Questions regarding my audio mixer design

[Gordito]

Hi, I've got a few questions about my design and i am open to improvements. At the moment this is just one channel (mono), i'll duplicate and use dual pots once it's done.

I've assembled it on a breadboard and works well as far as i can tell. The only thing i haven't done yet is run it from a 9v wall wart (get one tomorrow), i use a power supply.

I've used a sine wav generator and my mp3 player connected into my old koss mini-system and an older oscilloscope to check how well it behaves. The filter seems to work well.

The power is from a 9v wall wart, regulated with a LM7809 (IC1).
Inputs go into IC2A, the inverting summing opamp (single supply) with a gain of 14.
Summed signal goes thru the Baxandall filter around IC2B and goes to the line-out buffer (IC2C) and to the half wave peak detector (IC2D) which sends a vcc/2 biased voltage to the LM3915 logarithmic led driver controlling a led bar (3 green 4 yellow 3 red).

EDIT: The LM3915 is in dot mode.

Caps C1 to C7 are film type, resistors in Baxandall filter resistors are 1%.


Questions:
1)Biasing non inverting inputs for single supply opamps: What are the 'rules' for determining the values for the resistors and caps? The 100 nf cap i understand is for transients but how big (or small) must the larger one be? Resistors, why 10K or 100K?

2) Can i use one voltage divider / cap network so supply the vcc/2 bias for TWO opamps (on the same side of chip and the transient cap between pins) or should they really have each their own dividers and caps.

3) Output cap size: Is the frequency formula here the good old 1/(2*pi*R*C) where R would be the expected input impedance of the equipment we are plugging the mixer into? I believe that is normally 10k?

4) I've often read that 9v wall warts give 12v unregulated. With diode D1 (1n914) in the path, is this reliable enough for the LM7809? I guess i'll see tomorrow...

EDIT 5) I've sometimes seen a small cap (27pf - 47pf) in parallel with the feedback resistor in inverting opamps (R5 in my schematic). I presume this is to filter out very high freq. Is this desirable and if so what is the formula to determine the cap value?

Any improvements or preventive measures are welcome.

 

[audioguru]

The only thing i haven't done yet is run it from a 9v wall wart (get one tomorrow), i use a power supply.

No.
The 7809 needs an input that is a minimum of 11V and the series diode increases the required input. Use 12VDC.

The power is from a 9v wall wart, regulated with a LM7809 (IC1).
No.
Again, the LM7809 with a series diode needs an input of 12VDC.

1) Biasing non inverting inputs for single supply opamps: What are the 'rules' for determining the values for the resistors and caps? The 100 nf cap i understand is for transients but how big (or small) must the larger one be? Resistors, why 10K or 100K?

Each opamp has a different amount of input bias current that is supplied by the resistors. The TL074 has Jfet input transistors so its input bias current is almost zero and almost any resistor values can be used. 100k instead of 10k is fine. The 100nF bias filter capacitor filters out hiss from the voltage regulator IC and should be 3.3uF to 10uF.

2) Can i use one voltage divider / cap network so supply the vcc/2 bias for TWO opamps (on the same side of chip and the transient cap between pins) or should they really have each their own dividers and caps.

Yes, only one VCC/2 circuit is needed since the input bias current of the TL074 opamps is almost zero.

3) Output cap size: Is the frequency formula here the good old 1/(2*pi*R*C) where R would be the expected input impedance of the equipment we are plugging the mixer into? I believe that is normally 10k?

Correct.
The TL074 and most other high frequency audio opamps oscillate when they directly feed the capacitance of a shielded cable. Add 100 ohms in series with the output from C7 to prevent the oscillation.

4) I've often read that 9v wall warts give 12v unregulated. With diode D1 (1n914) in the path, is this reliable enough for the LM7809? I guess i'll see tomorrow...

I have a small "9V" wall wart that has 18.5V with no load but is 9V with its rated load. I have another 9V wall wart that has an output that is always 9V which is too low for your circuit.
Use a wall wart that is rated at 12VDC at 500mA.

EDIT 5) I've sometimes seen a small cap (27pf - 47pf) in parallel with the feedback resistor in inverting opamps (R5 in my schematic). I presume this is to filter out very high freq. Is this desirable and if so what is the formula to determine the cap value?

The extra tiny capacitor is not needed unless you have a radio station transmitter next door (then other capacitors are also needed).

 

[Gordito]

Thank you for your help.

The 100nF bias filter capacitor filters out hiss from the voltage regulator IC and should be 3.3uF to 10uF.

Don't the capacitors next to the voltage dividers do the job of filtering out the hiss (4.7uF caps C13 and C15)?

For the 100nF caps (C14 and C16) to replace with 3.3uf to 10uf, is there a rule to determine the proper size? And which type should be used (electro vs tantalum).

About the 10k vs 100k voltage divider values, if i understand correctly i would use a lower value if i were using an opamp with lower input impedance in order to have a 'firmer' voltage divider.


I also suppose i could use one single voltage divider (and caps) for all four opamps on a single chip but for additional chips it might be pushing it (due to noise possibly getting coupled onto the trace, non zero trace resistance, etc).

The extra tiny capacitor is not needed unless you have a radio station transmitter next door (then other capacitors are also needed).

A question about theory, suppose i did want to add a cap in parallel to the feedback cap to make a really simple filter, how would that be calculated? Again 1/(2*pi*Cfeedback * ( Rfeedback || impedance seen going into the next stage))? And whatrolloff would it give?

 

[audioguru]

Don't the capacitors next to the voltage dividers do the job of filtering out the hiss (4.7uF caps C13 and C15)?

No.
C13 and C15 are not needed since C11 keeps the regulator from oscillating and C12 keeps the opamps from oscillating. Noise on the power supply pins of an opamp does not produce noise in its output. C14 and C16 filter away noise at the inputs.

For the 100nF caps (C14 and C16) to replace with 3.3uf to 10uf, is there a rule to determine the proper size? And which type should be used (electro vs tantalum).

If the bias divider use 100k resistors then in parallel they make 50k. If the capacitor is 10uF then it will charge from the 50k resistance in (10uF x 50k)= half a second which is a pretty good filter and a reasonable charging time.

the 10k vs 100k voltage divider values, if i understand correctly i would use a lower value if i were using an opamp with lower input impedance in order to have a 'firmer' voltage divider.

The poorest opamp has an input bias current of 0.5uA so 100k resistors to bias it will be fine. You do not need 10k resistors.

I also suppose i could use one single voltage divider (and caps) for all four opamps on a single chip but for additional chips it might be pushing it (due to noise possibly getting coupled onto the trace, non zero trace resistance, etc).

One voltage divider with two 100k resistors can bias thousands of TL074 quad opamps. Use one filter capacitor at each opamp. But do not use thousands of 10uF capacitors because they will take weeks to charge.

A question about theory, suppose i did want to add a cap in parallel to the feedback cap to make a really simple filter, how would that be calculated? Again 1/(2*pi*Cfeedback * ( Rfeedback || impedance seen going into the next stage))? And whatrolloff would it give?

A capacitor parallel with the feedback resistor reduces the output level 3dB (x 0.707 times) at 1/(2*pi*Cfeedback * Rfeedback . The input impedance of the next stage has nothing to do with it. It is a single RC so the slope is 6dB (half the level) per octave.

 

[Gordito]

Thanks for the explanations.
Keeping a local copy of this, gonna be an helpful reference

 

[Bob Scott]

The circuit is OK except that it can be improved to lower noise hiss. If it were my mixer in a hifi or pro environment, I would give each input it's own op-amp amplifier stage and have a separate mixing stage. That will cut your noise gain by 12dB compared to the circuit as it is now. Summing amps are notorious for adding noise, so it is a good idea to sum the outputs AFTER the microphone signals have been preamplified. You can do that by just adding one quad op-amp, one stage/input.

Note: It is a good idea to make the first stage non-inverting. Most preamps have the noise go down if the input gets shorted to ground. If you use a non-inverting stage for the input, the amplifer noise increases when the input is shorted to ground.

 

[Nigel Goodwin]

Summing amps are notorious for adding noise, so it is a good idea to sum the outputs AFTER the microphone signals have been preamplified.

Where did he say it's a microphone mixer?, presumably it's just a line level mixer so perfectly fine.

It would be a pretty crappy design if it was supposed to be a microphone mixer

 

[audioguru]

The gain is only 15 so the vocals are probably screaming into the mics.

Oh, maybe the mics are inside drums?

 

[Gordito]

Sorry i didn't mention that i am using it for both line in levels and the headphone output from an mp3 player. The mp3 player at a reasonable volume (20 out of 30) outputs a 75 mVpp to 100mVpp signal which explains the max gain of around 14.

At these levels does it warrant the extra opamp(s) for doing the gain separately from the mixing?

The 12 dB figure is because i have 4 inputs, the noise is getting summed (4x = 12db). Correct?
Where is this noise coming from and is it also 12db if i have a single input connected?

My stereo jacks have NC switches for left and right (not shown in schematic). Should i ground these when a jack is not in use or would closing the circuit actually make it more noisy?

Thanks!

 

[Nigel Goodwin]

Sorry i didn't mention that i am using it for both line in levels and the headphone output from an mp3 player. The mp3 player at a reasonable volume (20 out of 30) outputs a 75 mVpp to 100mVpp signal which explains the max gain of around 14.

At these levels does it warrant the extra opamp(s) for doing the gain separately from the mixing?

No - personally I found it obvious that it was only a line level mixer, if you wanted to make it a mike mixer, simply add mike preamps before the pots.

 

[audioguru]

Sorry i didn't mention that i am using it for both line in levels and the headphone output from an mp3 player. The mp3 player at a reasonable volume (20 out of 30) outputs a 75 mVpp to 100mVpp signal which explains the max gain of around 14.

At these levels does it warrant the extra opamp(s) for doing the gain separately from the mixing?

No. The mixer is fine.

The 12 dB figure is because i have 4 inputs, the noise is getting summed (4x = 12db). Correct?
Where is this noise coming from and is it also 12db if i have a single input connected?

No.
A summing amplifier is not noisy. The noise (what noise?) is not increased because there are four inputs. The opamp has a gain of 15 so it simply amplifies its own input noise (plus a tiny amount of input resistors noise) 15 times. Of course if all four inputs are noisy then their noise is added together (+12dB) and is also amplified 15 times.

My stereo jacks have NC switches for left and right (not shown in schematic). Should i ground these when a jack is not in use or would closing the circuit actually make it more noisy?

Grounding the inputs will not affect hiss but it might reduce mains hum pickup if the wiring has poor shielding.