Fifteen Monarks
First year of school. The brief: take a patch built in Massive and rebuild it in Monark. My partner picked a pad. The pad was inharmonic. Monark can't do inharmonic. So I made fifteen of them.
The trap.
Monark is Native Instruments' model of the Minimoog: a subtractive synth. You start from a harmonic-rich waveform - saw, triangle, sine - and carve it down with a filter. Every move you make is about removing or shaping harmonics that are already there, sitting on the harmonic series, integer multiples of the fundamental.
An inharmonic sound is the opposite. Its partials don't land on integer multiples - they sit off the series, the way bells, metal, and detuned clusters do. That's what made the pad sound the way it did.
Which is the problem in one line: the single thing the brief needed was the single thing the instrument structurally cannot do. You can filter a saw all day, you will never move a partial off the harmonic series. There's no setting for it. It's not how the machine thinks.
The way out.
The way out is older than the synth. Any tone is a sum of sines - that's Fourier. A subtractive synth refuses to place partials freely, but additive synthesis does nothing else: you build the spectrum one sine at a time, and you put each sine wherever you want, harmonic series or not.
Monark can output a clean sine. So one Monark, set to sine, tuned to a single frequency, is one partial. The instrument that can't make an inharmonic spectrum can make exactly one point of one. Stack enough of them, put each sine where a partial sits, balance the levels, and the spectrum reassembles - inharmonic and all, because nothing is forcing the partials onto the series anymore. They go where the source put them.
I didn't read the partials off a number. I read them off a sonogram. The source pad's spectrogram lays the inharmonic content out as bright horizontal bands - that's the map. Then I put my reconstruction's sonogram right next to it and tuned by eye: detune each Monark until its band slides onto the target's, partial by partial, crossing them over until the two spectrograms read as one. Fifteen instances, fifteen sines, fifteen bands. Summed, they are the pad.
It's not elegant and it's not what Monark is for. That was the point. The instrument couldn't, so I turned fifteen copies of it into a different instrument.
A / B.
The two sonograms side by side - the source pad and the fifteen-Monark reconstruction - with each band detuned into place until they overlap. The proof is visual before it's audible: when the spectrograms match, the sound matches.
Why I still point to it.
Twenty out of twenty. The professor told me he didn't even read the documentation - he looked at the patch and that was enough. The patch said it before the doc could.
I was proud of it then and I still bring it up, for a reason that took me a while to name. It's the first time I solved a problem by refusing the frame of the tool instead of bending the problem to fit it. The brief assumed Monark's limits were the walls of the room. They were just the default. Everything I've built since - the toolkits, the modular rack, this site - is the same move, scaled up: when the instrument can't, you don't lower the ask, you rebuild the instrument.