LM386 – How LM386 Allows Input Past Negative Rail

Think of it this way, at low frequencies, the external series RC network is effectively an open circuit so the amplifier behaves as if the network were not present.

At high frequencies, the RC network is effectively just the 10k resistance.

Looking at the equivalent schematic in the data sheet, note that this 10k resistance is in parallel with the internal 15k feedback resistor.

So, instead of a 15k feedback resistor, there is, at high frequencies, effectively just a 6k resistor , i.e., there is more feedback at high frequencies and thus, reduced gain.

In fact, rather than boosting the gain at low frequencies, the addition of the RC network actually reduces the gain at high frequencies. The effect on frequency response is equivalent; the bass frequencies are amplified more than the high frequencies.

With the addition of "R17" (C17) before the digital pot, the voltages present on pin 5 of the pot, for any typical input signal, will settle to an average of zero volts since capacitors block DC. So the AC signal being fed into the pot, if it were 0-3.3v from a microcontroller, is now -1.65 to +1.65v, which

To quote the MCP4131 Datasheet,

The terminal A pin does not have a polarity relative to the terminal W or B pins. The terminal A pin can support both positive and negative current. The voltage on terminal A must be between Vss and Vdd.

So you are placing as low as -1.65v on a wiper pin with these maximums as defined in the datasheet:

Voltage on all other pins (PxA, PxW, PxB, and SDO) with respect to VSS .................. -0.3V to Vdd + 0.3V

Try it without C17 and it will likely work. If it still misbehaves, then keep C17, but bias pin 5 (add a resistor divider to it from +3V3 to Vss) to around +1.65v or slightly over. Doing so should prevent the signal from going negative into the pot.