Technique (John Phillips / Instructables): pause print after layer 3 (~0.6mm), embed copper tape, resume — plastic fully encapsulates it.

Copper tape size: tested at 20mm × 5mm. Larger = more sensitive, more noise.
Filament: PETG recommended over PLA — handles soldering heat near the access hole.
Solder access hole: 3mm × 5mm extruded to layer 3, seal with hot glue after.
Layer 4-5: print at 30–40mm/s, 100% solid infill to encapsulate cleanly.
Wire routing: keep short. Avoid routing near PWM, switching regulators, WS2812 data lines.
Result: invisible, durable, professional touch surface inside any 3D printed part.

Integer EMA formula: ema = ema - (ema >> shift) + (raw >> shift)
No floats, very fast. α = 1/2^shift.

shift=1 → α=0.5, very responsive, less smooth
shift=2 → α=0.25, responsive (default)
shift=3 → α=0.125, smooth, some lag
shift=4+ → α≤0.0625, very smooth, slow to react

Hysteresis: separate press/release thresholds prevent chattering at the boundary. Always set press > release (ESP32-S3 values go UP on touch).

Note: original library written for ESP32 original (values go DOWN on touch). Logic is inverted for S3 — press threshold must be HIGHER than release.

Observations so far (March 14 2026 — desk, loose wires, no enclosure):
• Insulated wire more sensitive than bare pin tip (more contact area + dielectric)
• GND guard removal seemed to have minimal effect — but only one test, one environment
• Baseline ~22,800–23,400 on this board (lower than spec'd 30k–60k)

Need more data: different electrode types, enclosure vs open air, different humidity days, grounded vs floating setup, long wire vs short wire.