Mixer Design

[yano]

Here is what I think the input stage is going to look like. It seems to me that using the op-amps will be easier to build than the JFET buffer. Correct me if I'm wrong.



Questions:
-Should I increase the value on the level pot? If so, what to?
-Is the channel listen jack wired properly?
-Are there any op-amps that lend themselves to this application, or can I simply use generic, like a TL072 or NE5532...? Does anyone know of an op-amp with 6 pairs of inputs?

Thanks.

[R.G.]

Comments on the schematics:

- the schemos imply that you will be using bipolar (+ and -) power supplies centered around ground; if that's what you're doing, great
- you will need some kind of resistor to ground from the noninverting inputs of the opamps to bias them; 1M should work fine
- think about putting a non-polar capacitor in series with the inputs anyway; any DC offset in the circuit coming in will be carried through the rest of the circuit, causing bad juju all around.
- It might be good to put a 10K resistor in series with the input jacks and reverse biased diodes from the signal line to the +  and - power supplies so that if someone plugs 120Vac into the input, a fuse will blow somewhere, but it will not smoke your whole mixer.
- You used some polarized caps in your first schemo; do not use polar caps where there will not be DC volts across them - they die on a 0V diet.

Input pot at 10K is fine; Channel jack is OK; NE5532 is a *great* opamp for this.

[yano]

Alright, more questions, more schematics...



Were the polarized capacitors you were talking about the ones in that box?



The questions I have for the input stage are in the image...

My power supply is either going to be a 9V battery or wall/small wart at 9V...but I'm considering 12V. I guess these are bipolar, I'm not really familar with the term. Thanks a lot, R.G. for your input.

[brett]

A few ideas.

If you want to power this off a 9V plug, you'll need to make a few changes.  As RG said, with a bipolar supply (+ and -9V) you'll be right with what you've got.

For 9V supply, you'll need that powers upply box.  It gives you a stabilised 4.5V supply that you'll definately need.  (that 4.5V is often called Vbias).

Put that 10k resistor after the 1M resistor, and put a 0.1uF cap (in-line) before the 1M resistor. Now, connect the 1M resistor to the 4.5V supply - the junction of R8 and R9 - instead of earth.

Also, shift that polarised cap to *before* the 10k "level" pot.  Having done those things, the signal path from input cap to output cap will operate happily at 4.5V, allowing you to power this from a single-sided power supply of 9V.  (ie you can hook up pin 8 to 9V and pin 4 to 0V (instead of -9V))

If any of that is unclear, ask again.

[yano]

So, what I'm understanding is that a bipolar power supply (in this case) has a -9V ground, rather than 0V?

And those mods that Brett listed, they are not neccesary if I'm using a bipolar power supply?

If I were to use a bipolar power supply, would that have any impact on the rest of the circuit (rather than just the input stages)?

How can I make a bipolar power supply, or where can I buy a +/- 9V wallwart?

Thanks in advance.

[gez]

You wanted a peak detecter set for 2V peak-to-peak, suggesting that you're using low level signals.  You're only dealing with unity gain in each stage of the circuit too.  You can get away with using a 9V supply no problem.

If you're using a regulated 9V wall-wart you don't need the extra circuitry Bret mentioned, i.e. regulator fed from 12-25V (or whatever it was).edit: if that is what Brett was implying (apologies if it wasn't).

[gez]

So, what I'm understanding is that a bipolar power supply (in this case) has a -9V ground, rather than 0V?

And those mods that Brett listed, they are not neccesary if I'm using a bipolar power supply?

If I were to use a bipolar power supply, would that have any impact on the rest of the circuit (rather than just the input stages)?

How can I make a bipolar power supply, or where can I buy a +/- 9V wallwart?

Thanks in advance.

With a bipolar supply 0V is still earth and not -9V.  The positive inputs of each op-amp connect to 0V so a divider for bias is unnecessary.

Using a bipolar supply gives you more headroom, but if you’re only dealing with 2V peak-to-peak signals it’s not necessary, you can get away with just a 9V supply.

[yano]

This mixer is intended for guitar level and line level inputs, straight from the instruments, essentially.

My plan calls for power to be supplied via a standard BOSS-type wallwart (for maximum flexibility), is this acceptably regulated, or should I use, say, a 12V wallwart going to some sort of regulation circuitry?

Also, I never intended to use a bipolar supply...mainly because I didn't even know what they were when I designed this

[gez]

Boss type wall-wart is fine...s'what I use!

[yano]

I've got something of a site going for my mixer project, its located here:

http://yano.lardpirates.com/electronics/mixer

Some of you mentioned earlier some things about polarized capacitors in the main mixing circuit...(http://yano.lardpirates.com/electronics/mixer/mixer.php)...what were you getting at?

Also, what % tolerance capacitors should I use?

And what of polyester film? Should I use this for the .1uF capacitor?

[gez]

I gather from your previous posts that you're running the whole circuit from 9V and not using a bipolar supply?  If that's the case you need DC blocking caps in the input stage and you need to reference the ground connections (except the jack grounds) to half the supply voltage (center of the 100k divider - though it would probably be wise to make the resistor values smaller and up the value of the decoupling cap).

The input stage as shown is for a bipolar supply.  If that was what you intended then it would be an idea to include input caps (non-polarised).

[gez]

In the mixer section the 'master level' pot has DC voltage across it so will probably crackle when turned.  Either stick it at the output AFTER the output cap OR connect the lug currently connected to ground to Vref (centre of the 100k divider).

Take a look at RG's "Quick & clean bias voltage thread" - might be an idea to provide the bias for all those amps?

You might be able to get away with replacing the 10k feedback resistor in the second op-amp with a 10k pot set up to go from 0 ohms - 10k as you turn it.  This would act as the master level control and saves you a resistor.  With a FET input amp there wouldn't be much leakage current through it (hardly any voltage across it) so it probably wouldn't crackle.

[yano]

I learned some interesting things today. I will require a voltage divider per channel. If I try and use the same 4.5V lead, the signals will mix prematurely and result in distortion of the actual frequencies. I think it will be more practical to use two 1M resistors rather than the LM386 way.

This is my rethought input stage:


C3 is included so that the channel level potentiometers will not interact, at least that's what I think they're supposed to do. Can someone tell me if they are neccesary or not?

Edit: Can I connect the channel listen jack between C3 and R3?

[R.G.]

I learned some interesting things today. I will require a voltage divider per channel. If I try and use the same 4.5V lead, the signals will mix prematurely and result in distortion of the actual frequencies.

I'm curious - where did you "learn" that? It's not necessarily true.

If you make a bias voltage from resistors only, then there may be some basis in fact. However, every bias reference for more than one stage uses a capacitor or some other means to reduce the impedance of the bias voltage to the point where it shunts all the signal to ground, producing a quiet spot where the signals are swallowed by ground.

The commonest way to do this is with a big capacitor to ground from the joint resistor spot.

The LM386 bias trick is even better than the R-R-C bias trick, because it *actively* reduces the impedance of the bias voltage, with the amp doing everything it can to suppress any voltage movement of the point. The signals can't mix to any noticeable degree and then get back out to the signal path.

[yano]

I've been expiramenting with pspice simulations, and the waveforms appeared to be quite distorted when the 4.5V was all connected to each other...

So you think the 386 "trick" will work?

I think at this point, my best bet will be to start breadboarding and expiramenting.

[brett]

With your pspice simulations, did you put a big capacitor (e.g. 100uF) from the junction of the bias resistors to earth?  If you did, there should be no interference between channels, even if they use the same resistors for the bias voltage divider.

Also, don't forget that with your 9V supply you'll also need to use the bias voltage for the mixing stage (and the associated input and output caps to isolate the stage with respect to DC).

[yano]

so you're saying I should bias the inputs of the amps in the mixing stage?

I assume there's no reason to unbias and then rebias the signal between op-amps.

[gez]

I think what Brett's trying to say is you've used a divider (haven't got the schematic in front of me but I believe it was two 100k resistors) in the mixer section and the positive inputs of the mixer amps connect to the centre of this divider.  You should be able to bias the op-amps used in the input stage using the SAME divider - connect a 1M resistor from each + input to the centre of the two resistors - PROVIDED the junction of this divider is decoupled by a LARGE value cap.

It would be advisable to make the value of the divider resistors smaller, or use RG's 386 idea (yes, it will work!).

[gez]

Ok, the divider resistors have  vanished since the I last looked!  

For the voltage ref (4.5V) either use a divider with a large value decoupling cap, or the 386 idea.

[12afael]

about the aoutput buffer why don`t you transform it to a gain stage you can use a gain control instead the volume.