You can see what it does and hear how it sounds here:
On the left oscilloscope you can see the input signal to the wave folder, on the right oscilloscope the output signal of the wave folder. The input signal stems from a Behringer Crave. It is the basic saw tooth wave. While I’m turning the knobs, in the background you can hear the accompanying sound. The top knob sets the dry/wet ratio. The 2nd knob the DC-offset fed into the system. If the input signal is strong enough, the LEDs light up and if the DC-offset is zero, both will be equally bright. The brightness depends on the signal’s amplitude. The switches are there to select either the red or green pair of leds that are used for the wave folding. There are 2 folding sections in the circuit, both have their led pairs.
The recording starts in the dry situation, so you see the input signal on the RHS at the start as well.Then I show the wet signal and fiddle around with the offset. Typically this leads to clipping. But the whole effect is very much dependent on the input signal’s amplitude, so off camera I now and then change the input voltage and adjust the mixer’s gain to end up with a somewhat constant volume of the signal. An example of that can be found around 2m02 when the sound changes quite a bit because I change the input signal’s amplitude.
The schematic diagram and STL file for the Euro Rack format front panel can be found at this Fold-O-Matic repository.
[Edit] Please check the rest of this thread for updated versions of the schematic and build.
On RV2 I think pins 1 and 3 are reversed. As shown you’d get zero CV when fully clockwise and maximum CV when fully counterclockwise.
Why are R6 and R7 present? That is, why a unity gain non inverting amplifier instead of a voltage follower?
The schematic labels the switches as two halves of a DPDT but from the photos it looks like two SPDTs instead, is that right?
Finally, can you tell me how the wet/dry control works? One gang of the pot has the dry signal at one end and the wet at the other, and I was thinking that was all that would be needed. But there’s a second gang in the wet output op amp feedback. Assuming the pin numbers are correct as shown and I haven’t gotten confused, when fully counterclockwise there is only dry signal and when fully clockwise there is only wet, but at the same time fully counterclockwise doubles the wet signal gain (if the gain trimmer is near maximum, more than doubles if it’s lower) compared to fully clockwise. So turning clockwise increases the wet to dry proportion but decreases the wet signal… is that right? What’s the purpose of the gain decrease?
I didn’t know about the clockwise and counterclockwise numbering convention of potentiometers. So this I will change if what you say is the rule. I will rotate Rv1 then as well and add a more meaningfull annotation.
That is, why a unity gain non inverting amplifier instead of a voltage follower?
I think the designer wanted to add up some signals but wanted to keep the phase of the signal without needing extra components. Had U2 been an inverting amplifier that would not be the case ( keeping 2 inversions in mind in the following section U1A, U1C and U1D ). But as always, there are more ways to implement this, so it might just be idiosyncratic.
As a KiCad noob I was happy to find the symbols I’m used to using for switches in the KiCad library and didn’t pay much attention to the labels. I’ll change the labels into ‘timbre’, because changing the diodes sort of has that effect on the sound.
The rational is that of a common wet/dry pot. But what I found is that the gain of the circuit is much higher for the wet signal than for the dry signal so the gain of the final stage is lowered when the amount of wetness is increased.
The gain should decrease if wet/dry is turned towards wet. So I think the gain pot should be wired the other way around (following the numbering scheme convention you mentioned). The trim pot is there to set a base gain factor. The setting is best done empirically / by ear I think.
Of course if you’re wiring pots it doesn’t matter because you can connect them any way you want, and for instance I’ve seen the same reversals (inconsistently) in Ray Wilson’s schematics. It becomes problematic when doing PCBs for board mounted pots though. And, as below with the wet/dry pot, the designer’s intentions may be misunderstood if the numbering is reversedl.
Ah, so this wasn’t your decision, okay. But a voltage follower doesn’t invert.
Still not getting it, why not just a fixed reduction in the wet signal gain?
I get this kind of thing backwards sometimes, but I think that behavior is what you had the first time. With the original connectivity, at full clockwise 2 is shorted to 3 and 5 is shorted to 6 so it’s full wet and wet gain is minimal, right?
The wetness is variable therefore the increase/reduction of the gain should be variable. It should sort of result in a more or less constant volume output.
Rats, you’re right. It was correct the 1st time. Something got me convinced it wasn’t. Can’t remember what that was.
Actually, that’s a gain of 2 (G=1+R7/R6 for a non-inverting amplifier). If you want a gain of 1, you need R6 to be infinite or R7 to be 0. Either case, you end up with a voltage follower, magic, isn’t it?
Oh duh, you’re right. I was thinking inverting amp gain. Thanks!
Also it seems the “wet” signal already has some dry in it due to U1C. Seems like one of the dry paths could be eliminated.
I’m puzzled by the gains. First stage gain is 2, second stage dry path gain is 1, third stage gain is as high as 20 depending on trimmer (but only when close to fully dry, still as high as 10 when fully wet)! Why so large?
Well, it depends. Not if you see the wave folder as a module that adds something to the original signal. Then the original signal should always be there. But you might want to blend out or bypass the effect entirely, that is the intention behind the wet/dry pot.
In stead of fixed gain factors I decided to make them variable using pots and vactrols. I did not find the official kicad symbols for the vactrols, nor did I want to spend any time in making a new one, so I drew LDR and LED separately and named them appropriately. This should be clear as well, I think.
The vactrols I bought via aliexpress. Their type is VLT5C1.
My KiCad installation has a number of optocoupler symbols such as this one
although a two unit symbol that lets you put the LED and LDR in different parts of the schematic would be a good idea. Maybe I’ll work one up.
… although how good an idea this really is, I don’t know; normally different units are for independent separate components within a package, which isn’t really the case here. Maybe it’s best to use a 1-unit symbol. Anyway, there’s a choice now.
Given that I already had printed a 5HP module panel, and there was no room for the extra pots and CV-inputs, I improvised and added a control panel, the so called FOMC (Fold-O-Matic Controller).
I was looking at the schematic for the NLC Timbre wavefolder, which is based on a Buchla design, trying to figure out how it works. It’s based on a stack of five cells like this
which I thought looked crazy, because Vk is connected to the inverting input and the non inverting input is connected to ground, so Vk should be zero. Then I found this paper (which that figure comes from):
which discusses how the circuit works and how to simulate it. The point is these wavefolders rely on op amps’ deviation from ideal behavior when you drive them close to their rails (and in the Buchla circuit the op amps are powered with only ±6V), in which case Vk does indeed become nonzero. Thought that paper might be of interest to anyone wanting to mess around with wavefolding.
I didn’t so much like the wet/dry pot in the module. It resembled more a feedback option in practice. So I changed that and changed the gain factors of the op-amps (U1A and U1B) as well. Because one is inverting and the other isn’t they can be used to partly cancel each other’s effect out, but because of the wide range of the gain factor you can choose the cancellation point and measure of cancellation along the I-V curve of the LEDs. The amount of variation one can achieve in this way combined with adding the dry signal and switching the LEDs is quite large. Yep, this is a keeper!
One more change, I swapped the output elco for a non polar capacitor because the signal at the output of the opamp can have a negative or positive DC offset.
Note: I left out the attenuation pot at the CV-input. This was because I didn’t have enough space on my panel. You can of course leave that in if you like. I will make a passive or active multi with attenuation in the future, so that function will be implemented elsewhere.
