Audio crossbar switching - question

[david77]

Ok, I'm planning to build a crossbar switch for a hifi setup. It is a bit more complicated than your run off the mill stereo.

I've got the switches, they're mechanical non-interlocking push-button switches.
Each input channel will be fed by a small op-amp based amp to get signal levels from all sources to the same level.
For the outputs I was thinking I'd feed them into an virtual earth type circuit per channel.

This picture may help to clarify what I plan to do:



Never mind the details, at the moment I'm only interested in the theory.

Let's say there are signals on all three inputs.
If I depress S1 and S4 I get the sum of input 1 + 2 on output 1, so far so good.
But what would happen if I depress S1 and S2? All other switches are off.
Does the signal split between output 1 + 2 evenly? Probably not, right?

Please enlighten me, I've been thinking about this for some time but have not been able to make up my mind.

[Psi]

If you depress S1 and S2 then the same signal just goes to both opamp inputs.

[david77]

Ah. So the two virtual earth amps don't interfer with one another?

[Psi]

I'm no opamp expert but the inputs are highz, so i can't see any input issue.

[amspire]

With the way you have drawn the opamps, you will get a maximum amplitude square wave out of the opamps.

The current circuit as drawn will produce a beautiful maximum amplitude square wave on output1 and output2 as no opamp has any feedback circuit.

If the input is via the inverting input, it means you are probably going to wire is up as a summing amplifier with a resistor (not shown) from the inverting input to the output.

If it is wired this way, what happens if S1 and S2 are closed?

The two opamps (output 1 and output 2)  will have different offset voltages and the result is one opamp will go near maximum positive voltage on the output, and the other will go to maximum negative output.

One of the two opamps will end up being saturated at maximum or minimum voltage, and the other will probably be just below saturation, so the input signal will end up being superimposed on the large dc offset on the output.

If the opamps are going to be wired differently, how?

Richard.

[HackedFridgeMagnet]

Tell me if I misinterpret you but I think what you want is:
 to set the gain of each input channel individually and then assign them to the output channels via an adder on each output channel.

As Richard said all your op-amps need feedback to control the amplification.

But I think another change necessary in your circuit is to put an op-amp just after the slider on each channel but before the adder resistor so that the slider doesn't effect the adding.
You might be able to drop the input op-amps as well, depends what is coming in.
If all the switches on a channel are off you probably don't want to leave the inputs to the output op-amps floating either

[david77]

Yes, of course. I forgot to draw the feedback resistors, I saw it right after I did not save the CAD file  .

I could wire them as standard non inverting (high input Z) amplifiers but then I'd get the draw back I wanted to avoid. The setting of the level trimmers would influence one another in case two inputs get switched to one output.

EDIT: I've added the feedback resistors in MS Paint  .

[david77]

@Magnet: That's why I wanted to use the summing (vrtual ground) amps at the outputs of the switch matrix.

What I need is this: I have many different sources CD, phono, tape, MP3 etc. both consumer and studio equipment. All this stuff needs to be hooked up to my stereo system. Now the consumer gear has wildly varying output levels, the studio stuff is better behaved but of course has balanced outputs.
The idea was to build something like this:

source ---> line preamp unbal. ----> switching matrix ----> fader ----> power amplifier
source ---> line preamp bal.     ----> switching matrix ----> fader ----> power amplifier

The line preamps will get trimmers to set the gain/attenuation as desired, that feeds into the switches, with wich I can route the sources to 5 fader channels.

[amspire]

Now you have added the feedback resistors, then my description of what will happen is correct.

What you probably want instead is to connect all the amplifiers as non-inverting high impedance amplifiers - perhaps just connect the inverting input to the output, and use the non-inverting inputs for the signal.

Even better - a non-inverting unity gain amp on the volume control outputs. Then the switches. Then the resistors into the summing amplifiers.

Richard.

[david77]

I've drawn up the circuit like Richard suggested with the unity gain buffer after the volume pots and a resistor after each switch feeding into a summing amp. This seems to be a good solution that I really should have come up with myself, thanks Richard!

One question I'm asking myself now is do I really need the unity gain buffers for each channel or could I leave them out and achieve maybe a maximum attenuation of -1...2dB? It would save a whole lot of opamps.

I had a think about it and the whole scenario is rather hypothetical, there's really no reason why someone would want to feed one source into two different channels, so a bit of attenuation wouldn't matter that much. It comes down to selection of component values, I'd guess. I'll have to breadboard that and find out.

[amspire]

I've drawn up the circuit like Richard suggested with the unity gain buffer after the volume pots and a resistor after each switch feeding into a summing amp. This seems to be a good solution that I really should have come up with myself, thanks Richard!

One question I'm asking myself now is do I really need the unity gain buffers for each channel or could I leave them out and achieve maybe a maximum attenuation of -1...2dB? It would save a whole lot of opamps.

I had a think about it and the whole scenario is rather hypothetical, there's really no reason why someone would want to feed one source into two different channels, so a bit of attenuation wouldn't matter that much. It comes down to selection of component values, I'd guess. I'll have to breadboard that and find out.

If you do not have the buffers, then when a switch turns on, it can cause a level shift and if something else is using it at the time, there will be a click.

If the pot is 10K and the resistor is 1M, then the level shift will not be audible. A 1M resistor has more thermal noise then a 10K resistor, but if you are switching at high levels, it may not matter.

The unity gain opamps can be replaced with a transistor emitter follower and a series capacitor on the output (or summing amplifier input).

Richard.