v10.5, 2026-11-09. The steering on both boats: the actuator (the powered ram that pushes the tiller arm), the tiller arm, the rear mount and the stop blocks. On Boat 1 (Prime Ranger) one Ewellix electric actuator, on the port side of the cockpit, pushes the old wooden tiller 13 in from the rudder stock (the shaft the rudder hangs from), driven by the central node (the small control box beside the battery that switches loads and triggers releases). The rudder rebuild (stock, blade, heel, tube and the failure hierarchy) is Build Guide sec 3.8. The idea: Boat 1 has one actuator and no backup, so the fitting protects it: no sideways load on its rod, a stiff mount, stop blocks it can never reach, and a feed that stays live if a control box dies. Secs 1 to 8 are Boat 1 only. Everything Boat 2 (Lone Ranger) does differently, its base bracket, pull-pin and aux rudder among them, is in sec 9; Boat 1 builds none of it. Steering numbers: System Settings Sheet sec 3. The drive's wiring in the central node: Compute Bay Housing and Wiring Plan sec 10.2. Boxes and holes: Electronics and Wiring Plan. Test order: Build Guide sec 11. Reasons: Design Decisions R-3.8B, R-3.8C, R-8.4B and R-8.4C.
Parts (Boat 1).
| Part | Status |
|---|---|
| Actuator: Ewellix CAHB-22-F4E-3000529-BAA0PT-000 (sec 2.1) | Bought 2026-09-10 at RS, $273.70 |
| Drive: Pololu G2 High-Power Motor Driver 24v13, Pololu item 2992 (sec 5) | Ordered 2026-09-27: , $79.99 |
| One 8-way 0.1 in screw terminal block for the G2's logic row: Hxchen 8-pin, 20 pack | , price not checked |
| Power-cut relay: HiLetgo 24 V 30 A single-channel relay module | On hand: six bought in three 2-packs (2026-08-16, 2026-08-18 and 2026-08-24). Three have jobs (this one, Release Node B's enable relay, the central node's release enable relay); three are spare |
| Eye pins: two 1/2-13 x 3-1/4 in 18-8 hex bolts, Bolt Depot 2628216. Not 316 clevis pins (Mike, 2026-09-26) | Bought 2026-09-24 |
| 30 nylon washers, 0.525 in hole. Not acetal | Bought 2026-09-24 |
| Tiller arm: the old wooden tiller in its bronze fitting, with a scrap 1/4 in 316 plate each side (sec 3) | Nothing to buy. Check the wood first |
| Rear mount (sec 4) | Built (test fit, Mike, 2026-09-28) |
| Two stop blocks faced with G10 offcuts (sec 3) | To make |
| Encoder A, an AS5048A on the stock (sec 5); its mount measured from the floor | Spec final |
| Silicone bellows boot (sec 4) | Open: whether a boot is still wanted, and which; Mike's word settles it. Not the YAOKE (Amazon B0C8Y526RB): sent back, refunded 2026-10-02 |
| Desiccant sized as one-time capacity; VCI emitters (rust-stopping vapor) near the actuator and encoder | No part named in this guide |
Not fitted, and not to be re-proposed (Boat 1 steering is single string by decision):
Slow is enough: about 1.85 deg/s at the rudder. At 2 to 3 kt the rudder's force comes from water speed, not swing rate; passage steering is small corrections minutes apart; in storm mode the rod parks and the drogue holds the stern. 13 in, so the stops sit outside the actuator's own ends (Mike, 2026-09-28). There its full stroke ends at about 27 degrees each way, so the blocks go at about 28-1/2. If the software fails, the rod runs into its own end and that force stays inside the actuator, not in the rear mount or the tiller plates. The cost: rod force about 8 percent lower and the rudder about 8 percent slower than at 12 in (Design Decisions R-3.8C). The old tiller is strong enough. The stop impact, about 295 ft-lb, is about what a hard pull on the 3 ft tiller gave it for forty years. Stiffness, not strength. Give at the rear mount blurs the steering (Design Decisions R-3.8B). The floor braces the tube top sideways at exactly the height where the actuator pushes on the stock; nothing else bears on the stock.
Ewellix CAHB-22-F4E-3000529-BAA0PT-000. Bought; nothing holds up the build. Figures from the Ewellix CAHB series catalogue (Reference/Ewellix CAHB Series Catalogue.pdf), page 24, speed and current at 68 F (20 C), unless marked.
| Spec | Value |
|---|---|
| Force | 2,250 lbf (10,000 N) rated, push and pull. About 3,150 lbf (14,000 N) maximum, capped by the clutch |
| Stroke | 11.8 in (300 mm) |
| Speed | 0.51 in/s (13.0 mm/s) no load, 0.40 in/s (10.2 mm/s) at rated force. The 11.0 in software sweep (25 degrees each way at the 13 in station) takes about 22 to 28 s |
| Supply | 24 V DC, two-wire motor, 14 AWG red and black leads; swap them to reverse. Fed on 14/2 from a 20 A breaker (Electronics and Wiring Plan sec 7.6) |
| Current | 8.5 A at rated load, 10.9 A when the clutch trips. The SKF CAHB-22 E datasheet for the 10,000 N version (read 2026-09-22) says 10.0 A rated. The 20 A breaker covers either |
| Duty cycle | 20 percent: 85 seconds on, 340 seconds off |
| Temperature | Full performance only from 32 to 104 F (0 to 40 C). Survives -40 to 185 F (-40 to 85 C) |
| Overheat | A built-in over-temperature switch backs up the node's stall sensing |
| Position feedback | Built-in potentiometer (a variable resistor), three 22 AWG wires: green the wiper, white the end reading 0 ohm at full retract, brown the far end. About 254 ohm per inch (catalogue p. 27: 10 ohm per mm, not 30), about 3,000 ohm over the stroke. Measure it on the bench (sec 2.3) and scale the node's input to the measured span, not to the 10 kohm the SKF datasheet names for the far end |
| Holding | Self-locking screw: it stays where it stops with no power and no brake; the rudder cannot push it back. No single holding-force number, only a graph against stroke |
| Bus | None. No CAN and no J1939 (data links used by industrial and vehicle actuators) anywhere on the steering |
| Service life | Ewellix publishes none. The boat works the rod at 10 to 37 lbf at the 13 in station, half a percent to two percent of rated, so no life limit forces fewer moves (Design Decisions R-8.4B). Do not quote a rod-travel life |
| Eyes | Plain, no ball joint, parallel to each other. Holes 0.516 in (13.1 mm), for 1/2 in pins (catalogue p. 28) |
| Backlash | 0.024 in (0.6 mm) |
| Sealing | IP 69K/66M. Salt-spray tested 250 hours to ISO 9227:2012 |
Pins and washers. The pins are the two 18-8 hex bolts, one per eye (sec 1, parts), fitted by rules 2 to 4. Fine if the smooth shank covers the eye (Claude's view); at the arm a sleeve or a shoulder bolt also works. A nylon washer above and below each eye keeps the eye's metal off the stainless: the catalogue says only "corrosion protected metal parts", and a magnet tells plated steel from aluminum. The washers carry only the actuator's own weight, about 11 lb split between its two ends.
Not fitted, and not to be re-proposed: a second, freewheeling actuator on the arm would share the arm, stock and blade with the primary. Not used: the Tolomatic ERD15/20 with the BNM10 ball screw, or Ultra Motion, with a Flipsky Mini FSESC 4.20 and a second stock encoder (none bought). The actuator is 24 V native; no hydraulics.
On the bench drive, the Ewellix must pass:
The drive electronics talk to the bench node: the Pololu G2 and its current sense (sec 7, bench step 3). No bench hold test with a scale (Mike, 2026-09-28): the hold is checked on the platform (sec 7).
One afternoon in the stern compartment, after the tiller wood check (sec 3) and the two checks of sec 4, with the parts in hand and before anything is drilled. Measure:
Open: whether the cockpit floor has room out to about 30 in to port of the centerline; this session's measurement settles it. At haul-out: the rudder's area, chord, span and stock-axis offset. The one measurement behind the storm-loads case: it bounds the wave torque (twisting force) that nothing relieves (Build Guide sec 3.8) and sets the authority schedule.
The arm is the original wooden tiller, cut short, in its own bronze fitting. Nothing to buy. The head on the stock is an upright tongue about 1-1/4 in wide with one horizontal hinge pin; the tiller's bronze end fitting hangs on that pin by a fork. Check first.
Make and fit the arm.
Not verified: whether the cast fitting takes the stop impact, the same load it saw in hand steering. Inspection: witness marks on the fitting bolts at refits; the aft camera's steering view covers the arm in service. Stops. With the rod eye at 13 in the actuator's full stroke ends at about 27 degrees each way, and the blocks sit just past, at about 28-1/2 (rule 5; why: sec 1). Keep the 25 degree software travel: big angles earn their keep in slow-speed work (the lake first setup, docking, the Build Guide sec 3.9A turn away from ice). Set the stop blocks with the actuator fitted (sec 4, step 7).
The block fastening. The stop-impact jolt never reaches the rod. The 500 lbf rating is conservative over the roughly 295 ft-lb stop impact: about 235 lbf where the arm meets the blocks 15 in from the stock center, less further out. Re-check it against the measured wave-torque bound once the rudder area and center of pressure are measured (sec 2.4). Loads on the arm. Rod force is the stock torque divided by the 13 in (1.08 ft) arm, about 8 percent less than a 12 in station gives; the rudder turns about 8 percent slower for the same rod speed.
| Condition | Torque at stock | Rod force at the 13 in station |
|---|---|---|
| Normal course-keeping (under 1 percent of the 2,250 lbf rating) | 11 to 18 ft-lb | 10 to 17 lbf |
| Rough-weather steering | ~25 to 40 ft-lb | 23 to 37 lbf |
| Drive authority limit | 110 ft-lb | about 102 lbf |
| Stroke used by the 25 deg software swing each way | 11.0 in of 11.8 in, 0.4 in to spare at each end; the full stroke ends at about 27 deg each way |
More rudder rate would be a mount move, not a purchase: an 8 in station gives about 3 deg/s at about 165 lbf and uses about 7.3 in of the stroke at the 27 degree stops. [IMAGE] Figure 1. Plan view: tiller arm, actuator station, stop blocks at the sweep limits, stock encoder below the head. An old drawing (before v10.0, starboard layout); the text wins where they differ: the arm is the old wooden tiller, not a bar; one actuator, the Ewellix at 13 in, on the port side; stops just past the actuator's own ends; no backup, no pull-pins, one stock encoder. The current top view is Reference/Steering Actuator Top View 2026-09-28.pdf (its Section B still shows the old angle bracket, not the rear mount as built).
The floor braces the tube top and is the datum (rules 7 and 8) for the rear mount, stop blocks, encoder bracket and bellows boot. The blocks, the encoder bracket and the boot all sit on the floor, above bilge water, with drip loops on all wiring. Two checks at fit-out prove it is stiff enough:
The actuator runs from the tiller arm to the port side of the cockpit (Mike, 2026-09-28), the mirror of a starboard layout.
| Rudder | Rod length, pin to pin |
|---|---|
| Fully in, about 27 degrees to port | 20.8 in |
| 25 degree software limit to port | 21.2 in |
| Centered | 26.7 in |
| 25 degrees to starboard | 32.2 in |
| Fully out, about 27 degrees to starboard | 32.6 in |
Small end clamped to the stock below the tiller head, big end over the rudder-tube top. It keeps spray out of the tube and off the actuator, and salt mist off the stock encoder. The tube top is the only salt source near the actuator, which sits above knockdown bilge level. Face the actuator's cable gland so condensation runs away from the actuator.
First.
On the bench.
In the boat.
[IMAGE] Figure 2. Steering platform, plan view: stop blocks and the closed load loop (dashed). An old drawing (before v10.0, starboard layout); the text wins where they differ: no glassed shelf, tube-top collar or bushing; one actuator on one rear mount, no backup, no pull-pins. The current top view: as for Figure 1.
One Pololu G2 High-Power Motor Driver 24v13 (Pololu item 2992) in the central node: an H-bridge, four electronic switches that run a motor either way off one 24 V supply.
How it steers. The computer in command (Pi A, or Pi B when it has command) sends the target angle over the control bus. The central node drives the rod through the G2, reads the Ewellix feedback, encoder A and motor current on every move, then stops. Between moves the self-locking screw holds, with no power. A move: set DIR, ramp PWM from 0 to full over about 1/4 s, hold full, and set PWM to 0 at the target reading. PWM at 0 brakes the motor and the screw holds; a dead Nucleo leaves PWM at 0. Stall stop. Starting numbers, Claude's guess, set at the bench (System Settings Sheet sec 3):
Encoder A (a sensor that reads the rudder stock's angle) is an AS5048A 14-bit magnetic absolute encoder on the stock end: a diametric magnet on the post (held by rule 10), the encoder on a fixed bracket. Its Belden 6504FE lead runs in one piece from the rudder stock to the central node (control bus), which reads its pulse output (PWM) on one wire and one timer pin, not the four-wire SPI link.
| Symptom | Meaning |
|---|---|
| Actuator moves but the stock does not | Sheared linkage |
| Commanded current with no stock motion | Jam |
| Stock turns, the boat does not | The rudder has lost its bite |
A jam is the actuator told to move and the stock not moving. Effort commanded with no stock-encoder motion declares a jam within seconds and cuts the power, because a jammed, still-commanded electric actuator cooks its windings in 10 to 25 minutes (NASA run-to-failure data). High motor current with no stock motion on the healthy sensor (encoder A, or the Ewellix wiper if encoder A is lost) is a jam; the node stops the drive and opens the power-cut relay. The Ewellix has no bus and no overload report, so this current sense is the only jam detector. The G2's own 30 A current limit is the hardware backstop.
When the control bus goes quiet, power drops to everything: the primary holds on its self-locking screw, its unpowered state; the motor ramps to zero; the wing feathers. No safety default depends on a message arriving (Build Guide sec 5.1 rule).
On the Iridium command whitelist (Build Guide sec 8.3), all signed, over Iridium and Starlink: current-limit retune, a new route, a keep-clear zone and remote-manual control of heading and rudder angle, the boat returning to its route when the link drops.
Software numbers, all in System Settings Sheet sec 3.
Every backup exists to keep the mission going (Mike, 2026-11-01; the rule: Build Guide sec 8.2A): when a main part fails, its backup must let the boat keep sailing toward the goal, not only keep it safe.
| State | Primary actuator | Power | Keeps sailing? |
|---|---|---|---|
| Normal steering | Drives; the screw holds between moves at 0 W | Measure at the lake trials; the power budget carries 45 Wh/day, the cautious end of about 25 to 45 (Build Guide sec 4.6) | Yes |
| Pi B in command (Pi A down or handed over) | Drives as normal: Pi B sends the target angle to the central node over the control bus | As normal | Yes |
| One rudder sensor lost (encoder A or the actuator's own feedback) | Drives on the healthy sensor (sec 5) | As normal | Yes |
| A false stall stop | Retried after a pause (sec 5) | As normal | Yes |
| Main side dead, on the backup battery | Drives from the backup converter in passage mode; the screw holds between moves | Converter on only for each move, node and rudder powered about 15 s at a time; about 145 to 165 Wh a day for the whole backup steer mode (a guess; Compute Bay Plan sec 5.3) | Yes while the converter runs. When it cuts off for low voltage the rudder holds where it is and the boat drifts; steering starts again by itself when it restarts |
| Total power loss | The screw holds the rudder at zero power, meeting the total-power-loss hold requirement | 0 W; rudder held | No: the boat drifts with the rudder held |
| STORM / JSD (Jordan series drogue) | Centered, held by the screw. No relief exists: wave overloads run into arm, stock and structure (the failure hierarchy, Build Guide sec 3.8) | 0 W parked | Held on purpose while the drogue steers; steering resumes after the release |
| Primary fails | The voyage ends (accepted, Mike, 2026-11-01); wing-as-rudder downwind only (sec 8) | No, beyond wing-as-rudder downwind |
The steering feed (line 5, 20 A breaker) sits on the backup-fed bar on the main bus board, with the central node, both drogue release feeds and the wing box (Electronics and Wiring Plan sec 7.6). With the main side truly dead, Pi B turns on the backup converter (the Victron Orion-Tr Smart 24/24-17 isolated, 17 A, run from the backup battery) by its remote on/off, only for each rudder move and for a wind reading now and then, and steers in passage mode. A 12 A move on the converter's 12/2 run drops about 2.7 percent. Steering pauses while a drogue horn burns. The steer-or-drift switch is the converter's own input low-voltage cutoff (about 30 percent left in the backup battery, a setting): it stops steering, and its restart voltage starts it again. In drift the self-locking screw holds the rudder where it is. The energy budget: Compute Bay Plan sec 5.3.
| Cadence | Check |
|---|---|
| Every move | Trend the motor current: a screw going bad shows months early, and the same trend tracks the rudder's breakaway force against its starting point, so a stiffening bearing shows early too |
| Weekly | Full-stroke exercise sweep |
| Seasonally, and after any storm where the encoders logged an offset or the rudder camera shows change | Hold-integrity re-check |
| At refits | Witness marks on the tiller fitting bolts (sec 3) |
| At haul-out | Renew the heel bearing and the rudder-tube bearings (the rudder's own; there is no tube-top bushing) |
The fallbacks. One rudder, one tiller arm, one actuator, then wing-as-rudder (below). The shared chain. The arm, tiller head, key, stock, blade and heel fitting are one chain with the actuator, accepted on the oversize-and-inspect rule rather than duplicated. The stock carries working loads of 11 to 18 ft-lb against about 400 ft-lb first yield in twist (about 22 times). Against the design overload the margin is near one, but that is first yield, not rupture (about 1,100 to 1,200 ft-lb), so the stock takes a measurable permanent twist well before any break. The failure hierarchy along the chain: Build Guide sec 3.8. Wing-as-rudder (Build Guide sec 8.4). A software-only emergency mode, and Boat 1's only fallback. Wing trim alone holds a downwind heading within plus or minus 10 to 20 degrees in 8 to 25 kt. It cannot beat, and cannot clear a lee shore. Authority schedule (Build Guide sec 8). Full deflection below about 8 kt; reduced slew and wider deadband when surfing; centered and held under JSD. Threshold provisional, pending the rudder geometry measurement at haul-out (sec 2.4), which also bounds the worst wave twisting force. Open: the Build Guide says the deadband is reduced when surfing, where this handbook says wider; settled when the authority schedule is set from the rudder measurements.
Boat 2 (Lone Ranger) steers with one braked Thomson actuator on a releasable pin, then a transom aux rudder, then wing-as-rudder. Its differences from secs 1 to 8 follow, grouped by the section each belongs to. Boat 2's older Steering Proposal Rev 5 (Reference folder) is background only; every Boat 2 fact a builder needs is here.
Not fitted on Boat 2 either: the backup actuator (sec 2.2), the Flipsky FSESC box with its phase-short relay and feeds, and encoder B. Boat 2's backup is the aux rudder (sec 9.9). All its actuators are 24 V native; no hydraulics.
Thomson Electrak HD HD24B026-0300CNO1EES, at the 12 in station: 24 V, 585 lbf, 11.8 in (300 mm) stroke, ball screw, J1939 CAN, IP67/IP69K, holding brake, rated to -40 F; about $1,500 to $2,500 from US distributors. Ordered on published specifications, not gated on a vendor answer. Chosen over the 382 lbf HD24B017-0300CNO1EES for about 8x the ball-screw life (Design Decisions R-8.4B).
Bench run (sec 2.3): the drive electronics talk over CAN to a bench Pi. Measurements (sec 2.4), before the Thomson is ordered, in one session: check Boat 2's tiller head and fitting (sec 9.3), measure the true arm radius range and the actuator mounting range, and confirm the 12 in station and the stop-block sweep at the arm tip. At haul-out, Boat 2's rudder too: its area, chord, span and stock-axis offset.
The arm. One arm, on the oversize-and-inspect rule (built strong and checked, not doubled). Look at Boat 2's tiller head before drawing its arm.
One base bracket, in place of the rear mount. Rev 5 put it and the stop blocks on the deck beams; this handbook measures Boat 2's pull-pin parking rest (sec 9.5) from the cockpit floor, the datum. Open: Boat 2's base and stops on the deck beams or on the cockpit floor as on Boat 1; Boat 2's geometry session (sec 9.2) settles it.
Boat 2 fits one pull-pin (a pin released on command, freeing a dead actuator from the arm), on the primary. When the primary fails locked or dead, the pin releases so the rudder trails and the aux rudder can steer: a trailing rudder beats a locked one for aux-rudder steerage. No steer-through: the Thomson brake grips harder under load.
Open: which release rule Boat 2's software follows; settled before Boat 2's release logic is written. The Build Guide says instead: unload before release (detect the lock, drive against it or wait for the wave load to reverse, confirm low pin load, enable, release), and never command a loaded pin, except with the actuator dead, where a trailing rudder beats a locked one.
On release the pin drops into its housing and the freed actuator droops a few degrees about its base pivot, jam or no jam (the jam is in the screw, not the mount), onto its rubber-faced parking rest about 1 in below the arm. The arm sweeps in clear air above it, full travel both ways, and re-engagement is impossible without lifting against gravity. A pin that lets go by itself shows within one correction cycle on the stock encoder. [IMAGE] Figure 3. BOAT 2 ONLY. Pull-pin station, side view: tongue under the bar, pin engaging upward from the spring and latch housing, gravity drop to the parking rest on release.
A small second rudder on the transom, lowered when the main steering is lost. With its own stock, blade and transom mount, it is the only path past the shared chain (sec 8) and shares nothing with it.
Open questions for this guide are in Pending Amendments v10.6.