About
After realizing that the Compact Clock Source has a lot of flaws, i decided to improve it. Firstly, the old design has two bad clocks, and i felt like i need one good clock. The crossmod system is a lot of fun and brings nice results, but it's patchable on a lot of other modules i have and not exacly useful for a 'main' clock source. The range of both halves was very decent, despite the range switches - i wanted something that covers the entire range at once. Finally, the XOR of the two clocks barely ever got used, so i wanted to replace it with something i actually lacked all this time - a small on-board clock divider.
After taking all these requests from me to me into account, i came up with a voltage controlled clock core that has an absolutely insane range, from ticking once every few minutes to literally going into sound range. It has the frequency knob, and the two CV inputs. One, called CV1, gets amplified by two internally when its attenuator is maxed out, so it can modulate the clock over its entire range. The other, CV2, is a slight modulation input which gets divided by 10 when its attenuator is full clockwise (and even more when you attenuate it further).
Below the frequency controls are the outputs. Main, divided by two and divided by four. If you think of the 'main' clock output as of 1/16ths, cause that's what people usually percieve their clocks as, then /2 would be your 1/8ths and /4 are 1/4ths - handy for your kickdrum heavy techno. The divided outputs also have inverted copies broken out to the panel. The /2 inverted out is normalled to CV2 input jack for and instant swing control.
For a design just a couple steps harder than SFP06, this clock source is times better, providing a huge range of clocks and great, almost exponential voltage control over the frequency. The archive downloadable through the link in the page header contains the schematic .png, as well as two .txt files of Falstad models - one for the initial theoretical idea, other for the actual stuff that I made - had to change one place up due to crosstalk of op-amps that happens in practice, but is not included in the model.
Schematic
This started off as a "what if i try to replicate the 40106 VCO's core without using a CD40106" idea, and ended up as a simple and efficient clock generator design. The core of the device is IC2 - a TL072 powered from ground to +12v, and most importantly IC2A in a peculiar configuration. The resistor from its output to its own non-inverting input, where it meets another two resistors to +12 and ground, replicate the 40106's hysteresis, but better. The diode to IC2A's inverting input and capacitors C1/C2 are what make the thing oscillate. Let's forget about T1/Q1 behind the core for a second, and suppose the output of IC2A is low (0v) and capacitors hold no initial charge. Then the voltage at IC2A's - is about zero, whereas at + it's about 4v or something, surely more than zero. IC2A flips up, charges the capacitor C1 (and C2, if the range switch is down and it's connected in parallel to C1, increasing the overall capacity). The diode has tiny resistance, so the caps almost instantly charge up and raise the voltage at the inverting input. It becomes more than whatever is going on the non-inverting input, which should be about 6v, and IC2A flips down, bringing the non-inverting terminal to about 4v again. The voltage at the capacitors-IC2A inverting in junction is above zero, so the diode now doesn't conduct anything, and the capacitors very slowly discharge through all possible paths (mostly op amp's voltage droop). IC2B compares the sawtooth that happens because of instant charging and slow discharging of the caps to a set voltage, and outputs a rectangular wave from 0 to about 12 volts - the clock signal that we need. The trimpot can be (and is, in my unit) replaced with two 100K resistors as a voltage divider from +12 to gnd - results in narrow, but usable and visible pulse.
To make this construction's oscillation frequency voltage-controlled, we can simply provide an alternate path for the capacitors to discharge through. T1 does exactly that - adds another path from the caps to ground, which can be controlled by messing with its base. Although it could potentially be left alone and i could call it a day - like in the 40106 design, it is controlled by Q1 for exponential response and better thermal characteristics. Q1 in its turn is controlled by a voltage that is the sum of the frequency knob, CV1 (amplified by two) and CV2 (affects the circuit a little). The summation is done with another TL072 - this one is fed from +12/-12v, unlike the first one, otherwise this design could be done using a single TL074. It does the usual double-inverting summation, which is what you should use instead of non-inverting single op amp summation, as the double-inverting way has the benefit of virtual ground at the summing node and almost completely eliminates crosstalk between the inputs. The result of the summation is divided throuhg R4/R5 by about 50 - this fits it to range acceptable for Q1.
Finally, the bottom half of the schematic is the outputs. This part is optional - you could just take IC2B's out and be happy, but having this simple addon instantly closes a few usual functions. Two halves of a CD4013 are both wired up as frequency halvers - feed it a pulse, and it outputs a 50% symmetry squarewave that's exactly two times slower. They are cascaded one after another to have /2 and /4 outputs. Another perk that comes with the CD4013 is its dual output, inverted and non-inverted. This way, we can have inverted and non-inverted subdivisions right on the panel. The inverted versions are not only percussively useful sometimes - for example, switching between inverted and uninverted 1/8ths while having steady 1/4ths yields a nice offbeat effect - but also there's an exploit: feed the inverted /2 output back as a CV for the clock, and you have a swing control by simply forcing every other clock pulse to be faster. I have done this internally by normalling the not/2 output to the CV2 input jack - instant swing control!
P.S. the NPN and PNP transistors T1/Q1 can probably be any general purpose ones, but i used 2N3904/2N3906, if that matters. Diode is 1n4148 - standard general purpose crap.
