A keyboard that uses a mircoprocessor to detect which keys are pressed at a given time, and outputs this information as a stream of digital data. The scanning keyboard was essential to the design of affordable polyphonic synthesizers, starting in the late 1970s. Prior to that time, most synths used keyboards which produced a control voltage using an array of diodes and precision resistors known as a resistor ladder. This worked well for monophonic synths, and clever designers managed to extend the design to produce control voltages corresponding to two pressed keys for duophonic synths, but it was unworkable for true polyphony.
Rationale[]
Some electric keyboard instruments (starting with the Hammond organ and Novachord) were implemented starting in the 1930s by making them fully polyphonic; there was a separate tone-generating circuit for each key on the keyboard, such that if the performer were to depress every key simultaneously using a board or some such, every note would sound. Combo organs, electronic pianos, and string synthesizers were also implemented this way. However, these instruments had very limited timbral capability compared to, say, a Minimoog. Designers realized in the early 1970s that implementing a fully polyphonic version of something like a Minimoog was impractical; it would be too expensive, too heavy, and the high parts count would lead to frequent malfunctions. There were a few attempts, such as the Polymoog and the Korg PS-3100. The Polymoog was only paraphonic if the performer switched away from using the presets, and the PS-3100 used top octave division so that it didn't actually have a VCO for each key.
Starting around 1970, designers began to realize that the way out of this track was to have a limited number of units of synthesizer circuitry, which eventually came to be called voices, each capable of playing one note with all of the capabilities of the synth. Performers generally would never have a need to actually have all of the notes on the keyboard sounding at once; being able to play some maximum number of notes suffices for the vast majority of music. Once this concept was arrived at, the next question was to figure out how to map keys to voices. Early attempts tried fixed mapping, dividing keys into groups each of which was assigned to a voice. However, these schemes had the limitation that certain chords and melodies could not be played, because there would always be some combinations where two notes would be assigned to the same voice.
The key was to be able to do the assignment dynamically. For each key pressed, the keyboard needed to find an available voice and assign that note to it. When a key was released, the keyboard needed to mark the associated voice as available for assignment. It soon became clear that this required digital logic interfacing with the keyboard.
History[]
Several entities went to work on the problem in the early 1970s, each apparently unaware of the others' work. In Japan, Yamaha was working on a concept that originally was applied to their Electone line of electric organs. Simultaneously, in the U.S., E-mu Systems and inventor Amond Pascetta were developing their own designs. These were called "scanning keyboards" because they used digital logic to examine each key, one at a time, doing so much faster than the performer can play (hundreds or thousands of times per second). When the logic detects that a key has been pressed, it assigns that note to a voice; when it detects a key release, it informs the assigned voice. The logic maintains a map of which keys are currently assigned to which voices, so that it does not assign the same key to two voices, nor does it steal a voice from a currently held key (unless it is supposed to; see below).
Yamaha's scanning keyboard design first appeared in the massive GX-1 synth in 1974. Because this very expensive synth was only made in small numbers, and was not marketed outside of Japan, few in the West were aware of it at the time. On the other hand, that same year, E-mu licensed their design to Oberheim for use in the Four Voice, the first well-known synth to use the voice allocation method. In 1977, E-mu introduced its 4060 scanning keyboard, an all-singing, all-dancing keyboard that included abilities to do splits, layering, support multitimbral synths (even though few existed at the time), and it included a digital sequencer. The performer controlled all this by connecting a computer terminal to an RS-232 data connection on the keyboard. That same year, E-mu licensed the design to Sequential Circuits for use in the Prophet-5. Royalties from the scanning keyboard patents were a lucrative source of income to E-mu for a number of years.
Pascetta, meanwhile, was doing custom work. He built a scanning keyboard for T.O.N.T.O. around 1975, and then did some units for other forward-thinking performers. Like the E-mu design, Pascetta's design supported keyboard splits, grouping and layering.
Later, the digital scanning keyboard design had the additional advantage of making it easy to add velocity sensing. And, once the cost of microprocessors dropped, it was cheaper to manufacture than the resistor-ladder keyboards. This was done by placing a second set of switch contacts under each key, a little lower that the first contacts; by measuring the time difference between the engagement of the first and second switches, the logic could infer the key velocity.
Design Considerations[]
A scanning keyboard has to do a number of things: (1) detect when keys are pressed and released, (2) keep track of which keys are assigned to which voices, and (3) mark voices as available after they have finished sounding. In this last, a consideration (which early designs had problems with) was that the voice is not available until after its envelope generator completes its release phase, which may be some time after key up.
The scan needs to be fast enough so that there will not be a perceptible delay between the performer pressing the key and the start of the voice sounding. At the same time, it needs to not be fooled by switch noise and "bounce". The more keys there are to scan, the faster the scan has to run. On scanning systems that scan multiple manuals or pedals, the scan speed may place an upper limit on the number that can be accommodated.
What happens if more voices are required than are available? Several approaches are possible, and all of them have been implemented in different synths. The most obvious approach is to ignore a new key press if no voice is available. This may, however, thwart the performer who plays fast runs. Another approach is to allow voice stealing. Typically, the voice assignment method will try to keep track of which voice has been active the longest, and steal that one to play the new note. More sophisticated systems will preferentially steal a voice whose envelope generator is in release phase, since such a voice is probably nearly done playing its note anyway and its abrupt disappearance is less likely to be noticed.
The way that the scanning keyboard actually interfaces to voices may vary. Early systems were interfacing with analog circuitry and necessarily had to use CV/gate interface channels. An issue with these is that they did not provide any information back to the voice allocation logic about the envelope generator status. Starting in the late 1970s, the trend became to send digital information to the voice, which contained its own D/A converter to produce the control voltage. Controller keyboards, work by sending MIDI messages to a remote synth. In this case, the voice assignment logic is actually in the receiving synth, not in the keyboard; it just issues Note On and Note Off messages and lets the synth figure out what to do with them.