Photoelectric encoder keyboards
Photoelectric encoder keyboards use light beams interrupted by shutters to encode keystrokes. Each light beam encodes a single bit of output. The output codes are defined by the presence or absence of slots or holes in the shutter for each light beam. The following illustration depicts a generalised idea of how such a keyboard functions:
Photoelectric (also optical or optoelectric, even “photo-optical”) keystroke detection dates back to before the days of electronic keyboards. US patent 2438825 “Selector” filed by Trans Lux Corp in June 1945 shows an encoder that appears to be intended for use with devices such as teleprinters or card punches. The complete unit is not depicted, but elements 7 in the patent drawings represent the bars attached to the keys; the patent refers to existing patents to show the kinds of devices with which it could be used. Monroe filed US patent 2641753 “Photoelectric keyboard” in 1951 for a photoelectric encoding keyboard for calculators.
The oldest confirmed dedicated photoelectric keyboard unit—not part of a teleprinter, typewriter or calculator—is the PK-144 from Invac. Their 1960 patent filing was for a photoelectric encoder assembly designed to integrate with a manual typewriter; this was followed by a 1961 patent filing for a dedicated keyboard unit with the same design of encoder.
Such keyboards are inherently physically bulky, but they could be manufactured with the barest minimum of electronics and were free of both bounce and radio-frequency interference (RFI). The only components required were one or more light sources, a back of photoreceptors and one or more amplifier chips.
Photoelectric keyboards are self-encoding: the output from each key is defined solely by the notch pattern on that key’s shutter. Rearranging keys requires only the shutters (with keycaps attached) be moved into the desired positions. For the most part, as with any other self-encoding keyboard, rollover is limited to two keys at most, although a patent exists for a design with multi-key rollover. Models with one or more incandescent bulbs are also vulnerable to bulb failure.
Each key is fitted with a shutter that runs from the front to the back of the keyboard. Each shutter contains one or more notches or holes, depending on the design. Typically there is only a single row of light beams, but patents show that some keyboards used two rows for additional capacity.
In Monroe’s 1951 patent, a single, long lamp provides the source of light for all the keys. Invac however chose to use a separate incandescent bulb for each light channel, combined with (depending where you read) photoresistors or photocells as the light detectors.
In their 1967 patent (US patent 3465099 “Optical encoder”, filed September 1967), Friden depicted a means of using only a single incandescent bulb. A parabolic mirror on the left side of the keyboard collimates the light into parallel beams that reach the right side of the keyboard. There, a vertically mounted PCB bearing photocells detects the light.
Collimation filed a patent of their own (US patent 3818485 “Keyboard apparatus”, filed March 1973 as Western Digital Systems); they continued the single-bulb and parabolic mirror used by Friden (hence the company name of Collimation), but replaced the vertical PCB with a horizontal one and added a planar mirror to reflect the light beams downwards.
Some implementations use a separate strobe channel that activates slightly later than all the other light channels. This ensures that all the data channels are sufficiently blocked as to be read correctly before the host examines the output. Differences in component positioning, shutter angle and component tolerances could lead to some sensors changing state before others, so the strobe prevents the output code being signalled until all light sensors have settled into their new state.
Active light beams can be treated as either a 0 or a 1 depending on preference. By default, all light beams are active, and this can be used as a self-test process to verify whether any lamps have failed. Invac arranged the incandescent bulbs in series so that a single lamp failure would deactivate the keyboard instead of produce incorrect output.
Keystrokes can only be read if a single shutter is depressed at a time. Depressing a second shutter puts the keyboard into an invalid state where the light beam pattern will no longer represent either of the keys pressed. Perhaps the only way to limit the risk would be a two-part shutter where the mask portion springs upwards as soon as the shutter has passed the actuation point, but no such keyboard or patent has been seen. In contrast, Invac’s keyboards used a solenoid-driven latch to keep the shutter in place long enough to register the output.
Invac took the classic approach of physically interlocking the keys. Each unit has a ball bearing tray that runs across the width of the keyboard. The tray is slightly wider than the sum of the ball bearings it contains, allowing exactly one shutter to be pushed down in between the ball bearings. With one key down, no further keys can be pressed as there is no space remaining that would allow the ball bearings to be pushed aside.
The keyboard supplied with Viatron’s System 21—of unconfirmed origin—used constant-weight code encoding. Each shutter blocks four out of the light channels representing the output code for the key. This approach ensures that the output of two or more keys pressed together is always invalid and thus not open to misinterpretation. The disadvantage is that this intermediate code must be converted to the output codes using additional electronics. The exact number of channels is not clear; the owner of the Viatron keyboard that was examined reports 4-of-8 codes, and the system schematics include a chart with 4-of-8 and 2-of-13 codes, but the shutters themselves appear to be 4-of-9 code.
Keyboards without any interlock should function reasonably well so long as the host reads the output promptly upon receiving the strobe signal and then waits for the strobe to clear. As the strobe is the first channel to be opened when the key is released, it is possible for a second key to trigger the strobe signal between the first key releasing the strobe and releasing the remainder of the light channels. Details on Collimation’s keyboards are rare, but one former owner did report that “ghosting and mangled keystrokes were almost impossible to avoid”, suggesting that the implementation (despite having a delayed strobe channel) failed to provide any form of electronic interlocking.
Self-encoding keyboards in general only support a single output code per key: the amount of data channels per key is not sufficient for anything further. Such keyboards are frequently able to wire up the modifier keys separately as required. Photoelectric encoder keyboards use optical sensing for all keys, and thus each modifier key requires its own light channel. In the simplest implementations, each modifier key has its own light channel and separate logic circuitry adapts the output codes accordingly. With ASCII this is generally a matter of bit manipulation, but other encoding schemes do not necessarily allow this. This also requires expensive logic circuitry that keyboard designers may wish to avoid, or for the host device to apply the bit manipulation in some form.
National Semiconductor’s MM5740 MOS/LSI keyboard encoder has a curious feature where bits 0–4 and 8 within the ROM look-up table are common to all modes, while bits 5–7 are mode specific. Although no photoelectric encoder keyboards are known to take this approach, it would provide a way to offer additional modes without too many extra light channels.
Friden’s 1967 patent, noted above, depicts a method of providing three fully distinct modes per key. There is a separate set of light channels for each of the three modes. The two modifier keys (called “shift” keys in the patent) selectively isolate one mode at a time, by allowing only one mode’s light beams to pass. Thus, each shutter has apertures for each of the three modes supported. Keys unaffected by one or more modifier keys (e.g. A–Z on an uppercase-only system) will repeat the same apertures for the normal and shift modes. Shutters could also selectively mask one or more bits in order to change the ASCII code. In positive mode (active beam = 1), the control key shutter on an ASCII keyboard would need only block bits 6 upwards.
In most instances, rollover is either impossible, or limited to two keys. Two key rollover is available when the keyboard is able to recover from clashes, which is possible with constant-weight encoding.
Standard Elektrik Lorenz AG filed a patent in August 1971 for a design that has theoretically unlimited rollover. This design, later covered in US patent 3750150 “Photoelectric keyboard for data input devices or the like”—filed in July 1972 by International Standard Electric Corporation in New York, with priority to the German patent—uses semi-transparent instead of fully opaque shutters. Each shutter reduces the amount of light reaching the photodetector rather than block it outright. Semi-transparent shutters alone would offer an interlock method, where the strobe would be valid when transitioning to or from a light level indicating that a single key is held.
The Standard Elektrik Lorenz design is however much more sophisticated. Threshold sensors detect when the light level at the photosensor decreases past a specified level, and each time this happens, a flip-flop is activated to store a data bit. Assuming a 50% shutter opacity, when a key is pressed, some light beams are lowered to 50% brightness. Assuming a 60% threshold on each sensor, each light beam that is dimmed reduces to 50% and in passing the 60% threshold triggers the sensor, causing the flip-flop to store a 1. Those light beams that are not blocked remain at 100% brightness and the flip-flop stays at 0.
After a brief delay, the output code is read from the flip-flops into a buffer and the flip-flops are all reset to 0. This step causes the sensor electronics to forget the first key that was pressed.
If the first key remains held and a second key is pressed, some light beams will be dimmed from full brightness to 50%, and some that were already dimmed to 50% will be dimmed again to 25%. The reduction in brightness from 50% to 25% will cross the (say) 35% threshold for the second stage sensors, setting those flip-flops to 1. The beams that reduced from 100% to 50% will trigger the first stage sensors and also set the corresponding flip-flops to 1. As all the bits set to 1 on the first keystroke were already cleared, the flip-flops will represent solely the encoded value of the second key to be pressed.
In the diagram above, you will notice that the bottom light beam has not triggered the sensor: the light level has remained steady and not crossed the 60% threshold again, so the threshold sensor has not triggered from the second keystroke. (The implication is that the threshold sensors are themselves based on flip-flops, and that they only reset once the light level is increased above the threshold; once reset, they can be triggered again by another keystroke.) The 60% sensor has triggered for the first beam as the second shutter alone has dimmed it, and the 35% sensors for the second and fourth beams have triggered as those beams are now being dimmed by a second shutter in conjunction with the first, causing the light level to fall below the threshold of the second stage of sensors.
Increasing the rollover requires higher sensor and lamp tolerances, and the patent notes that, in practice, three-key rollover is deemed sufficient.
The US patent transferred to Alcatel N.V. in the Netherlands in 1987. Thus far, no such keyboards are known to have been discovered.
Invac is the oldest company discovered to have patented a full photoelectric encoder keyboard, with their patent filed in 1961. Invac keyboards were mechanically interlocked. Model PK-144 was advertised as far back as August 1962, and models PK-144 and PK-164 were depicted and described in Computer Design magazine in December 1965.
Friden introduced their patent for a tri-mode all-optical-encoded parabolic mirror keyboard in 1967. By this time, Friden had already been aquired by Singer, and the follow-up patent was filed by Singer in 1970. No Friden or Singer keyboards are known to have been discovered.
Western Digital Systems, AKA Collimation, was later acquired by Applied Dynamics International. Their keyboard patent was filed in 1973. Collimation keyboards are very rare.
The manufacturer of Viatron’s System 21 keyboard is not known. It is presently the only known example of a photoelectric encoder keyboard with constant-weight code interlocking.