Ffrydiau is a Welsh word (prounounced a bit like vri-dee-eye) that translates in English as ‘Streams’.
Ffrydiau is a two stage ‘complex wave distortion’ module based on the ‘Simple Wavefolder’ or CGS52 by Ken Stone (Catgirl Synth). This circuit is significantly different from Stone’s original design (which in turn was derived from a circuit described by R.Lockhart Junior, sometimes referred to as the Lockhart Wavefolder) as it has various CV inputs, two distortion stages and a signal output that utilises phase cancellation to create harmonically complex waveforms. The input to this circuit should be a sine or triangle wave. The output stage is AC coupled and designed for audio signals. LFO signals can be applied but at lower frequencies (below 40Hz or so) DC blocking will mean the output will not be an accurate representation of the input, but don’t let that put you off trying!
This module is great for creating exotic and harmonically rich sounds. There are two distortion stages based on overloading BJT push-pull amplifiers to create non-linear crossover distortion. The two stages are summed, with the input signal, to a buffered amplifier before the main output. The input to each distortion stage is adjusted with a linear VCA, the first of which has a seperate output so it can be used independently (n.b. the output here is reverse phase), and both distortion stages have an individual output. The VCAs each have a bias control and CV input and each distortion stage also has an offset input which can be manipulated with CV or audio, this adds harmonic complexity to the output, try applying a slow LFO or stepped waveform to hear the effect. Finally, the second distortion stage can be selected to either take the input from stage one in series, or a secondary signal can be input here for additional weirdness! Happy patching!
A detailed description of the way this circuit works is included at the bottom of this page.
You can purchase a good quality PCB from me which has already been tried and tested through design revisions to work as described – coming soon
Hint: you can right click and select “View Image” to blow it up, alternatively there is a .pdf included in the repository:
- Supply Voltage: ±12V Supply
- Supply Current: TBA
- Inputs: Main signal input is intended for audio signals at ±5V, VCA CV is intended to be in the 0-10V range, but negative going signals will not hurt and you can set for this using the Bias controls. Offset inputs can be pretty much anything, but I would suggest applying signals in the ±5V range.
- Outputs: Main output is typically ±5V but applying the distortion is likely to push it to a greater amplitude, this is unlikely to be a problem within the ENB modular system, but you might want to consider this for connecting to external gear. Distortion (labelled ‘Fold 1’ and ‘ Fold 2’ on panel) outputs are similar. VCA output is representative of the input but is 180º out of phase.
Resistor numbers are found on top of the part due to the small amount of available space, this means that it can be hard to locate a part if you have soldered up the PCB – use the image below to help locate parts.
There is a layout to scale available as a printable .pdf in the repository. This image also includes the routing on the PCB so may be helpful if you need to diagnose faults.
This circuit can definitely be constructed on breadboard for testing, but it is a large design so I would recommend using two full size breadboards and to spread out the chips to allow yourself space to work. You will need a ±12V power supply at very least for testing.
I would recommend using 10K-100K potentiometers configured as an attenuator to ground for CV inputs – which will allow control over modulation depth. See the image for control wiring (right click and select ‘view image’ to blow it up):
A control PCB and front panel is also available (coming soon) – this is Eurorack panel size standard, and can take either banana sockets or mini-jacks. Be aware that the banana sockets are specifically intended to be Cinch type, as these have a good vertical depth that matches the potentiometers used. The mini-jack footprint is for the PJ398SM (Thonkiconn) sold by Thonk. Link to BOM below.
Layout (printable to scale)
The two VCAs in this design have a trimmer to adjust for DC bias in the output. As Iabc (current to amplifier bias pin on LM13700) increases, the output becomes slightly positively biased. To correct for this, you ideally need an oscilloscope to monitor the output. Make sure the oscilloscope is set to DC mode and connect the probe to R1 or R12 (I will add a test point for the next revision). Apply a sine wave to the input that is known to be correctly centred. Set the bias control to maximum, or alternatively apply a 10V DC signal to the CV input. Adjust trimmer R34 or R35 so that the output signal is centred around 0V. Do this for each amplifier stage.
A note on the output – the transistors that make up the push-pull amplifiers are pairs of NPN and PNP types, so cannot easily be matched. This means that the crossover distortion effect, which is the desired sonic effect in this design, will likely look different on the scope and sound different for any set of transistors used. It might be possible to match transistors for Hfe to get more replicatable results, however I have not aimed for linearity in this design (I may explore it further in the future) because I think it sounds great regardless. So the intention is to correctly bias the signal for DC offset at the VCA stage, and have the option to inject a DC offset with the specified input to do so, which creates interesting sonic effects. I definitely have an ‘if it ain’t broken, don’t fix it’ attitude here, but would welcome any suggestions on improving balance post push-pull amplifier stage and/or replicability. I think you will enjoy using this circuit regardless!
I recommend that you keep a copy of the schematic to hand when following this description.
This module can be broken into the following circuit blocks: the linear VCAs, the push-pull amplifiers, and the summing amplifiers to the output.
The linear VCAs use a very straightforward LM13700 application which is almost straight out of the datasheet. The LM13700 is a transconductance amplifier, that allows you to control gain with a current applied to the amplifier bias pin. It is one of the best chips still manufactured and widely available for building analogue electronic instruments for this very reason. Iabc is set against the negative supply rail, so when a voltage applied to the amplifier is around -Vsupply, the amplifier is fully closed. If the voltage connected to this point is higher, current is sourced from the pin. As the LM13700 is a current controlled device, we would normally discuss this operation in terms of current and not in terms of voltage, and current needs to be correctly limited to prevent damage to the amplifier. A range of 0 to 1mA is more than sufficient to get a wide control range, if we take our supply rails as being +/- 12V as is the case in this circuit, we could connect a resistor between the amplifier bias pin and +12V to set the amplifier gain. +12V is 24V of potential from -12V, so using Ohm’s law we can calculate and set 1mA of current using the equation R = E/I (or resistance = voltage/current) so R = 24/0.001 = 24,000Ω or 24KΩ. Doing this would pretty much fully open the amplifier.
To achieve voltage control for each VCA, a pair of op-amp inverting amplifiers are used. IC3C and IC3D make up an inverting summing amplifier and an inverting unity gain amplifier. The summing amplifier takes inputs from a potentiometer connected to the supply rails, and a CV input. Because both amplifiers are unity gain, the maths are relatively easy. The ‘worst case’ scenario at the output of IC3D is that the voltage is at +12V, which will source maximum current from the LM13700 amplifier bias pin. Resistor R39 is the current limiting resistor as described above, it is set at 47K, so maximum current through this resistor is I=E/R, current = 24/47000 = 0.00051 or 0.5mA, so well within the safe range of the LM13700 current control limit. I found that when the bias control was set to -12V, the amplifier would not be fully closed. A simple solution for this was to add resistor R37 to pull the amplifier bias pin closer to -12V.
The output of the LM13700 is high impedance, considered also like a current, and needs to be suitably buffered to be of much use in most applications. The LM13700 includes a Darlington pair amplifier for each transconductance amplifier for this purpose, but being a simple transistor buffer, they will introduce diode drop to the output. I usually ignore the buffer and instead used a JFET input op-amp which does a much better job. IC4B is effectively an inverting amplifier, there is no input resistor but R19 helps to reduce excessive feedback. R17 can be used to program gain and the output level – I have found 75K to be a good choice to get something close to unity (but note many other factors play into this, you could also attenuate the input or adjust Iabc to also effect the output). As a final word about this VCA design, the LM13700 will distort (in an interesting way, see Prism!!) if the signal input is anything more than a couple of hundred millivolts. But the chip also includes a ‘linearising diode’ pin for each amplifier which can be tweaked to allow signals with greater amplitude to be applied. In this case, the linearising diodes are biased with a current close to 2mA using R32 which I have found gives good performance, note that the input is also scaled down with the voltage divider made up by R33 and R34. R34 is a trim potentiometer, as Iabc increases, the output of the transconductance amplifier becomes more positively biased, this trimmer allows for adjustment to counter this bias at that output.
The VCAs each output to a push-pull amplifier made up by a NPN and PNP transistor pair. This is a useful configuration for a number of applications, but is subject to ‘crossover distortion’. Normally, crossover distortion is countered using negative feedback in some form, but this circuit exploits this characteristic for a sonic effect. The VCA stage sets how much signal is fed into the amplifier and once a certain amplitude is reached the transistors will begin clipping, but the clipping effect is very interesting and is almost like the signal folding in on itself. The transistors will not clip the signal equally, as they are not equally matched – and to match them would be difficult (perhaps futile?), so the distortion is particularly non-linear and will be a little different for every circuit built.
The output of the push-pull amplifiers connect to simple op-amp inverting amplifiers, the first stage distortion has a pair of inverting stages (IC1C and IC1D), the first amplifies the signal with a gain of -10. and the second re-inverts. The second stage has a single amplifier (IC1B) which inverts and boosts the gain by -10. IC1B and IC1D are AC coupled and sum into IC1A, a final inverting summing amplifier which is the main output of the module.
IC1A also has the main signal input summed in – so 180º phase inversion of the input is present at the output. When the VCAs increase the signal into the distortion stages, the outputs are mixed in by IC1A. Phase cancellation occurs here which results in the shaper distortion effect. IC1A is AC coupled because the push-pull amplifier outputs tend to incur DC bias, so this goes some of the way to correcting this problem. An alternative solution might be to build IC1C and IC1B as AC integrators, where a capacitor is included in the feedback loop – there are a couple of things to potentially be explored in the future which might improve performance here. The capacitors are specified at 4.7µF for a reasonable trade off between DC blocking and allowing for lower frequency signals like LFOs.
The last thing to mention in this description is the various inputs, outputs and routing in this circuit which haven’t already been mentioned. The linear VCA for stage 1 has it’s output broken out, and can be used independently – note that the output is 180º out of phase with the input. Each push-pull distortion stage has a seperate input which can be used to apply a DC bias – this means that the transistors will clip the signal differently, which can produce interesting effects particularly when modulated. Each distortion is broken to an individual output before the final summing amplifier – note that these are not AC coupled, so you might want to put a capacitor in series with the output if you want to block DC. The second VCA has an input selection, it can either take the output from the first distortion stage (to apply more crazy distortion!) or it can take an external signal – try exploring mixing up audio signals with LFOs for interesting results!
I think that about wraps it up, I hope find this information useful and have lots of fun with Ffrydiau! Happy patching!