|
|
|
|
Digitally-controlled oscillator
|
| |
|
| |
A digitally controlled oscillator or DCO is a hybrid digital/analogue electronic oscillator. The name is an analogy with "voltage-controlled oscillator". DCOs were designed to overcome the tuning stability limitations of early VCO designs.
term "digitally-controlled oscillator" has been used to describe the combination of a voltage-controlled oscillator driven by a control signal from a digital-to-analog converter, and is also sometimes used to describe numerically-controlled oscillators.
This article refers specifically to the DCOs used in many synthesizers of the 1980's.
Discussion
Ask a question about 'Digitally-controlled oscillator' Start a new discussion about 'Digitally-controlled oscillator' Answer questions from other users |
Encyclopedia
A digitally controlled oscillator or DCO is a hybrid digital/analogue electronic oscillator. The name is an analogy with "voltage-controlled oscillator". DCOs were designed to overcome the tuning stability limitations of early VCO designs.
Confusion over terminology
The term "digitally-controlled oscillator" has been used to describe the combination of a voltage-controlled oscillator driven by a control signal from a digital-to-analog converter, and is also sometimes used to describe numerically-controlled oscillators.
This article refers specifically to the DCOs used in many synthesizers of the 1980's. These include the Roland Juno-60, Juno-106, JX-3P, JX-8P, and JX-10, the Korg Poly-61 and Poly 800, and some instruments by Akai and Kawai.
Relation to earlier VCO designs
Many voltage-controlled oscillators for electronic music are based on a capacitor charging linearly in an op-amp integrator configuration. When the capacitor charge reaches a certain level, a comparator generates a reset pulse, which discharges the capacitor and the cycle begins again. This produces a rising ramp (or sawtooth) waveform, and this type of oscillator core is known as a ramp core.
The typical DCO design replaces the comparator in the ramp core with reset pulses generated from a counter or microprocessor. This gives the design digital frequency stability, whilst retaining a true analogue waveform output. Aside from the way reset pulses are generated, the typical VCO ramp core and the DCO are identical. Both produce a ramp waveform from which other waves are derived by waveshaping.
Historical context
In the early 1980's, many manufacturers were beginning to produce polyphonic synthesizers. The VCO designs of the time still left something to be desired in terms of tuning stability. Whilst this was an issue for monophonic synthesizers, the limited number of oscillators (typically 3 or fewer) meant that keeping instruments tuned was a manageable task, often performed using dedicated front panel controls. With the advent of polyphony, tuning problems became worse and costs went up, due to the much larger number of oscillators involved (often 16 in an 8-voice instrument like the Yamaha CS80 from 1977 or Roland Jupiter-8) from 1981 which adopted an auto-tune feature at least, first implemented on the Sequential circuits Prophet-5 in mass market.
This created a need for a cheap, reliable, and stable oscillator design. Engineers working on the problem looked to the frequency division technology used in electronic organs of the time and the microprocessors and associated chips that were starting to appear, and developed the DCO.
The DCO was seen at the time as an improvement over the unstable tuning of VCOs. However, it shared the same ramp core, and the same limited range of waveforms. Although sophisticated analogue waveshaping is possible, the greater simplicity and arbitary waveforms of digital systems like direct digital synthesis lead to most later instruments adopting entirely digital oscillator designs.
While the first production synthesizers (monophonic) to boast 'Digitally Controlled Osc.' on their front panel were the Crumar DS-1 (1978) and the more common improvement of it the DS-2, the same basic scheme of creating the actual oscillations that are *heard* (ie. the way the waveforms are generated; not their source of stability) was implemented earlier in synthesizers like the ARP Soloist (1972) and Roland SH-1000 (Japan's first, 1973). But these earlier units used an analog high frequency oscillator circuit to drive the divide down network that gives all the multiples of frequencies on tap. These can then be selectively added together to make an approximation of a sawtooth wave as well as offering footages of square waves that can then be duty cycle modified to create pulse waves as well. The Crumar however used 74LS221 Dual-Monostable multivibrator digital chips with capacitors and resistors (and an op amp summer; a circuit that is analog but has negligible drift contribution) to form the initial oscillation sources. (2 DCO's in each synthesizer) Resistors and capacitors can be much more cheaply selected for stability when compared with circuits bearing several semiconductor devices operating as something other than fast switches, like those inside the digital chips! In other words, it's not purely the presence of semiconductors that makes things unstable. It's whether or not they are used in the analog realm instead of digital.
So the control of the oscillators was analog. The pitch bender is analog. But the potentiometers in pitch benders, if kept clean, don't drift in value significantly. And digital key codes were converted to analog from the keyboard. But again only simple low drift parts involved so that the voltages are extremely stable. Then these anlog voltages control the operation of the digital multi-vibrator chips! So one is left to wonder why they didn't name it a 'digitally GENERATED Oscillator' possibly; a DGO. Or a VDO maybe. But in terms of drift, they were essentially perfect. That was the main point to be made. However in the Poly synth examples below, pitch wheel voltage was converted to digital I believe in all cases. And of course this results in a stepping which could be heard on some units and was functionally indiscernible on others. However the analog control on the pitch of the DS-1 for instance makes for perfectly smooth voltage transitions, assuming a perfect pitch potentiometer of course :-).
The same year the Prophet 5 synthesizer appeared on the market with its auto-tune features. A form of 'digital control by command'. You had to hit a button to tune it. And this became a standard feature for a few years, and especially nice on machines like the Memorymoog with its 18 CEM3340 analog oscillators.
In 1981 the first DCO based Polysynth appeared; the Roland JUNO-6 and in 1982 the Kawai SX-210 and Roland JUNO-60's; DCO synths with memory! The JUNO series used a computer reset sawtooth oscillator with compensation for the frequency it's trying to represent and many forms of this proceeded in other instruments. And to follow would be many variations adding features and changing sound qualities in various ways. The Korg DW series, Kawai K3, and the Ensoniq ESQ-1 (all in 1986) created waveforms from ROM data and in fact the K3 actually had some SRAM for creating a user waveform from the first 128 harmonics. And these digital waveforms then were filtered with analog filters. These are often referred to as hybrids. And actually these should be referred to specifically as NCO machines perhaps. Numbers represented in binary are simply being called up and spewed onto a data buss to be assimilated by a DAC. Nothing is really 'oscillating' except the clock that drives the cpu :-). NSO? Numerically simulated oscillator? In any case from the point the signal leaves the DAC the signal is modified by analog circuitry; hence the name 'hybrid'.
Of course in 1983 Yamaha completed the DX7 and entirely digital synthesizers would make all others seem obsolete in the eyes of many. Roland continued with the JUNO-106 which was a huge success and the JX3P and MKS30 rack, followed by other JX and Alpha Juno units til the later 80's. The advantages, particularly in polyphonic synthesizers, to having all control voltages processed digitally should be obvious. Analog voltages were still used with the bender for instance and portamento knob on the older units. But those voltages are interfaced to a multiplexer and Analog to Digital converter and the CPU uses that data to make decisions about how to control the DCO's. Newer all digital synthesizers commonly use encoders for entry knobs rather than smooth potentiometers.
Also the Crumar company under the 'Bit' name continued to make DCO polyphonic synthesizers as did the Siel company. The Siel Opera6 (known as the DK600 in America)of 1984 is worthy of mention though. While it says 'DCO on the label, again all control is actually analog. However they sacrificed Portamento/Glissando and here's why. (In fact the whole DK series is like this I believe; for sure the DK70 and DK80.) These machines really use an analog oscillator (SSM2031) to generate the high frequencies and the analog control voltages are quite in control of them. However these machines use divide down to generate the actual frequencies that form the sound source. But when a note is hit the frequency is locked into a 'synthesizer channel' and remains there until the note is 'stolen' by the note assigner. In the 'channel' up front is an integrator circuit which responds to the analog control voltage corresponding to the note value to create a sawtooth waveform option, so in this way it resembles many other DCO synthesizers, as well as use of a comparator to generate PWM. But since there are two analog oscillators they are free to drift ever so slightly about to each other, though the note relationships on each DCO hold tightly at all times. In this way the machine resembles the Polymoog (1975) which also uses two analog oscillators but digitally divides then reshapes the waveforms with analog circuitry as well. But Polymoog had a synth channel hard assigned to each note. So there were no dynamic adjustments in response to the note being played in the oscillations anyway on this machine. Rather all the frequencies were just waiting to be tapped by a key strike; already formed into the square and saw wave shapes. This type of scheme in fact dates back to the first significant synthesizer I suppose one could say; the Hammond Novacord. It used tubes to create a digital divide down in essence though the tubes of course added their own coloration to the 'square waves'. No wave is ever perfectly 'square'.
Operation
A DCO can be considered as a VCO that is syncronised to an external frequency reference. The reference in this case is the reset pulses. These are produced by a digital counter such as the 82C53 chip. The counter acts as a frequency divider, counting pulses from a high frequency master clock (typically several MHz) and toggling the state of its output when the count reaches some predetermined value. The frequency of the counter's output can thus be defined by the number of pulses counted, and this generates a square wave at the required frequency.
The leading edge of this square wave is used to derive a reset pulse to discharge the capacitor in the oscillator's ramp core. This ensures that the ramp waveform produced is of the same frequency as the counter output.
Problems with the design
For a given capacitor charging current, the amplitude of the output waveform will decrease linearly with frequency. In musical terms, this means a waveform an octave higher in pitch is of half the amplitude. In order to produce a constant amplitude over the full range of the oscillator, some compensation scheme must be employed. This is often done by controlling the charging current from the same microprocessor that controls the counter reset value.
Especially in earlier implementations of the DCO or DGO, there tend to be high frequency components on sounds where one would not expect them to be and sounds lose some of the warmth that is perceived on purely analog machines in low tones particularly.
Also an aspect of perceived warmth involves the drifting of oscillators around each other. In some later digital synthesizers this effect was modeled and could be varied by the user within the sound's parameters. Without this the digitally generated oscillations have a very precise but static nature to them of course.
I should add that someone did some research a few years back as to exactly what it was that gave the Minimoog its particularly distinct sound on some settings. It was determined that there was jitter in the pulse widths, causing a continuously changing tone in the signal! So often as early designers worked with what they had and struggled to find a way to just make things work, *artifacts* like this would find their way into the mix and sometimes wind up creating a dynamic of expression in the sound that really made the instrument in a way. The human ear can perceive many dimensions of complexity in the sound, but perhaps it actually loves to hear a battle taking place in the very fabric that the instrument is created from. Some complexity comes across as rather arbitrary. Other complexity comes across like a voice echoing from the depths of creation. Perhaps it can best be said that DCO's utterly reduce one potential aspect of complexity from the sound in order to turn attention to others without distraction.
See also
|
| |
|
|
