Balanced Stereo to Balanced Mono (Op Amp)

[jim_griff]

Hi everyone.

So, it's a long time since I've messed with op amps. I have been trying to create a balanced stereo to mono converter for use inside my passive volume control - to send a separate feed to the subwoofer.

I just wanted to check if I've made any serious mistakes. The circuit isn't finished yet and still needs things like input/output capacitors to isolate DC as well as some caps in the op amp gain to prevent DC bias.


The main thing is, am I right in working out the gain at each stage? My maths isn't the best!

1. The first stage (back-to-back op amps) is a strange one since it doesn't have the usual ground resistor to set the gain. I've put the formula for that one up, but is it correct or have I mixed it up with another type of op amp circuit?

2. The second stage I'm not sure if the gain is correct (both + and both - are adding together after the 1k resistors and terminating to ground through 100 ohms). Do I divide the 1k resistors by two (to get 500 ohm) or times by two (to get 2k) then work out the classic resistor divider formula? If a mono signal is put in, both previous op amps will be providing current through the 1k resistors, so I'm guessing it would be 500 ohm not 2k.

Please feel free to tell me that I'm a total idiot if I've really messed up the circuit somehow.

EDIT: Ignore the 750p caps on the outputs! Forgot to rename them to larger values.

Cheers!

[Zero999]

1) The formula looks right to me but I'd have to go and look it up to be certain.

2) Assume one op-amp output is zero and the other is at 1V. Without the 100R resistor, the output will be half the input, because you have a potential divider with two 1k resistors. Now add ther 100R resistor and I hope you'll see, it's effectively in parallel with the 1k resistor.

That 220µF capacitor has no DC bias, so will need to be non-polarised, which is inconvenient for such a large capacitor. You could use two 470µF aluminium capacitors back-to-back, I suppose.

The circuit isn't a true differential amplifier, because it doesn't reject the common mode, which appears on both outputs.

Your circuit doesn't have any gain. What's the output impedance? If it's high, then you might not need any op-amps at all. It could be done with passive components, especially if the input impedance can be lowered.

[jim_griff]

1) The formula looks right to me but I'd have to go and look it up to be certain.

2) Assume one op-amp output is zero and the other is at 1V. Without the 100R resistor, the output will be half the input, because you have a potential divider with two 1k resistors. Now add ther 100R resistor and I hope you'll see, it's effectively in parallel with the 1k resistor.

That 220µF capacitor has no DC bias, so will need to be non-polarised, which is inconvenient for such a large capacitor. You could use two 470µF aluminium capacitors back-to-back, I suppose.

The circuit isn't a true differential amplifier, because it doesn't reject the common mode, which appears on both outputs.

Your circuit doesn't have any gain. What's the output impedance? If it's high, then you might not need any op-amps at all. It could be done with passive components, especially if the input impedance can be lowered.

Thanks for your reply.

On point (2), I see what you mean about the 100 ohm resistor being effectively in parallel with the other 1k. So the signal panned full L or full R will be slightly less than half. To be honest, I think that'll be fine as the subwoofer is mainly there to amplify the Mid.

I may not need the next op amps as you say - it could be mixed passively. The input impedance to my subwoofer is probably above 100k (it's very voltage noisy compared to the studio monitors which are 20k input impedance). The output impedance of the NE5532 is 600 ohms, I believe, so plenty enough to drive the input to my subwoofer.

About the 200uF cap - Which way back-to-back would they be? I'm guessing negative middle: +||-  -||+

Which part won't reject common mode signals? I see the second part looks like it may not reject common mode signals. Didn't think about that. But then it will be in a shielded case and the input signals are differential, so it should be okay for what I need. I may just remove the second stage and make it all passive and use an H-bridge to attenuate the signal to whatever level I need instead.

Cheers!

[Yansi]

Your schematic has poorly filtered inputs. You forgot that the common mode component of the input signal has to be filtered too.  And it should be filtered with even more care than the differential mode!

I'd suggest summing using a summing amplifier (inverting). And using least amount of in circuit gain possible. Otherwise you are limiting the signal headroom, which is quite important feature for professional audio applications.

Do you need balanced output at all?

Output impedance of 5534 is NOT 600ohm. It is almost exactly zero at the AF frequencies.  You need to add a series resistor to NOT and never load the opamp capacitively (cable capacitance may and will make it unstable if enough of it present). Those output 750p caps are what? 750pF? Wtf?


Doesn't matter which way the caps will be back to back.  You could use probably a specifically bipolar electrolytic cap instead.
Even a bog standard polarized one would likely suffice in that place, assuming reasonably small DC offsets present.

[BrianHG]

Before I say more, looking in the past at Analog Devices reference designs for balanced output audio DACs and just googling now for balanced XLR drivers, most op-amp based XLR balanced line drivers have either a 100 ohm series resistor if not a 301 ohm series resistors on the outputs of the op-amps.  Isn't this supposed to be there to prevent the op-amp output from ringing or oscillating due to unpredictable cable length driving and also offer some degree of protection against short?

Q#1, Are you married to the NE5532?  If not, can you work with a single balanced in and out op-amp?
Q#2, Have you considered making the internal mix to mono, after deferentially buffering it, unbalanced?

[Zero999]

The output impedance of the NE5532 is much lower than 600R. When open loop (no negative feedback) it's 15R, but when it's closed loop, it's much lower. The data sheet specifies 0.3R, with a gain of 30dB, a 600R load and a frequency of 10kHz. The output impedance will be lower, at lower gains and frequencies, as there will be more negative feedback.

Why not use an instrumentation amplifier configuration, but with two output stages connected opposite to one another? Changing R3 will alter the gain of the entire circuit.
https://en.wikipedia.org/wiki/Instrumentation_amplifier

The attached will have double the gain of formula on the schematic shown on Wikipedia. The red plot is the common mode input voltage of 1V 2kHz, which we want to get rid of. The green is the voltage across the output terminals and the blue the common mode output voltage, which is very low.

[BrianHG]

I'm curious as to why you cant do it as my attached schematic?  It saves on op-amps and complexity...
Note that you can change the 600r to 100k for higher input impedance...

[Yansi]

Zero common mode suppression, huh?

[BrianHG]

That's one way to word it.  The inputs should appear as a 600 ohm connection to GND, though, not to each other.  But, differential error between the +&- of each input will be passed to the output.  I wonder if we can modify this so that the input load appears from + to - input without adding op-amps.

Also, R5&R6 in my schematic should be cut in half, changed to 300 ohm, if you want a mono output to be equal to a mono input.

[Yansi]

No, line inputs should NOT be 600ohm. They should be high impedance. 10-50k typical for line inputs,  2kohm typical for MIC inputs.

[BrianHG]

Ok, change R1/R2/R3/R4 in my schematic to 50k and change R5/R6 to 25k.

[Zero999]

I'm curious as to why you cant do it as my attached schematic?  It saves on op-amps and complexity...
Note that you can change the 600r to 100k for higher input impedance...

That would work, but I think he wants to be able to adjust the gain, with one variable resistor.

I forgot that he requires a mixer in my previous post, which is just a balanced differential amplifier.

This would do the job, but it's a heck of a lot more complicated.

EDIT:
It needs two variable resistors.

[Yansi]

I think you do not need to adjust any gain. Just use standard 1% resistor. These are precise enough for summing L+R channels for a subwoofer.

[Zero999]

The previous design, but with the inputs passively mixed, would achieve the desired result. It now has a gain of 1.

[Yansi]

Passive unbuffered mixing yields kinda low input channel separation and in this case even high noise. You should not use resistor more than a couple of kohms in the passive mixer networks when possible.

But all can be of course calculated beforehand, including the noise here. I did not run the numbers for this circuit, but I would strongly recommend against using passive mixing with a 100kohm resistors.

There's a nice book abotu such audio enginerding from Douglas Self, Small Signal Audio Design. Highly recommend that.

[Zero999]

Passive unbuffered mixing yields kinda low input channel separation and in this case even high noise. You should not use resistor more than a couple of kohms in the passive mixer networks when possible.

But all can be of course calculated beforehand, including the noise here. I did not run the numbers for this circuit, but I would strongly recommend against using passive mixing with a 100kohm resistors.

There's a nice book abotu such audio enginerding from Douglas Self, Small Signal Audio Design. Highly recommend that.

I agree. Passive mixing is messy and I certainly wouldn't recommend this scheme with low level signals, such as a microphone. Howver in this case it's a high level signal, so I don't see the issue.  By all means lower the input resistors to 10k, but it should be fine as is.

The alternative would be to use a differential amplifier with the inputs mixed, followed by an attenuator, before a single ended to differential converter.

EDIT:
Wait a miniute, that's easier. VR1 (a potentiometer simulated as two resistors) alters the output level.

[jim_griff]

Wow, thanks for the replies!

Yansi, yes the 750p caps were a mistake in the original schematic. I forgot to change the values.

BrianHG, I'm not married to the 5532 haha! I was going to prototype it with those since I have a few on hand

Your schematic has poorly filtered inputs. You forgot that the common mode component of the input signal has to be filtered too.  And it should be filtered with even more care than the differential mode!

Yeah I was going to get around to the filtering aspect at the end. I usually leave it till the last since the main bit is the amplification/mixing. Of course, it has no common mode rejection in the last op amps which is a giant mistake.

Output impedance of 5534 is NOT 600ohm. It is almost exactly zero at the AF frequencies.  You need to add a series resistor to NOT and never load the opamp capacitively (cable capacitance may and will make it unstable if enough of it present).

Doh! I knew I'd missed some resistors. Thanks.

I think you do not need to adjust any gain. Just use standard 1% resistor. These are precise enough for summing L+R channels for a subwoofer.

The gain for my circuit is a requirement because the subwoofer gain only goes from -6dB to +6dB on its input, referenced to its own amplifier inside. The studio monitors have an input swing between -20dB and +6dB referenced to their amplifier. So it's required to have the ability to set the gain between x0.5 (-6.02dB) and x2 (+6.02dB) at least, so I have extra room to play with.

There's a nice book about such audio enginerding from Douglas Self, Small Signal Audio Design. Highly recommend that.

I'll check it out, thanks. I've heard his name mentioned a couple of years back but will definitely give him a read.

EDIT:
Wait a miniute, that's easier. VR1 (a potentiometer simulated as two resistors) alters the output level.

Ahh, the output of the op amp goes into the inverted input of the other. I did that with an RIAA preamp I built a few years back to simulate balanced output. My brain must've forgotten! It is a long time since I've done op amp stuff and was never an expert to start with. Thanks for helping out with the schematic.


One question I have is: Do I specifically need capacitors in the gain stages of each op amp to prevent DC bias building up slowly? I've tried many circuits before and some op amps seem to stop working after a while as a DC voltage builds up on the inputs/outputs. Are some op amps just "better" than others in this respect?

Thanks for all your help guys. Really appreciate it! Top community.

[Yansi]

EDIT:
Wait a miniute, that's easier. VR1 (a potentiometer simulated as two resistors) alters the output level.

I think that's not gonna work correctly either. You are getting a pisspoor input isolation, but only on one of the input leads of the pair now.

Not to mention the inputs are no more balanced. One wire sees the 10k load, the other sees like 15k or so.

[jim_griff]

Here's a rough idea (attached). Please don't laugh!

Q1. I'm looking at OpAmp 3 (top right) and seeing the 10k potentiometer pulling -(cold) to ground, while the +(hot) is going directly into the non-inverting input. Is this likely to cause an imbalance in the inputs to the op amp? I'm unsure how to set the gain with both non-inverting and inverting inputs being used. Will I need to load the non-inverting input with the same resistor and capacitor values to ground?

Q2. Is it acceptable to use another op amp to "fake" the -(cold) signal for the output? I'd imagine slew rate, common mode noise and phase shift between the +(hot) and -(cold) shouldn't be a big issue for frequencies below 40Hz.

EDIT: Q3. It is okay to have 1k output resistors for each op amp stage while using 10k gain resistors? After looking at it for a bit, I can imagine one op amp will pull the other up or down depending on the voltage differential between the two. Although, the output impedance being low enough on the op amps, it should be able to cope. I think...

EDIT 2: BTW, The gain values at the top of the sheet are the total accumulated gain at that particular part of the circuit.

EDIT 3: I'm going to search for a differential op amp which is specifically designed for driving diff outputs rather than "faking" it the way I've done in my circuit. I think the input section will work properly (perhaps needs a capacitor on the gain resistor to prevent DC buildup).

Cheers!

[BrianHG]

Nothing to laugh about your circuit.

Note that my circuit, which only uses 1 single dual NE5532, the load is seen at each input as a resistor to GND, there is no cross mix since the - input of the opamp is always matching the + which is set to GND.  There is no crossing signal between the left and right input unless you remove the opamp.  In other words, if you want to test, make my design and only plug in the left channel.  Scope the right channel inputs and you will not see the left channel audio.  It should stay at 0v.  ***When routing this design, make sure R1/2 & R3/4 is as close to the - input of the opamp as possible.

If you want additional 6db gain, just match the feedback resistors R5&R6 with the load resistors R1/2/3/4.  For 10db gain, double the k of R5/R6.

[Zero999]

 jim_griff,
The circuit looks reasonable to me but I've not got much time at the moment to check it, at the moment.

Yansi,
I can assure you it does work. I've built an input stage like that before. It's a standard differential amplifier, but with two lots of input resistors and gives reasonable performance.

[jim_griff]

Nothing to laugh about your circuit.

Note that my circuit, which only uses 1 single dual NE5532, the load is seen at each input as a resistor to GND, there is no cross mix since the - input of the opamp is always matching the + which is set to GND.  There is no crossing signal between the left and right input unless you remove the opamp.  In other words, if you want to test, make my design and only plug in the left channel.  Scope the right channel inputs and you will not see the left channel audio.  It should stay at 0v.  ***When routing this design, make sure R1/2 & R3/4 is as close to the - input of the opamp as possible.

If you want additional 6db gain, just match the feedback resistors R5&R6 with the load resistors R1/2/3/4.  For 10db gain, double the k of R5/R6.

Thanks, I'll give your circuit a try in a bit. My circuit seems to have a problem (or perhaps a behaviour I wasn't expecting).



Problem with the input stage (attached).

Basically, I attached the output from my audio interface (through bipolar caps to isolate DC). A signal on the + input produces an in-phase signal on both + and - outputs, not differential output which is what I presumed it would do. *scratches head*

I tried connecting a signal to just the - input on the op amp (pin 5) and nothing happens. Zero output. Flat line.

I should have looked more closely at this specific op amp circuit before copying it as it's probably designed to take differential input and convert it to single-ended. Might try another design, but will give BrianHG's circuit a go first.

EDIT: Time to buy an op amp book and study more!!

[BrianHG]

Low 12V differential amp, 12v total: http://www.linear.com/product/LTC1992

High power differential driver designed specifically to drive 600 ohm XLR, super low distortion: http://www.thatcorp.com/1600-series_Balanced_Line_Driver_ICs.shtml

This is the one you want for +/- 15v support + availability: http://www.ti.com/lit/ds/symlink/opa1632.pdf

[jim_griff]

Low 12V differential amp, 12v total: http://www.linear.com/product/LTC1992

High power differential driver designed specifically to drive 600 ohm XLR, super low distortion: http://www.thatcorp.com/1600-series_Balanced_Line_Driver_ICs.shtml

This is the one you want for +/- 15v support + availability: http://www.ti.com/lit/ds/symlink/opa1632.pdf

Thanks! The OPA1632 looks perfect for what I need. I had a quick browse through Analog Devices and TI but there were so many different kinds, mostly for RF or high frequency data signals. Might order a couple and experiment after I've given your circuit a go.

Oh, I found where I got the idea for my first input stage: It's the first section of an "Instrumentation Amplifier" design, for ultra high impedance input and excellent common mode rejection ratio, but it requires its output to go into another op amp's +/- inputs to give a single-ended output. Although, I'm unsure why both outputs on my circuit are positive phase versus the input. Maybe they're meant to be different versus one another... I'm measuring relative to circuit ground. Doh!

[BrianHG]

Nothing to laugh about your circuit.

Note that my circuit, which only uses 1 single dual NE5532, the load is seen at each input as a resistor to GND, there is no cross mix since the - input of the opamp is always matching the + which is set to GND.  There is no crossing signal between the left and right input unless you remove the opamp.  In other words, if you want to test, make my design and only plug in the left channel.  Scope the right channel inputs and you will not see the left channel audio.  It should stay at 0v.  ***When routing this design, make sure R1/2 & R3/4 is as close to the - input of the opamp as possible.

If you want additional 6db gain, just match the feedback resistors R5&R6 with the load resistors R1/2/3/4.  For 10db gain, double the k of R5/R6.

Thanks, I'll give your circuit a try in a bit. My circuit seems to have a problem (or perhaps a behaviour I wasn't expecting).



Problem with the input stage (attached).

Basically, I attached the output from my audio interface (through bipolar caps to isolate DC). A signal on the + input produces an in-phase signal on both + and - outputs, not differential output which is what I presumed it would do. *scratches head*

I tried connecting a signal to just the - input on the op amp (pin 5) and nothing happens. Zero output. Flat line.

I should have looked more closely at this specific op amp circuit before copying it as it's probably designed to take differential input and convert it to single-ended. Might try another design, but will give BrianHG's circuit a go first.

EDIT: Time to buy an op amp book and study more!!

For your circuit, if you place a signal in the + in, it should be copied to the out +.
If you put a signal to the - in, it should be copied to the - out.
That middle 10k and 50uf AC cap, should add differential gain between the 2.
So, if you have a signal on the +, it should have a bit of a mirror, inverted version on your - output.

Do you have a wiring mistake?
A shorted output, bad bread board, semi-dead NE5532, are the + and - inputs biased to 0v and the NE5532 has +15 and -15v?

[BrianHG]

Note that my circuit expects a differential input.  Having a fake differential input means the opposite output will appear dead.

[jim_griff]

Nothing to laugh about your circuit.

Note that my circuit, which only uses 1 single dual NE5532, the load is seen at each input as a resistor to GND, there is no cross mix since the - input of the opamp is always matching the + which is set to GND.  There is no crossing signal between the left and right input unless you remove the opamp.  In other words, if you want to test, make my design and only plug in the left channel.  Scope the right channel inputs and you will not see the left channel audio.  It should stay at 0v.  ***When routing this design, make sure R1/2 & R3/4 is as close to the - input of the opamp as possible.

If you want additional 6db gain, just match the feedback resistors R5&R6 with the load resistors R1/2/3/4.  For 10db gain, double the k of R5/R6.

Thanks, I'll give your circuit a try in a bit. My circuit seems to have a problem (or perhaps a behaviour I wasn't expecting).



Problem with the input stage (attached).

Basically, I attached the output from my audio interface (through bipolar caps to isolate DC). A signal on the + input produces an in-phase signal on both + and - outputs, not differential output which is what I presumed it would do. *scratches head*

I tried connecting a signal to just the - input on the op amp (pin 5) and nothing happens. Zero output. Flat line.

I should have looked more closely at this specific op amp circuit before copying it as it's probably designed to take differential input and convert it to single-ended. Might try another design, but will give BrianHG's circuit a go first.

EDIT: Time to buy an op amp book and study more!!

For your circuit, if you place a signal in the + in, it should be copied to the out +.
If you put a signal to the - in, it should be copied to the - out.
That middle 10k and 50uf AC cap, should add differential gain between the 2.
So, if you have a signal on the +, it should have a bit of a mirror, inverted version on your - output.

Do you have a wiring mistake?
A shorted output, bad bread board, semi-dead NE5532, are the + and - inputs biased to 0v and the NE5532 has +15 and -15v?

I'l check the circuit again. It's exactly as it was, although the H-Bridge was still connected. I just removed the H-bridge attenuator.

Now it's like this:

Signal into +In = Positive phase signal on +Out (x3 amplification) and nothing on -Out.
Signal into -In = Nothing on either output.

The circuit is correct as is shown in the schematic. I'll double-check the pinout on the NE5532... Yep, perfect.

OH POO! +15V was connected to pin 4!!! I'm such an idiot. Must've counted the pins wrong instead of going anti-clockwise. No wonder only half of the op amp was working and the other half dead. I wonder how it got +15V into the top op amp. Probably blown it now haha.

EDIT: Working as intended! Differential signal being input and differential signal being output with x3 gain. Spot on! Now for the rest of the circuit...

[jim_griff]

Note that my circuit expects a differential input.  Having a fake differential input means the opposite output will appear dead.

Ah yes, I'm using the differential (balanced) output from my audio interface. Going to try common mode signal to see how good it is at rejecting it. It has a great differential output at the moment with excellent CMRR. It goes beyond what my analog oscilloscope can measure/see when adding both outputs together (add button on oscilloscope) and increasing the input sensitivity to almost max. Virtually nothing until it distorts from the input sensitivity being too high.

[jim_griff]

This circuit (attached) works perfectly for my purposes. Now using the H-Bridge attenuator as the gain control with a potentiometer. A bit easier to do it that way.

The first stage is set at unity gain instead. Got rid of the resistors and capacitors. Much simpler circuit now, yet not as elegant as it could likely be haha. I don't care as long as I have some bass in the studio.

The input gain on my subwoofer seems to be quite high and it only has a -6dB to +6dB control on the back, so the reduced gain in my circuit works.

Thanks so much to you all for your help. I really do appreciate it!

[Yansi]

Well that H bridge isn't symmetrical. It does not give symmetrical attenuation in this circuit, as the input impedance of the upper IC3 are not identical.

Maybe I did forget it mentioned here, but why don't you just convert the signal to single ended and work with that? I think there are many ways how to do it better using 6 opamps. This circuit is just too much complicated compared to what is needed.

As we already know, it is a signal summing amplifier for subwoofer.

Use one opamp to convert differential to single ended for each input. Just a simple single opamp differential amplifier. Make sure both inputs (the hot and cold wires) present the same impedance for the signal source (i.e. using 2x 20k and 2x10k resistors). Then use a proper summing amplifier using opamp.  BAM, done, using just 3 opamps in the signal path.  High channel isolation, good enough CMR for such application. If you that much need differential output (which I doubt), add 4th opamp to make the output differential.

[jim_griff]

Well that H bridge isn't symmetrical. It does not give symmetrical attenuation in this circuit, as the input impedance of the upper IC3 are not identical.

Maybe I did forget it mentioned here, but why don't you just convert the signal to single ended and work with that? I think there are many ways how to do it better using 6 opamps. This circuit is just too much complicated compared to what is needed.

As we already know, it is a signal summing amplifier for subwoofer.

Use one opamp to convert differential to single ended for each input. Just a simple single opamp differential amplifier. Make sure both inputs (the hot and cold wires) present the same impedance for the signal source (i.e. using 2x 20k and 2x10k resistors). Then use a proper summing amplifier using opamp.  BAM, done, using just 3 opamps in the signal path.  High channel isolation, good enough CMR for such application. If you that much need differential output (which I doubt), add 4th opamp to make the output differential.

That's a good idea actually. I was thinking about doing it single-ended before but must've got side-tracked and forgot to try it out.

I do need differential output as the subwoofer seems to have a fairly high input impedance - it's very voltage noisy with unbalanced cables longer than a few centimetres but virtually silent with balanced cables.

Thanks!

[Yansi]

it's very voltage noisy with unbalanced cables longer than a few centimetres but virtually silent with balanced cables.

I did never say you should use unbalanced cables.  You need always balanced cables.

And you need to know some old nasty trickery from hell like this

//More on that stuff in the Self's book

[jim_griff]

I did never say you should use unbalanced cables.  You need always balanced cables.

Ah sorry. I presumed you meant unbalanced because of using a single-ended circuit.

I just realised why I wanted to do it the way I did - I need a guaranteed high input impedance so it doesn't put a load on the passive volume control circuitry. Will keep it like this for now and may change it in future when I learn more from Doug Self's book - definitely going to get it soon!

[Yansi]

10-20k load for a passive volume control is of no issue whatsoever. (The potentiometer is to be expected 10k/log.)

[jim_griff]

10-20k load for a passive volume control is of no issue whatsoever. (The potentiometer is to be expected 10k/log.)

I am probably being a bit OCD about it heh. My PVC is about 1450 ohms input impedance. It's a stepped attenuator design with series resistors across each switch position, so it has fixed input impedance but variable output impedance.

The input to my monitors is 20kohm each. Any additional ~20k load causes a noticeable lowering of volume (I've tested it up to 100kohm additional load). 100k was about the point I could hardly hear any difference in volume.

[Yansi]

If you would be OCD about it, you would not drive cable runs with a passive volume control. That's just stupid to do. It just takes a bit over 5m of cable and a few kohms of output impedance to start degrading performance below 20kHz, due to cable capacitance. 

[jim_griff]

If you would be OCD about it, you would not drive cable runs with a passive volume control. That's just stupid to do. It just takes a bit over 5m of cable and a few kohms of output impedance to start degrading performance below 20kHz, due to cable capacitance. 

Haha true! Although, I purposefully made the cables as short as they could be. Too short, in fact. I had to make a new cable for one of the monitors because it was about 15cm too short

[jim_griff]

And this is what I get for making same length audio cables (attached) Stupid OCD.

[Zero999]

One question I have is: Do I specifically need capacitors in the gain stages of each op amp to prevent DC bias building up slowly? I've tried many circuits before and some op amps seem to stop working after a while as a DC voltage builds up on the inputs/outputs. Are some op amps just "better" than others in this respect?

Thanks for all your help guys. Really appreciate it! Top community.

That's only an issue if you have a massive DC gain or tiny signals. The circuits posted here all have a fairly low gain, so they can all be DC coupled. If DC really must be removed from the output, then add the capacitors at the end.

Use one opamp to convert differential to single ended for each input. Just a simple single opamp differential amplifier. Make sure both inputs (the hot and cold wires) present the same impedance for the signal source (i.e. using 2x 20k and 2x10k resistors). Then use a proper summing amplifier using opamp.  BAM, done, using just 3 opamps in the signal path.  High channel isolation, good enough CMR for such application. If you that much need differential output (which I doubt), add 4th opamp to make the output differential.

That's pretty much what I did in my previous circuit, except I used one op-amp for the differential+mixer function.

I've drawn up what you've described. Yes it will work and it really isn't much better than the schematic I posted previously. Try simulating them both. I suppose it only costs a few resistors, assuming the op-amps are in a dual or quad package, so is worth using any way.

[Yansi]

It is very different in terms of input to input isolation.

this one will have very poor signal isolation between positive input terminals, as I have already tried to point out.

I'd suggest using the latter one where two separate diff amps are used. I am still not liking the gain pot in the feedback path due to reliability issues, but you can't do much better, apart from adding one more op-amp and sticking the gain pot against ground, while having it fully AC coupled, to suppress any bias operation influence - scratchy pot having a bias flowing through will certainly be one hell annoying, as it will produce very nasty glitches at the output.

While trying to have only 4 opamps and the gain pot against ground, I'd suggest giving a try the semi balanced output circuit with one opamp. It usually works in most applications good enough.

[Zero999]

It is very different in terms of input to input isolation.

this one will have very poor signal isolation between positive input terminals, as I have already tried to point out.

I see what you mean. It's actually the case on both the positive and negative inputs. Still I don't see the issue here, just make sure the input resistors are much higher than the source impedances, which should be the case anyway, and it will work perfectly. If the source impedances are high and unmatched, then don't use either of these circuits. Use the instrumentation amplifier instead.

I'd suggest using the latter one where two separate diff amps are used. I am still not liking the gain pot in the feedback path due to reliability issues, but you can't do much better, apart from adding one more op-amp and sticking the gain pot against ground, while having it fully AC coupled, to suppress any bias operation influence - scratchy pot having a bias flowing through will certainly be one hell annoying, as it will produce very nasty glitches at the output.

While trying to have only 4 opamps and the gain pot against ground, I'd suggest giving a try the semi balanced output circuit with one opamp. It usually works in most applications good enough.

Yes, I agree about it being bad design to use the potentiometer in the feed-back path, certainly with a BJT input op-amp, such as the NE5532. You should be able to get away with it with a J-FET op-amp such as the TL074, or take the hit on adding more components and AC couple it, like you say.