simple JFET "adder" idea seems to work.... your thoughts?

[Boner]

What do you think of this "adder", its just a JFET wired up in common drain mode and an extra "input"

In multisim it works 
Just startng testing it....and I think it works?

[PRR]

> I think it works?

"Works" can be anything. It really is a half-adder: it can do part of 1+1 logic.

Do you mean "audio mixer"? That's easily done with two resistors.

For reasonable values of Vcc and Vss?

For what values of input voltages? 1mV? 10V?

Gain? (If you are gonna spend two 50-cent parts, you'd like some gain.)

[Boner]

Thank you for your input!


So how can you do legitimate current or voltage mixing/adding using discrete components? Without the virtual ground in a simple inverting op amp design, my mind draws a blank....

[GibsonGM]

Hi Boner, welcome to the forum.  Here is a SUMMING AMPLIFIER circuit for you, using an opamp.  http://sound.whsites.net/articles/audio-mixing.htm
The article should tell you most all of what you'll need to know about mixing.

The important parts are the R1, R2, R3's - that is what is doing the mixing.  The opamp adds the gain that Paul was talking about (recovery stage).  Could just as easily be a JFET, BJT....you can find many variations on the internet using the term summing amplifier, BJT audio mixer, and so on.

I think I know what you're getting at with your virtual earth reference...you don't need it unless you follow the R's with an active device, which of course the opamp here is.  So, you see R4, R5 between pin 2 of the opamp in the 2nd picture doing this job.    Just returns and biases the input of the opamp higher so it can swing along with the signal....passive resistors don't need this...but sometimes you want the recovery gain after the passive Rs have mixed your signal for you...make sense?    Resistors will cause loss to your input signal, and add noise - so they can't always be used alone to do the job (altho sometimes they are enough).   

[R.G.]

On mixers:
There is another technique for mixing that was once popular, but got lost when opamps came along. It uses the idea of current source mixing. Current mixing is actually what happens at the virtual ground of an inverting opamp: the incoming resistors convert the incoming voltages to currents, which are summed at the input. The summed currents are "cancelled" by the inverted-sense current provided by the feedback resistor.

You can do current summing another way. Use one discrete transistor, bipolar or FET, per input and use the high impedance output at the collector or drain to create a "current source", and sum the currents into a single, shared collector/drain resistor. Bipolars are near-ideal for this, excepting only the bipolar's lowish input impedance.

It's just one other way.

[PRR]

> how can you do legitimate current or voltage mixing/adding using discrete components? Without the virtual ground in a simple inverting op amp design, my mind draws a blank....

How did we do it before "amplifiers"? (Remember there were telephones long before there were tubes...)

Within some limits, you can put sources directly in series or in parallel. With classic telephones you just wire all the phones in parallel to rig a "conference call". If only one person talks, there is signal drop. If "all" persons talk at once, the signal level may be roughly the normal level.

This works with electric guitars. Except the impedances vary widely and interact badly.

The classic Fender g-amp input has two 68K resistors, and you can work two guitars at the same time. The 68K is a compromise mix resistor that works OK for casual or panic use.

It does not work with modern amplifier outputs. Two amp output connected together "fight", distort bad. Pad them out with resistors, it works fine.

"Two channel" (not just 2 input) guitar amps have two preamps mixed together to one power amp with resistors. It is very very rare to find a guitar amp that bothers with "Active Mixing". How "legitimate" do you need to be??

Apologies to "ESP" Rod for hacking his art:

It has "issues" but is often all you need for music sound mixing.

"Crosstalk" is often NOT an issue. All your inputs end up in the same space-- that's the point. (It gets different in radio network where two very different programs may pass through the same routing mixer to different destinations.)

Here's an older mixer from the archives. Three different mike preamps to one tape recorder. In this case it was known the elderly preamps wanted "600r" loading. The 220+220 makes 440r out of each preamp, and the 2-of-3-way mix bus looks-like 220r, making 660r=600r on all preamps.

[Boner]

Hi Boner, welcome to the forum.  Here is a SUMMING AMPLIFIER circuit for you, using an opamp.

Thanks, but I'm trying to figure out how to sum currents/voltages without the use of op amps.... discrete.

You can do current summing another way. Use one discrete transistor, bipolar or FET, per input and use the high impedance output at the collector or drain to create a "current source", and sum the currents into a single, shared collector/drain resistor. Bipolars are near-ideal for this, excepting only the bipolar's lowish input impedance.

It's just one other way.

So each input would be a buffer, tap the output at the drain and sum each "drain" to another buffer/amp?

What if I were to do something like this? If I understand correctly, its what youre saying but without the individual buffers

*edit*

but here I am tapping the source, not drain....

[Boner]

Arrrg.... I really need to learn to read. I think you mean something like this?

[GibsonGM]

Hi Boner.

I know you wanted something discrete.  Go read the Elliot page I linked to...you don't need active devices like opamps...Paul also discussed this a bit.   In the end, you will use a device, some time, but for a quick and dirty deal, you can use resistors. 

[R.G.]

Yep. For quick and dirty, use resistors.

Here's a bit more theory. Circuits with linear components only (that is, resistors, caps and inductors) are naturally ruled by the principle of superposition. That's a fancy way of saying that any sources fed into the special case of a resistor network will add linearly into any node in the network in the proportion forced by the resistor values. So any resistor network at all will function as a mixer. The problem is getting it to mix the way you want it to, and with the side effects of the resistor loading being acceptable to you.

All resistor mixers are lossy. They can't make any input appear at an output without some loss of signal level. So you're inherently going to get less signal out from any input than you put into it. That's OK if you have circuts before and after the resistor mixer than can make up the signal loss.

Worse, the resistors that cause the mixing are your only way to isolate sources, and the internal impedance of the voltage sources feeding the network are part of the mixing resistor network. When you say "mixer" you generally want some way of having sources not interact. For instance, if you have variable levels of signal to mix, you want having one input turned all the way down not significantly lower the other sources. In resistive mixers, the more isolation you get, the more signal loss you get.

And you're still at the mercy of the source impedances. How do you mix an opamp output from an effect circuit with source impedance of, say, ten ohms, with a guitar output? The opamp output looks like a ground to the guitar signal, so you have to insert mixing resistors between them to the summed node. The opamp output will be unchanged by any significant resistor loading, but the guitar will be heavily loaded down, and its treble dulled, by any load much lower than half a megohm. So the guitar needs at least isolation/mixing resistors of maybe 220K between it and the opamp output to not have big treble loss (and signal loss, but that's implicit). This is independent of the impedance at the mixing node itself. That could be anywhere between a virtual ground and many megohms, and the guitar is still loaded by the two mixing resistors. This situation only gets worse if you want to put pots in the mixer to let you have variable mixing.

That chain of reasoning is why active mixers became all the rage. You can buffer each input, or only the high impedance inputs, and then you're freed of having to actually know what you're doing with the mixing resistors and summing impedance. All inputs come out of a low impedance from the buffers, and get non-interactively mixed.

Virtual ground summers at opamp inputs were handy because the virtual ground "shorted" any cross-loading, eliminating it, and the calculations of loading were reduced to just the input resistance. It's not too common any more, but the input impedance of a bipolar transistor's base is quite low if it doesn't have an emitter resistor. So you can mix pretty well with higher value mixing resistors feeding into the bipolar's base as a "summing node" and get some of the value of an opamp virtual ground. If you set up the bipolar properly, you can still get amplification at the collector and not have a lot of signal loss from the summing. You can enhance this with a feedback resistor from the collector to base for biasing and gain setting. Gets a bit tricky to bias and get the right AC gain, but it's possible.

Series resistor mixers still have the issue of the series resistors having thermal noise. Can't be stopped,  only ameliorated.

[Transmogrifox]

A compound pair can be used to perform the mixing function in a manner very similar to an op amp.  Notice the resistor value ratio is not 1:1 like you would expect in an op amp.  This is related to the fact that you don't get as high of gain from the compound pair as you do from an op amp, but still this gives you a fairly linear mix.

The advantage here is you only need 2 transistors for gain to accommodate as many channels as you want to mix here.  Add a pot in front of each channel "INX" for individual level control.

I used this type of circuit in my home stereo system for being able to mix an external and internal source to a class D amplifier.  I refactored below to allow for the high input impedance you might want for guitar usage.

"INX" refers to an undefined number of inputs:  You can keep adding the resistors here to accommodate as many channels as you want.

Also note that general configuration of a compound pair behaves much like an op amp where the emitter of "Q1" behaves like the "-" input, and the base of "Q1" behaves like the "+" input.  Collector of Q2 along with "Rbias" are like the op amp output. 

R7 acts as an offset trim, and is not usually needed for small values of "Rfb". In this case, "Rfb" is pretty large and the offset needs to be compensated to keep "Vc" near 4.5V.



You could add a JFET into this mix, but some pains would need to be taken for biasing and also compensating Rfb for lower gain.

You could also do this with a single transistor but you would have higher output impedance and less head-room.  Above compound pair method gives you nearly rail-rail signal swing.

[ElectricDruid]

Here's a common-base BJT mixer I used once upon a time - works well enough for a single transistor:

http://www.electroschematics.com/2917/audio-mixer-with-one-transistor/

For my version, I reduced the supply voltage to 5V (that was what I had available) but a 9V version would be simple enough too. The common-base arrangement is something you don't see that often, so in my head that makes it an "interesting" solution.

HTH,
Tom

[Transmogrifox]

You could also do this with a single transistor but...

Here's a common-base BJT mixer I used once upon a time - works well enough for a single transistor...

And voila!  Somebody has done it.   Imagine that! 

The 2-transistor compound pair is a steroids version of Tom's simpler 1-trannie mixer.  The trade-off is 2-trannies can give you a wider spread between high input impedance an lower output impedance, or 1-trannie gives you high-ish input impedance and low-ish output impedance.  Definitely an improvement on a simple passive resistor network.

What's interesting is I think you're onto something with your JFET idea.  I can see there's something there, but requires 1 JFET for every input.