I've never been much into using the LM386 as distortion device. Mostly, the 50k cap on input resistance was what put me off, but also that using a power amplifier, albeit tiny, seemed wasteful, even if honestly the current draw at idle is modest indeed. Still, there are many designs and many people building them, with a very popular one consisting of two 386 in series.
Still, when a friend of mine was working on a distortion with it, I had one idea I wanted to try, which is to treat this as much as possible as an op-amp. After all, the device has two differential inputs, even if only one is normally used. Some important differences are that the inputs are much lower impedance, they bias themselves very close to ground, while the output stays at half supply. This is sometimes exploited to generate a powerful half-supply reference with no additional parts. The gain is normally limited between 20 and 200 by the internal feedback, but the datasheet says that it is compensated for closed loop gains higher than 9.
This is what brought me to this quick and simple design, which I dedicate to Tim Escobedo, the mysterious person behind many weird but cool guitar effects that have been on the internet for many years, and who I associate with using the 386 like this the most.
The input is AC coupled to the non-inverting input, and as usual the internal pulldown takes care of biasing. Then it gets more interesting: I really wanted to use the inverting input for negative feedback, that was the whole point. While I could have exploited the 50k resistor to ground on that pin, even a 1M pot, which I consider the upper limit of normality, would have resulted in the feedback signal being attenuated by only 20 times, resulting in a not very fun amount of closed-loop gain. No worries, a 1k resistor there bumps up that ratio to up to 1000.
Since the datasheet warns against trying to reduce gain below 9 (it's as if they knew!), the 10k series resistor ensures that (still above 10 with just 200 of open loop gain). Not sure if it would be so bad to go below that, but best to avoid oscillation.
The feedback is taken after the output capacitor so it's ground referenced. I've tried both with and without a capacitor to block DC and it seems fine, no ill effects either on the feedback or the volume pot.
The 33p cap is there to avoid oscillation, since this circuit has a lot of gain. A larger value will reduce treble to taste. This is after the 10k since the closed loop gain limit is still there! It will actually misbehave if you connect this capacitor before the 10k.
Finally, C3. Since I'm using my own feedback, I don't want any of that internal feedback to limit me. Shorting 1 to 8 is the usual way to get the gain from 20 to 200, but the feedback is still there, just less of it. I've tried dumping it to ground with this capacitor at pin 8, which was unimpressive, at pin 1, which was too much and oscillated no matter what, audibly or not, then by shorting 1 and 8 and then shunting the feedback to ground. This sounds just right. By having this capacitor here, I didn't feel like it's necessary to add a capacitor on pin 7, looking at the internal schematic, but you can try if you want.
Speaking of sound, how does it sound?