Program file format

Every program is one JSON file in presets/. The web app lists them, the brain view draws them, and the app sends the same JSON to the helmet over Bluetooth, where the firmware's own parser runs it. The JavaScript engine (webapp/src/engine) and the C++ engine (firmware/lib/engine) are tested against each other sample by sample.

Check a file before committing it:

cd webapp
node scripts/check-preset.mjs ../presets/my-program.json   # one file
node scripts/check-preset.mjs --all                         # every preset

Example

{
  "id": "obe-shiva-right-tpj",
  "name": "OBE: right angular gyrus (Shiva sites)",
  "category": "obe",
  "evidence": "anecdote",
  "origin": "Todd Murphy, 8 Coil Shakti OBE session",
  "uT": 1.0,
  "summary": "One or two sentences: what it is meant to do and where it comes from.",
  "notes": "Setting, timing, cautions; what is approximated and why.",
  "targets": ["right-angular-gyrus", "right-tpj"],
  "source": "Murphy, Shakti OBE page https://www.god-helmet.com/... ; Blanke et al. 2002, Nature 419:269",
  "layout": "custom",
  "placement": { "sites": ["FT7", "C5", "TP7", "T9", "FT8", "C6", "TP8", "T10"] },
  "fadeInS": 10,
  "fadeOutS": 15,
  "phases": [
    { "name": "Both sides, theta", "minutes": 10,
      "pattern": { "type": "sine", "f0": 7, "pointMs": 1 },
      "spatial": { "mode": "bilateral" } },
    { "name": "Right side stepping", "minutes": 20,
      "pattern": { "type": "pm", "f0": 40, "f1": 0.5, "depth": 3 },
      "spatial": { "mode": "cycle", "side": "right", "dir": "ccw", "dwellMs": 20, "weights": [0, 0, 0, 0, 0.6, 0.8, 1, 1] } }
  ]
}

Description fields

Placement

layout is one of:

• "temporal" (default): CH1–4 left FT7, T7, CP5, TP7; CH5–8 right FT8, T8, CP6, TP8.
• "ring": CH1–8 Fpz, F7, T7, P7, Oz, P8, T8, F8 (counter-clockwise seen from above).
• "custom": any 10-10 or 10-5 sites, given in placement:

"placement": { "sites": ["Fp1", "FT7", "TP7", "O1", "O2", "TP8", "FT8", "Fp2"], "order": [0, 1, 2, 3, 4, 5, 6, 7] }

• sites: 8 entries, one per channel; "" leaves a channel unused (it stays silent). Names follow the 10-5 system exactly (FT7, Cz, TP8, T10, FTT7h, ...); the checker rejects names that are not in the 10-05 set.
• Side comes from the name: odd number = left, even = right, z = midline. Midline coils take part in rotate and bilateral, not in left, right, cycle or alternate.
• order: the counter-clockwise rotation order (seen from above) for rotate and cycle; default is channel order. List only used channels.

Phases

Up to 12 phases, 60 minutes in total.

fadeInS and fadeOutS (default 10 s) ramp the whole session.

Signals (pattern)

All signals are s(t) in −1…1, multiplied by the coil gain, the phase level and the session intensity. pointMs holds each value for that long (sample-and-hold, like the Persinger/Koren point files); 0 = smooth. The sequencer ticks at 2 kHz, so 0.5 ms is the finest step.

Frequencies are 0–500 Hz, pointMs 0–1000, depth 0–20 rad.

Exact pulse tables

webapp/src/engine/generators.js builds published pulse protocols as tables; print one ready to paste:

cd webapp
node scripts/table.mjs list
node scripts/table.mjs itbs          # theta-burst, Huang 2005 timing
node scripts/table.mjs bhfs '{"hz": 6.55}'
node scripts/table.mjs cnp --meta    # US 6,234,953 Table 1 (Thomas/Prato/Persinger)

These copy the timing of clinical and laboratory protocols. Their field strengths were 100–100,000× higher than this helmet's; the docs say so wherever they are used.

Coil movement (spatial)

Dwell (time per coil): "dwellMs": 20, or a ramp "dwell": [start, step, end] in ms, e.g. [10, 2, 20] (Persinger's 10 → 20 ms in 2-ms steps) or [25, -2, 11] (Tsang 2004). crossfade: true blends neighbouring coils instead of switching. weights: 8 per-channel multipliers −1…1 for any mode; a negative weight drives that coil with reversed polarity (the 8 Coil Shakti wires each pair as one + and one − coil).

Persinger-lab notation. The lab wrote rotation schedules as base and change per solenoid, and the sources disagree on the sign. Persinger & Lavallee 2010 (and Lehman 2016) define "20+2" as 20, 18, …, 6 ms and "20−2" as 20, 22, …, 34 ms, and Tsang 2004 is 25 → 11 ms ([25, -2, 11]). The Koren software patent (US 6,312,376) and the configuration table of the 1999–2008 papers (e.g. Swann 2002 Table 1: 20/+2, 100/+20, 200/−2) read more naturally with the plain sign: the +20 row only works as an increase, and Cook 1999 gives rotation times of about 200–2000 ms. This project uses the authors' explicit definition where a paper gives one (2010 photon and yoked studies, Tsang 2004, Lehman 2016) and the plain sign for the configuration table ([20, 2, 34], [100, 20, 240], [200, -2, 186]); each program's notes say which reading it uses. Cook 1999 and Richards 2002 describe 20-ms changes per solenoid, not a fixed 20-ms dwell.

Persinger-lab signals as published. Burst-firing ("burstx"): about 230 points at 3 ms (0.69 s; 248 and 289 points in other theses) once every 3–4 s, point values unpublished: {"type": "burst", "sweepMs": 690, "gapMs": 3310, "pointMs": 3}. Thomas pulse: the exact table at 3 ms (node scripts/table.mjs thomas). LTP: a primed burst (primed). 7 Hz sine with 1 or 10 ms points. The lab's 8-bit values map 0…255 to −5…+5 V (patent: ±6 V).

Limits

• 12 phases, 60 min, at most 2 tracks per phase.
• Tables: 8192 points each, 16384 per program; the JSON must stay under 96 KB (Bluetooth request limit).
• Intensity: the app caps the program intensity at the calibrated maximum (hardware cap 4.54 mA per coil).