Si diode clipping sound differences

[Fancy Lime]

Hi there,

I played around with a few different Si diodes as clippers in the feedback loop of a high gain inverting opamp and was surprised how gig the differences were. Here my subjective sound impressions relative to each other:

1N4148: fine grained, smooth
9V1 Zener (BXZ55C), forward biased: brutal, coarse, jagged
1N4001: medium grained, piercing, synth-like

Of course, I have no idea what, if anything, my subjective impressions mean in terms of actual clipping behavior and how much of it is imagination. Do they match or go against other peoples impressions? What is the decisive difference between the diodes that is responsible for the difference in "grain"? The knee shape, capacitance, open resistance, and recovery time are quite different between 1N4148 and 1N4001 but the Zener data sheets I found do not list any of those parameters. I seem to remember, though, that Zeners have a sharp knee and substantial capacitance. Given the test circuit, I would expect that knee shape and open resistance would play rather minor roles but I may be wrong.

Cheers,
Andy

[Mark Hammer]

The thing I like to remind people about is that diodes were not developed for clipping guitar signals.  So, while they most assuredly have different knees and recovery times, etc., I have to ask whether the time parameters of that are relevant to the bandwidth of guitars, or whether those diode-to-diode differences are really only pertinent for high speed switching applications.  Conduction onsets of 20 vs 100nanoseconds doesn't really mean all that much for signals unlikely to have more than 6khz bandwidth.

That said, the forward-voltage differences DO matter.  And since guitar signals are not steady state, but dynamic, and varying in harmonic content as well as amplitude, that can result in different sonic qualities.

I'm a bit of a doubting Thomas in these matters, I'm afraid.

[amptramp]

The difference might be in the forward current capability of the diodes.  A 1N4148 is a small signal diode, so if a transistor or op amp stage starts to force current through it, the voltage may rise with the current so it will follow the sine wave slightly but more than a 1N4001.

A Zener diode is an abrupt-junction device so its forward conduction is mediated by the electrostatic force across a few atoms thickness of junction meaning that the current rise is abrupt, keeping the voltage flat during conduction.  I would imagine that VVC diodes (voltage-variable capacitance tuning diodes) which are also abrupt junction and Schottky diodes which are metal-on-semiconductor junctions would be similar to Zener behaviour.

[R.G.]

Try some germaniums and some schottky types just to round out the experience.
Then put a 10K pot, rigged as a variable resistor, in series with the clipping diodes and vary the pot, taking the signal from the signal (not diode) end of the pot.
Let us know what you find.

[Rob Strand]

A diode connected transistor (ie. B connected to C) has a *similar* shape curve to a zener but without the capacitance.

[willienillie]

Try some germaniums and some schottky types

And every pair of germaniums you try will sound different, even of the same type.  I've only barely tried schottkys (in sockets) a few times, haven't cared for them, but maybe they're nice in some circuits.  I'm stilll trying to find an agreeable set of clippers for the OD-855 I built last week, it's been frustrating.

[Fancy Lime]

The thing I like to remind people about is that diodes were not developed for clipping guitar signals.  So, while they most assuredly have different knees and recovery times, etc., I have to ask whether the time parameters of that are relevant to the bandwidth of guitars, or whether those diode-to-diode differences are really only pertinent for high speed switching applications.  Conduction onsets of 20 vs 100nanoseconds doesn't really mean all that much for signals unlikely to have more than 6khz bandwidth.

That said, the forward-voltage differences DO matter.  And since guitar signals are not steady state, but dynamic, and varying in harmonic content as well as amplitude, that can result in different sonic qualities.

I'm a bit of a doubting Thomas in these matters, I'm afraid.

That's what I thought, too. And that is the reason why I had never before bothered testing different Si diodes against one another as clippers and why I was surprised at the rather prominent differences when I finally did. Even slight differences in forward voltage will yield different output volumes, which can trick the ear into imagining different sounds rather easily. What surprises me, though, is that one does not simply sound vaguely "better" than the other, as is usually the result of different volumes, but that I hear differences in "grainyness". My best guess at this point, is that "fine grained" probably means more high order harmonics than "coarse grained", but I find it difficult to put my finger on the sound impression. Hence my idea that diode terminal capacitance or switching times may play a role. But true, even the 30μs recovery of a 1N400x should be inconsequential in the audio band. The 15pF of a 1N400x should also be negligible since there is a 680pF cap across the diodes in my tested circuit. For the Zeners, however, I could not find a datasheet that lists capacitance but according to this: https://www.onsemi.com/pub/Collateral/HBD854-D.PDF I would expect it to be somewhere in the hundreds of pF. So that may account for the difference. The question then becomes: why does it (seem to) sound different when using Zeners with a 680p cap compared to 1N4148's with a 820p or 1n cap? The fact that Zener capacity changes with the applied voltage and therefore throughout the signal swing may have something to do with it, but I really don't know.

Try some germaniums and some schottky types just to round out the experience.
Then put a 10K pot, rigged as a variable resistor, in series with the clipping diodes and vary the pot, taking the signal from the signal (not diode) end of the pot.
Let us know what you find.

Well, diodes with very different voltage drops sound a lot different. No surprise there. The effect of series resistance is also clear. And Ge and Schottkys have both. The point of this thread was to wonder why different Si diodes, with very similar forward voltage drops and also otherwise rather similar properties as per the datasheets, still seem to sound remarkably different. Especially the difference in what I would describe as "grainyness" is a bit mysterious to me.

Cheers,
Andy

[Rob Strand]

For the Zeners, however, I could not find a datasheet that lists capacitance

Look up BZX84. In particular ON-Semi for a graph vs Zener voltage and operating voltage.  Then NXP for tabulated data.

The (depletion) capacitance of all diode is non-linear ie. it a function of voltage.
The (depletion) capacitance is least with a large reverse bias, then reaches quite a large value at zero bias, then increases further at forward bias.  In forward bias silicon diodes have another component of capacitance called the diffusion capacitance which is a function of current.

[Fancy Lime]

For the Zeners, however, I could not find a datasheet that lists capacitance

Look up BZX84. In particular ON-Semi for a graph vs Zener voltage and operating voltage.  Then NXP for tabulated data.

The (depletion) capacitance of all diode is non-linear ie. it a function of voltage.
The (depletion) capacitance is least with a large reverse bias, then reaches quite a large value at zero bias, then increases further at forward bias.  In forward bias silicon diodes have another component of capacitance called the diffusion capacitance which is a function of current.

Thanks, Rob! In light of those graphs in the On-Semi data sheets and the numbers in the Nexperia sheets, it looks like the dynamic capacity may indeed play a role for the sound. I will test 2V7 Zeners for comparison as soon as the authorities let me back to my breadboard. So... in May, probably.

Andy