optimizing a single supply symmetrical opamp circuit (for audio)

[jmibk]

I like to have a symmetrical audio input infront an ADC that operates with a single supply of 5V.
I choosed the TL072 for that purpose, because it works with audio and I had it laying around.

Here is the schematic:


In theory, I lifted the positive path to a DC of 2,5V, that is produced by the voltage divider R57/R62.
R56, R58, R59, R60 (100k) are the regular resistors for the symmetrical amp.
R61 may not be needed here.

If I operate the thing with 1Vpp sine out of my function gen (positive input 0deg, negative input 180deg shift) I get the waveform below on the output.



If I lift the positive sine to 1V DC up, everything looks fine.
If I switch off one of the both input signals, the output vanishes (I expected half voltage on the output here)

Why is that, where I did a mistake by modyfing this circuit?

Thanks for your advices!

[Zero999]

What op-amp are you using? The TL072 or TSV7722? The TL072 will not work properly off 5V. Use the TSV7722.

[David Hess]

Douglas Self recommends the differential line input circuit shown below as a practical starting point.  R4 should be returned to a fraction of the supply voltage for single supply operation.

[Gyro]

I would like to see a rather larger value for your C76, decoupling the mid-rail bias point.

[jmibk]

Thanks for your answers.

The TSV7722 was planned before, but the price said to use another opamp - so I took the TL072 (and oversaw the fact, that the minimal voltage is 6V, OOPS)
Now I changed them to the LM358, and the signal gets more like it has to be, its limit is 4V.

Now I think about choosing the LMV358(A), that is a) capable to work with 5V and is an rail to rail opamp, that can be used from 0V to almost 5V - so I can use (in theory) a 4Vpp input signal without any reasonable distortion.

Who is Douglas and which whitepaper is that? Looks interesting.

[Zero999]

The LM358 is very noisy compared to the TSV7722 and has horrible crossover distortion. The LMV358 is still quite noisy but I don't believe it has as bad crossover distortion.
https://www.ti.com/lit/ds/symlink/lmv358.pdf
https://datasheet.lcsc.com/szlcsc/3PEAK-LM358-SR_C93149.pdf
https://www.mouser.co.uk/datasheet/2/389/tsv7721-2892142.pdf

[Benta]

Where's the AC coupling on the inputs? Without that, you just can't do this.
It may well be that your amplifier thinks that it's at 2.5 V DC bias, but I'm not sure your signal source agrees.

[strawberry]

LM358 run class A to reduce crossover

[Zero999]

The LM358 can only swing from 0V to just over 3V, when powered off 5V, so it's better to bias it to 1.5V, rather than 2.5V.

Yes, the input should be AC coupled and the output biased in class A to remove crossover distortion.

U13.2 = LM358

[jmibk]

Thanks again for your helpful hints and tipps.

Question: what means biasing the output in class A? Class A is for me a amplifier circuit that consists in theory out of one transistor and a resistor. The transistor is held into linear conductive area via some additional resistors on its base. The output of a regular opamp isn't one transistor.

[Zero999]

Thanks again for your helpful hints and tipps.

Question: what means biasing the output in class A? Class A is for me a amplifier circuit that consists in theory out of one transistor and a resistor. The transistor is held into linear conductive area via some additional resistors on its base. The output of a regular opamp isn't one transistor.

R61 is low enough to bias the LM358 in to class A, so only the high side transistor in the output stage conducts, which eliminates crossover distortion.

Look up LM358 crossover distortion.

[mikerj]

The output of a regular opamp isn't one transistor.

With a strong enough pull-down the upper transistor in the push-pull output never switches off, so you effectively have a class A output.

[Bassman59]

Thanks for your answers.

The TSV7722 was planned before, but the price said to use another opamp - so I took the TL072 (and oversaw the fact, that the minimal voltage is 6V, OOPS)
Now I changed them to the LM358, and the signal gets more like it has to be, its limit is 4V.

Now I think about choosing the LMV358(A), that is a) capable to work with 5V and is an rail to rail opamp, that can be used from 0V to almost 5V - so I can use (in theory) a 4Vpp input signal without any reasonable distortion.

Look into proper low-voltage single-supply rail-to-rail op-amps. While they are more expensive than an LM358, you'll get much better performance. Some examples off the top of my head are AD8602 and OPA2365.

Who is Douglas and which whitepaper is that? Looks interesting.

Doug Self has published some well-known papers about audio circuit design.

[TimFox]

The LM358 can only swing from 0V to just over 3V, when powered off 5V, so it's better to bias it to 1.5V, rather than 2.5V.

Yes, the input should be AC coupled and the output biased in class A to remove crossover distortion.

U13.2 = LM358

In this circuit, for good CMRR, the input coupling capacitors should match each other in value (the value itself is not critical and any cheap measurement tool should be adequate to judge matching to 1% or so), and the CMRR at frequencies above where the capacitance matters depends similarly on the matching of the four resistors.

[David Hess]

In this circuit, for good CMRR, the input coupling capacitors should match each other in value (the value itself is not critical and any cheap measurement tool should be adequate to judge matching to 1% or so), and the CMRR at frequencies above where the capacitance matters depends similarly on the matching of the four resistors.

It is difficult to match capacitors so the usual recommendation, and what the example I posted above does, is use a much larger value than necessary moving the transition band lower in frequency than strictly necessary.  This is important not only for common mode rejection in this case, but to prevent distortion.

Who is Douglas and which whitepaper is that? Looks interesting.

Douglas Self had written a couple of very good books on audio circuit design.  The example I posted is from Small Signal Audio Design (2015).

[TimFox]

For a critical application, I used a DE 5000 to match 10 uF polypropylenes to 0.1%.
I assume that the crummy capacitor measurements in DMMs can judge equality to 1%.

[David Hess]

For a critical application, I used a DE 5000 to match 10 uF polypropylenes to 0.1%.
I assume that the crummy capacitor measurements in DMMs can judge equality to 1%.

At least some DMM's do pretty good with capacitance measurement, and the capacitors need to be matched for equality and not absolute value which considerably relaxes the measurement requirements.

[Zero999]

In this circuit, for good CMRR, the input coupling capacitors should match each other in value (the value itself is not critical and any cheap measurement tool should be adequate to judge matching to 1% or so), and the CMRR at frequencies above where the capacitance matters depends similarly on the matching of the four resistors.

It is difficult to match capacitors so the usual recommendation, and what the example I posted above does, is use a much larger value than necessary moving the transition band lower in frequency than strictly necessary.  This is important not only for common mode rejection in this case, but to prevent distortion.

How much is necessary? C = 100nF and R = 100k gives a lower cut-off of 16Hz, which admittedly I selected because it's standard for audio,  but most of the common mode noise will be at the mains frequency and its harmonics.

[David Hess]

How much is necessary? C = 100nF and R = 100k gives a lower cut-off of 16Hz, which admittedly I selected because it's standard for audio,  but most of the common mode noise will be at the mains frequency and its harmonics.

With such a low value of capacitance, the only requirement is to use a good non-electrolytic capacitor.  But in the example I posted, electrolytic capacitors of a much larger value than necessary are used to avoid distortion and the low frequency cutoff is applied elsewhere.