Download FEMM (Finite Element Method Magnetics) from femm.info — free software. Model the Ring 3 cross-section first (largest, most consequential). Set up:

• NdFeB block with Br = 1.32 T (N45), µr = 1.05, coercivity Hc = 995 kA/m
• Copper layer with conductivity σ = 5.8×10⁷ S/m
• Aluminium layer with σ = 3.5×10⁷ S/m
• Air gap = 0.75mm (midpoint of 0.5–1.0mm target)
• Vary the compound-angle pole tilt from 10° to 25° and measure tangential force on a roller element

Target: positive tangential force (self-driving torque) at the design RPM. If force is zero or negative, adjust segment count or tilt angle before ordering materials.

Order rings and rollers as un-magnetised blanks in the final composite-fitted dimensions. Contact Eclipse Magnetics or Bunting with the FEMM model output as a spec sheet. Specify:

• Grade: N45 (Br ≥ 1.32T, BHmax ≥ 342 kJ/m³)
• Coating: Ni-Cu-Ni triple layer (corrosion protection)
• Status: UN-magnetised (critical — shipped unmagnetised for machining)
• Tolerance: OD/ID ±0.05mm, face flatness ≤0.01mm/100mm

Simultaneously order copper, aluminium, and nylon bar stock. Expected lead time for custom NdFeB: 4–8 weeks.

Machine on a CNC lathe. Sequence (inside-out):

1. Turn NdFeB blank OD to final dimension (external grinding with CBN wheel if needed — NdFeB machines poorly with HSS but accepts CBN/diamond)
2. Turn copper tube to press-fit over NdFeB: interference = +0.02mm on diameter. Warm Cu tube to 80°C for assembly, slide over cold NdFeB, cool-press to seat
3. Turn aluminium tube to press-fit over copper layer: same interference. Use 3M DP420 structural epoxy at interfaces for long-term bond security
4. Nylon outer shell: machine to clearance fit over Al (−0.01mm, sliding fit), bond with Loctite 406 cyanoacrylate for plastics

Check concentricity with dial gauge (TIR ≤ 0.05mm) before epoxy cure.

Contact a specialist magnetisation service with your FEMM pole-pattern spec. The required pattern is a 12-segment Halbach-inspired compound-angle arrangement:

• 12 sectors for Ring 3 (30° each), 8 for Ring 2, 4 for Ring 1
• Within each sector, the pole axis is tilted 15–20° off the purely radial direction, toward the axial direction (compound angle)
• Adjacent sectors have poles rotated 90° (NSEW pattern as viewed from the ring axis) — this is the Halbach rotation sequence

Magnetisation services: Laboratorio Elettrofisico (Milan, Italy) · Bunting Magnetics UK · Arnold Magnetic Technologies (US). Verify magnetisation with a scanning hall-array or MRI-style flux mapper before assembly.

After full assembly and magnetisation, charge the nylon surface electrostatic layer:

• Set up a corona discharge station: 20–30 kV needle electrode, 15mm gap from the nylon surface, in a dry room (RH < 30% — use silica gel desiccant if needed)
• Rotate each ring/roller at ~1 RPM past the needle for 3–5 passes
• Immediately measure surface potential with a static meter (Monroe Electronics 244A or similar). Target: 500V–2kV surface potential
• Store charged components in grounded Al-foil wrapping until assembly to prevent discharge from dust/humidity

Before winding, prepare a twisted bundle of the three filars:

• Cut AWG 18 wire (Filars A and B, blue and green) and AWG 23 wire (Filar C, orange) to 130 ft lengths each
• Twist all three together at 6 turns per inch using a hand drill with all three ends chucked. Mark filar identities with coloured heat-shrink at each end
• This "litzed" bundle ensures all three filars share the same flux window on every turn — critical for the Bedini trigger-coil coupling principle

Wind each toroid station manually (shuttle winding or a home-made winding jig):

1. Leave a 6-inch lead on all three filars at the start
2. Wind 200 turns of the Filar A/B pair (AWG 18 ×2) and 40 turns of Filar C (AWG 23) simultaneously as the twisted bundle. They will all lay in the same groove
3. Wind evenly, maintaining consistent tension. Overlaps are acceptable for the outer turns
4. Leave 6-inch leads at the finish end. Mark A1/A2, B1/B2, C1/C2 with coloured cable markers
5. Measure DC resistance: Filar A should read 2–5 Ω, Filar B 2–5 Ω, Filar C 10–20 Ω
6. Test inductance with an LCR meter (target L ≈ 1–10 mH depending on core µi)

Each stator ring has a non-magnetic (aluminium or G10 fibreglass) frame with coil stations spaced evenly around the circumference:

• Ring 3: 8 stations, 45° apart
• Ring 2: 6 stations, 60° apart
• Ring 1: 4 stations, 90° apart

Position each coil so the core's inner bore faces the roller path, 0.5–1.0mm from the roller's nylon surface. Secure with nylon cable ties and a small dab of hot glue (non-permanent, allows adjustment). Verify alignment with a feeler gauge.

Each coil has 6 leads (A1, A2, B1, B2, C1, C2). Route them through cable management channels in the ring frame to a terminal strip at a designated wiring hub per ring. Use colour-coded PTFE-insulated hookup wire (rated ≥1kV) for all connections in the IPC flyback path.

Build one circuit board per ring (3 boards total, each serving the coils of one ring). Use a 100×150mm fibreglass perfboard or a custom PCB (gerber files are straightforward to design in KiCad from the schematic in DWG-005).

Component placement order:

1. Mount the MJL21194G to heatsink with thermal paste (Fischer SK68). Pre-drill heatsink mounting holes; use M3 screws with insulating bushing (transistor is NOT isolated from case)
2. Solder 470Ω resistor and 1kΩ trimmer pot in series on the trigger line from Filar C
3. Solder UF4007 diode in the flyback path (cathode toward charge-battery positive)
4. Mount electrolytic caps (2× 22,000µF) with correct polarity. Use screw-terminal lugs for high-current connections
5. Wire LM311 comparator: non-inverting input (+) to cap bank via voltage divider (set Vref = 55V for 75V caps), inverting input (−) to reference (potentiometer from Vcc), output to IRFP260N gate via 4N35 optoisolator
6. Mount IRFP260N (dump FET) with heatsink. Drain connected to cap bank (+), source to charge-battery (+)
7. Place NE-2 neon bulb and 1.5KE200A TVS across BJT collector-emitter as protection clamps

Test each IPC board on the bench before installing in the SEG frame:

• Use a small (Ø25mm, 100-turn) test coil with a Bedini-style rotor (bicycle wheel + 8 ceramic magnets) to generate trigger pulses
• Connect Battery A (12.8V, 7Ah, LiFePO₄) as run battery; connect Battery B as receiving battery
• With oscilloscope, verify flyback spike at collector of MJL21194G: target peak ≥100V (>200V is better)
• Adjust the 1kΩ trigger pot until oscillation is crisp and repeatable (not intermittent, not continuous)
• Run for 1 hour. Measure input power (WT310E) and battery B voltage rise. Verify cap-dump fires correctly (LM311 should fire IRFP260N when cap reaches setpoint)

Programme the Teensy 4.1 with the battery rotation logic:

// Simplified battery swap controller logic
// Relays: R1=A→drive, R2=B→charge, R3=C→rest
void loop() {
  if (millis() - lastSwap > SWAP_INTERVAL_MS) {
    lastSwap = millis();
    rotateBatteryRoles(); // A→C (rest), B→A (drive), C→B (charge)
    logSoC();            // read coulomb counters
    delay(500);          // relay settle time
  }
}
// SWAP_INTERVAL_MS = 300000 (5 min) to 3600000 (60 min)
// Determine optimal interval empirically per battery chemistry

Connect Omron G7L relays to Teensy digital outputs via 4N35 optoisolators. Use MBR3045 Schottky diodes on each battery terminal in blocking polarity.

Fabricate a non-magnetic frame from aluminium profile (Bosch Rexroth 40×40mm T-slot extrusion) or 316 stainless steel (non-magnetic). The three stator rings mount concentrically on this frame. Design requirements:

• Rings must be coplanar to ±0.1mm over their full diameter
• Ring spacing (radially) must allow 0.5–1.0mm air gap to rollers
• Roller bearing housings must allow each roller to spin freely (ceramic bearings preferred, no magnetic interference)
• The whole assembly must be statically balanced (check with bubble level; dynamic balance after initial spin-up)

Mount the three stator rings onto the frame. Each roller rides in a carriage that:

• Holds the roller at the correct radial distance (0.5–1.0mm gap to ring face)
• Allows the roller to orbit the ring (roller carriage rotates around ring axis)
• Allows the roller to spin on its own axis (via ceramic ball bearings, ZrO₂ preferred)

Install Hall-effect sensors (SS495A) at one reference position per ring to measure roller orbital speed and provide commutation trigger for the drive ESC.

Fix the pre-wound toroid coil stations onto the ring frame at their designated angular positions. Final gap check: use non-magnetic feeler gauge (plastic or brass) between each coil core and the nearest roller path. Adjust mounting until gap = 0.75±0.25mm at all stations. Secure with M3 stainless bolts and Loctite 243 thread-locker.

Mount the three IPC boards on the frame (non-magnetic standoffs, away from ring flux). Run ≥1kV-rated PTFE wire from each coil station's Filar B and C leads to the board. Connect boards to cap-dump bus and to the battery management swap controller. Enclose all high-voltage connections in insulated terminal blocks (Phoenix Contact or Wago) rated ≥1kV.

• Confirm all IPC high-voltage paths are insulated and no exposed conductors are accessible during operation
• Verify cap-dump bleeder resistors (10kΩ, 5W) are in place across each 22,000µF cap — these discharge caps when power is removed
• Confirm all three batteries are at 50% SoC (not full — gives room to measure gain) and baseline-logged (terminal voltage after 1-hr rest, internal resistance)
• Set oscilloscope on Filar C output and BJT collector. Set Yokogawa WT310E to measure Battery A input power
• Ensure fire extinguisher (CO₂ type) is accessible. Venting batteries produce hydrogen — no open flames

Apply 3-phase AC drive signal from VFD to Filar A of all ring coils (connected in the appropriate 3-phase configuration for the number of coil stations per ring). Start VFD at 1 Hz and ramp at 0.5 Hz/second. Watch for:

• Rollers beginning to orbit their respective rings — first visible motion typically at 2–5 Hz
• Current draw should rise then plateau as rollers settle into magnetic equilibrium
• Any mechanical interference (chattering, grinding) → STOP immediately and re-check gaps
• Target ramp-up RPM: 300–600 RPM orbital speed of rollers around Ring 3 axis

Once rollers are orbiting at target RPM:

1. Verify flyback pulses are present on oscilloscope (Filar B output → UF4007 → cap bank). Expected: sharp spikes at roller-pass frequency × coil-station count
2. Verify cap voltage is rising (LM311 setpoint not yet reached during initial ramp-up)
3. Once cap reaches comparator setpoint, verify IRFP260N fires (dump to Battery B)
4. Adjust 1kΩ trigger pot on each board for maximum cap charge rate while maintaining stable oscillation

This is the critical Searl-effect test. At steady RPM with VFD driving:

1. Gradually reduce VFD output power while monitoring roller RPM
2. If the SEG self-spins (RPM holds or increases as drive is reduced), this is the Roschin-Godin threshold phenomenon
3. If RPM drops proportionally with drive reduction (expected in a conventional motor), document the drag-torque vs RPM curve for later analysis
4. At no point disconnect the drive abruptly — ramp down gradually to avoid transient voltage spikes

Once the SEG is running with battery swap controller active:

• Log all three battery terminal voltages and internal resistances after 1-hour rest each day (disconnect device, wait 1 hour, measure, reconnect)
• Log daily: V_A, V_B, V_C, R_int_A, R_int_B, R_int_C, accumulated coulombs per battery
• True self-sustaining operation = no net SoC decline in any battery over 7 days
• Gradual SoC increase across all three = CoP > 1 (excess energy charging batteries overall)
• Gradual SoC decrease = CoP < 1 (battery bank is being consumed — stop and analyse)

The following variables have the highest impact on measured CoP, per Perry's research:

• Pulse repetition frequency (PRF): Determined by RPM × rollers × coil stations. Target the battery's optimal PRF. Vary RPM in 50 RPM steps and log CoP change.
• Peak flyback voltage: Adjust via 1kΩ trigger pot on each board. Higher voltage generally improves CoP but risks BJT damage. Monitor temperature of MJL21194G with thermocouple.
• Battery capacity: Perry found higher-capacity batteries (larger "capture cross-section") show higher CoP at similar internal resistance. If CoP is below 1, try a 30Ah battery instead of 18Ah as Battery B.
• Swap interval: Perry's exploratory work suggested 15–30 minutes per role as a starting point for LiFePO₄. Vary from 5 to 60 minutes and log effect on net SoC trend.