Spad Electronics Board — Technical Summary

No public Spad-specific repo found. Lineage traced:

• MIDIsprout v0.21 (github.com/electricityforprogress/MIDIsprout) — base sensing/MIDI code, MIT licensed, no scale/menu selection.
• Sam Cusumano's “shield” firmware v026 (github.com/electricityforprogress/BiodataSonificationBreadboardKit) — adds button+knob menu system for live Threshold, Scale, Channel, LED Brightness selection.
• Leetronics fork (https://github.com/leetronics/midi-biodata) — ports v026's menu features onto the classic MIDI Sprout board layout.
• Spad Electronics — per third-party comparison (Gearspace forum), described as same hardware/software as the Leetronics version.

Two leaf/skin probe electrodes wired in series with a fixed 100K resistor (R2), into a 555 timer in astable configuration, with a fixed .0042µF timing capacitor (C1).

• Leaf resistance is the only variable in the RC network — this measures resistance (ionic conductance through leaf fluid), not capacitance.
• Higher leaf resistance → longer 555 oscillation period.
• Lower resistance (more conductive) → shorter period.
• Contrast this with capacitive sensing (SFCS/touch, digital-pin capsense) which fixes R and varies C instead

• 328p MCU interrupt fires on every rising edge of the 555 output (attachInterrupt(..., RISING)) — one timestamp per oscillation period.
• 10 consecutive period values collected into an array per cycle.

Once the array fills, computed over the sample window:

• averg — mean of the samples
• delta — max minus min (spread across the window)
• stdevi — standard deviation of the samples

Trigger condition (does a note fire at all):

''delta > stdevi * threshold''  (threshold default 2.3)

• Higher average resistance → higher pitch (not lower).
• %127 folds unbounded microsecond values into MIDI's 7-bit range — a modulo wrap, not a calibrated scale. Monotonic only within each 127µs wrap-band, not globally across the full sensor range.

Identical (line-for-line, confirmed by direct diff):

• sample() — byte-for-byte identical
• analyzeSample() — identical except one substitution (below)
• setNote(), setControl(), checkControl(), checkNote(), MIDIpanic(), midiSerial()
• readVcc(), checkBattery(), bootLightshow(), mapfloat(), freeRAM(), rampUp(), rampDown(), checkLED()
• scaleNote(), scaleSearch()
• Core constants: noteMin=36, noteMax=96, samplesize=10, threshMin/Max = 1.61/3.71, batteryLimit=3000

Added in v026 (UI/config layer only, sensing math untouched):

• Scale selection: v0.21 uses scaleSelect (one fixed array, compile-time only). v026 uses scale[currScale] (5 arrays: Chromatic/Major/DiaMinor/Indian/Minor), runtime-selectable via menu.
• Full button+knob menu system (checkMenu(), thresholdMode(), scaleMode(), channelMode(), brightnessMode()) — v0.21 has only empty stubs (checkButton(), knobMode()).
• Threshold: v0.21 sets continuously every loop directly from the knob. v026 only adjusts threshold inside the menu; persists after exit.
• MIDI channel: fixed constant in v0.21; runtime-adjustable in v026.
• LED brightness: fixed in v0.21; runtime-adjustable in v026.
• EEPROMex.h included in v026 but no read/write calls found — likely unused/vestigial.

Conclusion: v026 (and by extension Leetronics/Spad) = v0.21's sensing/analysis core, unmodified, with a menu/config layer bolted on top.

• Passive — no injected signal, no RC network, no 555.
• Reads the plant's own native millivolt-scale bioelectric activity (action potentials, slow-wave potentials) via AC-coupled high-gain instrumentation amplifier (e.g. AD8232).
• Electrode damage/longevity problem with penetrating electrodes → SWCNT nanofilm electrodes identified as best non-invasive option (2+ months stable, documented light-induced biopotential response).