Boss JFET Switching

[Rayman]

Does anybody know if there are "better" JFETs for electronic switching than those typically used by Boss and Ibanez like the 2SK30s.  By "better" I guess I was wondering if different on/off voltage characteristics would make a difference.

[R.G.]

Could you define the difference you're looking for?

The 2SK30 is just about perfect for the set of compromises required in that application in most pedals. Similar JFETs work about the same way. Some are: 2N5292, 2N5485, 2SK118Y.

All of these have a Vgsoff of less than 5V, an Rdson of a few hundred ohms, and like all JFETs an off resistance in the giga-ohms. The feedthrough is trivial at audio.

What sort of difference?

[Rayman]

Thanks R.G. - actually I think that answers my question.  I was wondering if the 2SK30s are optimal for switching circuit paths on and off, or if there are other JFETs that would be just as good or more optimal for switching, particularily with respect to any feedthrough when they are switched "off" (which you mentioned will be trivial).

[Mark Hammer]

DOD uses the J113.  Supposedly, they have a low "pinch-off" voltage (whatever that is) and present a fairly low on resistance.  Read all about it here: http://hammer.ampage.org/files/dodswitch.gif

[O]

Not to hijack the thread, but does anyone know what jfets are being used on the newer boss pedals? I have a TU2 stuck in the demo mode and the switch does absolutely nothing.

[R.G.]

Good tip, Mark. I'll go look at the J113.

For best off resistance, the biggest factor is how far below Vgsoff the gate is pulled and the relative size of the signal.

The way these things are used is that the source and drain are tied to a reference voltage of about 4.5V. The gate is let float for "on" and pulled to 0V for off. That's one reason the Vgsoff has to be less than 4.5V. If it's more than the bias voltage, tying the gate to ground does not turn the JFET off at all.

The source is tied to reference for DC, but the signal voltage rides on top of this, so the source also goes as high as Vref+Vsignal and as low as Vref-Vsignal. If Vref-Vsignal gets to being less than Vgsoff, then the JFET enters the resistive region and bleeds some signal through. In the standard circuit, a Vgsoff of about 3V is pretty optimal.

However, an even better thing is to take the gate much further negative compared to the source. You can be pretty sure that there's no bleed through if you get the gate to ten or twelve volts below the source. That's just difficult in the stock 9V and ground effects circuit. One possible way to work this is to not tie the JFET source and drain to Vref, but possibly to a decoupled voltage near 9V. Then the gate is pulled more in the off direction and isolation is better. This is only practical if you have both the source and drain of the JFET DC-blocked by capacitors - but it is possible. Usually only one end is blocked, so it might cost you one more cap and two resistors.

[tommy.genes]

It's interesting that this topic came up today since I was just messing with this last night.

Here is a small excerpt of the Boss DD-2 schematic.



Node 1 comes off the flip-flop (BA634) via the MODE switch. It measures at about 9V when the effect is on and 0V when the effect is off. Node 2 measures at 0V when the effect is off (expected), but measures only a few tenths of a volt when the effect is on (not expected). Wouldn't current flow through R36 and C27 to ground until C27 is charged enough that Node 2 is the same voltage as Node 1?

-- T. G. --

[R.G.]

Node 1 comes off the flip-flop (BA634) via the MODE switch. It measures at about 9V when the effect is on and 0V when the effect is off. Node 2 measures at 0V when the effect is off (expected), but measures only a few tenths of a volt when the effect is on (not expected). Wouldn't current flow through R36 and C27 to ground until C27 is charged enough that Node 2 is the same voltage as Node 1?

R36 is 1M. If your meter has a modest impedance, the meter itself pulls down the voltage it reads. It's tricky to measure voltage at high impedance nodes. Hook a JFET input opamp's + input to node 2, connect its output to its input, and make sure it's got a +9 and -9 power supply. Then its output will more accurately reflect the voltage at 2 because it's input impedance is much higher than R36's 1M.


On the schematic, if R37 and R38 go to Vcc, not 1/2 Vcc and node 1 switches from Vcc to ground, then you get the circuit I was talking about for more "off" in a JFET switch.

[tommy.genes]

It's tricky to measure voltage at high impedance nodes.

Thanks for the reply. I was using the cheaper of my two meters and didn't stop to think that might be the problem. If the input impedance of the other meter isn't much better, it shouldn't be too hard to put together a voltage follower like you describe - just a matter of setting up +/- supplies.

As for the 1/2 Vcc idea, I'll have to look at the board itself to see where they were pulling that from. It might not be too hard to substitute full Vcc in its place. It's not that there is a problem with bleed through now, but it would be fun to try it out. It doesn't look like that would result in any additional current draw from the battery (although all of my stuff is on a powered board anyway).

-- T. G. --