MODULES with corresponding design/inspiration. Devices that 'made the cut' into the current rig are highlighted.

Noise Generators
- MIDI VCO Oscillator - based on MDCO-3 Chip from MIDIMUSO
- CEM3340 VCO - Look Mum No Computer
- Electric Druid VCDO
- Tri-oscillator drone - inspired by Look Mum No Computer
- White Noise - Ray Wilson

Filters
- CGS30 State Variable BandPass Filter - Ken Stone
- Vactrol Controlled CGS30 Band Pass filter - based on Ken Stone Design
- Vactrol controlled 3-Pole filter - own design Sallen-Key filter
- Vactrol controlled 2-Pole BPF Filter with built in LFO - own design.
- Manual Control 4-pole LPF - own design
- Vactrol controlled 4-pole BPF with built in LFO - own design.
- Manual control Inductor Based LPF own design.
- Vactrol controlled 2-pole LPF with built in LFO controlling both filter and reso - own design.
- MS-20 style LPF/HPF VCF following Moritz Klein design.
- Diode Ladder filter following Moritz Klein Design.
- Vactrol Controlled Moog Ladder filter based on Eddy Bergman/Rene Schmitz/Kassutronics/Yusynth plus my own Vactrol controls.
- Vactrol Controlled Wasp filter, Rene Schmitz plus own design for Vactrol control.
- Manual control Using UAF42 , Texas Instruments design.
- MS-20 Style Filter with LM13700 , based on Rene Schmitz design.
- 2 * LPF filter based on the Korg PE-1000 design, with my own tweaks.

Sound Shaping
- Low Pass Gate - Synth DIY Guy
- Vactrol VCA. own design
- LFO/VCA - Nicholas Woolaston LFO in same panel as own design Vactrol VCA
- Attack-Release - Nathan Ramsden
- Reverb - PT2399 reverb using Ebay board.
- Sub-Oscillator ; 1 & 2 Octave Sub, based on Thomas Henry Design
- Ring Modulator - Ken Stone

LFO
- Ken Stone Dual Utility LFO design adapted for 'Positive Only'

Sequencer
- Baby 8 Drum Sequencer - based on Look Mum No Computer concept
- Kick Drum Synth - Ken Stone

Utilities
- 3 channel Mixer / EQ - own design based on Ken Stone
- Inverting / Non-Inverting Amp/Buffer - own design
- Splitter / Buffer - own design
- Uni-Gain Mixer - Richard Jones design
- 2 Channel Performance Switcher - own design based on Keith Emerson concept
- Portamento - based on Jesse Stevens design.
- Sample & Hold - Moritz Klein design.
- Foot Pedal CV Controller - own design.


Keyboards and Controllers
- Arturia Keystep 32
- Modded Casio SA-46
- Alesis Q49
- KORG SQ-1 Sequencer


FORMAT :
Voltage - plus/minus 15V
Jack Size - 1/4 inch
Power Adapter - 3 pin header
Audio Signal Level - 5v peak to peak
CV Signal Level - Positive 0 to 10V
Panel Size - Moog style 8.75 inches

In general, my idiosyncratic format has kept me away from kits or finished module purchases. As a result I think I've dived a bit deeper into how things worked with the aim of creating a playable musical device.

My key reference books are Ray Wilson Make:Analog Synth and Horowitz & Hill The Art of Electronics

Synth as of June 2024
Here is my rig as of June 2024 in its new portable case. The vibe is definitely grungy - no attention whatsoever has been given to appearance.

TEST RIG Synth Test Box

Here is my synth test box, featuring a triangle wave LFO, a manually controllable CV with digital voltmeter, a pair of +-15v power supplies, a momentary gate button, a square wave audio output & probe and a white noise generator.


OKAWA DESIGN UTILITIES.
Okawa Electronic Design Tools
I think one of the most important steps you can take in Modular Synth Building is to embrace the mathematics of electronics and break away from just building circuits you got from here or there on the internet. In fact once you get to trust the math, even modelling in Spice is often unneccessary. So a shout out to Okawa Electronics for publishing this excellent set of design tools, simple to use and a great set of output data and graphs.



VACTROLS
A word about Vactrols; basically any CV controlled synth module normally requires a form of voltage controlled resistance, which could be a JFET, an OTA, a diode but my preference is very much the Vactrol. as you can see from my module list above.
A Vactrol in it's simplest form is just a Light Emitting Diode (LED) illuminating a Light Detecting Resistor (LDR)....thus as the LED emits more photons, the resistance of the LDR falls. So why do I like Vactrols so much ?
1- Vactrols can be precisely measured & calibrated as to their resistance when a given current is applied, thus they give nice predictable frequency response in a filter situation and will obey the calculated values for the circuit.
2- Vactrols are very rugged and can switch large voltages and currents, avoiding complex scaling and post-filter re-amplification. In other terms Vactrols have excellent dynamic range and support a high signal to noise ratio.
3- Vactrols have an LED and LDR response time that to my ear is quite musical and can help smooth out the response to a jittery CV. Different choices of LED size, color and LDR type will certainly have an impact.
4- The LED circuit can be electronically isolated from the audio path, thereby eliminating any noise.

So I'm definitely a vactrol fan. I typically make mine using Red LEDs, because the forward voltage is low,about 1.7 volts and 5516 LDRs that run from from 5K to approx 800K ohms. IR Leds have even lower forward voltage so I'm about to make a batch of them and see how they work.
Here's an example of one of my vactrols - note identifier E. And also a little vactrol jam. Vactrol - note identifier E Push ME

ON PLAYING IN TUNE
I really like to just plug in and jam but with a traditional CV driven VCO it is just incredibly difficult to maintain good pitch and correct octave tracking. So my rig utilizes a MIDI audio sound source , a MIDIMUSO MDCO-3 twin oscillator. The MDCO-3 in produces two awesome square wave outputs that are always in tune and track perfectly across the octaves.

I do also have an Electric Druid VCDO that is driven by V/Octave CV, which is produced directly by the Arturia keyboard. It's got onboard quantization and can be tuned to agree with the MIDIMUSO with some effort but it does have a broad range of knarly waveforms, so its kind of fun.

FET Variable Resistor Investigation....

As you can see from the above my preferred variable resistor is the Vactrol, but Vactrols have one main disadvantage which is slower speed to light up and go off again. For a really deterministic type of response I've been looking at FET type solutions.
It's pretty well known that FETs have a so-called Linear or Ohmic region of amplification where they exhibit resistor-like functions for a certain range of gate voltages. But it turns out there are four commonly encountered types of JFETs and MOSFETs and each type will need a different pattern of CV to work effectively.

For example ; my Arturia Keystep produces a +5V Gate CV. If I wanted a VCA that passed signal only when the gate is active, my only option with the raw Gate CV is to use an N-Channel Mosfet. The N-Channel JFET and P-Channel MOSFET only respond to -Ve CV and the P-Channel JFET is fully conductive at zero gate voltage, which is the inverse of what I want.
Any other type of FET device would require additional circuitry to invert/bias/scale the CV signal.

Device Channel Mode Conductance at Zero Gate Volts Max Conductance Volts CV Type Example Devices
JFET P-Channel Depletion Max Zero +5 (min) to zero(max) 2N5460 2N5461
JFET N-Channel Depletion Max Zero -5 (min) to zero(max) MPF102 J112
MOSFET P-Channel Enhancement Min Nominal -12 V Zero (min) to -12 (max) IRF4905 IRF9540
MOSFET N-Channel Enhancement Min Nominal +12 V Zero (min) to +12 (max) BS107 IRF540


FET Operating Regions


The above diagram (for an N-Channel MOSFET) is similar to those seen in FET Datasheets and it is worth getting to understand how it works.
VGS is the Gate to Source Voltage and in Synth terms is effectively the CV.
VDS is the Drain to Source Voltage and again in Synth terms is the audio signal.
ID is the output Current and therefore within the 'Ohmic' region this is effectively the inverse of resistance - higher current equates to lower resistance.

Each of the curved lines corresponds to the Current-to-Signal (i.e. Id to VDS) at a specific CV (or VGS) voltage. The graph is almost like a 3D diagram in 2D, with the CV defining a surface extending into the page.
What this tells us is that once the Audio Signal level (VDS) exceeds about 2V, that the MOSFET will enter Saturation mode and emit constant current ; in other words for synth purposes
we will need to scale our audio levels to about 2V and use a gain stage to bring back to synth line level after processing.

In a synth situation the CV will be constantly changing and we'd like to get the most dynamic range of resistances, therefore using the standard synth range of about 5V seems absolutely fine.

So my conclusions so far are as follows.
1 - There are Four different FET types that look relevant to Synth makers. Each has a specific behaviour relative to CVs and so DESIGN is the key word to specifying the right FET device for a certain use.
2 - Approximately 5v CV and Audio less than 2V seems like the right kind of scaling.

So far I've built a basic VCA just as a trial run. My intention now is to build a VCA and a Sallen-Key VCF to prove I've got things about right.....watch this space. Simple MOSFET VCA Schematic

Here's my basic VCA schematic - it works like this. R1 & R2 create a voltage divider that attenuates the input signal to prevent the signal driving the MOSFET into saturation. R6 & R7 likewise form a voltage divider on the CV.
Q1 is a BS107 N-Channel MOSFET and per my chart above has maximum resistance at ZERO volts and the resistance then decreases as the CV increases in a positive direction.
The CV does not need to be buffered because the MOSFET input impedance is very high & it does not consume current.
Thus Q1 & R3 form a voltage divider....when the CV is low the effective resistance of Q1 is high relative to R3 and as the resistance of Q1 falls, more and more of the signal is 'seen' by the OpAmps - the resistors R4 & R5 define the output stage gain which restores the signal to synth line level.
Simple MOSFET VCA Output

The above trace shows how the VCA responds to a varying amplitude signal modulated by a triangle wave LFO. The MOSFET is probably still going into saturation at the top of the LFO signal as you can see a flattening at the top of the curve with the 'soft knee' tube-like distortion that MOSFETs are known for. This gives the module a sound which is rather like a tube compressor. When the signal amplitude gets to the 'knee' in the response curve the input waveforms start to get 'rounded off' leading to a pleasant harmonic distortion and a gentle compression. I found that my Arturia Keystep is producing a gate voltage of +12V and actually the MOSFETs can easily handle this and it gives audio headroom up to I think about 4 volts.
Some trimpots in the design could help fine tune the signal or CV attenuation. When I solder up the final version I will include those tweaks as well as variable output gain.

Having got a working MOSFET VCA I am moving onto a VCLPF Sallen Key design.

FET VCLPF Schematic
This is a tried and true LPF topology, and as with the VCA the concept is to attenuate the input signal, perform the filtering and apply gain to get back to synth line level.

So my basic design consideration for this filter is that I'd like it to sweep between the the upper practical end of human audible frequency (lets say 10K Hz) down to a few 10s Hertz at the low end.
Without putting addiional resistors in parallel or series with the MOSFETs, the conductance of the MOSFETs is essentially a given, for the range of about 0-12V CV. THerefore we are left with a suitable choice of capacitor. We know from the VCA build that the MOSFET Resistance full 'on' must be small relative to 1K Ohms and fully 'off' must be large relative to 1K.
So if we assume 500 OHms and 50K Ohms , we know that the equation for cutoff frequency is F = 1 / 2piRC.
And plugging in some capacitor choices we find that .031 muF gives us 10200 Hrz at fully 'on' and about 100 Hz at fully 'off'.
So that will be my start point in the prototype build but I'll use sockets to make it easy to swap capacitors.

I built the above schematic and started testing, initially on my test rig, but when I realized the sensitivity to both the CV and Audio signal I started testing on the real synth......learning points so far...
1 - the Mosfet gate MUST have a resistive route to ground, because otherwise it stores charge like a tiny capacitor and will tend to stay at a high CV level, effectively always ON.
2 - the synth amplitude appears to be capable driving the MOSFET right out of the Ohmic region and into saturation. This equates to approx zero ohms and cansequently the filter goes to a supersonic high cutoff frequency. Therefore I've put 500 Ohm resistors in series
with the Mosfet Sources and this definitely goes in the right direction, but I'd probably go with 1K in the final schematic. The problem is rooted in that the resonance circuit needs to feed back a reinforcing positive signal and this can easily drive the filter into saturation.
3 - the Mosfet (IRF540) turns out to have a threshold cutoff at about 2 volts where the resistance is Maximum, i.e. Fully Off. I am therefore designing a bias circuit to stop the gate falling below about 2 V, which should mean that the resistive behaviour
of the Mosfet is always under CV control.
4 - As you might expect the Resonance control is not perfect. I'll stabilize the points above and then try to fine tune the resonance as the last step to avoid the problem of pushing the envelope into saturation. 5 - In short the fixes to above push the component count up....it remains to be seen whether the result justifies the effort.

Push MEPush ME
Well, the 2 samples above probably are about as good as i can get with this design. The basic problem with the Mosfet filter, is that the input signal needs to be pretty low, say below 1 Volt and then re-amplified for the output.
This makes the filter quite noisy because any noise also gets amplified. Also, the resonance you want in a good filter is just positive feedback, which increases the signal through the Mosfet, and of course pushes it out of the ohmic region.

The good thing is the responsiveness of the Mosfet to using a gate signal to modulate the sound, which is really its one intrinsic advantage over a Vactrol.
I think that probably concludes the FET investigation.....





So this has led me down into exploring PWM and I'm looking at a PWM LFO specifically to create waveforms to drive Vactrols, especially in the shape of a VCF.

So I first created 2 Vactrols using Green LEDs and LDRs that go from about 1.5K (Fully Lit) to 400K (Fully Dark) and set up an Arduino program to apply PWM Voltages across the LED.
Now the Arduino allows a range of 0-255 to control the PWM duty cycle. I soon found that anything greater than about 85/255 meant the Vactrol was maxed out, i.e. minimum resistance, so I created a program to change the PWM from 85 to 1 in increments of 5 and back.
The equivalent analog voltage to a PWM factor of say 125 is therefore ((125/255)*5V) = 2.45 V. So it's at first sight quite suprising that even a PWM factor of 1, still results in a Vactrol resistance drop....how can this be when the equivalent analog voltage is just .02V ?
The answer I believe is the hysteresis of the LDR response, that rapidly drops resistance when illuminated but relatively slowly returns to high resistance when dark. This effect means that the LDR tends to smooth out or average the PWM signal and in fact I think this should help the PWM LFO VCF work well.