Mixer Design

[yano]

gez:
Yes i'm planning on going with the 386...do you think it can provide enough current for all the channels as well as the summing opamps? My guess is that'd i'd be safe with an LM386-3.
What do you think about C2 (2.2uF)? Do I still need it with that voltage divider gone?

12afael:
I'm reworking parts of my design...the input buffer (now that it has opamps in it) i will be adding pots for gain in the feedback loop for the input, but the outputs will just be on a volume control. In fact though, I could put a pot in the feedback loop for the 2nd summing opamp for gain as well...I'm not sure what all i want to do yet.

[brett]

About C2.  You need it because you don't want any DC voltage across the potentiometer.  Because that signal is at DC ground after the pot, you'll need to re-bias the signal in the mixing stage (ie isolate it with input and output caps and connect it to Vref/Vbias)

[gez]

gez:
Yes i'm planning on going with the 386...do you think it can provide enough current for all the channels as well as the summing opamps? My guess is that'd i'd be safe with an LM386-3.

Op-amp inputs are designed to draw very little current for their bias.  FET input amps don't require input bias current, but they do have leakage so you'll still see quoted figures for 'bias current' in data sheets.

Total it all together and it's pretty insignificant, but enough to cause loading of a divider if the resistor values are on the large side.  The 386 is an ideal solution and it'll provide a rock-solid bias to all the amps you use.

[gez]

Because that signal is at DC ground after the pot, you'll need to re-bias the signal in the mixing stage (ie isolate it with input and output caps and connect it to Vref/Vbias)

One thing I've done in the past is direct coupling from op-amp input buffers to a mixer.  Using amps from the same chip doesn't seem to cause offsets in the mixer.  Whether it would here though I couldn't say as there are more than four inputs so different chips would have to be used.  The pots coming off the output of each input buffer are  connected to V Ref too, so it might be possible (well, it would with just two or three inputs and a quad chip).

[gez]

What do you think about C2 (2.2uF)? Do I still need it with that voltage divider gone?

You've reference the 100k pot to half the supply voltage so theoretically you could directly couple everything.  But, as I've said above, using different chips might create an offset in the mixer.  I've only ever used simple mixers as part of larger circuits, so I couldn't say whether this would be a problem.

The ground connection for each input jack should connect to ground and not V ref.  Provided you leave the 100k pots connected to Vref you shouldn't need an input cap for the mixer.

[gez]

The last schematic you linked to has C2 as 10u and not 2u2.  Are you referring to a different part of the circuit?  Could you link it? (I'm a bit confused/slow!)

[yano]

I've been changing schematics a lot, renaming parts...i've got the jacks on ground now, i got a bit crazy thinking about virtual grounds.

Anyway, the latest schematics can be seen here:
http://yano.lardpirates.com/electronics/mixer/stages/input.php

I was referring to C2 in the mixing stage, which is a 2.2uF cap to ground, coming off of the power lead of my power input jack, i'm not sure its neccesary anymore

All the schematics can be seen from here:
http://yano.lardpirates.com/electronics/mixer/stages/

I probably won't be posting any schematic changes until i get this breadboarded out...but i'll only change them if i can't get them to work  :wink: .

[gez]

The 2u2 is just connected up across the rails.  If it’s there for filtering then it’s a waste of space.  You have a connection for ‘voltage reference generator’ shown in the mixer section coming off +9V.  This would go to the + power connection of your 386 and Vref is taken from the 386’s output (4.5V).

The ‘channel with +6dB gain control’ input schematic is a little off.  You’ve got the 100k pot wired with one leg to ground so it’ll screw with the DC bias of the amp as it’s turned.

If this were me, I’d replace R2 (100k resistor in the first schematic) in the feedback loop of the buffer with a 250k pot set up so that it’s resistance increases from O ohms as it’s turned.  I’d then do away with R3, the 100k pot at the output, as it would now be superfluous.

[gez]

By using the -ve input you’ve lost the high input impedance you could have had by using the +ve input, which means your ‘buffers’ are a bit of a waste of time.

Read back through this thread, look at some other schematics and hopefully it’ll click.

[yano]

Man, I'm confused.

The two inputs of an op amp are not equally high resistance?

[brett]

Hi Yano
Don't despair.  You've made a terrific journey just to get this far.

The +ve input of an op-amp provides a super-high input impedance.  So high that the effective impedance is usually set by the biasing resistor (between Vref and the signal line).  THis is often 1M or thereabouts.

The input impedance of the -ve input of an op-amp is set by the series resistor in the signal line.  To avoid resistor noise in the input and feedback resistors, the value of that resistor is usually kept low (often 10k or thereabouts).  

Does that make sense?  What it means in practice is that for guitar signals, etc, that need high impedance, you are better off using the +ve input of the op-amp for the input stage, and biasing the stage with a nice, big resistor (e.g. 1M or 2.2M).

[yano]

What about the LM386 biasing approach, how does that compare to using 1M or 2M2 resistors?

The reason for using 100K resistors was in order to reduce filtering of frequencies below 100Hz (with the 100k resistor, the filtering starts below 15Hz or so).

[R.G.]

What about the LM386 biasing approach, how does that compare to using 1M or 2M2 resistors?

Think of  it this way. With two-resistor biasing, you get an input impedance (that is, loading on the thing driving the input) of the parallel combination of the two resistors since the power supply is an AC short circuit. For 1M resistors, that's 500K, for 2m2 resistors, 1M1.

OK as far as it goes. However, there is a lot of voltage across those resistors and they are high valued, which is the ideal setup to make the resistor noise a prominent part of the input signal.

The LM386 bias source is a very low impedance source of bias. If you connect it directly to the inputs, it shorts out all of the input signal. So you put a high value resistor in series with it, maybe the same 1M or 2m2 values. Now the input signal at the input sees the high value resistor as a load, and you get low loading of the signal. The resistor has a low DC voltage across it, which minimizes its noise contribution.

[yano]

In the schematic shown here (linked from http://europa.spaceports.com/~fishbake/mixer/linemix.htm )



What is the function of the .1 and 100uF caps on the power supply lines?

[Edit 03/01/04--I answered my own question]