This is a circuit I've made a little more than a year ago, as an answer to a very popular distortion pedal based on preamplifiers of the old days (tip: that one is "broad", this one is "narrow"). While another circuit that qualifies as being derived by said desk inputs is none other than the old Fuzz Face, I want to give credit to the claim because what we are dealing with is a transistor pair with a feedback network wrapped around it, of the kind that was very popular before op-amps to get close enough to the ideal amplifier. You can learn more here.
There are many things about the original I wanted to change, starting the use of the transformer of which I strongly object the necessity:
it's not providing isolation because the cold wires are connected;
it's not providing impedance or voltage benefits because it's 1:1 ratio;
it's not providing bandwidth limiting or significant distortion because it's used with a suitable source impedance and level. A worse transformer than the one used might have more of an effect.
You don't have to take my word, I took hold of a "10k" 1:1 transformer to test if there was any difference and recorded it. I won't tell you which is which both because it's a blind test and because I honestly don't remember.
Other than that, the inclusion of a charge pump, the values used and a somewhat excessive complexity asked for action. What I didn't want to change was the sound, people seem to like that part. That's how I ended up with my version:
Actually there's two of them, one with NPN and PNP transistor like the original, and the one I actually built with two NPN transistors. Other than polarities, the other difference is that Q10's collector needs to be biased close to ground to correctly bias Q2 in the NPN version, close to Vcc in the PNP version. I leave you free to experiment with different transistors if you deem it time well spent, the 3904 is a stand-in for any generic NPN transistor and anything should work without much trouble, as explained later about bias stability.
Altough different at first glance, like the broader brother this is a series-shunt feedback amplifier (feedback is in series with the input signal, in parallel with the output). The feedback is AC-only.
Base bias is taken straight from the collector, instead of a tap on the collector load. Since the 1k is bypassed, the collector load is made closer to the unbypassed 5.6k. This network doesn't load the amplifier significantly. The 220k matches the original bias resistor, and together with the increased feedback (straight from the collector instead of attenuated by the tap) using only two resistors here resulted in a stable bias even with different hfe and voltages. This feedback is DC-only, so there's no Miller effect, and this also means a quiet bias.
On Q1's emitter, the parts count is reduced without changing function. The 5.6k is bypassed by the parallel capacitor and doesn't affect AC gain, so I removed both. While this does mean less degeneration for the DC bias, I've found that the increased amount of DC feedback from the collector more than makes up for it. The 1k resistor is kept as is. The rest of the components are rearranged but in synthesis, you have 43 ohms of unbypassed emitter resistance in "high gain" mode and 470 in "low gain" mode, compared to 68 and 378 in the original.
Q2 is about the same as in the original, with the exception of the polarity swap in the NPN version. Q2's collector has been biased at about 2.5V in the original, which is 2V "colder" than center bias. I've kept this bias for the PNP version, while for the NPN it means 6.5V, since the polarities are mirrored. The input sensitivity, or the peak headroom, is the same as in the original. One addition is C5, which limits gain at high frequencies and prevents oscillation, the main issue I had with the first attempt at this circuit. It doesn't affect the frequency response otherwise, so I guess without it you're relying on poor bandwidth to begin with and even my germanium transistors were too good.
The feedback network isn't changed much. V1 couldn't have C11 or I had oscillation no matter what, with any transistor, with or without transformer.
The output is just a 25k volume pot (you can use 50k or 100k if you prefer), which is about the same load as the original -3dB attenuator, transformer and volume pot (15k+33k//25k=29k). The taper is logarithmic.
In general, I've used E6 values where I could, while maintaining the same AC and transient response. This, the lower parts count, the possibility to use almost any transistor and the lack of an expensive transformer make this a very appealing alternative for ease of building.
Bias stability
One very nice property of this circuit is that, even if the global feedback is AC only, it has great bias stability due to the emitter degeneration (on both stages) and feedback bias. In practice, this means that different hfe and different supply voltages result in the same bias without much effort. This is also why I didn't explicitly include a charge pump: you can run this circuit from any voltage up to the 27V reached by the original, as long as the component ratings are up to it, whether taken from a random power supply, batteries in series, boost converters or charge pumps. I personally don't recommend it, because for me it mostly meant it couldn't distort as much, given that the gain is the same and the headroom isn't, but I guess it means you're using less feedback for the same amount of clipping, if that's something you want to try.
In simulation, Q2 collector went from 6.5V at 9V to 19.5V at 27V, which is exactly three times as much, just as the supply. This tells you how well it works.
Frequency response
I've simulated the frequency response and gain of the circuit compared to the original for both low and high gain modes, at various settings of the gain pot, and you can see that they're very similar except for the Narrow having an higher maximum gain, so it should be easy to see by interpolating that you can get all the same sounds as the original in this regard. I've also run some transient analysis, mostly to see if the signal was clipped by the same amount on both peaks.
Layout
Here's a compact, verified layout for the NPN version. The PNP one should be easily derived.
Demo
Here's how mine sounds:
Final thoughts
Other than gain, the main difference between the two modes is that the input impedance is lower in "High" mode, resulting in the familiar guitar volume interaction and a mellower sound. This is a nice circuit that covers all needs of vintage distortion sounds, from boost, to bright crunch, to a classic fuzz, also thanks to the effective bass control.