Op-amp unity gain buffer introduces distortion in the LM386 circuit with high gain factor

Question

I'm building a simple mixing stage, lowpass filter, and final amp in sigle supply (9v battery). I started from the LM386 circuit on the data sheet. It had a lot of distortion which was solved with the decoupling capacitor (fig.1). As a second step, I introduced an op-amp unity gain buffer to isolate the mixing stage from the filter stage and be able to calculate easily the cutoff range of the pot. I put the input capacitor (C11), matched the input pots with the input resistors (useful?), and used C13 to remove the bias, since the LM386 wasn't working with the biased signal (fig.2).

In fig.1, everything works fine: I tried many input sources and the amp works well from a gain factor of 20 to 200.
In fig.2, it works until the gain factor of LM386 becomes very high. At this point, if I open the filter and let the high frequencies pass there is a significant distortion.
The input and the speaker I used for the test are the same.

Any solution to have the circuit in fig.2 withstanding the same amp gain without distortion as in fig.1?

Answer 1

A 9 V battery has a high internal resistance, as soon as the LM386 starts to drive the speaker, it pulls current from the battery resulting in ripple on the battery voltage.

The DC bias you apply to the input signal is derived directly from that supply including the ripple voltage by R16 and R17. So you have some feedback, this output signal feeds into the input as well, that can create all kinds of strange results.

I would add a 10 uF decoupling cap in parallel with R17, that would attenuate the ripple. Then feed that filtered DC bias to the TL071's input via a 10 k resistor.

And as commented you need a better power source than a 9 V battery, it does not have sufficient power to drive an LM386 + speaker. At least not for a reasonable amount of time.

Answer 2

1. TL071P has minimum recommended supply of 10V, but it depends on the chip manufacturer which you don't mention.

2. The speaker needs a lot of drive current which a 9V battery is unable to provide without significant voltage drop.

3. What are your signal voltage levels? You do realize that with the 10uF setting the gain to 200, you must limit audio voltage to 45mVpp or it will clip. With gain of 20, the input must be 450mVpp, except that LM386 has absolute maximum of +/- 400mV at the audio inputs before it gets damaged. It is possible that op amp output has driven larger voltages than what the LM386 input can handle.