen. No model numbers or product names are yet discovered. US patent 3653038 “Capacitive electric signal device and keyboard using said device” was filed in February 1970 for this design.

Matrix scan

The majority of capacitive keyboards are matrix scan, just as with conductive keyboards. Matrix scan allows the capacitance measurement electronics to be centralised. All that is required at each key position is a variable capacitor operated by the key. The control circuitry activates each key in turn (or each row in turn) and measures the capacitance to identify which keys are pressed. In the most basic sense, the circuit can be thought of as shown in the diagram below:

Although the diagram above is effectively how Key Tronic portrayed the circuit, in practice, the circuitry is typically more complicated. Some designs, including those of Key Tronic and BTC, run a separate ground track between the capacitive pads on the PCB to limit the idle capacitance between the pads of inactive keys. The diagram above represents the ideal condition, but adaptations are required for reliable sensing.

Several switch designs have been devised for capacitive matrix scan. In most cases, the foam pad approach has been used. An example foam pad switch is shown below, based on Key Tronic’s DIN-compliant implementation:

Most matrix scan keyboards are built around a printed circuit board, although membrane-based types also exist. Such PCB-based keyboards generally involve a movable element that forms one side of each of two variable capacitors. In this arrangement, the movable element is moved either closer to, or further away from, the PCB when the key is pressed. A pair of pads on the PCB provide the stationary plates of the variable capacitors, and collectively these parts behave as though each key has a single variable capacitor. In the example above, the movable element is a piece of metal foil, coated with plastic to prevent it from shorting out the pads on the PCB. (Alternatively, or additionally, the pads on the PCB may be coated with solder mask or—in Univac magnetic separation keyboards—a Kapton sheet for insulation.) The + and − symbols in the diagram show the polarity of the charges formed, although in the diagram there would be little to no charge because the key is in its released position. As showing the key in the actuated position is too difficult (the foam pad would cover over the PCB pads), the charges are shown where they would form during actuation.

dmaone@github identified five capacitances in such a switch:

These capacitances are illustrated in the diagram below:

C1 and C2 operate together as the variable capacitance of the key.

Pressing a key typically increases its capacitance, although in some designs (e.g. IBM beam spring, Digitran’s DIN-compliant KD series) the capacitance is reduced upon actuation, with the movable element pushed away from the PCB instead of towards it.

A naïvely-designed capacitive matrix is subject to ghost current during capacitor charge and discharge in exactly the same manner as a conductive matrix without diodes. The difference with a capacitive matrix is that a ghost pathway passes through three variable capacitors instead of three conductive switches. Chaining capacitors reduces the effective capacitance rather than increase it, so the ghost keys would register less charge and could be ignored on a threshold basis. However, if inactive rows—those that are not being scanned—are grounded by the driver circuit, the ghost keys do not charge and are not detected. Consequently, capacitive keyboards have N-key rollover.

The specific details of the various capacitive scan processes and implementations is far beyond the scope of this page, and requires considerable background in electronics to understand.

Various matrix scan capacitive sense switch designs are described briefly below.

Conductive plastic

Some capacitive designs use a conductive plastic block as the variable element. IBM’s beam spring and Model F designs both take this form. The US patent application for IBM’s beam spring keyboard appears to have never been granted, but the overseas patents reference the 1971 application.

Foam pad

Foam pad—or “foam and foil”—switches have a foam pad fitted to the bottom of the plunger, affixed onto which is a thin layer of metallic foil. The distance between foil layer and the PCB affects the capacitance registered at that key position. The compressible foam pad allows the plunger to continue being pressed past the point that the foil layer reaches the PCB, to provide overtravel. As such, the complete assembly will not operate without the use of an extra moving part (the foam pad) beyond the plunger itself. Worse, the foam pads are a weakness in the design: as foam can stiffen with age and lose its flexibility, some older keyboards fail due to the foam pad failing to re-expand after a key is struck.

US patent 3965399 “Pushbutton capacitive transducer” filed in March 1974 indicates that the reason for the flexible foil coating on a foam pad is so that the foil coating can be pressed uniformly against the PCB:

A substantial change in capacitance between the members is effected when the thin dielectric material is disposed across a gap between the capacitance forming member. In order to provide a consistent change in capacitance, a gap of uniform width is preferably maintained. This is provided by the use of a resilient backing which causes the flexible metallic and dielectric films to conform to surface irregularities of the capacitance forming member and its two elements.

The foam pad can also be referred to as an “overtravel” pad: it permits the key to be pressed further after the foil layer makes contact with the PCB. The foam pad design, just like the conductive plastic approaches, appear to require the movable object to rest against the PCB for detection to occur. Topre’s conical spring method is a full analogue sensing method that allows for firmware-configurable pretravel and electronically-defined hysteresis.

Foam pad keyboard manufacturers include:

Digitran amusingly referred to foam pad keyboards as “sponge-on-a-stick” in an advertisment in Computer Design in January 1979; this was a slight against foam pad designs, as compared to their Golden Touch leaf spring design.

Foil-in-dome

Foil-in-dome switches place the foil disc inside a flat-topped rubber dome. This arrangement is best known from Brother typewriters and keyboards.

Leaf spring

Leaf spring capacitive sensing uses a metal leaf spring that is pushed down onto the PCB. One side of the leaf spring is attached to the PCB, and the plate hinges down towards the PCB under pressure from the plunger. Overtravel is achieved by a special prong on the leaf spring. The best-known leaf spring capacitive design is that of Digitran’s Golden Touch. Cortron filed a patent for their own design, but in both the one known CP-4550 keyboard and the CP-4550 advertisements, the design is the same as Digitran’s.

The exact date of introduction of leaf spring capacitive keyboards is unclear. CP-4550 appeared around 1981. Golden Touch keyboards have been found as far back as 1976, and advertised only as far back as 1978.

Membrane

Although the vast majority of membrane keyboards use conductive sensing, some manufacturers produced conductive membrane keyboards. Manufacturers of capacitive membrane keyboards include Micro Switch with SC and ST Series. Micro Switch’s approach appears to date back to around 1981.

The Hewlett-Packard C1405A keyboard used a capacitive membrane system with membranes supplied by W.H. Brady Co. who held the patent on the sensing process.

Conical spring

Conical spring switching uses a small, low-force conical spring placed under a rubber dome. As the key is pressed, the rubber dome compresses the conical spring. Topre have made such keyboards since 1983, and continue to produce them to this day. They introduced APC—Actuation Point Changer—to permit the pretravel of the keyboard to be set to any of 1.5 mm, 2.2 mm or 3 mm; this is made possible by the inherently analogue nature of capacitive sensing. See the Realforce website for details and illustrations of Topre keyboards.

Originally a design exclusive to Topre, in recent years this design has been copied by other brands such as Noppoo. Owing to the secretive nature of Far East manufacturing, the specific details of who manufactures what are intentionally not revealed by the companies involved.