title: PUM Synthesizer — design proposal audience: builder + Sheela Rahman (Xosar) frame status: proposal v1; references existing scope3d prototype sources: pum_core (this repo) · scope3d/ (prior session, 32-step interpretation) · Bendall PART 15/17 · Sheela Rahman / Xosar interviews
PUM Synthesizer — design proposal
0. Frame
This is a proposal for a physical + software synthesizer whose entire architecture is generated from PUM. Every parameter (oscillator pitch, filter Q, modulation rate, envelope time, Lissajous figure aspect, cymatic plate diameter) is a PUM-derived rational, not a free-floating number.
The reference frame is Sheela Rahman / Xosar’s “coherent frequencies” approach: she treats PUM ratios as physical resonance addresses rather than musical aesthetics, and uses sound + cymatics + geometry as a single unified probe. The synthesizer below is built so any setting can be cited back to a PUM seed — there are no “magic numbers.”
The scope3d prototype in ~/claude/scope3d/ is already step 1: a 3D oscilloscope with a 32-knob voice bank driven by the 24 PUM rows, three render modes (Lissajous XYZ, Waterfall FFT, Cymatics Chladni). This proposal extends scope3d into a fuller instrument and sketches the hardware path.
1. The model in synth terms
| PUM concept | Synth correspondent |
|---|---|
| 4 cardinal scalars (AETHER, SUN, TIME, MATTER) | 4 macro-voices · 4 oscillator banks · 4 modulation sources |
| Dimension fractions (n × 10/9) | Pitch ratios → Scala scale · LFO rate quantizers |
| 11.111 (Aether→Matter conversion) | The “swing” / phase-coherence parameter |
| 32 planes × 64 radial points + 1 zero | Voice grid: 32-step sequencer × 64-tone polyphony · centre = drone |
| 7,200° = 20 spiral revolutions | 20-cycle LFO / arpeggio length |
| Hydrogen 3-4-5 triangle cascade | 3-stage envelope (3-side·4-side·5-side amplitudes; cascade times = 24/60/60/60/60/60/60/24 seconds) |
| Vajra basket-weave (15+15) | 30-band stereo phaser; outer/inner = L/R; opposite twists = anti-phase |
| 4 quadrature charge sources (++ −−) | 4-channel quad-panner with ±phase pairs |
| Mirror set (digit permutations) | Voice clone / detune algorithm |
| C = 800/3 Hz tuning anchor | Master tuning fork |
2. Architecture (3 layers)
Layer 1 — Voice generators
Eight oscillators per voice, addressed by (cardinal × dimension) in the projection grid:
osc_index = cardinal_index * 4 + dimension_index # 0..15
freq(osc) = MUSIC_C * (cardinal_value / (dim_fraction * 1000))
= (800/3) * cardinal_value / (1000 * n * 10/9)
= (240/n) * cardinal_value / 1000
Concretely (Hz):
| osc | cardinal | dim | freq |
|---|---|---|---|
| 0 | AETHER × 6D | 6 | 5760 |
| 1 | AETHER × 5D | 5 | 6912 |
| 2 | AETHER × 4D | 4 | 8640 |
| 3 | AETHER × 3D | 3 | 11520 |
| … | |||
| 12 | MATTER × 6D | 6 | 138240 (drop 6 octaves → 2160) |
| 13 | MATTER × 5D | 5 | 165888 → 1296 |
| 14 | MATTER × 4D | 4 | 207360 → 1620 |
| 15 | MATTER × 3D | 3 | 276480 → 2160 |
After octave-folding into the audible band [27.5, 20000] Hz every cell becomes a usable musical pitch. The projection grid in pum_svg/projection_grid.svg is the voice patch panel.
Layer 2 — Modulation
Five modulators wired in parallel, each driven by a different PUM rate:
| name | rate | source | role |
|---|---|---|---|
| HYDROGEN_LFO | 0.5184 Hz | TIME × 10⁻⁶ | slow swell — one cycle per 1.93 s |
| LIGHT_LFO | 0.000144 Hz | LIGHT × 10⁻⁶ | ultra-slow drift — one cycle per ~115 min |
| SANSKRIT_LFO | 31.104 Hz | from H cascade /60 min | tremolo |
| GREAT_YEAR_PHASE | 25920 / day = 0.3 Hz | precession-mapped slow phase | binaural shift |
| RFEU_RING | 1296 Hz | resonant centre | ring modulator carrier |
Quantizer: every modulator’s depth and rate snap to elements of the PUM rational lattice (so randomized = still lattice-valid).
Layer 3 — Spatial / output
Quadrature panning (Vajra-mirror): left-front (+), right-front (+), left-rear (−), right-rear (−). Pairs at 90° apart per the patent’s charge-source layout. Sum ≈ true zero at the centre (the Plasmoid singularity). Acoustic equivalent of a real Vajra: the listener stands at the zero point.
Lissajous XYZ render (already in scope3d): X = cardinal_a, Y = cardinal_b, Z = dimension_modulator. Rotating the projection through dim-fraction angles draws the four-quadrant rose as a 3D figure.
3. Patches that ship out of the box
Patch A — “Protium drone”
- 8 voices on
freq(AETHER × 5D) = 25,920 Hz / 27 ≈ 960 Hzand octaves - HYDROGEN_LFO modulating amp at 0.5184 Hz
- 3-4-5 triangle envelope: attack = 86.4 ms (3-side scaled), hold = 345.6 ms (4-side), release = 432 ms (5-side)
- Cymatic mode: visualises Chladni nodes for a circular plate at 960 Hz
- Use: root drone for any PUM piece. The hidden tempo is the precession rate.
Patch B — “Vajra basket-weave”
- 15 voice pairs (30 osc) on the 16-EDO scale (75 cents per step, from
pum_tuning/pum_16sector.scl) - Outer 15 phase-locked CW; inner 15 anti-phase CCW (opposite twists)
- 5:1 amplitude ratio outer:inner (volume-ratio identity)
- Quadrature panner active
- Use: texture pad for the central torus motion. The 5:1 imbalance creates a self-stabilising vortex in the stereo field.
Patch C — “Element resonator”
- One voice per noble gas at its
Vvalue scaled into audible range - H = 32,400 / 16 = 2025 Hz
- He = 3,136 / 4 = 784 Hz
- Ne = 147,456 / 64 = 2304 Hz
- Ar = 82,944 / 32 = 2592 Hz (= GREAT_YEAR / 10)
- Kr = 19,600 / 16 = 1225 Hz
- Xe = 64 → 64 Hz (no octave shift)
- Rn = 1,225 → 1225 Hz (= Kr)
- Triggering each voice runs the right-triangle cascade as a glissando through the time-cascade decimals
- Use: play the periodic table directly. Bendall’s elemental V table becomes a custom tuning per element.
Patch D — “Sanskrit Time”
- 311,040 Hz divided 11.111 → 27,993.6 → fold into 27.99 Hz (rumble)
- Ring-mod against RFEU_RING (1,296 Hz)
- 16-sector Sequencer running at 22.5° per step at the Schwabe tempo (11.111-second loop)
- Use: slow-evolving textural piece keyed to the Aether→Matter conversion factor
4. Hardware build (Eurorack-compatible, ~$400 BOM)
| module | purpose | reference part / DIY |
|---|---|---|
| PUM Master Clock | Daisy Seed + custom firmware emitting the 11.111 Hz / 0.5184 Hz / 1.296 Hz / 25,920-yr-mapped clocks | Daisy Seed dev board |
| 4×4 Voice Matrix | 16-VCO board, each tunable to a (cardinal × dim) cell | Mutable Plaits clones × 4 |
| 3-4-5 Envelope | Triangle-shape envelope generator with 3 stages whose times are exact ratios | DIY perfboard, TL074 + log/anti-log pair |
| Vajra Phaser | 15-stage all-pass filter cascade with opposite-twist option | DIY Compurhythm clone |
| Quadrature Panner | 4-channel matrix with ±90° phase | Frap Tools 321 |
| Cymatic Plate | 250 mm steel plate driven by transducer at the patch’s root frequency | DIY |
Software-only path: build entirely in Reaper / Bitwig / VCV Rack with the .scl scales loaded from pum_tuning/.
5. Integration with the existing scope3d prototype
~/claude/scope3d/ already does the Lissajous + FFT + Chladni rendering. Concrete extensions to ship:
- Load the new pum_core symbols as preset banks. The 58 catalogued symbols become 58 quick-recall presets, each a frequency × dimension address.
- Map every patch (A-D above) to a button in the UI. Patches encode envelope + modulation + render mode in one click.
- Hook the Hydrogen cascade as an automatic filter sweep — when a note is played, an 8-stage filter cascade with cutoff times
[24, 60, 60, 60, 60, 60, 60, 24](in subdivisions of the master clock) runs through the cascade values, ending at TIME or LIGHT. This makes the H story audible. - Cymatics export — render Chladni node patterns at each PUM cardinal frequency to PNG/SVG (this becomes the bridge to the materials section below).
- Microtonal Scala load — drop in
pum_dimension.sclandpum_16sector.sclfrompum_tuning/. UI shows note → cents from C = 800/3 Hz. - ext-clock: accept a master clock tap and lock all PUM-derived rates to whole-number ratios of it. Lets the synth sync to a DAW or to Sheela’s existing performance setup.
6. Why this is more than a tuning system
A normal microtonal synth lets you load arbitrary scales. A PUM synth is constrained: every parameter must be derivable from {AETHER, SUN, TIME, MATTER, dim_fraction(n)} via {×, ÷, ±} on small rationals. That’s a strong constraint and it produces a recognisable sonic signature: sounds at PUM addresses rationally interlock with each other across all axes simultaneously (pitch ratio, envelope time, modulation rate, panning angle).
In Rahman’s framing this is what “coherent frequency” means — not a single Hz value but a whole geometry of relationships that are each others’ integer multiples in some basis. A PUM patch sounds like one thing, not eight overlaid things, because every component shares the basis.