Two-sensor topology modes

Two-sensor topology modes

_Five distinct ways to deploy a Mega + Bolt pair. Ranked by recommendation,
with CAD impact, software impact, calibration effort, BOM cost per unit
shipped, and aesthetic readout._

Mode 1 — Sibling units (Bolt-bar = production, Mega-bar = development)

TL;DR — recommended primary path. Two complete bars, identical except
for the front camera module. Mega-bar stays as Mac4 dev rig. Bolt-bar
becomes the production-aesthetic flagship.

### Geometry
- Bar 1 (Mega): existing 500×120×85mm bar with current Mega cradle.
Pod attached. K11 in pod. Wall-mount via pod VESA.
- Bar 2 (Bolt): NEW slim bar at 500×120×60mm. Bolt cradle redesigned for
39mm depth + USB-C only. Pod attached. K11 in pod. Wall-mount via pod
VESA.

### CAD impact
- Net new Bolt-cradle module in `lume-shell.scad`. About 60 lines of SCAD.
- New `lume-config-bolt.scad` overriding LUME_DEPTH=60, FEMTO_W=140 etc.
- Bolt-bar shell halves regenerate at smaller dims (300×30×120, 200×30×120).
- Pod stays identical between the two bars (K11 same, dims same).
- Estimated effort: 1 day SCAD + half-day STL regen + half-day reslice
in OrcaSlicer.

### Software impact
- Zero. Both bars run the same wire format, same Unity components,
same software. Two LUME bars = two unrelated installations from the
software's perspective. Each has its own Mac running Unity, its own
publisher, its own audio sidecar, its own user.

### Calibration
- Each bar calibrates independently via the F12 LumeCalibrationPanel.
K11/sensor pairs calibrate once, persist to JSON, never need to talk
to each other.

### BOM per unit
- Bar 1 (Mega): existing BOM (~$1,100 Mega + ~$900 K11 + plastic + screws)
- Bar 2 (Bolt): ~$500 Bolt + ~$900 K11 + plastic + screws (~$1,400 total)
- Bolt-bar saves ~$700 per unit vs Mega-bar.

### Aesthetic readout
- Two distinct product silhouettes. Bolt-bar is 30
reads as the "Pro" or "production" SKU. Mega-bar reads as the "dev" SKU.
- Marketing-friendly: "Bolt is what we ship to customers. Mega is what we
develop on."

### Open questions
- Do we sell both? Or is Mega-bar internal only and Bolt-bar the only
customer-facing SKU?
- Does the Bolt-bar pod get smaller too (no Mega Jetson hump to accommodate)?
POD_D could drop from 130mm to 100mm.

Mode 2 — Twin-sensor single bar (wide-angle stereo fusion)

TL;DR — most ambitious. Both sensors in one bar at angles. Software
fuses depth fields for 120°+ horizontal capture. Maximum capability,
maximum complexity.

### Geometry
- Single 500×120×85mm bar (existing dimensions or slightly wider).
- Mega centered, NFOV (65°) facing dead front.
- Bolt offset to one side, angled outward 30°, providing wide-angle
context for the periphery the Mega misses.
- Both sensors mounted in the front half. Pod for K11 in rear.
- Or alternative: symmetric pair, two Bolts (or Mega+Bolt) at ±30°,
each covering half the room.

### CAD impact
- Front shell needs a second sensor cutout + cradle.
- Cable routing for two USB feeds (Bolt USB-C to K11 USB4, Mega Ethernet
to K11 RJ45 — different ports, no contention).
- Bar may need to widen to 550-600mm to fit both with proper FOV separation.
- Estimated effort: 2 days SCAD + 1 day prototype print.

### Software impact
- Significant. Two simultaneous depth feeds need separate identification.
Options:
- Two ports: Mega on :9700, Bolt on :9710. Existing wire format unchanged.
LumeUdpReceiver instantiates twice with different port bindings, two
LumeDepthReprojector instances, two _FlowField textures.
- One port, device_id field added to LUMD header (breaks wire format
backward compat, requires receiver patch).
- Unity-side fusion: two depth textures need to be combined into one
point cloud. Either screen-space stitching (project each into world
space and append, dedupe overlapping points) or proper sensor calibration
(extrinsic transform between Mega and Bolt frames) for a single fused
cloud.
- New SCAD config to capture extrinsic between sensor 1 and sensor 2.
- Estimated effort: 3-5 days software + 1 day calibration tooling.

### Calibration
- Major. Need a calibration target (chessboard or ChArUco) visible to
both sensors simultaneously. Solve for the rigid transform between
sensor 1's frame and sensor 2's frame. Persist as
`software/demo/femto_extrinsics.json`.
- Recalibrate on every install (since the sensors are mounted at angles
in a shared chassis, they should be rigid relative to each other,
but bar warpage / shipping shock can shift them).

### BOM per unit
- ~$1,100 Mega + ~$500 Bolt + ~$900 K11 = $2,500 per bar (vs $1,400 Bolt-only).
- 78

### Aesthetic readout
- One bar, more visible "tech" via twin lens cutouts. Reads as "Pro"
or "Plus" tier above the single-Bolt baseline.
- Looks more like a media-bar with security camera DNA — could be a plus
or minus depending on positioning.

### Open questions
- Is the wide-FOV gain visually meaningful for LUME's use case? The
particle/fluid effects render in a single screen-space, so wider depth
capture mostly helps with multi-person scenarios.
- Do we need two Megas (no audio sidecar workaround) or Mega+Bolt (mixed)?

Mode 3 — Front+rear single bar (depth front, spatial-context rear)

TL;DR — niche but interesting. One sensor faces the user (depth),
the other faces back toward the wall/space (spatial awareness).

### Geometry
- Mega on front (depth/RGB user-facing).
- Bolt on REAR of bar pointing backward — captures wall behind, ambient
light, room geometry.
- Could feed into "context-aware" effects: brightness adapts to room
light, particle dispersion adapts to wall distance, etc.

### CAD impact
- Modest. One additional rear-facing cradle. Adds ~30mm to bar depth
for the Bolt, OR mount Bolt OUTSIDE bar on the pod's rear face.

### Software impact
- Unity needs a second LumeUdpReceiver+Reprojector instance for the rear
feed. The rear feed doesn't drive particle position — it drives
ambient parameters: rear-distance to a "wall proximity" scalar,
rear-RGB-mean to a "room ambient color" Vector4, rear-depth-variance
to a "room complexity" indicator. ~2 days software.

### Calibration
- Minimal. Rear sensor calibration only matters for room context, not
for fused capture.

### BOM
- $2,500 per unit (same as Mode 2).

### Aesthetic readout
- Hidden second sensor — invisible from front. Marketing story:
"the bar knows the room around it."

### Open questions
- Is "spatial context" a feature people will pay for? Probably not as
a launch SKU — useful for Wave 7+ ambient-aware modes.

Mode 4 — Interchangeable cradle (one bar, two front modules)

TL;DR — modular cradle system. Same bar accepts either Mega or Bolt
via a swappable front insert. Customer chooses at order time.

### Geometry
- Bar built around the LARGER sensor envelope (85mm depth for Mega).
- Front shell has a removable cradle plate held by 4 M3 screws.
- Two cradle plates manufactured: `cradle_mega.stl`, `cradle_bolt.stl`.
- Customer or installer swaps cradle to match sensor.

### CAD impact
- Modest. Add a removable-cradle interface to existing front shell.
- Net new: cradle_mega.stl, cradle_bolt.stl. About a day.
- Bar shell stays at 85mm depth (sized for Mega) — Bolt fans get extra
airflow space, no harm.

### Software impact
- Zero. Single sensor at a time, current pipeline.

### Calibration
- Per-sensor on cradle swap. Calibration JSON keyed to sensor type so
swapping back to Mega restores the prior calibration without re-running.

### BOM
- One bar, one sensor, choose at order:
- Bolt config: ~$1,400 per unit
- Mega config: ~$2,100 per unit (the more-capable one but pricier)
- Customer can upgrade or downgrade post-purchase.

### Aesthetic readout
- One product, one marketing line. Two cradle SKUs. Reads as
"future-proof" — can swap to next-gen sensor without replacing the bar.

### Open questions
- Does the bar's 85mm depth read as "too deep" once Bolt becomes the
default? Customers may expect a thinner bar.
- Print queue grows by 2 cradle SKUs.

Mode 5 — Two bars, master/replica pair (stereo install)

TL;DR — two physically separate bars on opposite walls of a room.
Each captures its own field of view. Software treats them as two
independent installations OR fuses for "everywhere in the room" coverage.

### Geometry
- Two complete bars (could be Bolt-bar each, or Mega+Bolt mix).
- One mounts on east wall, one on west.
- Each runs its own K11, its own Unity, its own audio.

### CAD impact
- Same as Mode 1.

### Software impact
- Each bar runs independently for the simple case. Mesh networking
between K11s via Tailscale for the optional fused-room mode.
- Optional Wave 8+ feature: shared particle field across both bars.

### Calibration
- Per-bar independent.

### BOM
- 2× whatever per-bar SKU we sell.

### Aesthetic readout
- For commercial installs (gallery, retail, restaurant), this is the
"Duncan HOLOVIS at-home" story — multiple bars enclose a room.

Recommendation matrix

Strong recommendation: Mode 1 + Mode 4 in parallel. Ship the Bolt-bar
as a complete product (Mode 1), AND make the cradle interface modular so
the Mega-bar uses a Mega cradle and the Bolt-bar uses a Bolt cradle that
share most of the front shell geometry (Mode 4). One CAD pass produces
both bar SKUs because the shell is identical except for the cradle plate.
Mode 2/3/5 are post-launch ladder rungs.