Trying to make a kill-switch stomp-box - without pops!

[tss]

I'm trying to do a kill-switch type effect that will mute the signal while I press a switch. So far no luck... what are the recommended muting methods for killing the signal while doing "switching"? I guess those could be used here as well. I have seen many talking about this but the end I could not find a good link with schematic and that was said to be working.

[bluebunny]

Short the signal to ground?

[R.G.]

The only way that's good and simple is with a JFET used as a shunt switch to ground and with its gate drive arranged to let the JFET resistance fade the signal over a few milliseconds. Even this requires making sure that the signal have absolutely zero DC component.

Signal switching with hard-contact switches carries with it little ticks from the suddenness of switching, having the signal change from full peak signal voltage to zero instantly. That discontinuity is heard as a [tick]. The problem is familiar to old-style organ makers where the key contacts readily cause this unless the keying is done with one fader circuit per key.

I like P-channel JFETs, such as the J175. They work with both source and drain at 0V. With the gate at 0V (or open) they conduct like a resistor of maybe 50 ohms. With the gate pulled up above 5-6V, they are off. As a shunt switch to ground, you tie your signal to the drain, ground the source, shunt the gate to ground with a 0.1uF ceramic and drive the gate with a 0V/6V or more signal through a 1M resistor. 0V is signal off, high on the gate is signal on. I've used a similar arrangement in several pedals that have been or are being manufactured.

Note that JFET switching is really only good for signals smaller than the gate drive voltage minus Vgsoff. Any bigger than that and the signal itself turns the JFET partly on at peaks.

[tss]

R.G.,

Lets assume that the JFET will be controlled by a digital circuit. This means that the voltage will change with a very quickly, which might couple into the analog signal path through the PCB tracer / wiring or perhaps through the JFET itself? I can think about using a large cap together with the 1M resistor you've mentioned to smooth out the transient however this sounds like something that would work well in simulation and less well in real life

I'll dig around my parts bin and see if I have a P channel JFET (don't think so). I do have a H11F3 which might help as the gate is galvanically isolated?

[R.G.]

Actually, it works very well in real life. All Boss and Ibanez pedals use some variety of this, albeit with N-channels biased up to about 5V and the gates pulled to 0V for "off" on the JFETs.

There is always some issue with hard-edged transients being coupled, but even modest attention to good layout makes this circuit about noiseless. One big help is that the 1M resistor can be placed over with the digital stuff and the cap to slow down the gates placed by the JFETs. The intervening trace only carries the slowed-down voltage step, not the fast one from the logic that is deciding the turn-on. But in practice, that level of concern has not been needed. There's also not much current involved, so even the transient current from the ground side of the cap is not big, which is another place you can get transients.

The H11F3 is good, but there is no practical way to slow it down. The exact transition point of the internal photo-FET is not specified, and especially not specified in terms of LED current; and the LED varies in how much light you get per ma of current. But it works. I've used it in several commercial pedals. Try it. Might do what you want. If you have it in hand, that's another point in its favor. Even if it doesn't, it may do 90% of what you want while you wait for mail order JFETs.

You can rig the same kind of thing by biasing an N-channel JFET up at several volts positive and pulling the gate to several volts below the drain and source to turn it off. This is what Boss and Ibanez do. I prefer the P-channel simplicity.

Another problem with the H11F3 is that it can't necessarily turn off bigger signals, and that what "bigger" means varies from unit to unit. But you're probably good with guitar or line level signals.

One big issue is signal impedance. It's really hard to not introduce clicks into raw guitar signals terminated in the traditional 1M resistance. The line impedance is high and almost any charge coupled to it causes a transient. If you do a buffer in front of the signal you're trying to shunt and get the source impedance down in the 10K range (perhaps with a 10K resistor after a buffer!) then the amount of charge needed to "tick" the line is remarkably higher than for raw-guitar signals or other high impedance lines. This doesn't help with ground-shift ticks, unfortunately, because the "tick" impedance is already quite low. Those have to be wire-isolated.

[tss]

I've found several J176 !!!

[tss]

So I've tried to simulate this... the switch doesn't do anything. Thoughts about this?

[R.G.]

So I've tried to simulate this... the switch doesn't do anything. Thoughts about this?

Ohm's Law.

OK, that's a bit too subtle.     In a simulator, voltage sources are real **voltage** sources, and they have an output that can supply infinite current to any load, which is another way of saying that their internal source impedance is zero. This is done s you can insert other simulated components, like resistors, caps, inductors, etc. to simulate real world components.

Your simulation has that ideal voltage signal source feeding the JFET and a 100K load in parallel. The voltage source does not care that the JFET is going to a few ohms when turned on, it just supplies more current. The cure is simple - insert some simulated source impedance. Put a 1K to 10K resistor after the signal voltage source and before the scope connection.

[tss]

Agreed, an ideal source is not the way to go. So I added a 10k resistor in series. Nothing happens

[R.G.]

Agreed, an ideal source is not the way to go. So I added a 10k resistor in series. Nothing happens

Lemme go simulate it and see what the problem might be. Back soon.

[R.G.]

That was a longer soon that I intended. It works. I had some simulation issues too, not least of which was that the model for the TO-92 part number only said "DUMMY MODEL" when I looked at it.      The SSTJ76 had a valid model, though, and it worked in (simulated) circuit.

I used a 10K series resistor between the AC signal voltage source and the JFET and 100K output resistor. The JFET's drain connected to the signal line, the source to ground, and the gate was fed with a 0V to 5V step through a series 1M and a shunt 1nF cap.

[PRR]

> the switch doesn't do anything.

As R.G. says, the SPICE voltage source is a "perfect source", it can power LA and LV and NYC without the least sag. It doesn't even notice a bitty JFET, it just drags it along.

Shorting-out a perfect source may be a poor way to turn off the lights in LV. Burn-up all the wires.

Shorting-out a guitar works fine. There's many thousands of ohms of impedance, hardly any power, it can be shorted forever. But you MUST emulate the guitar resistance.

My sim problem is that there's no P-type in my package. I flipped the battery and it worked fine; but as R.G. says that polarity is awkward in pedal-land. Looking in the model I found an "N", changed to a "P", and it did the right thing.

To keep the action on-screen I shortened the 0.1u cap to 1n. That's too abrupt for smooth audio. 0.1u gives a smooth fall-off (with asymmetry due to JFET Vgs action and high signal level), though perhaps a bit slow for rapid work.

[tss]

I've looked through the DB and indeed all of the J1** parts are dummy parts!!! I've looked for the 1st P-channel that I could find with SPICE model and it worked as expected. I'm not sure why someone would bother adding dummy parts to a library. Thanks!