Light and shadow in DCO (Part 1): clock and divider

I am reproducing the voice of the Crumar Bit One on a breadboard. The deeper I dive into it, the clearer my verdict becomes: it seems to me that Crumar tried to copy some of Mario Maggi's original ideas from the ELKA Synthex into the design to make something unconventional (which part of it is responsible for that unique sound remains to be seen). Yet, it’s like an experiment that was stopped or – rather – forced into production in full flight. The design flaws are so severe, that I can only imagine what the go-to-market management decisions must have been like. In this series I want to show you some of the outstanding features and pitfalls of DCO concepts. For this I compare the Roland JX8P (serves as a classic example of common DCO synth design) and the Crumar Bit One (unconventional design).

If you’re interested in the electronics of DCO synths, I recommend Stargirl Flowers’ essay: The Design of the Roland Juno oscillators. In relation to the Bit One I also learned a lot from documentation by Tauntek and by Andrej from Synthronics.

The common DCO design

DCO synth oscillators can be thought of simply as derivatives of a (master) clock source through a divider (deriving the basic tone frequency), a ramp builder and a waveshaper, as shown below.

The clock provides a medium to high frequency pulse, i.e. 2-8 MHz, depending on design. The divider is a programmable component that divides this frequency to the selected register and then to the target pitch (i.e. the note that is being played). The ramp builder is a technical component that creates a more useful shape (a ramp) from this which is then used in the waveshaper to obtain different final wave forms, such as sawtooth (which is in essence the same as ramp) or a pulse with different pulse widths.

The “D” in DCO refers to the fact, that the divider component must be digitally controlled by a CPU. For achieving a certain pitch, the CPU “tells” the divider the correct divisor in digital form, for example a 16 bit number.  Maybe I should say "closest" instead of "correct", though. In one of the upccoming parts of this series I will go into more depth of what is meant by "division" and how that relates to pitch.  

The three decisive advantages of the digitally controlled oscillator design over the use of a voltage controlled oscillator (VCO) are:

1.     The master clock source provides a perfect reference frequency which does not depend on temperature. DCO synths typically provide the luxury of letting you choose the tuning in steps of 1 Hz around 440 Hz.

2.     The divide mechanism is digital, so that the creation of the basic tone frequency from the master clock is independent from any differences in material and circuit components. Once the master clock is tuned, then the derived frequencies across all octaves are automatically also in tune. 

3.     All oscillators which derive the frequency from one clock are always in tune

Remember: in a synth run with voltage controlled oscillators, you need to take care of tuning for each voice across all relevant octaves. By contrast, when you switch on a DCO synth, in principle it is immediately in tune and will stay so whether you play in a cold basement or in bright sunlight on an open-air stage.

The figure below illustrates how the CPU would control the oscillator through instructing the divider and the waveshaper.

In the JX8P, for example, the CPU controls the note frequency, the portamento and pitch modulation by LFO and ADSR in real time. As a note on the side, the JX8P does not support a static pulse width parameter, let alone pulse width modulation (PWM). In the basic design pattern this would have been quite straight forward and there are many other similarly constructed DCO synths (especially from Roland) which offer that. It really is a pity that Roland did not implement it.

Analogue controlled DCO

One limitation of the DCO design is that the control actions on the divider are digital by nature. They can only be changed at a limited resolution (digital increments, e.g. 16 bit) and only at the maximum frequency at which the digital components can be reliably programmed. It is not that much of limitation, though. For playing and performance, the digital update speed is more than sufficient. However, you may reach limits on envelopes for very short times and with LFO at high frequencies. Still, it works fine for most applications and it is also not a concept that only applies to DCO synths: from the 80s on also VCO synths use CPUs and digital control methods. The main difference to the DCO divider logic is that in VCO synths the digital controls are converted to analogue control voltages using digital analogue converters (DACs).

Now, the Bit One is quite different in its control concept. It uses the same programmable divider chip as the JX8P (this is the 8253, originally by Intel) but relies partly on non-digital controls. Crumar achieved this by introducing a voltage controlled master clock as opposed to a super accurate crystal clock as in the JX8P, and in fact all other DCO synths I know well enough (with the notabe exception of the ELKA Synthex which will be explained in an upcoming part of this series). This is shown below.

Of course, the first thing you notice is that with this design you lose the first of the three advantages of DCO architectures: this design requires tuning because the clock frequency depends on the control voltage in a material specific way. Consequently, the Bit One has a little knob on the rear panel which changes the overall pitch by about +/- 2 semitones. Incidentally, it has also 4 trimmers on the main circuit board which calibrate the clock frequency.

So the immediate question is: why on earth would you want to choose such a design?  In essence this design comes from the ELKA Synthex. It is only speculation but I would imagine that copying this design was meant to distinguish the Crumar Bit of from the Roland, Oberheim and other DCO synths at the time. Therefore creating a synth with a twist, that makes pitch related controls fully analogue. A second aspect is that the voltage controlled clock (or rather: the two clocks, one for each of the two oscillators per voice) is not fully stable and has small random variations which adds to the liveliness of the Bit One. Then of course, there may also have been lesser known technicians who worked for ELKA on the Synthex and later joined Crumar. It's all within a few kilometers around Ancona anyway.

Finally, for me the most important feature of the Bit One voice design is that allows for spectacular oscillator detuning (there are two oscillators per voice in the Bit One). In that department the Bit One totally outperforms the JX8P and other popular DCOs including the Juno series, all of which heavily rely on their chorus effects instead. I will get deeper into this in this blog.

Epilogue on part 1: great idea with equally great flaws

The modulation of pitch by ADSR is probably the most useless feature of the Bit One. Observant readers may have spotted this drastic design flaw already: since the pitch modulation by ADSR operates on the clock directly, it affects all voices which derive their frequency from this clock. Even if we put aside the unfortunate fact that the ADSR pitch modulation in the Bit One can only be activated but not controlled in intensity, it is clear that this is not what musicians would expect. It is basically paraphonic, i.e. when this modulation is active and a new a note is played, all currently sounding voices are pitch modulated as well. Listen closely next time you switch on your Bit!

I guess, the designers got carried away by the idea of having an analogue modulated DCO but the logical limitation becomes painfully evident when you realize that it is one of the key success factors of DCO synths that one clock drives all voices. Or is it not the only way?