Switch design characteristics
The following are only brief notes. At present I do not have the technical data to assess these in more detail. These are simply notes to give a brief introduction to alternatives.
For individual switch functions such as alternate action, see switch operating functions.
The vast majority of keyboards can be divided into two categories: conductive and contactless. The majority of keyboards in existence rely on some means of physically switching electrical current to detect keypresses; this is true of most rubber dome and membrane keyboards, as well as mechanical keyboards. Contactless keyboards are still electronic, but no physical switching occurs.
Most keyboards function as an arrangement of pushbutton switches. The switch contacts may be solid metal, or they can be screen printed onto thin plastic sheets. Such keyboards all share a particular characteristic: they cannot switch current cleanly. Physical switch contacts (be they sprung metal or plastic membranes) will rebound after they make contact, and it takes a small fraction of a second for the contacts to settle into their closed position. This behaviour is described as “contact bounce”, and it typically lasts no more than 20 milliseconds, with reed switches coming in around 2 milliseconds with their smaller contacts and subtle and gentle contact operation.
Without special processing, contact bounce will cause a single keystroke to be detected as a series of brief pulses before the switch remains in closed position until it is intentionally released. “De-bouncing” is the process of dealing with contact bounce such that the keyboard, keyboard or control panel only reports when a switch is actuated after the contact bounce has completed. Contact bounce has not generally posed a problem in decades in correctly-designed products, but as switches age or fail, the degree of bounce can exceed design tolerances and cause a single keystroke to register multiple times in rapid succession. This failure mode is referred to as “chatter”, although in Japanese that term (spelt as “チャッタ”) refers to bounce.
By 1971, Cherry had introduced a keyboard design with electronic circuitry capable of handling de-bouncing. However, in the 1960s and even into the 1970s, contact bounce was considered a significant challenge when using logic circuitry (such as transistor-transistor logic, or TTL). This led keyboard manufacturers to find ways to detect keystrokes without the need for physical switch contacts. Micro Switch adopted reed switches for keyboards in 1966 (with KB), but shortly afterwards in 1968 they introduced solid-state contactless switches based on Hall effect (SW series). Licon followed with ferrite core inductive switching by 1970, and RAFI introduced their magnetoresistive solid-state switches around the same time. Mechanical Enterprises opted to stay with conductive switching to keep costs down, but replaced the solid metal contacts with a flexible mercury-filled tube that would be pinched closed when the switch was released; their Mercutronic keyboard switches appear to have been on the market in 1969.
Micro Switch PB series included pushbutton switch units with integrated contact buffers: flip-flop–based circuits paired with NO+NC switches to render the contact bounce undetectable. Such an approach will only work with switches that have both normally-open and normally-closed contacts.
Encoding switches self-encode their identity, such as the ASCII code representing the character typed. Such switches are wired into a bus with one line per bit of the output code.
Switches may be illuminated for one of several reasons. This is often to provide status indication, such as for the caps lock key. Keys can also be backlit by the switches.
The source of illumination was originally an incandescent lamp, which by the 1980s were largely replaced by LEDs. A lamp placed in the centre of the switch is good for lighting the entire keycap, while a lamp placed in the corner can be used to light a window in the keycap. Generally a switch series would only offer corner or centre lighting, but Cherry offered both options between their US and German factories, with Cherry USA providing corner-lit M4 switches and Cherry Germany providing centre-lit M71 switches.
With a static lamp, the illumination brightness can vary by key travel. To avoid this, some switches place the lamp directly into the plunger, and the lamp moves with the key. Omron’s B2H and B2R switches offered this functionality, as did Alps KCC and Clare-Pendar S950, as well as Fujitsu’s taller reed switches. Such designs are more complex and often more fragile, as the lamp or LED requires sprung contacts.
Some switches offered two lamp positions: this is true of both Cherry M4 (one or two lamp positions depending on model) and Hi-Tek Series 725.
Buckling rubber sleeves
“Buckling rubber sleeves” are discrete rubber domes with a hole in the top, which sit around the keystem and add tactility to the switch. Buckling rubber sleeves are most common in contactless switch systems where it is physically impossible to have inherent tactility in the design. Inverted sleeves are also widely recognised from their usage in Mitsumi’s KKQ and KPQ low-profile membrane types.
US patent 3767022 from Singer Co depicts a buckling rubber sleeve design as early as 1970. This was used with their optoelectronic switch system, which as a contactless system has no means of offering tactility of its own.
The shell, body or housing is the enclosure of the switch module, that guides and retains the plunger, spring and switch contacts. This is typically one or two pieces. One-piece types can be open at the bottom (in particular Micro Switch SW), or the plunger can occupy the entire top of the switch, as with Hi-Tek Series 725.
More commonly, the shell comprises a lower piece and an upper piece. The relative proportions vary; first-generation Futaba switches typically have a base cap that close off the bottom, while Mechanical Enterprises T-5 has a small clip-in retainer for the plunger.
Modern switches generally snap together, but it was quite common in the past for switches to be ultrasonically welded together. This was the case with Micro Switch SN (the fully-enclosed equivalent to SW), Datanetics DC-50 and Cherry M4, M5, M6 and M7. The rationale for ultrasonic welding is not known. This process required specialist equipment, and it was not without problems. Peter Cherry and Günter Murmann (both formerly of Cherry) reported that small pieces of plastic could break off (possibly sputter) and these could wedge between the switch contacts and render the switch inoperative. This was at least one of the factors for Cherry to switch to snap-together shells. Ultrasonic welding, along with heat staking, has a long-term disadvantage of making switches unnecessarily difficult to service in ageing equipment, which is a particular problem when the supply of spare switches has dried up.
Ultrasonic welding and heat staking fell out of favour over time, with a shift towards snap-together shells. This was not done to improve the ability to repair switches, as they were cheap enough to replace outright. Rather, snap-together shells are a more cost-effective assembly technique, and the keyboard industry has for most if not all of its life faced pressure to reduce costs.
Discrete-switch keyboards require some means of securing each switch in place. Important goals include ensuring precise key alignment so that the keycaps are all lined up correctly, protecting the switch terminals from excessive force (especially where they are formed from thin, flexible metal), holding the switches in place during the soldering process, and ensuring that the forces involved in typing (and in some cases keycap removal) do not crack or sever the solder joints to the PCB. Not all switches provide for each of these objectives.
In the 1960s, it was not uncommon for keyboards with discrete switches to have the switches held only by the PCB. Raytheon’s KBSM, KBSR and KBFR switches and Cherry’s reed keyboard switches had specially-shaped terminals that would lock into the PCB and hold the switch. Some other manufacturers produced switches that could be screwed to the PCB. This included Datanetics DC-50, introduced in 1973, where each switch was attached separately. Key Tronic’s and Maxi-Switch’s ganged switch modules were also affixed with screws, as were the little retainer covers in Maxi-Switch’s ultra-low-profile dome units. Micro Switch SD also permitted screw attachment, but a switch secured in this matter has yet to be observed.
Another 1960s approach was Micro Switch KB, where the switches are mounted to metal rails that run the width of the keyboard, a practice shared by other Micro Switch series. Notches in the rail combined with lugs on the switches guarantee correct spacing and prevent the switches from sliding along the rails. This design was copied by RAFI and seen with their magnetoresistive and RC 72 keyboards introduced in the early 1970s. Alps Electric used their own mounting bar arrangement in their reed switch keyboards.
With SW Series, which was introduced in 1968, Micro Switch decided to attach each switch using a pair of thin sheet metal loops formed from a single pressing spanning multiple keys (often a complete row); these pressings were riveted to the PCB. The use of these mounting frames supposedly helped prevent keystroke impact from being felt by the PCB (which would jeopardise the solder joints) but how this is possible is not known.
Of those designs that attach directly to the PCB, it is common for them to have fixing pins moulded into the base. These are studs that fit through holes in the PCB. These pins can grip the PCB prior to soldering, and they prevent the switch from twisting. An alternative approach used especially with the smaller Alps, Cherry and Mitsumi switches, is additional solder points. This can be by way of an internal jumper, or in the case of Alps SKFL and the miniature Mitsumi switches, pins formed from the metal lid of the switch.
Micro Switch SN Series switches have one of the most unconventional methods: each switch is fitted with a self-adhesive strip that secures the switch onto the PCB. These switches are intended for control panels rather than keyboards, but they have been found in the occasional industrial keyboard.
From the start of the 1970s, the idea of plate mounting took hold, with many manufacturers adopting it. Here, each switch snaps into a hole stamped into a sturdy sheet of steel or (in rare cases) aluminium. A pair of plastic clips moulded into the shell of each switch retains the switch in the mounting plate; by releasing these clips, the switch can be removed. Keystroke impact is isolated from the PCB, and the overall assembly process is significantly simpler than the alternatives. The plate also serves the purpose of maintaining precise keycap orientation: none of the switches can be mounted at an angle or twist out of position afterwards. Cherry adopted the plate mount method somewhere around 1971 with 261 and 262 keyswitches (later renamed M6), and Elec-Trol advertised their own Snap-Lock Datakeys plate-mounted reed switches in 1970. Micro Switch would adopt plate mounting in 1975 with SD Series.
Another cost-cutting approach was to retain the use of a full mounting plate, but mould it from plastic instead. Various manufacturers took this approach, including Tai-Hao with their APC keyboards, Nan Tan Computer with their KB-6250 range of keyboards, and Acer with some variants of their semi-mechanical keyboards. The use of plastic plates is likely to have come about somewhere around the start of the 1990s, although details on this are not clear.
The use of a heavy steel plate was felt by some manufacturers to be an unnecessary expense. Finding an alternative method to keep all the switches correctly aligned and reliably held would bring cost savings. Around 1973, Hi-Tek Corporation introduced their switch array system retroactively named “Hi-Tek High Profile”. Here, all the switches are formed as a single moulding. This design was later copied by Stackpole around the end of the 1970s. Mechanical Enterprises would later produce a broadly similar design, Sabrecoil, introduced somewhere around 1979 (the exact date is not known). Both Hi-Tek and Stackpole still produced corresponding discrete switches, available in single units or blocks, including arrangements such as 1 × 4 and 4 × 4 key stations. These moulded arrays were then secured to the PCB using screws, with (in some cases at least) alignment studs that fit into holes in the PCB (as found with the IMSAI IKB-1.
The disadvantage of the single-array moulding method is the need to set up a fresh moulding arrangement for every new keyboard layout. An alternative option is interlocking switches: each switch slots into those adjacent to it, forming in effect a solid block of switches. The oldest-known example of such a design is that of Hathaway Instruments, who filed a patent for interlocking reed keyboard switches in 1971. Hathaway’s design offered horizontal interlocking only, with end pieces to secure all the rows together.
RAFI filed German patent 2516527 “Keyboard with illuminated keys - has each key square housing provided with dovetail grooves and ribs parallel to key motion” in April 1975. No details of the switches are shown: the patent is solely for their interlocking system. RAFI’s design has both row and column interlocking. The design as patented only allows half-unit stagger, making it unsuitable for computer keyboards that require quarter-unit stagger.
Hi-Tek Dovetail Series uses a design very similar to that of RAFI’s, but with support for quarter-offset stagger. Hi-Tek’s design, the first patent for which being filed in 1983, is well-known from keyboards. Hi-Tek’s design was a compromise: the main keyboard area was still supplied as a single moulding, with additional keys affixed to it using the dovetail assembly. This way, the majority of the keyboard could still be a single moulding, but additional key stations could be added without the need for any additional mounting hardware besides more screws. Keycap alignment was assured, as was the rigidity needed to protect the switch terminals, as these additional keys were locked firmly in place against the main array.
Stackpole’s KS-200E series is another interlocking type, although in most cases KS-200E keyboards used a single moulded array with no interlocking joints. Unlike Hi-Tek’s design, the interlocking variant is seldom seen, and it’s not clear how the interlocking was meant to be used. The KS-200E advertisement shows a keyboard that could have a main array with edge interlocking (as per Hi-Tek Dovetail Series) or that could be composed solely of single interlocking switches.
It has always been a common expectation that keys towards the rear of the keyboard should be higher up than keys towards the front. Modern desktop keyboards typically have retractable feet to tilt the keyboard accordingly, and some laptops offered the same, such as Apple’s early PowerBook models. This raises the question of what angle the tops of the keycaps should present. Manual typewriters used a stepped or “staircase” profile, with the top of each key level (horizontal). With keyboards, this was at the discretion of the manufacturer or the customer.
Where a typewriter-like stepped profile file was required, a means was needed to angle the keycaps in the opposite direction to the slope of the keyboard itself. Tilting the keyboard means tilting all the switches with it. A common approach was to then tilt the keystem on the plunger by the same or similar amount in the opposite direction. Typical keystem angles fall in the range of 10–12°: 10° for Cherry, 11° for Clare-Pendar and 12° for Datanetics DC-50. Tilting the keystem allows for the same keycaps to be used regardless of keystem angle. It was also possible for the upper surface of the keycap to be sloped relative to its base, allowing for the same switches to be used regardless of keyboard angle.
Various manufacturers chose instead to angle the entire switch: the bottom of the switch could be flat (parallel to the top and perpendicular to the sides) or sloped, such that with the keyboard tilted, the switch would be vertically upright, with its angled base resting flat against the raised-up PCB. Typically the use of an angled base was an optional choice. Manufacturers offering angled-base switches included Navcor (who filed US patent 3311210 “Sloping panel keyboard mount” in 1965 for the idea), Raytheon with their KBSR and KBFR reed switches, George Risk Industries with their Series KB reed switches, and Fujitsu with FES-5 (introduced around 1972, but the tilted base is so far only seen in a 1975 article). There is also an unidentified Soviet-era switch with the same principle. The use of a tilted base is so far only known from reed switches.
Current keyboard switches (Cherry MX, Omron B3K, Matias) all follow the same design principle that there is no distinct movable contact piece. The movable contact’s terminal is formed from the same piece of sprung metal as the contact itself. This is why it is very common to receive a pack of switches with one or more bent terminals. (It is not clear how this is addressed in automated assembly, where machinery would be less impressed with incorrectly-shaped parts.)
A number of mechanical types are notable for having found a way to have both terminals be stiff and not prone to being bent; examples include:
- Alps SKCC, SKCK/SKCM and SKCP series: the “switchplate” contact assembly uses a foil sheet as the movable contact; one terminal bears the stationary contact, and the other is pressed against the foil sheet.
- Futaba MD, MR and ML series: the terminals screw into the switch, and push through holes in separate pieces of metal that form the two contacts.
- MEI T-5 series: the movable contact is a piece of spring wire wrapped around one of the terminals.
- Omron B3G-S and Alps-style: these use a “switchplate”-like contact assembly where the movable contact appears to be a phosphor bronze leaf spring that is attached to one of the terminals
See sealed terminals on the Deskthority wiki.
Data Interfaces advertised switch models DI-100 and DI-104 are having sealed terminals in order to allow soldering. While this is not a very precise explanation, it is a suggesting that sealed terminals protect against some aspect of the soldering process.