I was thinking about this 5V-output buck converter I dumpster-dived. A plausible-looking datasheet says the input is 8–48 volts and the output is 5V at up to 2.1 amps. It has a feedback pin which it uses to adjust the duty cycle of the 150kHz PWM signal it uses to generate the output; you’re supposed to hook that up to the output side of the inductor so it can regulate the output voltage properly.
It occurred to me that this sort of implies that if, instead of hooking up the feedback pin directly to the output side, you hook it up to the output side through a voltage divider to ground, you should be able to get it to act as a 6V or 8V or 10V regulated power supply instead of a 5V one. Maybe it won’t compensate for load changes quite as quickly, but I don’t think it should oscillate.
But what if, instead of using a voltage divider to ground, you use a voltage divider to a variable reference voltage? Maybe you could make a high-power class D audio amplifier! Say you use a high-impedance 4:1 divider, so that a 1-V shift in the “reference” voltage (really the input) requires a 4-V shift in the output in the opposite direction to compensate. It’ll be superimposed on a 20VDC level that you probably want to block with a capacitor, but 4 volts RMS across 8Ω should give you 500 mA RMS, which is 2 watts, which is pretty loud. Filtering out the 150kHz (and harmonics) carrier should be pretty easy.
Also, you might be able to use this as an AM radio transmitter. AM radio is in the range 540 kHz to 1610 kHz, so there ought to be some harmonics of the 150kHz PWM signal in there (at least the fourth through ninth or so), and their power ought to vary with its duty cycle. Some experiments reveal that the second harmonic of a PWM wave is zero at 0% power, 50% power, or 100% power (which is the same as 0%), but maximum at 25% power; the third harmonic is zero at 0% and I guess 33⅓% and 66⅔%; the fourth harmonic is zero at the same places as the second harmonic and also at 25% and 50% power; the fifth harmonic at multiples of 20%; and so on. So lower harmonics should modulate more reliably. If you’re trying to modulate the fourth harmonic, you’d probably want to use a duty cycle on the steepest part of this slope, like in between the maximum at 12½% and the minimum at 25%, around 18.75%, which means you want to regulate the nominal output voltage to be around 18.75% of the input voltage, and vary that percentage according to the signal you’re trying to modulate. Then you “just” need to filter out the other harmonics (a bit tricky since the actual frequency could be anywhere from 125kHz to 175kHz, so the desired fourth harmonic could in theory be anywhere from 500kHz to 700kHz, while the undesired fifth harmonic could be anywhere from 700kHz to 875kHz), hook it up to an antenna, and tune a radio to it.
The thing isn’t designed for an output duty cycle of over 62⅕% (⅝), but maybe using the region between the maximum at 37½% and the minimum at 50%, like around 43.75%. For its nominal 5V output, this would mean an input voltage of around 11.43 volts, which is more manageable than the input voltage needed to get 18.75%.