WaveForge · the buoy demonstrator · concept fixed 2026-09-17

Where This Is Going

Everything on this site so far is a board, a sensor or a bracket. This is the thing they are all parts of — the target for the next few months.

The demonstrator

Concept render: sprung buoy hull inside a braced hexagonal frame, vertical shaft carrying an eccentric mass on a linear carriage — ⚠ **CONCEPT RENDER — AI-generated, not a photograph.** Nothing in this picture has been built. It is an accurate drawing of the intended architecture and nothing more.
Ninth iteration, 2026-09-17. Six corner floats (small waterplane → quiet reference), inner sphere with a sealed dome and no rotary seal, thrusters submerged on outriggers, glazed bays, pitched solar roof.

Flotation upgrade — the fender ring

Concept render: the same buoy with one continuous segmented hexagonal float ring acting as both flotation and fender — ⚠ **CONCEPT RENDER — AI-generated, not a photograph.** Under consideration, 2026-09-17. One continuous segmented float ring in place of six corner floats.

Why it is probably better. It is flotation and fender in one part. The ring stands proud of the frame on every face, so nothing — dock, boat, log, trailer — can strike the aluminum without hitting the float first. The frame’s corners, which are what would otherwise meet a dock, no longer exist as corners.

Segmented into six, one per bay, with bolted flanges. That keeps the frame modular and flat-packable, and a damaged segment is replaceable without rebuilding the float.

Motion	What a big ring does	Does it matter?
Heave	Large waterplane → the frame follows the wave up and down strongly, where six small floats would not	Little. Frame heave relative to the inner body mostly just works the springs vertically. It is not what drives the mass.
Tilt	Wide footprint → the frame stays comparatively level while the small inner sphere follows the local wave slope and heels	Yes — this is the one. Relative tilt is the power input, so a tilt-stiff frame increases output. The ring may be the better choice for this machine, not merely an acceptable one.

⚠ A scaling trap that must be written down before any number is taken. A wide frame only averages the wave slope if it spans a meaningful fraction of the wavelength. In the pool, with wavelengths of a metre or two, a 1.2 m ring genuinely averages — it sits flat while the sphere rocks, and the machine will look excellent. In the ocean, at 50–100 m wavelengths, that same ring is a speck on the slope and tilts with it exactly as the sphere does. So the demonstrator can show strong relative tilt arising from a geometry effect that does not survive the scale-up. Record relative tilt against frame diameter ÷ wavelength, not against float size alone, or the bench number is an artifact. → ledger C22.

⭐ ARCHITECTURE CHANGE — 2026-09-28

The sealed inner hull is gone. No dome, no torus float, no springs, no bumpers, and — the headline — no part of the mechanism is under water. A frame on floats, a vertical shaft at its apex, and a rail cantilevered off that shaft with the working mass hanging from it. Everything in air.

Compare that to what it replaces: an acrylic dome, a sprung inner body, a rocking hull needing clearance and bumpers, and a sealed enclosure to keep water out. This is less to build in every dimension, and there is nothing left to keep watertight.

The machine, 2026-09-28

Amber rotates; teal is fixed. ⭐ The bearing at the base carries the weight and the thrust; the guyed swivel at the top only holds the shaft centred. Two supports a whole machine-height apart turn the eccentric mass’s bending moment into two small opposing loads — the best geometry available, and better than the cantilever we had drawn.

What it is

A buoy that converts a passing swell into electricity, and that can be tuned to the sea it is sitting in. The frame follows the wave; gravity acts on an off-center mass; the mass turns a shaft.

An outer frame on floats, aluminum extrusion, with rigid A-frame members rising to an apex. Small waterplane at the floats keeps it a comparatively quiet reference.
A vertical shaft at the apex, in TWO pillow blocks. ⚠️ Not a swivel and not a cable sling. A cantilevered rail with an off-center mass puts a bending moment into that joint, and a single-point bearing cannot restrain a moment — it would simply tilt until the center of mass hung directly beneath, which destroys the eccentricity the machine runs on. Two bearings, spaced, turn that moment into two opposing radial loads. Same lesson as the pendulum arm on the shake table, stood on end.
⭐ The battery IS the mass. It hangs from a linear rail cantilevered off the shaft. That weight has to be carried regardless — making it the thing that drives the generator means the mass does double duty instead of being ballast.
⭐⭐ A second, smaller counterweight rides the same rail, positioned by a stepper and belt. Two masses on one rail give two tuning variables, not one: moving the counterweight changes the net eccentricity (how hard the wave drives it) and the moment of inertia (its natural period) — and they do not move together. With a single mass you could only trade them off.
Everything electrical rides on the rail — generator, battery, controller, stepper, sensors. The generator charges the battery, the battery runs the controller, the controller drives the stepper. A closed loop on the rotating side.
The frame side is its own closed loop — solar, thrusters, their own pack, their own ESP32. ⭐ Data crosses by ESP-NOW, not copper. Same pattern as the two-ESP32 shake table.
Two wires through a through-bore slip ring for the one link that must be physical. Even a single wire crossing a continuously rotating joint needs one, or it winds up and tears out. ⚠️ Do not let the slip ring's 12.7 mm bore size the structural shaft — put it on a stub extension above the top bearing where it carries no load, and size the shaft for the bending it actually sees. Parallel circuits for current: 6 wires gives 3× capacity on two conductors.

⭐ Why the split keeps the measurement honest. What the generator makes charges a battery nothing else touches, so it can never be confused with what the sun made. C18 stays clean by construction rather than by discipline. ⚠️ The consequence to accept: the rail battery is charged by the generator or by hand between runs — which is also the cleanest possible proof the machine runs its own brain off its own output.

⚠️ The float guidance REVERSES with this architecture

The 2026-09-17 note below concluded that a small waterplane makes the frame a quiet reference, and that the difference between a still frame and a wave-following inner hull was the power. That was correct for the two-body machine. It is wrong for this one.

There is no inner body any more. The driving torque is m·g·r·sin θ, where θ is how far the axis has tilted — so tilt IS the input. A frame that averages the wave slope and stays level produces nothing.

⭐ What you now want: the frame should follow the local wave slope as closely as it can. Narrow is good. The 27 in hull is tiny against any wavelength, so it rides the slope rather than averaging it. ⛔ The wide fender ring and the widely spread floats would have been actively wrong here — they were designed to hold the frame level, which is now the one thing that kills the output.
⚠️ Ledger C22 still applies but its sign flips: record relative tilt against frame diameter ÷ wavelength, and expect more tilt from a smaller frame.

⭐ The weight traverses the FULL rail, centred on the shaft

His, 2026-09-28. The rail is centred on the shaft and the weight rides a belt over its whole 32 in length — so it can sit on either side of the axis, or exactly on it.

⭐⭐ Centre is an OFF switch. Weight on the axis = zero eccentricity = no driving torque, no output, and no rotating load on the bearings. That is a storm mode and a transport mode with nothing to latch and no brake to fit — just drive the carriage home. A shutdown that costs no parts.
⭐ It doubles as attitude control. Off-centre one way and the hull leans that way; combined with the rotation, the lean can be pointed anywhere. Useful for trimming out uneven loading, and for deliberately provoking a response during testing rather than waiting for the right wave.
Eccentricity in both directions through zero, rather than one-sided — roughly double the usable range for the same rail.
⚠️ The cost: with one weight traversing through centre, eccentricity and inertia move together again — sliding out increases both. That coupling is exactly what the counterweight existed to break, so keep the counterweight as a hand-set block and retain the option. Ledger C26 is the test that says whether independent control is worth the part.

⭐ OBSERVED — spring orientation makes it circular

His, tested in the wave pool: “it creates a very good circular motion when you line the horizontal springs parallel to the incoming wave.” This is the first real data on this machine — observed, though not yet quantified.

Why it works. Springs parallel to the wave make the system stiffer along that axis than across it. Driven at a frequency between the two resulting natural frequencies, the two axes respond about a quarter cycle apart — and perpendicular motions 90° out of phase trace a circle. ⭐ The asymmetry is the mechanism, not a defect. Symmetric springs would give linear sloshing; the mismatch is what makes it rotate.

⚠️ This may reverse the heading rule. The 2026-09-14 rule says the thrusters should spin the buoy so the springs are PERPENDICULAR to the swell — written for the pendulum-arm machine, where free fore-and-aft pitch in the arm’s swing plane was the goal. This machine is a rotator, and parallel is what was observed to work. Same thrusters, opposite instruction. → ledger C30, to settle before the control logic is written.
⚑ Next step on C29: two MPUs and a log, to put an angle and an amplitude ratio on what is currently an eyeball observation.

Numbers as of 2026-09-28

Quantity	Value	Note
Rail	32 in, 6 lb — uncut	SBR20 at full length. Torque is linear in radius, so the extra inches are free output; 6 lb is 12% of budget for a part carried anyway. Travel set by the SBR20 stop collars already on hand — no cutting hardened shaft.
Hull width	27 in at the widest	Rail exceeds it, as intended.
Payload budget	50 lb to the half-inch mark	⭐ 44 lb left after the rail for battery, gearbox, generator, counterweight, shaft, bearings, stepper, controller and slip ring. Every pound in the moving mass earns power; every pound of structure does not.
Gearbox	10:1, vertical	On hand, same as the shake-table demonstrator. Turns a slow swing into generator RPM with no extra parts and nothing off-axis.

⚑ PLANNED, not measured — first-order estimate at 10° of tilt, battery at 12 in radius, generator at 200 RPM:

Moving mass	Torque at rail	At generator	Power
10 lb	2.3 N·m	0.23 N·m	~5 W
25 lb	5.8 N·m	0.58 N·m	~12 W
40 lb at the full 16 in	15.9 N·m	1.6 N·m	~33 W

⚠️ Every figure above is computed, none measured. The rail’s actual rotation rate is the biggest unknown — an eccentric mass on a near-vertical axis accumulates rotation rather than reversing, so it is not locked to wave period and may run faster than the sea pumping it. That rate is the first number this rig should produce.

What got decided 2026-09-17

The concept went through four versions in an afternoon. Three real design decisions came out of it:

Question	Answer	Why
How is the arm supported?	Two pillow blocks on one shaft, above and below the arm — not a tether from an overhead swivel	A tether only pulls, goes slack at the top of the swing and then snatches. Two bearings make the shaft simply supported. A single pillow block cannot do it — the insert is self-aligning by design and carries no bending moment.
Vertical shaft or horizontal?	Vertical	The shaft exits through the top of the dome, above the waterline. A horizontal shaft would penetrate the hull at the equator, at or below the waterline, where a rotary seal is a genuine leak path. Driven by hull tilt, not by heave.
Where do the thrusters go?	On outriggers, fully submerged	A thruster at the surface ventilates — it draws air down the duct and loses thrust, the failure his 2026-09-14 note called out (axis at least two duct widths under the working waterline). Hanging them below on outriggers fixes that and buys turning moment at the same time.
Does the shaft exit the enclosure?	No — it terminates inside	Kills the rotary seal entirely. Earlier versions ran the shaft up through the top to a bearing bracket, which needed one. Everything now lives inside the sealed dome and only a static cable gland leaves it.
One rigid body or two?	Two — inner body free to rock, sprung to the frame	A single rigid platform would work (that is essentially the Wello arrangement) but it throws away the sprung-mass tuning that the filings rest on. Two bodies keep spring rate as the mode selector.
How does power cross between the two bodies?	It doesn’t — two independent buses, ESP-NOW between them	Any cable spanning a body that rocks continuously is a fatigue failure waiting to happen. Generator and its pack inside the hull; solar, thrusters and their pack on the frame. ⭐ Side benefit: the solar bus and the wave bus are physically separate, so what the generator made can never be confused with what the sun made — C18 stays clean by construction.
How long can the arm be?	About 200 mm of radius	The sphere has to stay inside the sealed dome, so the dome’s diameter caps the arm. The 800 mm rail will be cut down. Costs torque, which scales with radius — but not the ratios this rig exists to measure.

The honest part

The render shows a finished machine in open ocean at golden hour. Here is the actual state of it.

Piece	State
Hull — two bowls, gasket, clamps	On hand and float-tested at the Willamette Park ramp, 2026-09-14. 8 lb per inch of immersion, ~70 lb of reserve. Gasket seeped at the end of the test — going to thicker neoprene and eight clamps.
Generator, gearbox, couplers, pillow blocks	On hand. The printed pillow-block housing was the one missing part.
Rail, carriage, stepper, belt	On hand. SBR20 rails and blocks, NEMA 17, GT2 kit.
Thrusters, sonar, lidar, MPUs, battery packs	On hand.
Control boards	Sensor hub working — seven halls and the MPU live through Blynk as of 2026-09-16. Stepper board v1.1 at the fab.
The frame	Not built. Not drawn either — no dimensions yet, only the concept above.
The clear dome	Does not exist. The real hull is opaque. A clear dome is right for a demonstrator because the mechanism should be visible; it is wrong for the ocean, where acrylic crazes, scratches and fouls.
Anything in the water producing power	Has not happened. No number on this page has been measured from this machine, because this machine does not exist yet.

The standing rule. Every feature gets a row in the validation ledger: the test that proves it, the number the test yields, and its status. A feature with no measured row is an idea and gets described as one. Six new rows were opened for this machine today — all of them planned, none measured.

The dates

2026-09-30 — wave simulation on the bench. Sensors done, ten jumpers, first Z motion, steps per millimetre on Z only, open-loop replay from SD. Six-axis closed-loop control is explicitly not on the path; that is the trap.
2026-11-30 — the demonstrator characterized. Not “working” — characterized, which means curves: coast-down, output against load, output against wave input, and now a fourth, output against weight position. That last one is the tuning map, and it is the whole argument of the machine on one chart.
A few months out — the thing in the picture. Frame built, hull sprung inside it, arm turning, thrusters holding heading, in water deep enough to matter.

Every part of it except the frame and the dome is already on the shelf. That is the point of the way this has been built — the expensive version of this machine was never ordered, it was accumulated.

Open — not yet decided

✅ RESOLVED 2026-09-17 — six floats at the frame’s corners, and the choice turns out to matter more than "it needs to float." ⭐ A small waterplane makes the frame a QUIET REFERENCE. Wave force at the surface scales with the area presented there — the principle behind spar buoys and semi-submersibles. So the two bodies end up with exactly the right relationship: the inner sphere has a large waterplane and follows the wave (the mover), the frame has a small one and stays comparatively still (the reference). The difference between them IS the relative motion, and relative motion is the power. Float sizing is therefore a design parameter to sweep, not just a load calculation.
Practical: size from a real weight estimate with reserve, then make them no bigger — six 300 mm discs half-submerged is only ~40–45 kg of buoyancy and the frame, glazing, solar and thrusters could eat most of it, so weigh parts as they arrive. And slim at the waterline, fat below: putting the volume deeper and piercing the surface with a narrower column makes the reference quieter still, and keeps flat discs out of the slam zone.
✅ RESOLVED 2026-09-17 — the frame’s panels earn their place, for a reason I had missed. I had argued for deleting them since the inner dome does all the sealing. His argument is better: the inner body is the measuring instrument, and its tilt is the signal. Wind on the frame is a nuisance the thrusters absorb; wind on the inner body is contamination, because gusts are turbulent and cannot be separated from wave-driven motion after the fact. Shielding means the only thing touching the instrument is water — which is what C18 needs. They also stand between a knock and the brittle acrylic dome. ⭐ The arrangement that gets both: glaze the UPPER bays, leave the LOWER bays open. Wind lives above the water, slam lives at it — so panels where the wind is, open where the water is. Keeps the shielding and the protection, drops the wave-slam load case. ⚠️ Honest counter: panels raise the whole assembly’s sail area and its center, so more leeway and more overturning moment. The wide waterplane should carry it, but measure the heel angle in a stiff breeze before trusting a dataset taken in one.
Brace the thruster outriggers. As drawn they are cantilevers crossing the waterline, which is the worst place for an unbraced member — that is where slam lands, cyclically, forever. A diagonal from each arm back to the frame’s upper ring turns bending into tension for the cost of one strut each.
The thrusters are now the deepest thing on the buoy — first to touch bottom at a ramp, first to catch kelp or a stray line. Wants a skid or guard below each, and the draft number written down before launching anywhere shallow. Run the cables inside the struts, or clamped hard along the back edge; a cable hanging in the flow flutters and work-hardens at the gland.
Bumpers. A body free to rock inside a fixed cage will touch it in any real sea, and acrylic against aluminum chips the acrylic. Generous clearance and soft replaceable bumpers on the frame’s inner faces, sized so the hull meets rubber before anything rigid.
The sphere is steel in the render, and the generator sits directly beneath it. The no-steel-in-field rule applies — check with the magnetic viewing film once assembled; if the field reaches the swept circle it becomes lead or brass.
Prior art search, before filing. An eccentric mass on a near-vertical axis driven by hull motion is an existing architecture — Wello’s Penguin is the first to check, ISWEC the related gyroscopic version. If the base architecture is prior art, the novelty sits in the within-cycle inertia modulation and the swell-group scheduling, not in the rotating mass.

⚠ Not for distribution. The mechanism visible through the dome — specifically the moving carriage and what it is for — is unfiled and undisclosed. It is not on the public site, not in the animation, and not in any of the four provisionals. Keep this page and this image inside the log until the fifth is filed. There is no grace clock running on it, and that is worth preserving.

Private build log · WaveForge / StabilityCore · concept render 2026-09-17