Some way or another I found myself salvaging a lot of ULN2003 and similar chips. Description says "Darlington driver", and while I try to be boring and do things properly most of the time, I have to admit that when I desolder random components the first thing I sometimes ask myself is "can it fuzz?".
Turns out this one can fuzz. Aside from the parasitic diodes and the "common" diodes, this chip basically contains 7 Darlington pairs with accessible base and collector pins and some included resistors. This imposes some limitations, since the emitter is grounded (unless maybe you're only using one stage and lifting the ground pin with emitter degeneration) and the bias resistors included force to work with low impedances.
But wait! If less gain is desired, one can add an input resistor! This works as expected if one were to build this same circuit with discrete components, treating it as an inverting amplifier with its limited performance, but for some reason the IC doesn't like too big of an input resistance. It can do amplification somewhat well, and even be as linear as you might expect, but you can't really lower the gain that much before it starts to sound very bad. I couldn't see any oscillation in that case, so I'm not entirely sure what's the cause, but it makes other fun uses like a summing amplifier or filters less appealing, although maybe still doable.
Anyway, as proof of concept, here's two simple fuzzes for you:
I've included the equivalent circuit of a single stage so that it makes them less weird than they seem once you analyze them.
Both feature an emitter follower at the input. This was necessary because of the forced low input impedance, which results in just too much signal loss with a guitar. It's still somewhat unusual: first, it doesn't have a resistor to ground because the input of the stage already has a DC path to ground; second, it's biased with DC feedback from the collector, which was the only way I found to make it bias not only reliably, but at all since it's DC coupled and there's a lot of DC gain otherwise. The AC feedback is shunted, otherwise the Miller effect would drop the input impedance back where we started. A more boring solution would be an AC coupled emitter follower and the two stages biased separately. This would also mean extra parts for the follower bias unless you want to take the divider directly from the power supply and couple its noise in the input.
From the second stage on, it's very simple: aside from the 3.3k collector load, there's a 22k shunt feedback and bias resistor from collector to base. This biases the collector close enough to center (I didn't strive for perfection given the context) and shows one strength of this chip: you need very few parts to get an amplifier going. The other face of the coin is that you can't go very far from these values since the path to ground is fixed and the emitter is grounded.
Other than that, the ULNa is two stage and the Radius is three, with a few value changes to taste. You can in theory keep stacking stages and gain, but since you can't add input resistance to decrease gain much before it stops working, I had to stop here before it was too much (or maybe it's already too much for someone's taste heh) and I had to give up things like T notch between stages or "clean" blend because of this limitation. Still, I think they're fun. The ULNa is tamer and has good cleanup (maybe because the output impedance of the buffer depends on the source impedance of the guitar?), the Radius is thicker and more extreme.
Feel free to mess with the idea and find other uses. Just one warning when changing values: because of the dc feedback bias, if you want to increase the coupling capacitors to 10u or so even for the second stage, you also need to increase the 10u feedback shunting capacitor or you'll get a low frequency peak that can result in oscillation. As it is, I've checked for absence of both low and high frequency oscillation.
Here's demos of both: