Hall effect keyboards
The Hall effect is a tiny voltage present across (instead of along) an electrical conductor when a magnet is placed in proximity to it while it conducts current. In an electrical generator, the conductor is moved inside a magnetic field in order to produce current. Instead of moving the conductor, you can instead move the electrons directly: by passing current through the conductor, the electrons are pushed through the magnetic field, which generates a voltage across the conductor as the electrons travel along the side of the conductor. This Hall voltage produced is very small (measured in microvolts), requiring amplification in order to detect it.
The diagram below depicts the general principle. The sensor depicted is a HFO B 461 G, and the diminutive 2 × 1 × 3 mm magnet is the type used by RAFI with these switches in their RS 76 C Hall effect keyswitches alongside the B 461 G.
Note that the physical form of the sensor element in the image above is purely illustrative and not the exact form that the sensor takes. In the case of Micro Switch, the Hall element within the sensor integrated circuit is normally angled at 45°; this can be seen in Ed Nisley’s microscope images of a 201SN1B1 switch, although the form of the sensor element varies noticeably between sensors, even within the same switch model, and the sensor element can be rectangular instead of square. Honeywell’s Hall Effect Sensing and Application book suggests in one diagram that the terminals of the Hall element are at its four corners.
Hall-effect sensing is one of the oldest methods of constructing a computer keyboard, being introduced by Micro Switch in 1968 with SW Series. It has a number of distinct advantages and disadvantages compared to other sensing techniques, as detailed below. A plain Hall sensor (or “Hall generator”) is an analogue circuit, producing a voltage proportional to the proximity of the magnet. However, for keyboard purposes, it is normally paired with a Schmitt trigger to provide logic output or to function similarly to a conductive switch.
Hall-effect keyboards require a Hall sensor IC at each key position and a permanent magnet within each plunger. In some designs, the plunger assembly can be removed from the keyboard while leaving the Hall IC in place (allowing for a damaged switch to be replaced without the need to open the keyboard or desolder a switch), while other switch designs are self-contained and the Hall IC is integrated within the switch body. Magnet sizes and quantities vary; Micro Switch tended to use fairly large magnets, while RAFI and Omron used extremely small magnets. The size of the Hall sensor IC package also varies considerably.
Hall effect switches are “true” solid state: the sensing portion of the switch comprises only a small integrated circuit. The moving portion of the switch requires only a plunger, a return spring and a permanent magnet. Consequently, Hall effect switches have the potential to offer extremely long lifetimes. Micro Switch reported a testing failure in 1973 of a bond wire at 99 million cycles, while RAFI give the lifetime of their 4 mm momentary Hall switches as 100 million cycles. There are no switch contacts to oxidise or become obstructed, nor any complex switching mechanisms that could wear out or become damaged.
Hall switching is entirely bounce-free. The perfect switchover from off to on (guaranteed by electronic hysteresis, even if the operator hovers the key at the actuation point) means that Hall switches can directly drive TTL circuitry without the need for a contact buffer circuit. (Micro Switch’s PB family of switches included various models with integrated contact buffer, and these were bulky circuits.) For gaming purposes, removing the need to debounce switches increases the keyboard’s response time by up to 5–20 milliseconds over conventional mechanical and membrane switches. The lack of bounce also greatly simplified the circuitry of computer keyboards, as switch output could be read directly; this advantage was largely lost once keyboards were operated by on-board microcontrollers, but switch bounce can still exceed the controller’s debounce time, especially if switches go out of condition due to age or design or manufacturing flaws.
Hall ICs have complete control of the switching output entirely in solid state. This means that switches can reliably provide pulse output, where the switch is only “on” for a brief fraction of a second (typically under 100 µs). This is one way to achieve a form of N-key rollover, by ensuring that—in normal usage—no two switches are ever conducting electricity at the same time. Micro Switch PB also provided pulse output entirely mechanically.
Digital Hall sensors using a Schmitt trigger provide hysteresis as standard. Alternatively, Hall sensors can be used to provide an analogue response, to allow a customisable actuation point.
Finally, as with other solid state types, the lack of physical contacts allows for switches to be extremely smooth, with only a minimal level of friction from the plunger itself.
Hall ICs require power to operate, making Hall-based keyboards more power-hungry than simple conductive keyboards. This is significantly mitigated by having the Hall sensors power up only when they are being sensed (only a single row in the matrix will be powered up at any time).
The requirement for a magnet and microchip inside every key raises the price of the keyboard considerably. This expense is only justifiable now for specialist equipment, although Hall effect has seen a revival in the enthusiast market and has made an appearance on the gaming market for its ability to provide full actuation point customisation. In the gaming market it competes against optoelectronic sensing.
The lack of mechanical sensing means that any tactile feedback (and mechanically-generated audible feedback) will not be tied into the actuation of the switch. Few switch types guarantee that the feedback corresponds with the actuation, and Hall effect is not one of those.
SPST contact switches (be they metal contact or membrane) are all largely the same in terms of operational characteristics. There are differences in supported voltage and current ranges, and contact bounce time, but beyond this, they can all be considered to function the same. Hall keyswitches each contain an integrated circuit, giving rise to a number of distinctly different types of switch, each suited to different uses. Even within computer keyboards, there is no one single type selected: a single keyboard may have two or more different Hall sensors within the switches.
The longevity of Hall effect switches is not clearly stated. RAFI’s RC 72 C and RS 76 C families are both rated at 100 million cycles for momentary switches. When these series were introduced in the early 1970s this was a decent lifetime, compared to mechanical switches which were more likely to have a lifetime of 10 million or less. This is a figure that is now achievable with mechanical switches, being the rated lifetime for modern Cherry MX (up from 20 million in 1985) and at least one variant of Omron B3K (up from 70 million). Datanetics DC-50, introduced in 1973, offered the same 100 million in a mechanical switch, but this was highly unusual.
By 1973, Micro Switch reported that one of their keyboard switches undergoing continuous testing had suffered a bond wire failure in the sensor chip, at 99 million cycles. This is the only verifiable figure of keyboard switch lifetime from Micro Switch. Their testing rig had accumulated 20 billion cycles, which seems to be the overall total. With 90 switches in the unit, this suggests that each switch had reached over 200 million cycles by this point. Honeywell’s Hall Effect Sensing and Application book claims “30 billion operations in a continuing keyboard module test program”. This document was published in 1998, suggesting that the 30 billion cycles by this point is likely to be the per-switch figure.
A long life is not guaranteed, of course. The friction within the switch will gradually wear down the plastics, with dust from the air acting as an abrasive, and how long the switch lasts physically will depend on the design and manufacturing quality. Even Micro Switch switches are known to suffer from physical wear. Nonetheless, the sensing portion of the switch will outlast the mechanical portion if it is manufactured to a high standard.
Analogue and digital output
Hall sensors are analogue devices: they report the magnetic field strength. Many applications, including pushbutton switches, expect discrete state: the Hall output should be either on or off, and not any intermediate state. Hall effect pushbutton switches uses digital Hall sensors, which incorporate a Schmitt trigger to provide binary (on/off) output.
Analogue sensing is still relevant to keyboards, as it allows for the actuation point (operating travel or pretravel) to be customised by the user on demand. Analogue keyboards can also be used as game input, to allow for variable character or vehicle movement speed.
Level and pulse output
A typical digital Hall sensor will respond according to the presence of the magnet: the output will be enabled when the magnet is detected, and disabled when the magnet is removed (according to the sensing and hysteresis thresholds). These sensors are described as “static” (in Germany) or “level” (by Micro Switch). Some sensors only enable the output momentarily, typically under 100 µs. As the magnet reaches the detection threshold, the output is briefly enabled and then disabled. The switch will not respond again until the magnet is removed and reintroduced. These sensors are described as “dynamic” (in Germany) or “pulse” (by Micro Switch). Pulse switches appear to have multiple uses (Micro Switch provided pulse output for non-keyboard switches), but in keyboards they were used to permit N-key rollover in scanless encoding keyboards; this is described further under two-of-N encoding below.
The diagram below shows the response in output for both level and pulse types:
Pulse switches are unsuitable for any key which must be held or latched. Modifier keys (such as control and shift) and alternate action keys will always use current level switches.
Current sinking and sourcing
Hall sensors have one or more “output” terminals. The term “output” can seem misleading. The outputs are divided into two types: current sinking (also known as open collector), and current sourcing. Current sourcing outputs supply current when active.
Current sinking outputs draw current when active. For the purposes of sensing, they can be treated the same as a normal contact switch connected to ground. Because the current flows into the output, this seems to contradict the directions treated as “in” and “out”, but it would be wrong to call this terminal “input” as it does not process the signal, but rather supply it.
In Electronic Design vol. 21, no. 1, 4th January 1973, page 136, Micro Switch advertised 12SW and 16SW Hall effect numeric keypads. These were current sinking, with the stated reason of providing compatibility with TTL (transistor-transistor logic) and DTL (diode-transistor logic).
The following diagram illustrates the two types. A single “application circuit” represents the circuit that is being controlled by the Hall keyswitch. The Hall keyswitch is represented as little more than a pushbutton switch, while in reality it is a more complex circuit.
(Note that while switching diagrams on this site use red for live current and black for inactive circuit pathways, here red denotes input current and black denotes output current.)
A Hall sensor only requires a single output terminal. However, many Hall sensors offer two. Although it may seem that this is for redundancy, Micro Switch intended dual output to make encoding easier. In [ED1976-FOK] page 126, the reason for the dual isolated outputs is given as “to reduce encoding-logic complexity”. In Micro Switch’s Solid State Keyboards brochure from 1973, they noted:
A unique two-of-n code is developed from the two isolated outputs available from each key switch. This code is used to address the keyboard encoder, thus eliminating the need for complex and costly scanning techniques.
This is explained further under two-of-N encoding below.
In Electronic Design vol. 21, no. 1 (page 136), Micro Switch indicated that the dual outputs provide greater fan-out. Additionally, they note that each output is capable of sinking 3.2 mA (in current sink switches), and that up to 6.4 mA of current can be sunk with the outputs wired in parallel.
Panasonic’s datasheet for their dual-output DN837 does not seem to make any mention of dual outputs as a feature. It only notes that the measurements for output apply to both outputs.
Conductive switches only draw current when they are being sensed. The current requirement is based on the contact resistance, which can be very low: as little as 20 mΩ. Hall sensors however require an amplifier to sense the minuscule Hall voltage, and the electronics inside the switch draw current continually when the switch is released. Micro Switch cited the standby (idle) current for their logic scan sensors as 3.5 mA maximum. For a keyboard of 100 keys, current draw could be as much as 350 mA.
Micro Switch resolved this around 1979 by creating a sensor type that could be powered down when not in use. Rather than toggle the VCC line, the ground level is switched between low and high, or connected and disconnected. With the ground disconnected or pulled high, there is no potential difference across the switch and the entire sensor is shut off. This is their three-terminal (E) sensor type, widely used in SD Series keyboards. There is no known explanation of why this cannot be done with the older sensor types: the E type is an alternative to their older S-type logic scan sensor. It may be that the settle time for the circuitry in the sensor is too long to allow reliable readings with the switch only powered up to take a reading.
All other Micro Switch sensor types are fully powered up during keyboard operation. The same is true of many other brands, including Siemens, TESLA and Panasonic. Siemens did also introduce a matrix scanning variety of sensor, that was copied by HFO in their B 461 G and B 462 G sensor types. These types have a terminal called “enable input” (“Freigabeeingang” in German). This terminal connects to the voltage control portion of the sensor:
The HFO circuit diagram in Mikroelektronik Information Applikation Heft 9 from October 1982 shows that the enable input line is connected to the base of a transistor whose emitter is connected to ground: when enable input is high, current is allowed to flow out of the circuit at that point. With enable input low, it seems that part of the circuit is still drawing some current: the RAFI and HFO specifications appear to indicate that the switches draw up to 5 mA when enabled and pressed, up to 3 mA when enabled and released, and up to 0.5 mA when disabled. This is not as power-efficient as Micro Switch’s three-terminal type, but is a huge improvement over switches that draw several milliamps when idle: the same 100-key keyboard with HFO sensors would only require at most 50 mA of current for the switches.
Because Hall effect switches contain ICs, there is a limit on the voltage and current that can be applied. This is not a hard limit, as different models of Hall IC provide different voltage and current levels. This is similar to the voltage and current limits of mechanical switches, which in some product ranges depends on which of the available contact materials is chosen (e.g. silver versus gold).
When Micro Switch pioneered Hall effect keyboards, they designed a keyboard system with direct encoding. This techique uses two-of-N codes. These are a variety of constant-weight code where two and only two bits are always set, and all other bits are cleared. Each of the two outputs from the switch will set one bit, resulting in a number that identifies the switch that was pressed. Consequently there is no need to scan the matrix, as the encoder will directly receive the scancodes as keys are struck. The use of pulse output avoids clashes when two keys are held at once.
The circuit pathways will be wired such that each switch is connected to two and only two data lines into the encoder. This is illustrated in the diagram below. Here, the power and ground lines into each switch are shown. Conventionally these would be omitted, but in this instance they are included for clarity.
Notice how all the switches are powered up at all times. This approach is simple but it does require considerably more power than a scanning keyboard. When a key is pressed, its two outputs enable two bits of data input into the encoder:
Thus, the scancodes of these keyboards are sparse: between the lowest and highest values, most intermediate values are unused.
When two keys are pressed simultaneously, the input into the encoder is illegal:
Here, the two-of-N code has three digits set, which is not permissible, and the keyboard encoder will not recognise the input. Thus, it is critical that such a situation seldom if ever arise during normal typing. The solution chosen was that the switches would only emit output briefly: the pulse type described above. This is also why pulse switches were chosen instead of diodes: as shown in the diagram above, there is no wrong direction current flow, and thus adding diodes would not achieve anything. This arrangement makes it impossible to determine how long a key is held; while this limitation seems harsh compared to modern keyboard technology, the proliferation of ASCII keyboards in that era meant that it was quite normal for keyboards to output character codes rather than press and release notifications.
Other brands of Hall sensor also feature dual isolated outputs (including Siemens, HFO, TESLA and Matsushita/Panasonic), but their reasons for doing so have not been determined. Dual outputs may have additional benefits beyond keyboard encoding.
Most Hall effect keyboards appear to use matrix scanning. Matrix scanning requires that only a single row be active at any one time. The dual-output sensors used with two-of-N coding are permanently active, and wiring them directly into a matrix will not work. Consider the following diagram, which shows that scenario:
Here, one output from each switch (output 1, O1) is connected to the corresponding matrix column, and the other output (output 2, O2) serves no purpose and is not connected. The switches are not connected to the matrix rows, as there is no means by which to do so. When a key is pressed, the column becomes active, but there is no way for the keyboard to isolate the row containing the switch:
One solution to this problem is to power up only a single row at a time. This can be seen in the Zbrojovka Brno 262.12 keyboard with TESLA MH3SS2 dual-output source level sensors. The circuit for that keyboard is based around the following arrangement:
Here, VCC is supplied to a bank of transistors, with the emitter of each transistor connected to a row. The controller enables each transistor in sequence to power up all the sensors on that row.
Another arrangement is to provide each switch with an input terminal. The input terminal enables the switch only when the row in question is active.
The diagram above depicts the arrangement used with Siemens and HFO sensors, which use the input terminal (marked I) as a power switch for the sensor itself. Micro Switch logic scan sensors are similar, but the input terminal is pulled low instead of high to enable the switch. The Siemens and HFO matrix scan sensors are only enabled when the row is being scanned, and they enter a low power state when the input terminal is low, which like with Zbrojovka Brno’s arrangement, greatly reduces power consumption.
Micro Switch’s original logic scan sensors still draw power when not enabled, and with a hundred of them powered up at once made the keyboard power-hungry. It seems that Micro Switch did not deem their SW and SD “S”-type logic scan sensors to be suitable for being switched on and off during scanning, so they later produced a separate “E” sensor type—with only three terminals—where the ground terminal is itself used as the input:
The input terminals is marked “I” in the diagram above, but functionally it is the ground terminal. The terminal is named “input” because it is used to control the state of the switch, but if it is permanently grounded then the switch functions as a simple source level switch:
By switching the ground on and off on a per-row basis, a single row of sensors can be powered up at a time.
The following manufacturer list is in very approximate chronological order, based on the limited information about when some product ranges were introduced to the market.
Micro Switch were the pioneer of Hall-based sensing in keyboards. SW Series Hall effect keyboards and switches were introduced in late 1968; this series offered a wide variety of switch types and output options. SW Series was followed around 1976 by SD Series, which reduced the switch size and introduced plate mounting to their range. SN Series standalone Hall switches (based on SW Series) were also offered. Micro Switch remained in the high end keyboard market for over three decades, with their website finally removing all mention of these products around the year 2000.
Micro Switch’s Hall switches were bulky, with SW and SN switches requiring two large barium ferrite–filled PVC magnets joined by a metal shunt. SD Series was smaller, but still a fairly large form factor. SD Series did however only require a single, smaller magnet. Micro Switch did not produce a Hall effect series to meet DIN compliance.
Siemens produced a range of Hall sensor ICs with quite a variety of characteristics. These chips seem to be fairly uncommon, and they are seldom found in keyboards. The only known instance of a keyboard switch being found to contain one is in an unidentified, loose RAFI RS 76 switch, but RAFI appear to have made extensive use of Siemens chips originally. The point that Siemens introduced these products remains unknown.
RAFI’s first solid state keyboards, introduced around 1970, were magnetoresistive. These were soon followed by their RC 72 C Hall effect switches, assumed to use Siemens Hall sensors. RC 72 is likely to have come onto the market around 1973. Shortly after that, around 1975, they brought out miniaturised RS 74 C switches. These diminutive parts were drastically smaller than anything Micro Switch offered, with only a single 2 × 3 × 1 mm magnet. Travel was 2.5 mm. The next year, travel was extended back up to 4 mm with RS 76 C, effectively the same design but enlarged, offering greater travel and higher potential protection against liquid and particle ingress.
Of all the Hall effect types introduced in the 60s and 70s, this is the only type still in production. The sensors at some point changed to HFO B 461 G, which is all that is offered now. RAFI still offer Hall effect keyboards, but these are not commercially available to consumers or businesses.
In the 1970s, Fujitsu focused on reed switches. However, for the purpose of driving TTL circuits, their FES-8 switches could be fitted with a Hall sensor instead of a reed. FES-9 continued to offer this option. It seems that Fujitsu developed its own Hall sensors; the details on these in the Fujitsu magazine are extensive but presently only known to be available in Japanese.
Matsushita (now Panasonic) also produced Hall sensors. The last-time buy date for their PS Hall sensors was the 31st of March 2009, and it appears that they no longer produce Hall sensors. Matsushita Hall sensors were used in Omron’s keyboard switches.
Details on Omron’s high-end keyboard switches are extremely scarce. The B2x range seems to have included both high-profile and low-profile Hall keyswitches, but the only observed model is B2H-F7W low-profile clicky type with a Matsushita DN837 Hall sensor.
Very few details have been recovered for Sasse Series 25. What is known is that, in 2003, one of the photographs on their website showed a Series 25 switch containing what appears to be a Siemens or HFO Hall sensor.
TESLA included Hall sensor ICs in its product range. These were used in central European switches based on Micro Switch SW and SD Series. The manufacturer of the switches based on these sensors remains a mystery.
A FEPER keyboard with PCB-mount Hall effect switches has been discovered. This “Junior XT” keyboard uses a Hall sensor not seen elsewhere. The IC is very compact, similar in size and shape to those from Matsushita.
HFO produced a range of Hall sensor ICs seemingly derived from Siemens’s SAS range. These remained on the market longer than those of Siemens, and were adopted by RAFI. RAFI may have a stockpile of these, as presumably they have long since stopped being manufactured.
Although Hall effect keyboards have never gone out of production, Chinese manufacturer Ace Pad Tech reinvigorated Hall effect keyboards, targeting the gamer or enthusiast markets. Their keyboards are touted as being waterproof. Ace Pad report on their website that they started working on Hall effect keyboards in late 2014, and the earliest English-language reviews of them seem to be from 2016.
For a brief period, XMIT Keyboards partnered with Ace Pad to make custom Hall effect keyboards. This partnership did not last, but the keyboards were for a short time in production.
The SteelSeries Apex Pro and Apex Pro TKL are gaming keyboards with analogue OmniPoint sensing. The actuation point of each individual key can be set between 0.4 mm and 3.6 mm. These keyboards are expensive (£199.99 and £189.99 respectively in April 2020) and yet only the alphanumeric area is Hall effect. The remaining keys (function row, navigation cluster and—on the full-size model—numeric keypad) use standard MX clone switches.
Input Club has announced their own fully-analogue Hall effect keyboard, the Keystone. As of April 2020, it has not yet gone on sale. The switches are manufactured by Kailh.