Explore

Research

I use these playgrounds to explore how Erv Wilson’s harmonic lattices connect to melodic scales. I can change the assumptions and see what happens to the tuning, the fit, and the harmony.

Loading playgrounds…

A guide to the playgrounds

I use these JavaScript experiments to explore specific tuning questions. Their results depend on the chosen parameters, tolerances, and search limits, and I read them alongside the dated research notes. The browser tools are separate from the installable Wilsonic software.

  • Lattice Recurrence

    I seed recurrence relations with a just lattice and explore how they converge to a generator, then compare the resulting scales with a direct MOS fit.

  • Lattice MOS-Fit

    I fit moments of symmetry (MOS) to a just lattice by varying the generator, note count, and mode, and compare tone errors with the triads that survive.

  • Comma Dial

    I start with one generator per prime and temper commas with a dial to see how the lattice’s rank drops and which scales close.

  • Diophantine Harmony

    I compare a generator chain’s fit to harmonic targets with its continued fraction, and explore how the two readouts respond when the targets change.

I keep research notes and the source repository available so the assumptions can be inspected. I welcome questions and corrections through my public contact routes.

The Ear Bench

The experiments behind these playgrounds score scales with numbers I trust as measurements: how many proportional and subcontrary triads survive in a chain, and how far each tone sits from just intonation. Those numbers are proxies, because they tell me where to listen and they cannot tell me what I will hear. When I listened to gate G-031, the two candidate chains differed by at most 2.4 cents on any step, and I still heard one as clearly more melodic, which surprised me and convinced me that the arpeggio I was playing could not answer the harmonic question the gate was asking. A computer can search far more scales than I could ever play, so my listening time has become the scarce resource, and I built the Ear Bench to spend that time well.

For each research gate the bench composes four passages from the just structure that the two candidate scales approximate: the scale run up and down one octave, a tune that walks from chord tone to chord tone, a chorale that holds every triad of the structure in turn, and the tune over the chorale. Each passage plays in both scales as X then Y, and which scale is X is hidden and drawn fresh every time the page loads. I pick X, Y or Same, and only after the pick does the bench reveal which was which. A live strip shows which of X and Y is sounding and which chord is held, so I always know what I am hearing without knowing which scale it is.

Open the Ear Bench

Anyone can judge these scales, and I would like you to. The bench runs right here on the site, your picks and notes save in your own browser only, and the Copy verdicts button at the bottom of the bench exports them as JSON. My method and the record format are written down in JUDGING.md, and every judgment I have made is public in judgments/marcus-w-hobbs.jsonl in the research repository.

If you run the bench and make your own picks, copy your verdicts and send a pull request to the research repository that adds experiments/ear_bench/judgments/<your-name>.jsonl in the same format. I will read every one, and where someone hears something I missed or builds a better way to listen, I will adopt the idea and credit the person who had it by name.

The bench's code is MIT licensed, like the rest of the research repository, and my judgments are released as a public dataset under CC BY 4.0, so you can reuse them for any purpose as long as you credit Marcus Hobbs and link back to the repository.