Question About Diode Clipping Sound

[GibsonGM]

I think we need a new thread!  This isn't about diode clipping any longer.  It might be more on-topic to discuss the ways a signal is clipped to produce distortion, and what the characteristics of each type are.  WHY they sound the way they do.

My answers are purposely limited because if you are new (Agung, Tham...), you may become overwhelmed with too much information and not enough foundation to really understand it.   So you probably are not getting the answers you want.

Harmonics:  http://www.teachmeaudio.com/recording/sound-reproduction/fundamental-harmonic-frequencies/

Harmonics have much to do with how distortion sounds, but also very important is how rapidly the signal enters and exits clipping (among MANY other things, as Teemuk wrote about above).   This may be part of your observation, Tham....

[ashcat_lt]

The thing is that the subtleties in the curve of the knee really don't amount to much.  The transfer curve through a diode-to-ground clipper looks a lot like an atan, which is really close to a tanh, but that's not much different from a sin, and that's not really all that far off from the power curve you get from a tube and is really close to a what a transistor does.

And I've played around with all of those curves and more in digital and the differences in tone are really tough to actually hear or even "feel" while playing.  You can hear the difference between an actual brick wall "zero knee" clipping and something with a curve to it, but the difference between actual curves we get from analog devices aren't as important as people seem to think.

Symmetry does make a difference.  I have often wondered, but never found a definitive answer, how different the curves look at either end of a tube (or transistor) operating range.  Like, if saturation has a much more gradual curve than cutoff.  ()  That could make a noticeable difference, but probably not much different from other forms of assymetrical clipping.

[Mark Hammer]

Look, diodes conduct when the signal exceeds a certain voltage/amplitude.  Where they become important in dictating tone is a function of the fact that strings work like a percussion instrument and decay once plucked/strummed.  The level of the signal reaching the diode changes, and with it, the prospects for pushing the diode into conduction.

Not only that, but the frequency content of a plucked string also changes over time, such that what is pushing the diode into conduction is changing, not just whether it pushes it into conduction.

So where the diode choices start to have an impact on tone is in terms of how much of the note's lifespan will prompt clipping, and what the frequency content of the signal to be clipped consists of during that timespan.  All of which leads me to declare that the forward voltage of whatever diode complement one is using (and it could be any sort of symmetrical or non-symmetrical arrangement, comprised of any number of any type of diode or diode equivalent) accounts for the lion's share of what you get when applying diodes for clipping purposes.

The gain structure and tone-shaping, both before and after the point of clipping will account for 99.9% of whatever is not dictated by forward voltage.  Do diodes conduct at different rates and have different "knees"?  Sure, but none of that applies to the 6-8khz bandwidth that most of us get from our amplifiers and speakers.  If we were designing 5mhz switching systems, then definitely it would matter, but not at 6khz and below.  What matters is whether the diodes are clipping for only this much of the note, or that much.  If it clips for a longer % of the note, then you get that much more harmonic content for that much longer.  Indeed, I like to distinguish between "fuzzes" and other types of distortion on the basis of how long the clipping continues for.  If the clipping continues until you mute the string with your hand, then it is a fuzz.  If it is nice and buzzy and bright, but that occurs mostly just after you pick/strum, then it's a distortion.  The persistence of the buzz/fuzz is what makes the audible difference.  Many "distortions" can be transformed into a "fuzz" by adjusting the gain of the signal such that it's amplitude is guaranteed to remain well above the clipping threshold for as long as you want, or until you mute the strings.

[bool]

...

Symmetry does make a difference.

...

Sure does, man.

[teemuk]

So where the diode choices start to have an impact on tone is in terms of how much of the note's lifespan will prompt clipping...  All of which leads me to declare that the forward voltage of whatever diode complement one is using ... accounts for the lion's share of what you get when applying diodes for clipping purposes.

I figure that is the case only when input signal's magnitude to clipping diode cell is "fixed". With input signal of, say, 1Vpeak a diode with 300mV forward voltage will clip larger portion of the signal than a diode with, say, 600mV forward voltage. Naturally. But when things aren't "fixed" like that.... Well, a diode with 600mV forward voltage clips the very same proportion of a 2Vpeak signal than a diode with 300mV forward voltage would clip of a 1Vpeak signal. ...And most circuits do have -gain controls- to vary input signal levels to the diode cell just so that one can vary how large proportion of the signal is clipped, and the amount of harmonic distortion that proportion of clipping introduces.

So assuming that input signal levels are "normalized" before the clipping cell ,Vf is an inconsequential parameter and what really affects the clipping are the -knee characteristics- in how the diode turns from forward bias to reverse bias.

One of my favourite graphs concerning the topic:

See how knee characteristics of 1N34A (germanium) and 1N914 (silicon) diodes are actually pretty much alike, with only difference in lower forward voltage of germanium diodes? They both actually "hard clip" like generic silicon diodes.

Yes, I do realize that many people deliberately DO employ effect of lower Vf to clip signal proportionally more than a diode with higher Vf would clip, but if input levels are normalized so that proportion of clipping does not really change then harmonic distortion introduced by clipping doesn't actually change at all and the end results are identical.

Now compare those graphs to another germanium diode, AA112, that features a less steeper slope, yet is loosely categorized as low forward voltage germanium diode. It starts to forward conduct at around 200mV input but where 1N914 is pretty much "fully saturated" at 600mV input the AA112 still continues to progress gradually towards full saturation. With normalized input levels the proportion of clipping distorted signal is much, much higher because there's much more non-linear area of "dynamic resistance" between reverse biased and -fully- forward biased states. Now THIS will impart a totally different harmonic pattern to clipping, and much more "gradual" overdrive characteristics, "softer clipping" to speak, even with normalized input signal levels.

Of course one can easily achieve that characteristic simply by introducing some external series resistance so you don't need to waste your time hunting for diodes with perfect characteristic curves. These are attributes that can be easily tweaked with additional components in the circuitry.

[Mark Hammer]

I think we agree. (a lot to sort through  ).
We certainly absolutely agree that  Vf differences can always be compensated for by gain.  WE may know that, but there are a lot of folks "out there" who don't see gain level and structure when they look at a schematic (if they look at a schematic).  What they see is diode type, and attribute tone to it.  Meanwhile, a few changes to gain structure would allow one type of diode complement to sound like another.  That works in both directions: trying to achieve a more severe clip and trying to achieve a milder clip.  Newbies are more likely to want to change the diodes, however, than tinker with the gain.  I suppose the fact that one requires some math and the other doesn't has at least a little to do with it. (Not a scolding; merely a reminder that a little math will take you farther.)

The natural variation in both guitar signal level and spectral content is, for me, a big reason why we can often have such a hard time mimicing the distorted sounds of many classic 60's recordings.  Many recordings included studio musicians who would bring in their big jazz boxes with floating wooden bridges, and the producer might request a fuzzy sound.  Part of what gives them their characteristic "fat" sound, when played without effects, is that the floating wooden bridge provides a quick decline in amplitude, post-pick, followed by a protracted sustain of mostly the fundamental, without much harmonic content.  Not the same at all as a solid-body, with an all-metal bridge firmly anchored into the body.  The input signal to any fuzzboxes is very different, and the resulting tone is different.  Attempts to nail something you heard on an old 45, by attempting to perfectly cllone the physical design of the pedal, may disappoint because what you're feeding the pedal is not the same.

Again, one should not misattribute the tone of a clipping circuit to  JUST those electronic components you can see or were told about.

[ashcat_lt]

Look, diodes conduct when the signal exceeds a certain voltage/amplitude.

I agree completely with this post, but I don't like this line.  It's a common way of putting it, and not actually wrong, but I think it obscures the truth.  I think a lot of folks think of a diode as a switch, almost like opposite of a circuit breaker that is completely closed until the voltage across it gets high enough and then it opens up completely.  But that's only true for an ideal diode, and it doesn't really help us understand the diode clipping deal.

I prefer to think of a diode as a resistance that is dependent on the voltage across it.  This way we can see how the diode-to-ground clipper is really just a voltage divider.  "Everything useful is a voltage divider".

[teemuk]

I don't think you can explain the difference between "fuzz" and "distortion" tones solely with different instrument inputs. The characteristic "fuzz" distortion tone is result of intermodulation distortion at wide bandwidth. Therefore it has that distinct "fartyness" and mediocre note sensitivity. I think this type of effect was conceived by natural means: You just overdrive a mixing console, gain stage, or whatever to the point of clipping distortion. The clipping compresses the signal envelope and provides more sustain while harmonics of distortion alter the instrument's natural timbre. Heck, you could even punch holes to loudspeaker cones. No further thinking required. Creating an effect that does that "fuzzy" distortion deliberately was easy once folks figured out that such effect as "distortion" even exists.

But "fuzz" is not a useful tone for everything, largely because of that poor note sensitivity and unclarity in the bandmix introduced by high orders of IMD. So what folks figured out next is that you can improve the note sensitivity, and how well the tone "cuts through the mix", by using a -Treble Booster- effect. A treble booster is basically a hi-pass filter, and the reduction of lower frequencies it introduces before clipping distortion will greatly reduce intermodulation distortion imparted by those lower frequencies. With that simple filtering effect the distortion no longer sounded like "fuzz" but had an entirely different timbre. One can call it "distortion" or "overdrive" but I generally think those are poor names for obvious reasons.

If you look at history you can see how many guitarists adopted a treble boosting effect of some sort to their arsenal pretty much around those times when generic "fuzz" tones went more or less out of the fashion. This was already a pretty good sign of where things would evolve next...

The third step was that "treble boosting" (or distinct hi-pass) was implemented as permanent element of the device in question. Amps were designed for "brighter" response (look at Marshall history for example) so that they would sound less "farty" and "fuzzy" when overdriven and without one needing an external pedal for that effect. Distortion effects, like Tube Screamers and alike, were introduced and they employd drastic amounts of hi-pass filtering. With appearance of "high-gain" channels the hi-pass filtering was simply included to the design.

Typical "distortion" pedal may hi-pass at around 1kHz while those "fuzzing" effects typically have bandwidths that extend to lowest frequencies of the instrument (e.g. 80 Hz). That. IME, is the biggest reason why they sound different, even with whatever you plug into the effect. Guitar through a fuzz effect sounds "fuzzed" whether its an acoustic guitar or solid body electric, despite their obvious differences in spectral content and timbre, and those "modern" high-gain tones are still as distinctive whether you plug in an acoustic or electric guitar. Yes, you can definitely hear a slight difference in timbre, like you hear with single-coil and humbucker pickups, but overall the timbre of the instrument will not effect THAT much the inherent timbre of the effect.

[robthequiet]

I would try building a collection of basic circuits, i.e. tube screams, big muffers, fuzzfaces, red llamas, ROG Fetzer preamps, etc, and listen closely to what each circuit does. You might find that an overdrive FET circuit works better than diodes to ground for certain things, and pay attention to your coupling caps and tone caps. If you have a Mesa Mk  II.V, does it have a slave out? Maybe you need a cabinet emulator to go into your soundcard, or a direct box.

Lots of great theory in this thread that comes from guys who have been at this for years. Worthy read.

[merlinb]

The clipping diodes are not nearly as important as you've been led to believe. In a distortion pedal what matters is:

1: Compression/bias shift, e.g.. changes in clipping threshold with signal level. In most pedals it is constant which is why they don't sound much like real tube OD.
2: EQ
3: Symmetry of clipping
4: Hardness of clipping
5: Wire colour.

In that order.

[Mark Hammer]

I always use white for inputs, and blue for outputs, in my builds.  Anything else and I tend to insert the plug into the wrong jack, and they don't sound nearly as good as a result. 

[GibsonGM]

The clipping diodes are not nearly as important as you've been led to believe. In a distortion pedal what matters is:

1: Compression/bias shift, e.g.. changes in clipping threshold with signal level. In most pedals it is constant which is why they don't sound much like real tube OD.
2: EQ
3: Symmetry of clipping
4: Hardness of clipping
5: Wire colour.

In that order.

1 thru 3 were what I wanted to discuss.  Perhaps it's just me, but other than Vf (Ge vs. Si), I can't tell 1 diode from another, really.  There isn't one that you'll put in a typical distortion and voila - TUBE SOUND YAY!!!   Isn't going to happen.  It will take a lot more work than that.   

I'm intrigued by #5, wire color - I have no idea what the formula is for choosing wire color!!   I never was very good with permutations and combinations...   

[merlinb]

I'm intrigued by #5, wire color - I have no idea what the formula is for choosing wire color!! 

Ever heard of the brown sound? Now you know. 

[EBK]

I'm intrigued by #5, wire color - I have no idea what the formula is for choosing wire color!!   I never was very good with permutations and combinations... 

Just remember that red is lower in frequency than violet, and you'll be fine.  You can make a nice all pass filter with a piece of black wire. 

[GibsonGM]

Please, Merlin, no gratuitous selfies on the forum, geez!!   


ROY  G  BIV  -  I got it now!!

[samhay]

You can make a nice all pass filter with a piece of black wire. 

Indeed, but you have to be careful to get the wire the right way round.

[Agung Kurniawan]

Of course one can easily achieve that characteristic simply by introducing some external series resistance so you don't need to waste your time hunting for diodes with perfect characteristic curves. These are attributes that can be easily tweaked with additional components in the circuitry.

Did you mean place resistor series with the diode?

[Mark Hammer]

I'm intrigued by #5, wire color - I have no idea what the formula is for choosing wire color!!   I never was very good with permutations and combinations... 

YMMV, but I think of it this way.  There are usually only 10 different colours available: red, black, green, purple, brown, orange, grey, yellow, blue, white.  Red and black are obviously reserved for power and ground, though occasionally green joins them.  Purple and grey an be used for toggles or pots (I tend to use purple for resonance and grey for LFO rate).  As noted, I use white for input and blue for output.  I tend to save yellow, orange, and brown for treble, mids, bass, respectively, where I use such controls.

I'm dead serious about this.  And I adopt this personal standard because I tend to build much more than I box up (boxes being the most expensive component on the majority of builds, and stompswitches next down on the list).  As a result, there are a lot of stuffed boards with pots, jacks, and toggles hanging off of them by wire.  The wire colour helps me to quickly identify which jack/switch/pot is which.  Very helpful when one only gets back to a board a year later.

[EBK]

Mark, you diffused all the other jokes I had about wires and the electromagnetic spectrum with that.  I was going to point out how purple doesn't really exist, except as interference between red and blue (making it too noisy to use), but you brought me back to practical reality, and I'll probably end up being more systematic like you (and buying fewer enclosures so I can build more boards).   

[Mark Hammer]

I suppose the other alternative is to buy a bunch of different coloured knobs, and use knob colour to identify what the function is on one's unboxed builds.  Once upon a time, you could write the function on the back of the pot with a Sharpie, but as pots get smaller, that gets harder to do.