JFET switching with bipolar supply

[FiveseveN]

My brain seems to be particularly clunky today so I'll just give up on thinking and ask a quick question:
In The Technology of JFET Switching in Boss and Ibanez Pedals, R.G. stated "The capacitors at the input and output side prevent the DC conditions before and after the JFET switch from preventing it working as a switch.". Provided a bipolar supply eliminates the need for biasing resistors (R_bias in the article), would I still need to use these coupling caps?

[R.G.]

What is important is that the DC voltage be the same on both the drain and source sides of the JFET at all times. If they are not, you will get a pop as it switches.

The caps and bias resistors are there to force that to be true. If you have a situation where the thing driving the JFET has a stable output voltage of 0V and the thing receiving signal from the JFET is biased at 0V by something other than the JFET, then you don't need either the caps or bias resistors.

Note carefully the "if" in that statement.

[FiveseveN]

Thanks for the clarification, the brain-mist is starting to clear. It would be driven by an opamp buffer and loaded by a volume pot, so I shouldn't have a problem controlling the potential on both "sides".

[R.G.]

Using a bipolar supply and driving a switch with the output of an opamp buffer is one of the few places where offset voltage on opamps matter. This gets you into the secondary considerations of DC accuracy on your opamps. AC coupled opamp circuits don't have this issue. You're probably fine, but if there's a small click, this may be the cause.

Opamps have both internal DC offsets and bias current offsets. Modern opamps are quite good compared to the dark ages of 709 and 741 opamps. The cheapo TL07x series is about 10-15mV offset voltage, and has a JFET input which all but eliminates bias offsets.

If you're not using a FET input opamp, either JFET or CMOS, then there is a measurable bias current into (or out of) the inputs to bias the input devices. This is small, but has to be thought about for DC offset reasons. The way to make this come out right is to make the resistance to the bias voltage be the same for both inputs, since both are nominally equal currents in a monolithic IC. The resistor to bias on the + input is easy, but the resistor to "bias" on the inverting input is the parallel combination of the feedback and input resistors.  This is one reason that DC followers with no "Ri" per se sometimes have a resistor between output and inverting input - it balances the voltage drop from the input bias currents to both inputs, and produces better DC accuracy.

The internal offset tends to come out as just the internal offset, but bias offsets are multiplied by the gain, so for high DC gain stages, it can get out of hand quickly. For applications like analog computing where the DC accuracy is critical, there is usually some way to add an offset balance control to balance out the internal offset.

Again, probably not going to bite you, but worth considering if you run into trouble.

[merlinb]

You may find this useful:

http://books.google.co.uk/books?id=PvKPEFu2PVkC&lpg=PP1&dq=small%20signal%20audio&pg=PA408#v=onepage&q&f=false

[FiveseveN]

Hey, I actually have that book. Hadn't got to that part yet. I'll try to keep you posted with results (error voltage, pops, crosstalk) as soon as I find some FETs in this barren land.

[PRR]

> I shouldn't have a problem controlling the potential

The opamp always has a ~~milliVolt error. You can control this: put an offset trim pot on it. But why work too hard? The R-C networks taught by R.G. and used in MOST FET-switching (even costly recording consoles) _naturally_ tend to zero error.

It costs very little to add them in advance, a LOT to jam them in later "if" pops are a problem.

[FiveseveN]

I guess it's easier to offer the whole application for review. I'm trying to develop a 1-of-3 channel switcher (with individual volume controls, as you see). It would be controlled by shorting SW1, 2 or 3 to the negative rail (which is digital "ground") with pushbuttons.

Link to image

As I said, I'm using a bipolar supply (+/-15 V rather than the +/-9 V pictured, but that shouldn't matter) because that's what I have at hand.
Should I tie the BJT's collectors to the positive rail instead of ground, would that make a difference? Am I missing something or have I gone about it completely backwards? Could it possibly be simplified?

[FiveseveN]

Oops, I mixed up the (most basic) logic. So yeah, trannies connect to !Q instead. That's what I get for working at 4:00 AM.
Updated schematic: http://img35.imageshack.us/img35/3024/switchingi.png I also labeled voltages correctly for clarity.
LTSPICE said it would work, so I guess I'll put it together and see.

[FiveseveN]

It works fine, in case anyone else was wondering . No pops, even without low-passing the switching. Only thing is I ended up using a much higher (100 uF is what I had handy) cap to get it latching channel A on startup.