Cross-Pollination

Player uses a magnetic wand at a Pollination Cabinet to touch 9 numbered specimen boxes across 5 trials: a pair, a set of 3, three sequential single stages, three donor-recipient crosses, and a fixed 4-step sequence. Completing all 5 trials triggers a full cabinet light payoff and one of 3 Hybrid Index endings, picked at random and announced by audio — no physical hybrid is built in the cabinet itself. Players then go to the off-cabinet examination table and Field Guide to continue. See the Player Test Guide for the exact box sequence.

Each of the six cabinets is autonomous. Five run the identical standard firmware and SD card. Unit 6 is the exception: its boxes were photographed on 2026-10-07 and its clues were rewritten to point at what is really in its boxes, so it has its own answer path, its own build (1.10.1-u6) and its own SD card. Networking (the ESP32 show-network modem) is additive and never required for local gameplay.

Cross-Pollination cabinet.

Source: platform/tools/cross_pollination/cross_pollination.ino (it replaced the old mega_dumb_slave.ino in late September; the audio still lives under platform/tools/mega_dumb_slave/audio/). Builds and flashing go through Understudy's firmware list, never by hand.

Latest firmware

What changed since the first bench build (plain language)

• Sensors you can trust (1.7.3–1.7.4): a touch has to hold for 50 ms over 3+ readings before it counts, so electrical glitches no longer play the game by themselves. The hall pins use the Mega's pull-up, so a loose sensor wire shows as a dead box instead of phantom touches. Every 10 s the unit prints how many glitches it filtered per box.
• Reset button (1.7.0, Mega pin 12): tap = reset the game; hold 2–6 s and release = box test on/off; hold 6 s = photo mode.
• Idle timeout: 60 s mid-game with no touch and no narration playing gives a warning (lights flash), then the game resets 30 s later.
• Attract look (1.9.2–1.9.3): at idle every square glows its own color (no two neighbours alike) with occasional tube-style flickers and slow color changes.
• Game look (1.9.5, 1.9.7): during the trials each box uses its own color; boxes nobody has picked are off, so a lit box clearly means something.
• Finale (1.9.4): a slow color glide across all nine boxes for the whole finale audio, then two white flashes, then back to idle.
• Show standby (1.8.0+): when Understudy sends standby (night mode / blackout) the cabinet stops the game and goes completely dark; resume returns it to a fresh idle.
• Photo mode (1.9.8–1.9.9): hold the reset button 6 s: every square full white, blacklights stay on, game and audio frozen, so each unit's real box colors can be photographed. Tap the button to leave.
• Trial 4 opening (1.10.0): on "Now examine the other specimens" the other six stay lit with the three broad leaves until "First, find the broad leaf" narrows back to the three.
• No accidental re-select (1.10.1): after a box is picked it ignores the wand for 1.5 s and until the wand has been lifted off it for a full second. Other boxes answer at once.

Supplier part numbers are Placeholder until sourcing is locked. Components and quantities below are real (from the working spec).

Mega pin map (checked against cross_pollination.ino 1.10.1, 2026-10-12)

Hall sensor / box numbering (real — 9 digital sensors, one per box)

Box numbering: front view, top-left = box 1, reading left-to-right then top-to-bottom — box 1 is always the lowest Mega pin number.

Box 1 (pin 22) | Box 2 (pin 24) | Box 3 (pin 26)
Box 4 (pin 28) | Box 5 (pin 30) | Box 6 (pin 32)
Box 7 (pin 34) | Box 8 (pin 36) | Box 9 (pin 38)

RGB: three data chains, one per box-row (1–3 / 4–6 / 7–9), each chain a 3-panel daisy-chain on its own Mega pin; distribute 5 V in parallel — never rely on matrix-to-matrix power transfer. Common ground with the controller.

The firmware reverses all 3 chains: matrixChainReversed[] = {true, true, true}. For every chain, wire the daisy-chain with the higher-numbered box's panel first and the lower-numbered box's panel last (e.g. chain 0: box 3's panel first, box 1's panel last; the middle box in each chain — 2, 5, 8 — is unaffected either way).

Real provenance, so you know how much to trust each one: chain 0 (boxes 1–3) and chain 2 (boxes 7–9) were directly, live-verified on 2026-09-25 — a solid-color test per panel with Zach reading off which physical box lit which color, confirming the swap in both cases. Chain 1 (boxes 4–6) was not independently tested (wasn't lit during that session) — it's reversed in firmware only by inference from the identical hardware/wiring pattern on the other two chains. If you're wiring boxes 4–6, this is the one to double-check with the same direct per-panel color test before trusting it blind.

Separately, if a panel doesn't light at all (not wrong box, just nothing): before suspecting firmware, check the panel's own DIN/DOUT — this exact symptom on this exact project was once traced to DIN/DOUT wired backwards on a hand-soldered panel (data going into the panel's OUTPUT pin, not its input). The two pins often look identical on cheap panel silkscreen; worth confirming with the panel's own datasheet/labeling before assuming a deeper problem.

Audio output jack (5-pin 3.5mm breakout)

Real, bench-confirmed: DFPlayer's audio output → jack pin 2 only. No separate ground wire needed — the speaker gets ground through its own USB power connection. Confirmed working live, 2026-09-25.

Real 5-trial state machine. Full trial-by-trial box sequence and answers (standard cabinets and unit 6): Player Test Guide.

State machine

IDLE (attract look) → TRIAL_1 (pair) → TRIAL_2 (set of 3) → TRIAL_3 (stages A → B)
→ TRIAL_4 (3× donor→recipient) → TRIAL_5 (fixed 4-step sequence)
→ FINALE → HYBRID_INDEX (random 1 of 3) → IDLE

• Idle (attract): blacklights on, every square glowing its own color with tube-style flickers. The first valid wand touch on any box starts play (that touch is not an answer); the squares flicker on in twitching colors and Trial 1 begins.
• During the trials, unpicked boxes are off; the narration blinks the boxes it talks about (timed to the words), and a correct answer flashes only the correct boxes (square and blacklight together). Blacklights stay on.
• Each trial has its own correct/incorrect cues and a wrong touch never clears progress already made. Trial 3 is two stages (A, B); Stage C was removed for good in 1.9.0 because its lines were never recorded.
• Trial 5's 4th correct touch starts the finale: a color glide across all nine boxes during the finale audio, two white flashes, then one of 3 Hybrid Index endings at random, announced by audio. Then back to idle.
• Idle timeout, reset button, show standby and photo mode: see Firmware above.

The full light rules Zach set on site (and any rebuild of the light logic must keep) are in GAME-INFO.md. Source of truth: platform/tools/cross_pollination/cross_pollination.ino.

Folders on the show SD card (DFPlayer numbering). The script of every clip: platform/tools/mega_dumb_slave/audio/sd-card-root/AUDIO_SCRIPT.md; the per-track mapping: TRIAL_AUDIO_MAPPING.md in the same folder.

Official SD cards

Five official standard cards (#1–#5, made 2026-09-30 to 2026-10-03) were written with provision_sd_card.sh and every file checked against the repo (57 files each). The register is platform/tools/cross_pollination/official-sd-cards.md. A card without folder 01 or 10 is an old-layout card: the box test will be silent on it.

Unit 6 needs its own card: Understudy audio set cross-pollination-sd-u6, version u6-2026-10-09-lavender (clues re-voiced for what is really in its boxes, e.g. "orange fronds", the lavender in box 6). A standard card in unit 6, or the unit-6 card in a standard cabinet, gives clues that don't match the boxes.

1. Flash every unit to 1.10.1 (unit 6: "Flash other firmware" → 1.10.1-u6) through Understudy.
2. Photo-mode photos of each unit, so the other cabinets' clues and colors can be checked against their real boxes the way unit 6 was.
3. Target stickers over the hall sensors: options A–E drafted in print/targets/ (compare-v2.html), not yet chosen; the narration may then mention the target.
4. Ceiling lights above the cabinets (Zach's 2026-10-10 list).
5. Decide the Welcome clip's (folder 08) trigger point; record the real idle-warning clip (folder 10).
6. Confirm the boxes 4–6 chain order with a per-panel color test (inferred, see Wiring).

Open design item: the analog Seed Station second phase (Field Guide, worksheet, capsule) is not yet specified electronically.

Photos of unit 6's nine boxes and the script review against them: boxes/README.md.