I've just posted a Youtube video of a circuit I built -- a signal modulator based on an RF unbalanced diode mixer. This circuit uses the small non-linear response area of a single diode to create combinations of sum and difference frequencies of two input signals (or one input signal containing multiple overtones). Radio designers use this type of circuit to "downconvert" received RF signals to a lower intermediate frequency, which makes it a lot easier to design the radio's signal processing circuitry. We can use the same circuit for electronic music to generate non-harmonic overtones. (In the RF circuitry literature, there is a class of related circuits that all use diodes to do frequency mixing functions. What we call a "ring modulator" originated as a more sophisticated version of the circuit presented here.)
Note that this circuit is not a clipping circuit. We are not allowing the input signal to make excursions to either the fully-off or fully-on regions of the diode's response. All of the effects you hear on the video stem from the signal being in the diode's non-linear response corner. If you watch the scope in the video, you will notice that there is no clipping visible.
Here is a schematic of the circuit, as built. Double click to see a larger version:
How it works: The two 33K resistors serve as a passive audio combiner (okay, a mixer, but don't let that usage get you confused). The 100K pot is used to reduce the signal to a low level; the circuit is designed to take signals at up to modular-synth voltage ranges. The pot has to trim the input level down to the neighborhood of 0.2V peak-to-peak. In the center of the circuit, the 22K and 220 ohm resistors and the 5K pot form a voltage divider, which adds an offset voltage to the signal passing into that part of the circuit through the leftmost capacitor.
The 5K pot is used to trim the voltage divider to about 0.5V, which puts it in the center of the "corner" area of the diode's response, which ranges from about 0.35V up to the 0.65V level at which the diode is fully conducting. The input signal, if properly attenuated, will be 0.4 to 0.6V peak-to-peak when added to this offset. It then excites the non-linear response range of the diode, generating various sum and difference frequencies.
The opamp is just to boost the signal back up to modular voltage levels. The cap on the right keeps the DC offset out of the opamp's input. The two resistors in the feedback loop set the opamp's gain at 10. The 100K resistor on the input is to bleed off the opamp's input leakage current, which would create a DC offset in the output otherwise (found this out the hard way). There is no particular reason why I used a LM358, other than that's what I had; a TL07x would probably do just as well.
There are a couple of things that I would have done differently if I had had the parts available. Both of the pots had little useable adjustment range, and the input attenuator really needs an audio taper pot. Before I built the circuit, I had calculated 50K for that pot, which might still be too much. The bottom end of the voltage divider was supposed to consist of a 400-ohm fixed resistor and a 1K pot. A 220-ohm resistor was the nearest value I had for the fixed resistor; with that value, the pot is right at 1K for 0.5V, and some of the pot's range is above the diode's conducting voltage at 0.65V. The 4.7 uF DC blocking caps are a bit large; when the voltage divider is adjusted, it takes the circuit several seconds to settle down.
Here's the breadboard. Signal flow is from right to left:
Next step: panelize this. I'll built it on a piece of stripboard, and figure out a way to mount it to a panel. The panel will have two inputs and one output. Controls will consist of the input attenuation and voltage divider adjustment; I might put in a bar graph display for the latter.