Just how bad is it? Audio mixer with headphone amp.

[paulca]

Okay, so backing things down to being a bit more simple.

A single channel input pre-amp.
As per the annotation:  Gain of 0 - 3.128



I did consider messing around with a series resistor to try and get 1.000 gain in the middle of the pot, but I think it's a fools game.

When I can get this right I will move add one mono channels to the mix down amp and see if I can get that right.

I do like the idea of the inverting summing amp for mix down.  I like that the inputs will not interfere with each other, although now I have the volume pot before the pre-amp I'm not sure it matters.

I'll review the other comments before attempting the mix down, as I think my Tesco delivery has arrived

[Bassman59]

Take a look at the schematic for the Rane MLM82S. Everything you want to implement, except for the headphone amp, is there.

One could argue that you should use THAT 124x differential receivers for the line-level inputs instead of the op-amps, and INA163 for mic-level inputs.

For the headphone amp, the OPA551 has enough drive to clean the wax outta your ears.

[Audioguru]

Your idea of using an opamp for each input with a useless gain of only 3.1 makes no sense. You can barely hear a level change of only 3.1 times.
The input level control must be logarithmic, not linear so when it is turned down to half the input level is 1/10th and when it is turned to maximum the input level is maximum.
Since the input level control set to halfway cuts the level to 1/10th then you need gain of 10 following it. The gain of 10 can be in the mix opamp then the input level control can feed directly into the mix opamp through a series resistor that the inverting mixer needs.

Carvin and Rane use 3 opamps or an instrumentation amplifier because they have balanced mic inputs that you don't have. Rane even has balanced line inputs using a differential opamp but you do not have balanced line inputs. Then you do not need input opamps.

The 32 ohms headphones need about 2V RMS to produce 0.125W to be loud. The old NE5532 (about 41 years old) is spec'd to drive 10V RMS into 600 ohms which has a peak current of 23.6mA. With a peak current of 23.6mA the voltage in a 32 ohm earphone is 0.53V RMS producing 88mW which is loud but maybe not loud enough. Your circuit uses two NE5532 opamps each with a 47 ohm output resistor so the output power into a 32 ohm earphone is enough when  the power supply is 6V or more.

   

[BrianHG]

Your idea of using an opamp for each input with a useless gain of only 3.1 makes no sense. You can barely hear a level change of only 3.1 times.
The input level control must be logarithmic, not linear so when it is turned down to half the input level is 1/10th and when it is turned to maximum the input level is maximum.
Since the input level control set to halfway cuts the level to 1/10th then you need gain of 10 following it. The gain of 10 can be in the mix opamp then the input level control can feed directly into the mix opamp through a series resistor that the inverting mixer needs.

Carvin and Rane use 3 opamps or an instrumentation amplifier because they have balanced mic inputs that you don't have. Rane even has balanced line inputs using a differential opamp but you do not have balanced line inputs. Then you do not need input opamps.

The 32 ohms headphones need about 2V RMS to produce 0.125W to be loud. The old NE5532 (about 41 years old) is spec'd to drive 10V RMS into 600 ohms which has a peak current of 23.6mA. With a peak current of 23.6mA the voltage in a 32 ohm earphone is 0.53V RMS producing 88mW which is loud but maybe not loud enough. Your circuit uses two NE5532 opamps each with a 47 ohm output resistor so the output power into a 32 ohm earphone is enough when  the power supply is 6V or more.

 

He is going from line level to line level at the next stage.  He is not amplifying a microphone, he is not looking for serious gain, nor does he want any of his stages to clip by turning up to source volume to the max.  If he wants minimal hiss and the best possible frequency response of the opamps, the minimal gain of 3 at that stage and another gain of 4, approx 10db at the mid stage isn't too terrible a choice.  Maybe to get a gain of 6db after the 3 source mix, he can choose a gain of 6, but going to a gain of 10, with a maximum line level source of +/-1.25v * 10, he hits his +/-12 supply rail, though, he can go to a +/-15v supply.  Placing some gain (maybe targeting to 6 instead of 3)  at the input stage, next the 1/3rd level loss due to the 3 way mixing, then adding some gain in that middle opamp as an additional boost, say x2 or x4, is a strategy where he can use with his 3 way all positive audio design mixer, generating some additional signal gain without any clipping risk anywhere.

Remember, if he was doing a negative mix in the middle feeding the -input, he would not need that extra voltage swing at the first opamp stage outputs.  But now, those opamps will need to invert the signal as well to make positive audio in the end.

[Audioguru]

The original circuit has no overall gain, it has loss.
1) Input opamp gain= 2.
2) Passive mixer gain= no gain, the output is 0.33 times because a 47k mix resistor is loaded with two 47k resistors in parallel from the other two inputs.
3) Mix buffer, no gain but its output is divided to be 0.52 times.
4) Output buffer= no gain.
Then the output level is 2 x 0.33 x 0.52= 0.34 times with an input level control at maximum.

My idea has the mix opamp a true inverting active mixer with a gain of 10. The mix opamp feeds directly to the output with no loss and it also feeds the headphones amplifier.
The logarithmic input levels at halfway produce the same output level as an input level and the level control can boost a low level up to 10 times or reduce it to zero.

If true phase is important then an inverting opamp can be added to each side with no change of gain.

[paulca]

The original circuit was wrong.  It was a best effort with a few bits of critical information missing from my concept.  I'm learning.

The Carvin and Rane as well as the mixers I have used have gain on the inputs to normalize them to the same level before mixing them.  The only exception was a £35 DJ mixer which assumed identical turntable decks.  Even that one (actually a 3 channel with 1 line input) was fairly useless as the line input was too low and needed the mix fader planted to MAX and the turntables turned down and then master raised to get a decent mix.  Bad gain structure = noise. 

If you search on ebay for "Audio mixer" all but the cheapest and dodgy-ist have pre-fader input gain on all channels.  Including line only channels.

I could go for more gain at the input, a mixing desk often has -10db / +60db and will clip.  The channel faders often have gain too, into the mix, though less than the input gain.  The master rarely has gain and usually 0db is the max stop.

I know your way will work AudioGuru, but I would like to aim for sending already gained or attenuated channels into the mix amp.  To be more honest to the mixing desk idea I should have another pot after the input amps.  But it's overkill for the purpose of a desktop audio box.  Though I expect you believe the input gain/attenuation is overkill too.

I accept your point on using log pots.  I also intend to not get caught into audiophoolery but I hope to buy branded 1% resistors and decent caps and log pots.  This is why I'm also trying to limit the number of resistor and cap values.

I'm still unsure of whether I should go with the inverting summing amp, post input opamp or not.

Anyway, I'm going to try and clear a breadboard tonight and start testing these things with the scope and some waveforms.  I only have a cheap signal generator kit thing and a soundcard, but it will do.

The only hurdle is not having a +-12V or +-15V supply.  I'm wondering if I can series my two RD Tech PSUs.  Can't find anything with google, but they pass they are definately not earth referenced as one is running off +- DC and the other off a battery.

[BrianHG]

For an op-amp based mixer, you need nothing more than a wall-wart 15v/12v AC out adapter.  2 1N400x diodes, an optional series 1w resistor to soften the rectification current peaks, 6 caps and a 7815/12 and 7915/12 regulators for your power supply.  This project wont suck more than 200ma total.  The AC out wall wart guarantees transformer isolation from the mains.  No stinking Y capacitor means no looping ground hum and the low power means you pretty much will survive an accidental short-circuit.  That is unless for some reason you use an above 500ma wall-wart.  I wouldn't go above 1 amp.

Makes the device safe to the touch as well, with approved wall-wart of course.
Stay away from switching supplies here.  You just don't need em.

[paulca]

I know I can do that and probably will for the breadboard.  I still think I can series my PSUs and take the mid point as ground to get true +/-15V.

I want to avoid a virtual ground in the real thing.  It tends to cause issues when connecting to other equipment, especially hot plugging.  Suddenly you are raising a piece of equipment ground by 6V, they tend to get upset, especially 5V devices which I found brown out and crash.  Not a bit deal for headphones though, but connecting it to an Arduino to display an LED level meter... brown out crash.  A bigger coupling cap would raise the ground slower, but maybe not slow enough. 

I also crashed the USB Host in the PC doing similar.  It did not like it's ground being ramped up to 6V at all.  "dmesg" output showed the OS considered the host as "dying" and disabled it.  I had to reboot to get it back after 20 minutes of trying to send the reset code to the device /sys

[BrianHG]

I want to avoid a virtual ground in the real thing.

[paulca]

Oh, hang on, I think I follow you now.  You mean to open the AC/DC wall wart and tap the transformer to get 15VAC then filter it to get +/- 15V DC?  EDIT: or more likely +- 6V or 9V or something similar after caps filters the half sines.

I actually bought a 12V wall wart off RS destined for that experiment exactly.

[Audioguru]

When you have too many level adjust pots then one or two of them always end up in the wrong position.

Everything you said about gain and noise applies to a microphone mixer that you are not making. A line level mixer does not need gain here, there and everywhere and makes inaudible noise.

[paulca]

When you have too many level adjust pots then one or two of them always end up in the wrong position.

Everything you said about gain and noise applies to a microphone mixer that you are not making. A line level mixer does not need gain here, there and everywhere and makes inaudible noise.

So every audio mixer is wrong.  Every sound engineer friend I have is wrong. 

Set the master fader to the bottom of the sweet spot region (e.g. -5 dB).
Set the channel faders for each input in the sweet spot, usually at the 0 dB point.
Unmute the inputs and dial up a rough mix using the preamp input gains as if they were faders. Let the preamp gains fall where they may.
Now, forget about the preamp gain knobs and focus on mixing the show using the faders. All the faders are now near zero dB for maximum mixing finesse.

http://www.fohonline.com/current-issue/74-tech-feature/10445-back-to-basics-gain-structure.html

I can find hundreds of references if you want.  Gain is added at the input.  Always is.

With your approach, if one input changes, all inputs need to change to balance the output.  This might be fine in my case where I might have one or two inputs active at a time, but if you have a 24 channel desk and one line feed from a keyboard is low it would be completely ludicrous to alter the mix of the other 23 channels to rebalance the mix.

[paulca]

The other consideration is noise.

Every gain stage amplifies noise.  If you attenuate a signal you do not lower the inherent SNR of the equipment.  Attenuate it enough you will lower it into the noise.  When you then amplify that signal you amplify the noise as well, therefore you have just cut your SNR dramatically.

So the approach of attenuate then amplify is backwards and ruins you signal to noise ratio.  You want to lift your signals away from the noise ONCE then attenuate them to the desired level.  Running a desk with the faders above 0db suggests a bad mix, for that reason.

This master mix output is not gained again until it goes to the power amps or the recording equipment.   In a recording studio the line and mic inputs will be gained only once.  (excluding guitar pre-amps, amp heads, active mics, electronic instrument line level output amps).

[Audioguru]

With your approach, if one input changes, all inputs need to change to balance the output.  This might be fine in my case where I might have one or two inputs active at a time, but if you have a 24 channel desk and one line feed from a keyboard is low it would be completely ludicrous to alter the mix of the other 23 channels to rebalance the mix.

Nope. When the mix opamp has a gain of 10 times then most input level controls are set to halfway on their logarithmic level controls. If one input is low then simply turn up its input level control without touching the other input level controls.
The boost of up to 10 times is 20dB which is plenty.

Audio opamps are low noise and are used with a gain of 200 times or more for microphones. With a gain of only 10 times in the mix opamp then its output noise level is 20 times less (-26dB). Inaudible.

[dmills]

Dangerous Audioguru...

In a virtual earth summing stage the noise gain is proportional to the number of inputs, so while not a problem with a small number of inputs the summing amp on a large desk can very easily become  a noise problem, to the point that actually breaking the desk down into a tree of summing stages can actually result in lower total system noise.

I would note that with even halfway respectable opamps (Which the 5532 easily are), operating in a suitable impedance environment, the noise from the RESISTORS usually exceeds the noise from the opamps.

I highly recommend that the OP obtain a copy of "Small signal audio design" by Douglas Self for more then they ever wanted to know about how to design audio doings.

Regards, Dan.

[Audioguru]

This simple mixer has only 3 inputs, not 24.

[BrianHG]

I think that paulca's design, though not necessarily some of our chosen topology/techniques for a 3 channel line level mixer, with some of the changes already made above, like the volume at the RCA inputs & most of everything else remaining the same will function excellently for his intended use.  The ne5532, even with some gain, will have such a quiet noise floor that he wont hear a thing so long as the PCB is well laid out.

As for the power supply, with an AC out wallwart.  Just use power wire A direct to GND, then the other goes through the optional resistor in series, to the 2 diodes (both tied at that point, just 1 forward, 1 backward).  Add your power filter caps there, I like something like 10000uf 25v each, or higher. (This is already overkill, but I like my audio filter caps huge and you are doing half phase rectification if you power supply is done like this...)  For a 12v AC adapter, you should get around +16v on the forward diode and -16v on the reverse diode.  Put those 2 voltages through your 7812 and 7912 and you will have a clean +/- 12v.

[paulca]

Thanks.  I don't have a negative regulator to hand

As per my other thread I have two of my bench supplies in series, but I'm too scared to lift them above 1V.  I don't what they to go bang.

One is powered off a battery, the other off a LED Lighting DC supply.

[paulca]

On the amp design.  I am waiting bread boarding a few things before getting back into the design.

I hope to start with one input amp.  Test it with a scope.  Then try a second.  Then try a mix amp.

I'm sure at that stage I will have messed something up!

[BrianHG]

if you have 1w or 5w approx 12v zener diodes, you can make a second R/C filter stage at your power input caps and parallel the zeners in parallel to limit voltage to 12v on each cap.  That will suffice.

[paulca]

So tonight's mission was successful.

The input amp worked (mono).  Gain looked good.  Response, for as far as I could tell was grand on sine and "fair" on square.  I have waveform caps I can share later.  I tested at 1Khz, 20Khz and for some fun all the way up to 1.2Mhz.  Things go ropey above about 600Khz, but it wasn't that bad.

Square wave at 20Khz has quite a slow rise/fall time and a little overshoot, but I don't think it's that big an issue.

It was also successful in that the two RD Tech power supplies did a grand job of providing +12 0 -12V rails.

Also successful in revealing just how important decoupling caps on the power rails is.  I had huge noise ripples all over the place without them.

Also successful in confirming the cheap chinese kit build waveform generator is crap.  After decoupling the noise off the rails, I was depressed at the noise on the output until I hit "HOLD" a few times on the wave and confirmed 100% the noise was on the input and not being added by the amp.  Little oscillation pulses at a much higher frequency flowing long the waveform.  They existed on both input wave and output wave, so I ignored them.  (Signal gen was running on a completely separate battery, so clean power, though it did share the ground from the PSUs).

Finally, "successful" in finding out my laptop power supply has no damn discharge what-you-ma-call-ems on the main caps.  Bastard thing zapped me after I unplugged it and it hurt.  I do make an effort not to touch the pins when I pull a plug as I have been bitten before, but it got me.  Grrr.  Nothing like a little 120/240V zap to piss you off.  It has a proper UK plug with part insulation, so there is no way to touch the pins while they are connected to the mains, so it was definately the caps on the mains side discharging into my finger

[paulca]

if you have 1w or 5w approx 12v zener diodes, you can make a second R/C filter stage at your power input caps and parallel the zeners in parallel to limit voltage to 12v on each cap.  That will suffice.

Not sure I follow, but... oh, voltage regulation.  It's fine the bench supplies stepped up to the task.

My finger hurts  and it twitches still

[paulca]

So I had another session with the prototype.  Learnt a few more things.

My PC is foul for noise.  There is noise between 300mV and 1V peak to peak up around 40KHz though the scope was having trouble with it.  It's most likely the switch mode power supply or some other EMI generated in there.  I spend about 30 minutes ruling things out until I had to insight to connect my phone to the input.  Wham, no noise.

The PC noise was so intense it was impossible to determine what the amps where actually doing.

That sorted out, using the mobile phone as a source, with 2 of the preamp circuits above and an inverting sum amp it was working.  I was able to mix two channel together.

The gain structure is a bit off though.  What is the natural gain of a inverting amp?  I looked up the basics and it said if R1 = ... = Rn = Rf I would have a gain of 1.  But while I had a gain of 3 on the preamp the gain after the inverting amp was only about 30%.

I also had PSU noise to deal with, an oscillating ripple pulse appearing around 500Hz with the oscillation around 20Khz.  The waveform looked like a bass drum does, if that makes sense.  I expect it's the PSU's output mosfet ringing. 

Filters...

On the HF PC input noise.  I know it won't bother me as it's way outside of human hearing range, but is it worth trying to put a low pass filter in to kill anything over about 22Khz?

On the PSU noise, I had two switch mode bench supplies in series, probably not the best clean audio supply, but even with a 100uF and 100nF across both rails AND 100uF and 100nF across each rail to ground, I could not get rid of it.  Assuming the PSU I get for the project (probably an isolated encapulated PSU) has similar noise, how do I kill it? 

[Audioguru]

Since your pc generates so much noise then it might cause the opamps to amplify it and produce clipping distortion of the audio. Filter out the ultrasonic noise.
The gain of an inverting opamp is Rf/Ri. If the two resistors have the same value then the gain is exactly 1 regardless of how many input resistors there are. The opamp must be able to drive the value of Rf and most opamps have trouble driving less than 2k ohms including the output load. A non-inverting opamp is a passive mixer where the inputs affect each other.

Why aren't you using logarithmic level controls? If a linear control is turned down from max to halfway then the signal level drops only a little, then drops more and more (like an on-off switch) as it is turned down to zero. A log pot drops the level properly and evenly.

[paulca]

Thanks.

I haven't decided on pots yet.  I expect I will end up going for some 1% Alps log pots from RS.  They may not play on the breadboard.

At the moment I am using stuff out of my component stock and as I only have 1 50K pot, I was using a 100K 10 turn trimmer on the second input channel.  It's a bit rough and ready.

Also it's on a breadboard... my feedback resistors are like noise antennas.  I can get all manor of pretty scope pictures by touching them.

That's why I'm not being too fussy.  I just want to prove a few concepts to myself before I consider actually building it properly on possibly strip board.

I expect I need a bit more practice and study on PCB layout to make the end device I want to build.  So I'm spending time doing digital circuits on PCB to learn before I move to an analogue board and all it's sensitivities.