Diode clipping, Frequency Vs. Time ~questions

[petemoore]

  I'm imagining what the diode sees in say a Dist+:
  An amalgum of frequency content...which all adds up to 'a' voltage threshold, and the diode switches relatively cleanly [ie quick] and 1 time per 1/2 wave.
  ie...are all frequencies switched to the ground shunt at exactly the same time ?
  and perhaps some BBing to cause the HF's to be clipped at a different point in time than the LF's [a slight offset of when the clipping is applied to the input waveform].
  What brought this on is the thread about 'Distortions knocking the body out of your tone'.
  Answer would be of course yes, and I like that sometimes...compressed and where the attacks make more of a frequency shift [more HF hash comes out during the onset of hard pick attacks] than a dynamic one.
  [Sonic Maximizer mentioned only as a reference for this concept...re-alignment of timing HF's to LF's [IIRC the SM pedal indexes the HF's wavepeaks back a '1/2 notch' in time compared to the Bass...a Hf to Lf peak re-alignment].
  None too sure what or even if this is the big idea...separately distort Highs, move 'em back  1/64th notch or so in the time constant with the [somewhat less distorted?..while your splitting prehaps allow bass to be a bit more pristine] Bass portion of the source.

[Mark Hammer]

The "knee" of diodes that is often referred to, represents the speed at which the diode begins conducting.  Think of it like the plug in your sink or tub.  Initially, there is a brief period where some of the water can drain out, but there is still an obstacle until the plug is removed completely.  Water will start draining faster from a tub  where that plug comes out easily, than from one where yu have to wiggle it to get it out.  Similarly, the diode will begin conducting when the critical voltage is approached/reached, but that conduction is not instantaneous from what I understand (hence the chatter about fast-response diodes some time ago, and some of the continuing chatter about silicon vs germanium differences). 

Of course, the fact that it does take time to fully conduct is separate from how long it takes.  For instance, if it took 1 picosecond to go from a nonconducting to fully conducting state, that's fast enough that it would have absolutely no bearing on the extent to which the diode favoured one frequency over another within the range of guitar frequency content.

If I could interpret the time-course properties of diodes from the datasheets, I'd offer more of an opinion.  Perhaps one or more of the EE types here could step us through the relevant information in a posted diode datasheet as an example.

[R.G.]

Diodes have a junction capacitance, a forward recovery time, and a reverse recovery time.

The capacitance is what it sounds like, except that it is non-linear and depends on both the physical setup (i.e. stray capacitance from leads, etc.) and the degree of bias on the junction. Varactors are diodes which are designed to have a variable capacitance in reverse bias and are used for electronic tuning in radios. The variable capacitance is on the order of ones to tens of pF. The very common 1N4148 diode has a junction capacitance of 4pF when measured at 1MHz and 0V bias. That would go down with reverse bias as the edges of the depletion region are moved apart by the bias voltage.

The recovery times are related to the semiconductor physics of the diode junction, and are best thought of in terms of filling and emptying the junction of charge carriers, and carrier transit time across the junction. For small signal switching type diodes like the very common 1N4148, the forward time was too small for the datasheet to spec back when they were first released, and the reverse recovery time is 4nS from 10ma forward to 1ma reverse with 6.0V of reverse voltage. About all you can take away from that is that it's far, far faster than audio, several orders of magnitude.

One can simply consider the diode turn on/off times to be so fast that it's instantaneous compared to the glacial speeds of audio signals.

[Mark Hammer]

Then is there ANY practical advantage to the use of "fast-response" diodes?  Or rather, are there any musically relevant special contexts where it is of practical advantage?  Or are we really talking about faster and slower response as being pertinent only in the realm of RF signals?

[R.G.]

An amalgum of frequency content...which all adds up to 'a' voltage threshold, and the diode switches relatively cleanly [ie quick] and 1 time per 1/2 wave.
  ie...are all frequencies switched to the ground shunt at exactly the same time ?

There is a problem in your understanding of time domain and frequency domain.

Real signals are single-valued. That is, they have one and only one value at any instant of time. This is the world that resistors, capacitors, pickups and diodes live in. Frequency domain is multi-valued, and cannot be restricted to one instant of time. You must specify a period of time to even compute a frequency spectrum.

Diodes only see the single-instant voltage and current of the signal in the time domain. A diode turns on when the voltage across it is in one direction for long enough (i.e., a few to tens of nanoseconds) and off when the voltage is in the other direction for long enough. The time scale is way faster than audio, so fast that it can usually be taken for instantaneous. (This last is an approximation I make here to keep things straight without getting into a never-ending spiral over slew rates.)

Then is there ANY practical advantage to the use of "fast-response" diodes?  Or rather, are there any musically relevant special contexts where it is of practical advantage?  Or are we really talking about faster and slower response as being pertinent only in the realm of RF signals?

To the best of my knowledge, the response speed can be ignored, except for its side effect of changing the forward V-I curve. Audio is quite slow, at least compared to the time for diodes to turn on and off. There are exceptions having to do with the combination of diodes and opamps not being able to slew fast enough for accurately rectifying some kinds of signals, but for the purposes we're discussing here, it would be hard to come up with a situation where diode on/off speed is slow enough to make a difference.

There can be certain situations, such as the use of fast-response diodes in power bridges to avoid pulsed-RF ringing when the diodes cut off, and the resulting RF buzzing in audio on that power supply, maybe a few others. But in simple clippers, the audio's not fast enough to distinguish the diodes.

[petemoore]

  This all makes sense.
  About all you can take away from that is that it's far, far faster than audio, several orders of magnitude.
  Resolute enough for my curiosity, I must have phrased the question well enough, I understand the answers, thank you !
  Which boils the rest of it down to a possible...
  Split frequencies, treat treble [or what have you] to hard clipping, bass to less [so as to not muck things up and just have more solid, less distorted low end 'effect'].
  Perhaps play with the phase alignment of Hf to Lf...[ via delay ] as a more involved project.

[Nasse]

There was a small circuit in Elektor magazine years ago, a clipper or limiter for speech. The circuit used diodes in the feedback loop of an opamp. The idea was to dynamically roll off bass when level got up and more clipping happened (less intermodulation distortion and more intelligent signal). But it was not diodes capacitance but resistance and impedance  change when diodes started conducting, with input cap doing the filtering. The opamp was very high impedance 3140 or 3130 so the cap could be small. Never tried it but thought it could maybe be useful for guitar distortion. Never tried it anyway.