The series-diode gate clipper for audio peak limiting.

[brad]

I've literally got about fifty 12AL5 tube diodes sitting around, and it would be cool to be able to actually do something with them.  I found this circuit in an old tube book, and it looks pretty interesting...however I'm not exactly sure what it would sound like.  After some searching, I found someone who seems to have built it, but the page really only adds to the confusion:

http://www.mines.uidaho.edu/~glowbugs/clipper.htm

Can anyone tell exactly what this circuit actually does? 

[R.G.]

The easiest way to think about this one is to replace the tube diode with solid state ones.

When there is no signal, the clipping level control sets up a DC current through R1, R2, and R3. That is, R1 and R2 divide the current that R3 and R4 let flow. So both input capacitors are charged to a DC voltage. This voltage will be maintained by caps C1 and C2 charging to the average DC level of the input and output signals, which we can just as well presume to be 0V - it won't matter the exact number as long as it's steady.

Looking at the input, as long as the input is pulling downwards on C1, the first diode is forward biased, and passes signal through to the second diode. This happens for all positive-going signals for peaks less than the DC level across R1. When the positive going peak exceeds the DC level on R1, it reverses the polarity on the first diode, and the diode turns off. No signal current can then flow through to the second diode or to the output.

This happens the reverse way at the second diode. As long as negative going peaks are less negative than the positive DC point on R2, then the second diode is forward biased and signal passes to the output on C2. When the signal passed from the first diode gets to two DC bias points negative, it turns off the second diode and no signal goes through C2 to the output.

The thing is exactly as described - a series diode clipper. It clips when the diodes turn off, not when they turn on. Either diode turning off keeps signal away from the output pin, and this happens at the opposite polarity for each diode.

R4 sets the amount of current going into the resistors and therefore the DC levels on R1 and R2, and from that, the clipping level.

[brad]

I see...thanks for the explanation R.G 

So...if the diodes are biased so they cease to conduct as say, 6 volts, the signal will be zero whenever it swings above that that threshold and won't return until it swings below it again?  That sounds like it would produce a weird looking waveform...a bit like devil horns.  Interesting!

Sounds like it could be easily tweaked for stompbox voltages too.

[R.G.]

I'm sorry - if I could just speak English as well as my native... um... oh, never mind, English is my native language. No excuses. 

When the diode turns off, the output doesn't fall to zero, it stays at the clipping level. It's held there by the charge on the other-side cap. So if you put in a sine wave, nothing happens until you hit the clipping voltage, then the sine flattens off on top.

In this kind of setup, the diode characteristics don't matter much because the signals are so big compared to the diode drops. That thing clips at several to many volts. The clipping looked ruler-straight on the circuit simulator.

Scaling it down to lower voltages might be interesting. You'd have to keep the resistances high, and maybe you could get to where the diode knee impedance was comparable for some rounding. It's worth looking into.

[brad]

Oh now I understand.

Even though the clipping is hard, it would still be an interesting effect to built just for its oddball design.  I bet no one else would have a series diode gate clipper based on a tube diode in their guitar rig 

[R.G.]

Go for it. Let us know how it comes out.

Hmmm. You know, you could soften the hard edged clipping by paralleling the diodes with some big-value resistors. That way when the diode slammed shut, the resistor would leak a little signal past. Should round the top a bit and make it a bit softer.

[lovric]

out of memory: joe davisson tried it and someone else here called their version 'black scorpion' or 'buddhist dorje' and even gave a link to a book scan of a tremolo circuit that exploits the mentioned method.

[R.G.]

The tremolo circuit looks similar to the clipping circuit, but is not really the same.

The tremolo circuit uses the forward resistance of the diodes as a variable resistor. That is, a diode goes from a "resistor" of very large, not conducting at all, to only a few ohms as you raise the current through it from 0 to 0.7V. So one way to do a tremolo is to put a signal into one end of a diode, take it out of the other end, and change the diode resistance by changing a current fed through the diode.

As you might imagine, there are some limitations. First, the signal has to be small compared to the diode voltage. In fact, it has to be small compared to the knee of the diode's forward voltage, because that's where all of the changing-resistance is hidden. Generally, 25mV is what you can do with reasonable distortion, maybe less than 1%. And the bulk voltage on the diode changes too, so you have to keep that out of the signal path somehow. That change in voltage is bigger than the signal by about 10-50 times. Also, the usable changing resistance is low, on the order of 10 ohms to a few K.

So you can pull the trick used in a diode attenuator and use two diodes and two resistors. You put two diodes in parallel, anode to anode, then put resistors to ground from each cathode. You put signal into one cathode/resistor, take it out the other side and control the attenuation by a resistor dropping current into the joined anodes. This setup still has the low signal and bulk voltage problems, but is more predictable.

Even better is to use a third diode. You take three  diodes and six resistors. You put a large resistor in series with each diode's anode, and with the cathode. You connect all three of the anode resistors together and drive that with your modulation LFO. You tie two anodes together, and put signal into and out of the two cathodes. The third diode is a source of reference voltage. If all three diodes are matched, and so are the anode resistors and cathode resistors, then the same voltage due to the LFO appears at the cathode of all three diodes. You use a differential amplifier opamp to subtract the reference voltage from the output cathode, and what you're left with is just the modulated signal. How well this works depends on how good your matching is and the quality of the opamp.

Next step: a diode ring. You hook up four diodes as follows: large resistor to two connected anodes. Each cathode connects to the anode of another diode. The cathodes of the last two diodes connect together and to a matching large resistor. You couple signal in at one anode/cathode junction, and out at the other. You drive the large resistors top and bottom with matching + and - LFO voltages. The diode resistances change with current, and conduction of the signal is from the input, through both anodes and cathodes symmetrically, and out the other anode/cathode. The driving voltage cancels if you did everything right. This is the circuit used in the Thomas Vox amplifiers like the Beatle and Royal Guardsman. It works well. It's noisy because you have to use small signals and then amplify them back up.

Using precision resistors and matched diodes could make this work well.

[lovekraft0]

Next step: a diode ring. You hook up four diodes as follows: large resistor to two connected anodes. Each cathode connects to the anode of another diode. The cathodes of the last two diodes connect together and to a matching large resistor. You couple signal in at one anode/cathode junction, and out at the other. You drive the large resistors top and bottom with matching + and - LFO voltages. The diode resistances change with current, and conduction of the signal is from the input, through both anodes and cathodes symmetrically, and out the other anode/cathode. The driving voltage cancels if you did everything right.

That sounds suspiciously like a ring modulator, right? 

[R.G.]

That sounds suspiciously like a ring modulator, right?

The description does, yes - but no, it's not a ring modulator.

The diode ring is arranged differently.

A ring modulator bridge is set up so that by flipping the polarity of the modulation, you flip the polarity of the carrier. In the diode ring attenuator, the modulator flow is all one way and the signal flow is bidirectional. It really is acting as a current variable resistance.

Diode ring modulators tend to be binarily modulated - the modulation flips the phase of the carrier signal entirely or not at all. Linear ring modulators like multiplier cells such as the 1496 are more linear - the amplitude of the carrier is also varied between maxes. So you start with one carrier phase, and as the ring modulates it the amplitude goes down but still in the original phase, becoming 0 amplitude at 0 modulation, and then increasing amplitude in opposite phase as the modulation goes negative.

It's that phase inversion through 0 that separates amplitude modulators from balanced modulators.

[slacker]

out of memory: joe davisson tried it and someone else here called their version 'black scorpion' or 'buddhist dorje' and even gave a link to a book scan of a tremolo circuit that exploits the mentioned method.

Here's the thread Joe started, http://www.diystompboxes.com/smfforum/index.php?topic=48507.0 unfortunately the circuit he posted seems to have disappeared.
I copied the tremolo posted in that thread for my slack trem for August's FX-X http://www.diystompboxes.com/smfforum/index.php?topic=48815.0

[Nasse]

http://www.elektor-electronics.co.uk/Default.aspx?TabId=27&year=2001&month=10&art=50173&PN=On

That elektor project was interesting for "soft" clip ability claimed. I think the basic circuit is perhaps not similar with the tube circuit in the first post. But does it sound cood that is another thing.

[George Giblet]

Fender have a patent for tube clippers.  From what I can remember your circuit is a little different - I suspect it clips harder.