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

[paulca]

I have learnt a lot about OpAmps, but I have a lot to learn still.

Be gentle, but just how bad is this first attempt at a schematic?

The idea I'm aiming for is individual line level pre-amps with a gain of 2, each attenuated by a pot before joining together and the output being buffered.  Effectively a 3 channel stereo mixer.

An additional buffer for the Aux output to prevent impedances affecting the power amp.

A parallel pair of op amps as a power (headphones) amplifier.

Thing's I'm not really sure about.
1.  I copied the power amp schematic.  Why are there 100Ks to ground on the + inputs?
2.  How do you choose the capacitor and resistor output values?  Is this to set the impedance seen from the next device along?
3.  Is this even remotely likely to work?
4.  Have I calculated my gain correctly?
5.  Do I need caps and resistors on the AuxOut?
6.  Same question for the inputs.


Note I know my connectors are not wired right yet.

I did attempt to test a few things in LTSpice but it keeps crashing on me today

[Audioguru]

Most of your opamps have no DC input bias voltage reference. Since you have a plus and minus 12V supply then 0V is usually the reference that must be applied with a resistor from 0V to the "+" input of the first opamps then they need an input coupling capacitor to block any DC from the signal source.

A mix opamp is usually inverting then adjusting one or more input levels does not affect the levels of the other inputs like your non-inverting mix circuit does. The input level controls in your circuit load down all the others so the output level from your mix opamps will be fairly low.

Why do you have input opamps? Feed the inputs into level control pots then they feed an inverting mix opamp that has a gain of 2.

You do not make a power amp by paralleling opamps. The 100k resistors to ground give the paralleled opamps their 0V DC input bias voltage reference that I was talking about.

A simple formula (one divided by pi x f x R) is used to calculate a coupling capacitor value.

Like most opamps, your output buffer opamps will probably oscillate when driving the capacitance of a shielded output cable. They need a 47 ohm to 100 ohms series output resistor to feed a cable.

[paulca]

Thanks Audioguru!

So it would just be the first op-amps that experience incoming signals that need to remove DC bias?  Once they have removed it the rest do not need 0V reference?

I was a bit suspicious of the mixdown with 3 stereo pots basically connected together.  At the same time I thought that putting them before the input opamps.  I figured that would be attenuate then amplify, which sounded backwards.  Though I note the headphone amp does just that.

How does an inverting amp not get pulled down with other resistance on the input?  Can you explain how that works?

I choose to have input opamps as an attempt to prevent different input impedance messing around with each other.  When I connect two audio sources to one output I invariably get one being very quiet while the other loud and things like that.  I thought putting a buffer (or preamp with a small gain) on each input would separate the impedances.  Besides every mixer circuit I looked at had input amps.

For the coupling caps, can I just carry on the trend of using 1uF and 47K's?

If I put 48KHz as the frequency and 47K as the resistor it gives me a 140pF. 
If I put 20Hz as the frequency and 47K as the resistor it gives me a 340nF. 
What do I do with that?  Aim high, middle, low?  Sample schematics online seem to just use 1uF.

[Audioguru]

I said that you do not need the input opamps. Feed the input signal to the grounded level control pots through series coupling capacitors then the sliders of the level control pots feed through series coupling capacitors through series resistors to the inverting input of a mix opamp that has its + input biased at Vdd/2. All opamps must have their + input biased by Vdd/2 or by the output of another opamp that has its + input biased at Vdd/2.

When a mix opamp is inverting then the inputs do not affect each other when you change the level  or impedance of them. Because the - input of an opamp that has negative feedback is a "virtual ground" with no measurable voltage signal. An input current through a series resistor to the - input simply causes the same current of the opposite polarity in the feedback resistor, not in the other input resistors. When the feedback resistor has current in it then the output voltage of the opamp changes according to Ohm's Law.

Most mixers have input opamps because they are microphone preamps.
1uF into a 47k load passes audio frequencies down to the earthquake frequency of 3.4Hz. Audio goes down to no lower than 20Hz so a 330nF film capacitor is fine for 10.3Hz since two coupling capacitor at 10.3Hz cause a cutoff of frequencies below 20.6Hz.

A coupling capacitor passes all frequencies above its calculated cutoff frequency. Then if it passes 20Hz it passes all audio frequencies and higher frequencies.
Why do you have an ultrasonic frequency of 48kHz that nobody can hear? Humans can hear no higher than 20kHz but when I was young I heard ultrasonic frequencies from burglar alarms that nobody else could hear. You need coupling capacitors to pass frequencies down to 20Hz then two RC couplings need to be calculated to pass 10Hz.

[hamster_nz]

It is only a minor point, but consider connecting the pots for the power amp to the outputs of the mix op-amps, the other side of the 47k resistors.

Otherwise changing the power amp volume settings will change the loading on "mix buffer" op-amps, reducing headphone volume (slightly) as power amp volume goes up.

With the power amp at min volume the headphone signal is between 47k and 50k, so about 50% of the mix amp output. With the power amp at max volume, it is between 47k and (100k in parallel with 50k, or 33k), so about 36% of the mix amp output.

You most likely wouldn't notice it in practice....

[paulca]

I said that you do not need the input opamps.

The other reason, of course is I want independent positive gain on the inputs.  If one input needs lifted above it's current level without input amps I would have to turn the others down and then raise the output volume.  That's just cheesy.

I have used 'real' mixing desks, line or mic they have input amps and gain on the input channels.

Why do you have an ultrasonic frequency of 48kHz that nobody can hear?

Sorry, was getting confused with the sample rate on the digital side 

You need coupling capacitors to pass frequencies down to 20Hz then two RC couplings need to be calculated to pass 10Hz.

The coupling caps on every amp?

Also, what are the two RCs for 10Hz for?

[BrianHG]

In keeping with your existing topology and op-amp count setup, to simultaneously get rid of the possible input DC bias and provide a visible termination load at the audio inputs, here is what I would change in your schematic:

Move and change the 1uf at the out of the op-amps to 10uf np to the audio input, then move the adjustment volume control right after that and feed the + input of each op-amp.  As for the outputs, just leave the 47k series resistors.  I would lower these to 10k just to lower impedance in the circuit.  This will not affect the op-amps ability to drive the audio output, but, lower the impedance to the next stage.

Now, there are other ways to mix all your inputs without cross-talk eliminating 4 op-amps, but, I would keep the input section as is for understanding and simplicity sake.

Next, remove the 47k series resistors at the output of your mix buffers.  You will also need to add gain on your mix buffers if you want the original sound levels of 1 single input source to 1 output.  I like to add a few db in my mixers, so I would make a gain of 2 over here too.

Once again, there are other ways to do this mix of all the source channels using inverted inputs, but, the way you set it up is very functional and good for basic understanding.

As for all the gain resistors, 100k is a little high.  I personally would use a lower impedance feedback, like 10k or even 1k.

[Audioguru]

Level controls are usually logarithmic so when they are set to halfway their output levels can be dropped to nothing or increased 10 times. Have the gain of the mix opamp set to 10 times or 20 times then set all the input level controls to halfway. Then you can turn one up or down a lot without affecting the others. No input opamps are needed.

The input to each level control needs an input coupling capacitor to block any DC that might come in from a signal source and it needs an output coupling capacitor to feed the input resistor of the mix inverting opamp. The Aux Out opamps can be removed because the output from the mix opamps do the same thing. The outputs of the Mix opamps need output coupling capacitors to block their DC. Then you have three RC lowpass filters made from the coupling capacitors.

[Audioguru]

You need low impedance feedback resistors for high frequency video circuits, not for audio.
Here is an inverting mixer that uses a single polarity supply:

[BrianHG]

Now, for the headphone amp.

I know that doubling the output current of a small op-amp may appear to be good to drive headphones, and, for cheap low volume and unpredictable current limiting points at higher volume with unpredictable headphone loads, ok.  There are op-amps better suited specifically for driving headphones, or, if you don't mind a little extra current on your power supply, I would use 1 of your same op-amp, driving a heavily biased 2N3904/2N3906 emitter follower configuration making a good strong output drive.

[BrianHG]

Question1:  Why are you swamping Left & Right between the pins of the stereo RCA input jacks and the RCA output jacks.

Question2: Why is pin 2 the GND on your headphone jack.  Most data sheets for these connectors has pin 1 as GND.

[Brumby]

Question1:  Why are you swamping Left & Right between the pins of the stereo RCA input jacks and the RCA output jacks.

Question2: Why is pin 2 the GND on your headphone jack.  Most data sheets for these connectors has pin 1 as GND.

Note I know my connectors are not wired right yet.

[paulca]

Question1:  Why are you swamping Left & Right between the pins of the stereo RCA input jacks and the RCA output jacks.

Question2: Why is pin 2 the GND on your headphone jack.  Most data sheets for these connectors has pin 1 as GND.

Note I know my connectors are not wired right yet.

The plan is to find, import or create a proper symbol for a stereo jack and possible an RCA Phono gang.

[paulca]

Now, for the headphone amp.

I know that doubling the output current of a small op-amp may appear to be good to drive headphones, and, for cheap low volume and unpredictable current limiting points at higher volume with unpredictable headphone loads, ok.  There are op-amps better suited specifically for driving headphones, or, if you don't mind a little extra current on your power supply, I would use 1 of your same op-amp, driving a heavily biased 2N3904/2N3906 emitter follower configuration making a good strong output drive.

I know I could use a push pull set of mosfets. 

Here's the thing.  I have that headphone amp already built, it came as a kit and it drives my headphones perfectly.  The only minor annoyance is that if the signal source is low volume, a quiet track or quiet YouTube video I just don't have the gain.  This lead to the idea of adding an adjustable pre-amp to the input.

I know AudioGuru is trying to steer me away from it, but I'm still getting dragged back to pro audio mixers and practices. 

1.  Set your input gain on the channel for 0db normalisation.
2.  Set your filters, EQ, balance (if these are after gain, repeat 1)
3.  Repeat for all your input channels.
4.  Set your send levels.
5.  Set your return levels.
6.  Set your mix levels.
7.  Set your master level.

To do that right, and it is tempting, but possibly OTT, I would have a 0db SIG and 12db CLIP LEDs on the inputs and have pots before and after the input pre-amps.  The LEDs would be after the pre-amp gain.  So I can normalise the channels to approximately 0db (relative to a reference I assume).

I keep hearing "Gain structure" when we consider just putting unnormalised signals into a mix amp, boosting it and then attenuating it down.  It just feels wrong.

I know it will "work", I'm not arguing that you are wrong, but it still feels wrong.  I don't mind making it over complicated either, a lot of this project is about learning stuff.  I could just go out and buy a desktop amp block with 3 inputs and 2 outputs and be done.  Very few will have independent input gain though.

Ill try and update the schematic tonight or tomorrow if I get time.

Thanks again, everyone for helping   I'm learning.

[IanMacdonald]

As mentioned a gain of 2 (6dB) will be only just noticeable on audio, but not much use for boosting low signals. 20dB (10x gain) is probably a more sensible figure to aim for.

The way this works is that 10dB is ten times the sound power, and an apparent doubling of loudness. However, since power is proportional to the square of voltage, that's just over 3x the signal voltage.

The minimum loudness change the human ear can reliably detect is 3dB, which is roughly a doubling of sound power.  Sounds surprising, but it's true.

Doubling the voltage thus gives about twice the minimum change the human ear can detect. it's still not much though.

2x: 6dB (approx)
10X: 20dB (exact)
100x: 40dB
1000x: 60dB

A mic input needs about 60dB gain to cope with relatively quiet talkers. Though I've heard some opera singers claim that they can produce a volt from a microhone.  (Put your glassware away whilst testing this)

[paulca]

Thanks Ian.  So a little more gain on the input amps.

I was searching around and getting annoyed with the usual "All about circuits" type random audio circuits and found this:

https://carvinaudio.com/pages/mixer-schematics

And a single mono input channel.

http://carvinimages.com/schematics/DX%20Series%20Channel.pdf

Seems to be the actual schematic for a single input channel on a mixing desk.  It's a bit hard to calculate the gain as they are doing funky stuff with the XLR input and the Line, though the Line passes through a voltage divider and the XLR doesn't R9/R10.

A lot of it is beyond me, but I can identify stages and components.... mostly because they are labelled LOL.

Another interesting thing is that the variable gains on the filters at least seem to break a supposed rule of not using feedback as gain.  I remember in High school the first opamp circuit I seen had a pot on the feedback loop.  It has always seemed a bit odd to me that people either boost then attenuate or vice versa, instead of just altering the gain of the amp directly.  I'm sure there are reasons, but this channel circuit uses variably amp gain.

I also count a whole 4 dual op amps in the input stage for mono.  To replicate that for a stereo channel that would be 8 dual op amps and they haven't even mixed it down yet.  Granted most of the opamps seem to be to do with active gain filters, LOW, HIGH and a sweep-able MID (not sure f it's fully parametric with Q or just sweeped).

I don't need anything this complicated.  It would be nice to have Low,Mid,High but maybe next time.   I just thought I'd add it to see what real mixing schematics look like. 

[dmills]

The auxout buffer should have 47 ohm or so resistors between the opamp output and the socket to reduce the chance of oscillation with a capacitive load.

100k is a bit on the large side for the 5532 (Which is a bipolar opamp with relatively low noise impedance), 10k or even 2k2 would be more like it, also lower value resistors are generally quieter due to lower intrinsic thermal noise.

If I was doing this, I would dump the input amps and do the gain in the summing amplifiers at the moment you have a per input gain to the output of Av = 2/3 so you actually attenuate.
Make the summing stage inverting and give it some gain, make the pots maybe 5 or 10k (To give yourself space to have the summing input resistors be large enough not to disturb the taper too much) and invert a second time at the output buffers.


Where is the decoupling?

State variable filters (Which are what that mixer input strip uses) consume about 4 opamps per band, so swept mids are expensive on parts count.

Regards, Dan.

[BrianHG]

Don't go too crazy on the gain.  You'll begin to add hiss.  Also, having gain on you mixer stage means the primary input opamps wont have to go all the way to your rails to provide the same 2vrms output since their output is divided by 3 in the mixing.  Unless you go to Audioguru's negative scheme at the mix stage (when done right, it's used quite often in mixers), stick with my recommendation of having a gain of 2, or 4 if you want more at the input amps and the same gain again at your mixing middle amps (unless you choose 3x deliberately to compensate exactly for the mix) & don't forget to shrink those overly large sized 100k feedback & GND gain resistors.

Also remember, the higher you make the gain on your op-amp, the more lousy the high frequency response.  Another good reason to divide the gain between the input amps and you mid mix amp.

[paulca]

If I was doing this, I would dump the input amps and do the gain in the summing amplifiers at the moment you have a per input gain to the output of Av = 2/3 so you actually attenuate.

Av=2/3... how?  A 1:2 divider on the feedback side should equal a gain of 2.

Anyway, as stated, while it may be more complicated than it needs to be, I feel I want normalised inputs.  I have played the gain structure game before and hated it.  To turn one channel up you have to turn the output gain up and then turn down the other channels.

It's all very well when you connect 2 or 3 channels of equal or nearly equal output together, but when you get one channel that is 1/10th the power of the others it gets really irritating pulling all the other channels down to that level and then amplifying all channels along with all channels noise.

As a random example, I was watching a YouTube video on OpAmps and the presenter was quite interesting, but he was also mumbling and the level on the video was barely audible.  I had YouTube at max, PCM at max, master at max and the headphone amp at max and if I had decided to play music at the same time I would have had to wind the music input right back and there still wouldn't have been enough gain on the output amp, if there was it would mostly be amplifying the noise.

Granted this is a for a desktop headphone amp to combine all my sources and outputs into one box, so, no I probably don't 'need' input pre-amps, but I have my mind set.

[BrianHG]

If I was doing this, I would dump the input amps and do the gain in the summing amplifiers at the moment you have a per input gain to the output of Av = 2/3 so you actually attenuate.

Av=2/3... how?  A 1:2 divider on the feedback side should equal a gain of 2.

Anyway, as stated, while it may be more complicated than it needs to be, I feel I want normalised inputs.  I have played the gain structure game before and hated it.  To turn one channel up you have to turn the output gain up and then turn down the other channels.

dmills may be speaking as after you make my recommended changes a few posts up here: https://www.eevblog.com/forum/beginners/just-how-bad-is-it-audio-mixer-with-headphone-amp/msg1424158/#msg1424158
Which solves this problem, as well as a bunch of others...

Same goes for all my subsequent posts, I expect you have updated you schematic to my recommendation.

[paulca]

Don't go too crazy on the gain.  You'll begin to add hiss.  Also, having gain on you mixer stage means the primary input opamps wont have to go all the way to your rails to provide the same 2vrms output since their output is divided by 3 in the mixing.  Unless you go to Audioguru's negative scheme at the mix stage (when done right, it's used quite often in mixers), stick with my recommendation of having a gain of 2, or 4 if you want more at the input amps and the same gain again at your mixing middle amps (unless you choose 3x deliberately to compensate exactly for the mix) & don't forget to shrink those overly large sized 100k feedback & GND gain resistors.

Also remember, the higher you make the gain on your op-amp, the more lousy the high frequency response.  Another good reason to divide the gain between the input amps and you mid mix amp.

Thanks.  The plan is to maybe up the gain on the input amps a little.  I am still tempted to use feedback gain control there.

Hiss and over-gaining is why I want to avoid this pattern being suggested of attenuating the inputs then boosting them up at mix amp only to attenuate them back down again at the headphone amp.  I want a small amount of normalisation gain on the inputs, then use AudioGuru's inverting mixing amp with unity gain. 

In normal operating conditions with good inputs the input amps should be running in unity.  Another reason I'm considering adding variable gain feedback instead of amplify then attenuate.

Finally enough gain on the output amps to make my headphones clip....  I'll add a third 5532 is I need to   Or a quad of mosfets if I have to.

Ideally I would like the input level pots to be unity gain at centre. x2 or x3 fully clockwise and 0 fully anti-clockwise.  That might be tricky though but a log pot of the right size might work.  I have to review these gain figures in light of the DB gain calculations some placed above.

The mixing circuits seem to all have variable gain on the inputs and then mostly attenuate across the whole desk (except filters which allow positive and negative gain).  Having your channel faders about 0db is a mix smell.  Master faders usually don't even go about 0db.

[paulca]

Same goes for all my subsequent posts, I expect you have updated you schematic to my recommendation.

I'll try and make the updates tonight to clear up confusion.

What are the pros-cons of variable feedback?  I suppose finding the correct pot/resistors might get expensive?

[BrianHG]

Do not use variable feedback.  The opamp feedback and gain resistor on the negative input is the most sensitive part of your circuit.  It responds to any RFI with ease.  Only use a fixed gain with 2 resistors as close as possible between the resistor and the - input of the opamp.

My directions were to place your series 10uf caps and volumes right at the RCA audio inputs.  This simultaneously protects your circuit from an audio source with a DC offset and makes a constant load on the RCA inputs matching the full series resistance, use 50kohm as this is standard, of the volume control.  The wiper goes to the + input of the opamp.  The opamp should have a fixed gain of 2 through 4 depending how much boost you want.

The opamp outputs of each channel should each have a series resistor feeding the center mix opamp which you can set a fixed gain of 3x to normalize the audio.

Doing things Audioguru's way means you need to invert the audio at each stage, and you still need now to further strategically place the mix resistors at the middle mixer stage right at that op-amp's - input to prevent noise.

[Audioguru]

The Carvin mixer uses three opamps at the input to make an instrumentation amplifier instead of buying an IC that does it (they were not available years ago). The input of an instrumentation amplifier is balanced (microphones are balanced) and for lots of gain for a microphone. Your inputs are unbualanced line levels so you do not need the three input opamps for easch of your inputs. The Carvin circuit does have an input gain control that you need for each of your inputs.

The Carvin circuit has normal bass, mid and treble tone controls but does not have the high frequency cut filter that is in your circuit. All three tone controls and the Fader use variable feedback to adjust the gain.

You said that you have a headphones amplifier that you made from a kit but it does not have enough gain. It is simple to replace the feedback resistor with a higher value resistor for more gain.

EDIT: Here is an instrumentation amplifier:

[paulca]

You said that you have a headphones amplifier that you made from a kit but it does not have enough gain. It is simple to replace the feedback resistor with a higher value resistor for more gain.

Yep.  I am borrowing the schematic for it into my design.  I believe the gain calculates as 3.13 but I could be wrong.  Opamps always surprise me.  It is the lower right corner of the schematic.

I can adjust the feedback resistor, but I need to check that I don't bump into the current limit of the parallel pair.  Of course, using 3 is an option or drop the parallel and use an output driver circuit.  I have to bear in mind that my headphones are fairly high impedance, around 34Ohms I believe, but if I connect a lower impedance set to the amps in the future I might bump into the current limiter of the 5532.

On the gain, it currently only has one input and with a good quality input I have enough gain, just about, it won't clip the headphones, but it comes close, distortion becomes present at full level, but barely noticable.  It is when there is a quiet source I don't have the gain.  In my experience in audio, yet not electronics, the place to add this gain is at the input, not the output and that becomes more important when there are multiple inputs.

EDIT:  Also if you note in the Carvin schematic both XLR and Line are amplified at the input.  Mixing desks have a "Gain" knob at the top of each channel.  It has 0db at the 12 o'clock position.  Anti CW attenuates, CW amplifies.  This is before the filters, channel fader etc.  They often have a +48V phantom voltage switch and a line level db selector..