I adopt a different approach to "hard"-vs-"soft" clipping. Consider an op-amp based distortion of any design, powered by 9V. The maximum voltage swing of the op-amp will be around +/-3500mv (i.e., a 9V possible range, minus 1V at each end). What is the amplitude of a typical guitar signal? It's obviously going to vary with the pickup and playing style, but in a great many cases it will be around roughly +/-100mv. How many times can +/-100mv "fit" into a +/-3500mv space? About 35x. What is the typical gain applied in an op-amp based distortion? Amplification MUCH greater than that is common. The Rat is a bit of an exception, with amplification in the thousands, but something like the MXR Distortion+ has a maximum gain of 213x. Easy to see that increasing the Gain control to only a fraction of its entire range will produce signal amplification that exceeds the chip's headroom. Even without the diodes in that circuit, the resulting sound will rarely be clean, unless the gain is set to minimum and one strums in a restrained way. In effect, what we call "hard" clipping is really *double* clipping. That is, the signal is clipped once, within the op-amp itself, and the output of the op-amp is then clipped *again* by the diodes. The fundamental rule of clipping circuits is that the quality of clipping, and spectral content produced, is *always* a product of the spectral content of the signal input. So if you feed a clipping circuit with a preclipped signal, it *will* sound different. I suppose one *could* theoretically power the op-amp with a higher supply voltage, like 18V, that would allow for that +/-100mv input signal to be amplified more before hitting the maximum voltage swing (headroom) of the chip, and use diodes with a lower forward voltage (like Schottky type with a forward voltage that is often in the sub 200mv range). In such a hypothetical case, the result would be somewhere VERY close to what we call "soft" clipping, because the op-amp output is still clean, and those diodes are the only thing doing the clipping. What about "soft" clipping? Diodes in the feedback loop of an op-amp, will conduct the negative feedback of its output when the signal reaches the forward voltage of those diodes. So, assuming use of a pair of plain vanilla silicon diodes, like a Tube Screamer circuit, once the input signal is amplified to be somewhere around +/-660mv, there is no further amplification, because those diodes start to conduct the negative feedback of the op-amp. And +/-660mv is still a long way from the headroom limitations of the op-amp. In short, "soft" clipping is really clipping that happens *once* in the feedback of the chip, instead of twice, within the chip and at the output, which is why it can sound different. Again, one could theoretically arrange for circumstances in which the power supply and diodes could conceivably still result in the signal bashing up against the headroom limits of the chip. For example, use a quartet (2+2) of red LEDs as the clipping diodes. These generally have a forward voltage in the range of 1500-1700mv, and sometimes more, depending on batch. But they would only clip when the signal is amplified to be > +/- 3000-3400mv. Drop the supply voltage from 9V down to 8V, and the clean headroom of the chip is now around +/-3000mv. Chances are very good the op-amp will clip before the LEDs do. Hope this makes sense.
In synthesizers, another common type of distortion is wavefolding, where the signal peaks “reflect” off the limit values. But I haven’t heard of any guitar pedals that do this. Any idea why? (Or do they exist but I just don’t know about them?)
I guess the question is whether wavefolding would add something not already achieved by "regular" distortion pedals. Keep in mind that wavefolders presume a fixed waveform with essentially fixed harmonic content, that can be dynamically varied, adding harmonic content. Guitar signal is anything BUT a fixed or predictable waveform. Probably the closest thing to wavefolding that comes in guitar form is octave fuzzes, which often use a rectified version of the input signal, "folded over" to yield frequency doubling.
illustrations of "ideal" waveforms are nothing like the waveform one will see on a scope trace. noone ever mentions forward or reverse recovery times, and how that affects the actual "knee" of the clipped waveform. its the "knee" that produces the harmonic (disharmonic?) content, as it gets closer to a square wave... if you run through several diodes and check the curve as it clips versus the curve as it "unclips"... theyre different. this is utterly different to the simple "symmetrical versus asymmetrical" clipping of diodes in anti-parallel... assymetrical clipping automatically gives even order harmonics, symmetrical gives odd order harmonics, and the knee determines the number of overtones present overall...
Love this series!
I adopt a different approach to "hard"-vs-"soft" clipping. Consider an op-amp based distortion of any design, powered by 9V. The maximum voltage swing of the op-amp will be around +/-3500mv (i.e., a 9V possible range, minus 1V at each end). What is the amplitude of a typical guitar signal? It's obviously going to vary with the pickup and playing style, but in a great many cases it will be around roughly +/-100mv. How many times can +/-100mv "fit" into a +/-3500mv space? About 35x. What is the typical gain applied in an op-amp based distortion? Amplification MUCH greater than that is common. The Rat is a bit of an exception, with amplification in the thousands, but something like the MXR Distortion+ has a maximum gain of 213x. Easy to see that increasing the Gain control to only a fraction of its entire range will produce signal amplification that exceeds the chip's headroom. Even without the diodes in that circuit, the resulting sound will rarely be clean, unless the gain is set to minimum and one strums in a restrained way. In effect, what we call "hard" clipping is really *double* clipping. That is, the signal is clipped once, within the op-amp itself, and the output of the op-amp is then clipped *again* by the diodes. The fundamental rule of clipping circuits is that the quality of clipping, and spectral content produced, is *always* a product of the spectral content of the signal input. So if you feed a clipping circuit with a preclipped signal, it *will* sound different.
I suppose one *could* theoretically power the op-amp with a higher supply voltage, like 18V, that would allow for that +/-100mv input signal to be amplified more before hitting the maximum voltage swing (headroom) of the chip, and use diodes with a lower forward voltage (like Schottky type with a forward voltage that is often in the sub 200mv range). In such a hypothetical case, the result would be somewhere VERY close to what we call "soft" clipping, because the op-amp output is still clean, and those diodes are the only thing doing the clipping.
What about "soft" clipping? Diodes in the feedback loop of an op-amp, will conduct the negative feedback of its output when the signal reaches the forward voltage of those diodes. So, assuming use of a pair of plain vanilla silicon diodes, like a Tube Screamer circuit, once the input signal is amplified to be somewhere around +/-660mv, there is no further amplification, because those diodes start to conduct the negative feedback of the op-amp. And +/-660mv is still a long way from the headroom limitations of the op-amp. In short, "soft" clipping is really clipping that happens *once* in the feedback of the chip, instead of twice, within the chip and at the output, which is why it can sound different.
Again, one could theoretically arrange for circumstances in which the power supply and diodes could conceivably still result in the signal bashing up against the headroom limits of the chip. For example, use a quartet (2+2) of red LEDs as the clipping diodes. These generally have a forward voltage in the range of 1500-1700mv, and sometimes more, depending on batch. But they would only clip when the signal is amplified to be > +/- 3000-3400mv. Drop the supply voltage from 9V down to 8V, and the clean headroom of the chip is now around +/-3000mv. Chances are very good the op-amp will clip before the LEDs do.
Hope this makes sense.
Wow, very nice video. Subbed.
Love this content. Please do more.
Enjoyed this series, cool stuff!
Banger series, can't wait for more!
In synthesizers, another common type of distortion is wavefolding, where the signal peaks “reflect” off the limit values. But I haven’t heard of any guitar pedals that do this. Any idea why? (Or do they exist but I just don’t know about them?)
I guess the question is whether wavefolding would add something not already achieved by "regular" distortion pedals. Keep in mind that wavefolders presume a fixed waveform with essentially fixed harmonic content, that can be dynamically varied, adding harmonic content. Guitar signal is anything BUT a fixed or predictable waveform. Probably the closest thing to wavefolding that comes in guitar form is octave fuzzes, which often use a rectified version of the input signal, "folded over" to yield frequency doubling.
illustrations of "ideal" waveforms are nothing like the waveform one will see on a scope trace.
noone ever mentions forward or reverse recovery times, and how that affects the actual "knee" of the clipped waveform. its the "knee" that produces the harmonic (disharmonic?) content, as it gets closer to a square wave... if you run through several diodes and check the curve as it clips versus the curve as it "unclips"... theyre different.
this is utterly different to the simple "symmetrical versus asymmetrical" clipping of diodes in anti-parallel... assymetrical clipping automatically gives even order harmonics, symmetrical gives odd order harmonics, and the knee determines the number of overtones present overall...