# Analysing AC-Coupling differential opamp configuration

I built an audio DSP board based on CS4272 and used its reference design for the output section. The signal is getting extremely distorted when summed signal exceeds ~0.7Vpp. (Rail voltages 0V-5V and OPAMP is OPA1642 rail-to-rail type, so nearly 5Vpp should be possible.)

Now I want to modify the output circuitry but I don't really understand how to analyse and dimension this. As far as I understand it's a differential opamp configuration but extended with some filters.

1. Can someone help me analyzing this circuitry and how to modify it to reduce the gain? How to calculate high pass characteristics?

2. Question: Why is it necessary to have additional frequency filters when we already have a 22uF AC-Coupling capacitor in the end?

• The values are very imbalanced. Is this a vendor-recommended circuit? Commented Jun 3, 2019 at 12:11

The common mode input voltage range for the OPA1642 is: -

(V–)–0.1 (V+)–3.5 V

This means that if you have a power supply of 0 volts and 5 volts, the input range is -0.1 volts to +1.5 volts i.e. it will heavily distort the input signal because it is biased at 2.5 volts DC.

OPAMP is OPA1642 rail-to-rail type

No it isn't. The output is specified as being "rail-to-rail" but the input range certainly isn't.

Can someone help me analyzing this circuitry and how to modify it to reduce the gain? How to calculate high pass characteristics?

My recommendation is that you acquire a free simulation tool (like micro-cap) and simulate the circuit in order to further your understanding of how it works and what the frequency response is.

Why is it necessary to have additional frequency filters when we already have a 22uF AC-Coupling capacitor in the end?

There may be some low pass filtering required and certainly, the 470 pF and 1.5 nF, offer some of that.

• Yep. Now I learned that rail-to-rail output doesn't automatically mean rail-to-rail input. Thank you very much. This explains the distortion issue. I will pick a suitable opamp and test with it. However, I still would like to understand how to analyze this circuit. I already simulated it in LTSpice and can adjust the gain changing the resistor but they influence the freq. filter responses. Empiric setting of those values is not the solution. And analysis doesn't seem to be trivial. Commented Jun 3, 2019 at 9:30
• There is no single resistor value you can change to properly give a gain change. Also, given the balanced nature of the AOUT+ and AOUT- and, that they are fairly low output impedance, I see no persuasive reason to have a circuit that is so imbalanced in the values. I.e. 4.42 kohms versus 1.33 kohms is a little bizarre. Clearly the RC times constants are identical i.e.6.8 nF and 1k33 is the same time constant as 2.2 nF and 4.42 kohm. So, if you want to change the gain I would go for a secondary gain stage. Commented Jun 3, 2019 at 10:14