Impedance question -- summing outputs with transistors

[midwayfair]

Let's say I've got this goin' on with a couple JFET drain followers:

Q1 output with 5K drain resistor > C1 (1uF) > Output ->
                                                           |
Q2 output with 5K drain resistor > C2 (1uF) -|

Circuits are in phase. I'm using the same values for simplicity.

Is there a low-pass filter formed by C1 in series with C2 back through the drain resistor of Q2 ... and vice versa?

I mean, I'm pretty darned sure there is. But what can I do about it?

[R.G.]

Principle of superposition applies.*




*That means "yes". Each drain sees a load equivalent to its own drain resistor, then through its own cap, then through the second cap to the other drain, then through the drain resistor to ground.

If this is for mixing, why not connect the two drains directly together and mess with the drain and source resistor values to get unity gain mixing without two caps? Drains and collectors are high impedance points for signal, so mixing currents directly in to one resistor gives you good mixing all the way to DC without either affecting the other. Of course you'll still have to worry about JFET variability giving you different gains, but that you have always.

[midwayfair]

If this is for mixing, why not connect the two drains directly together and mess with the drain and source resistor values to get unity gain mixing without two caps? Drains and collectors are high impedance points for signal, so mixing currents directly in to one resistor gives you good mixing all the way to DC without either affecting the other. Of course you'll still have to worry about JFET variability giving you different gains, but that you have always.

Interesting. I didn't know you could do that. The drains have trim pots on them. Would I be using a single drain trim instead, and then I'd have to match the fets for the same drain resistance if I'm looking for a specific drain voltage (6V)?

The circuit I'm messing with is a harmonic tremolo (yeah, still ...), and the capacitors are actually difference values. I wanted to keep things simple for the example. Let me muddy the waters a bit more -- In your version of the HT, you used some rather large resistors in series with the outputs of the drains, even though the FETs were matched, so I'm guessing that those resistors were to prevent backwashing into the drain at the mixing stage. (The problem was obviously that they also attenuated the signal.) So if I've got an LDR with a minimum resistance of about 10K in series with the 1uF caps; no matter how bit the caps got, I'm looking at a voltage divider that's never worse than 5K/10K regardless of the frequency only when both LDRs are on (minimum depth), right?

[teemuk]

Interesting. I didn't know you could do that.

You can. It's basically an adaptation of an old tube circuit working on the very same principle ("common plate load mixer").



Not a typical circuit to bumb into in guitar-related applications (amps, effects and such) but I do recall seeing this circuit somewhat often in older hifi amps. It's mostly becoming unforgotten with transistor technology and I think for obvious reasons.

[midwayfair]

Interesting. I didn't know you could do that.

You can. It's basically an adaptation of an old tube circuit working on the very same principle ("shared plate mixer").



Not a typical circuit to bumb into in guitar-related applications (amps, effects and such) but I do recall seeing this circuit somewhat often in older hifi amps. It's mostly becoming unforgotten with transistor technology and I think for obvious reasons.

Interesting. Do they have to share a source/emitter resistor? Or could I put a trim on the source resistors to balance the drains? Or is it best to just match the transistors? What about a source/emitter bypass cap -- still okay?

This is making me happy. It not only saves a couple capacitors on the output but if I move the LDRs to the inputs instead of the outputs, it would most likely save me from needing an output buffer.

[Bill Mountain]

I noticed this in a V4 schematic.  They mix channels 1 & 2 with tube common cathode amps sharing the same plate resistor.

http://www.drtube.com/schematics/ampeg/v4pre-jp.gif

Look at the second half of V1 and V2 right before the tone stack.

One of these days I'm gonna build a Rat/V4 in a box and use this same mixing method with some JFETs.

To match the FETs you could possible use source bypass caps and adjust the source resistor to bias.

Or use a slightly higher voltage (18V or so) to minimize the variances in VGS.

[R.G.]

The main channel mixer in all multichannel Thomas Vox amps is the transistor version of this.

The biggest problem with JFETs is that they vary. A lot. Yes, you will have to do something to get them both to conduct a similar portion of the drain current, then you'll have to do something to get them to have similar gains, as the individual gain on each will come out to Vin*yfs*Rd. Yfs will be different for each JFET, even of the same type.

It only takes running into a few - or a few hundred - of these issues to get to the point that JFETs are difficult to deal with because of their wide variations.

If you have room, it's a lot simpler to use the JFETs for source (not drain) followers and then mix in something simpler, like the transistor version. Or just resistors. Or an opamp. You very quickly run into more components and more pins to connect than doing it with an opamp.

Which is why opamp are so very, very popular with analog signal design folks.

[midwayfair]

If you have room, it's a lot simpler to use the JFETs for source (not drain) followers and then mix in something simpler, like the transistor version. Or just resistors. Or an opamp. You very quickly run into more components and more pins to connect than doing it with an opamp.

Re: Source followers -- you mean just as the input buffer, then do the amplification with other devices, right? Not that using JFETs as source followers solves the low-pass filter problem, right?

Re: Transistor version of the schem Teemuk linked to -- I tried this last night and it does indeed work. Is there something special going on with the way the transistors work, though, that means that they don't need to be matched? Or is it just that a BC549C is going to be more consistent overall than any JFET on the planet? (The transistors also seem to have better bandwidth; still deciding if I like that better than a bit of distortion with the FETs, which also sounds good and is my main reason for using them).

Op amp: I'm a little more familiar with op amps than I was two months ago now, but I still have a hard time understanding some aspects of them. In what way does an op amp produce a better mixer? It seems to me that if I mix two decoupled outputs, then the output of either will then go through the other capacitor to 4.5V; won't that still produce a treble cut all the same?

[teemuk]

Unless you are building a very "strange" sounding distortion effect then trust me, you want the least bit of distortion from this circuit as possible. Choose devices accordingly. Oh, and to make that worse, I recall that at least the tube version had an extremely limited input signal range. Don't know if it applies to other devices but I have a hunch that it could be a "circuit thing"...

[midwayfair]

you want the least bit of distortion from this circuit as possible

Does something in particular happen? Crossover distortion or something like that?

[R.G.]

Re: Source followers -- you mean just as the input buffer, then do the amplification with other devices, right? Not that using JFETs as source followers solves the low-pass filter problem, right?

Yes, that's what I meant. When you have a complicated set of things you're trying to do, then there is an obvious choice between doing all of them in one or a few devices (i.e. solutions) that take care of all of them, or doing each function with one device or thing for each of the set of things. It gets complicated in terms of many devices to have each requirement handled separately - but it gives you back individual control of the solution for that one thing it solves. It is elegant to have the insight to have one or a few devices do it all - but this usually costs you in terms of it not doing all the necessary things well at the same time.

So yes, as buffers, to isolate the high input impedance requirements, and to also isolate the JFET variability at the same time. Gain set to whatever you want is much easier to do in other than a JFET. And JFET inputs may or may not "solve" the low pass filter problem. Bigger caps are almost always cheaper than fancy circuits where you're trying to keep costs down. And once you get down to a low pass in the range of the bottom of audio, the problem is solved for all further issues. So there's a limit to what "low pass" is an issue.

Re: Transistor version of the schem Teemuk linked to -- I tried this last night and it does indeed work. Is there something special going on with the way the transistors work, though, that means that they don't need to be matched?Or is it just that a BC549C is going to be more consistent overall than any JFET on the planet?

Both. The BD549 is going ot be more consistent overall than any JFET on the planet, AND there is something special going on - but it goes on that way with almost all bipolars. Bipolars are inherently more consistent by the nature of the semiconductor physics involved.

(

The transistors also seem to have better bandwidth; still deciding if I like that better than a bit of distortion with the FETs, which also sounds good and is my main reason for using them).

... and JFETs have nicer distortion. It's almost like Mother Nature WANTS to complicate things for you. My solution to this conflicting set of requirements is to showcase the devices that do the distortion I like, by isolating them in a "protected workshop" part of the circuit to do their distortion, then taking care of everthing else in more controllable ways.

Note well: as a designer, you want consistency far, far more than you want isolated great performances. Think about it.

Op amp: I'm a little more familiar with op amps than I was two months ago now, but I still have a hard time understanding some aspects of them. In what way does an op amp produce a better mixer?

The classical inverting opamp circuit forces the inverting input to a "virtual ground" where the voltage at the input doesn't vary much at all. If you put signals into that virtual ground through resistors, the currents mix into that virtual ground without interacting with each other - which is what mixing is all about, independent addition or subtraction without one input diddling up what happens with another input.

It seems to me that if I mix two decoupled outputs, then the output of either will then go through the other capacitor to 4.5V; won't that still produce a treble cut all the same?

But you mix two outputs into an opamp mixer by feeding them through a resistor. The low end rolloff becomes the coupling cap and the input mixing resistor, which is not variable. You have control and predictability. This may be a larger cap value, because noise considerations mean you rarely ever want really high input resistor values, but the rolloff is F 1/(2*pi*R*C) and there's no wondering about what else is affecting it.

[midwayfair]

Note well: as a designer, you want consistency far, far more than you want isolated great performances. Think about it.

Indeed. I'm trying to get better about such things. I designed a couple things last month with op amps that I actually liked and it cured me of some of my dislike of them, but I still prefer the sound of discrete components more overall. I did a mockup using a pair of inverting op amps, I'll see how that goes.

Teemuk: Distortion does seem unusually harsh when the BJTs break up with this. I can get more volume out of them before they distort than I can out of the FETs, but the FETs just sound way better if/when they clip.

Thanks for the help as always.