So, I've been working on PRS stuff again and finally to a place where it might be worth sharing and also so I do not forget ...again. The PRS (and maybe AVC?) entry I've been designing is entirely drive-by-wire with custom steering actuators and independently powered/controlled, fail-safe, hydraulic disc brakes ...because I've just gotta take it way too seriously. Also hoping to use modified NEMA23 stepper motors for drive with custom Arduino/Teensy-compatible motor controller with up to four half-bridges. Comments and advice welcome.
Reworking surplus NEMA23 stepper motors for higher power (720W+) propulsion.
The surplus motors I bought several years ago have stators that are 32mm long with a 35.5mm inner diameter for a 35mm rotor (unless machining the stepper sub-teeth out of the larger stator teeth for an extra 1~2mm rotor diameter) and came wound with 22 turns of 2 strands of ~26AWG per tooth (8 teeth total). The teeth have cutouts for wire with ~5mm radial depth and can accommodate ~5mm of wire height between teeth. Each tooth is ~11mm wide at rotor interface with 3mm minimum width of material per tooth between interface and exterior. Planning to rewinding with 27mil thick copper plate cut into strips 4.8mm wide and insulated on one wide face with 6mm kapton tape (excess folds up along edges to insulate sides) which should result in ~12AWG worth of copper, up to 6~7 turns per tooth, and should provide better packing efficiency than magnet wire (also easier prep than measuring and cutting many strands of wire and keeping them from getting tangled while winding each tooth; and much cheaper than buying many small spools). In theory, this could result in a kV as low as ~450 RPM/V, but might wind up closer to ~900 RPM/V depending on assembly quality.
Replacing the rotor of an in-runner is a bit problematic without arc/cylindrical magnets because of basic geometry (rectangular magnets will always have larger gaps on outer diameter than inner diameter). Might get a quote for custom arc magnets from supermagnetman which would negate the need for any adapter plates pressed onto the 8mm shaft and give better coverage than rectangular magnets (thinking 4mm IR, 17.4mm OR, 55 degree arc angle, 30~32mm length). If using rectangular magnets, have to get CNC cut plates to hold the magnets onto the shaft. Would prefer Big Blue Saw waterjet cut 0.25" Al-6061 or cold finished steel, but probably going with ponoko 5.5mm hardboard or 4.9mm delrin. Lots of options for rectangular magnets, but most are not inexpensive and easy to assemble. A halbach array of 24 0.125"x0.125" magnets would be nice, but not the greatest coverage while a plain array of 26 0.125"x0.125" magnets gets better coverage. Unfortunately, I've not really found any source of long and inexpensive magnets in this size with lengths that multiply to 32mm+/-2mm ("rect0650" N50 magnets are 0.625" long with 48 or 52 magnets at $3.50 each; "rect0250" N45 magnets 0.25" long with 120 or 130 magnets at $0.30 each). Other options with not so great coverage are: a 4-pole-pair (4N+4S) halbach array of "rect0526" N50 magnets (0.5"x0.25"x0.125"; $0.80 each) on a 64mm long rotor with two 32mm long stators stacked, bonded, and wound as one 64mm long stator; and a 5-pole-pair (5N+5S) array of "rect1250" N40 magnets (1.25"x0.25"x0.25"; $2.20 each) on a 32mm long rotor.
Basically, just four N-FET half-bridges (eight TO-263 (D2Pak) 80V MOSFETs) with driver ICs (IRS21867) and current sensors (ACS711-EX plus two 1210 resistor footprints). Current design has six ETQ-*-YFC/YGC inductors and 12 SOT-669 (LFPAK/Power56) P-FETs for resonant snubbers between any two phases, but may never populate those since they are for playing with Zero-Voltage Switching on the half-bridge N-FETs. Intend it to be flexible enough to handle anything thrown at it with a bit of reprogramming and/or selection switches: four 2Q-driven brushed DC motors, two 4Q-driven brushed DC motors, one 4Q-driven brushed DC motor with two power channels, one 2-phase bipolar stepper motor, one 3-phase motor, one 4-phase motor, etc. Want to make it an all-in-one board on a single 100mm x 100mm PCB, but KiCad sucks and I can't really afford the commercial version of eagle layout+schematic right now. Fortunately, eagle does not care about anything except component pins/pads, so any vias and traces can still extend beyond the 101x81 limit. Will probably wind up splitting it into two boards: one for power electronics, driver ICs, and current sensors, and one for logic. Intend to make some little copper plugs to press-fit into the bottom of the board under the drains of the N-FETs; they get mechanically attached to larger copper bars after assembly+reflow to act as both heat sinks and electrical connectors.
Hypocycloid gearboxes (for steering and brake pumps).
The housing end plates are 60mm square with 18 3mm pins on a 46mm mounting circle (fits within a 2"ID tube of any material that can be used instead of standoffs to better prevent twisting of end plates and protect from debris) and the cycloidal discs have 17 pins resulting in a 17:1 reduction per stage (r=[18-17]/17). Each stage is composed of dual cycloidal discs 180 degrees out of phase for balanced rotation and better grip on output pins (hopefully less backlash). Going to try making them from laser cut delrin via ponoko, 3mm diameter O2 tool steel rod from McMaster-Carr, 6mm hex rod for shafts, a few screws and standoffs for holding it together, and a few more screws for mounting to NEMA23 motors. The design is split into two P1 plates: "must be delrin" P1 plate should be able to produce 10 cycloidal discs, 5 output discs, and 19 drive cams/bearings; and "could be hardboard" P1 plate should be able to produce 6 end plates to hold pins and 608ZZ bearings, 3 intermediate plates for pin support, and 3 spacer discs for keeping the cycloidal discs and output discs axially separated (and keeps the hex shafts from mechanically linking). The two plates in the design are ~$50 each (~$100 total) if using only 4.9mm delrin, but should provide 5 full stages (plus 9 spare cams) and three housings.
Each wheel gets a small brushless motor or stepper motor with hypocycloid gearbox driving a positive displacement pump. The brake pads are spring loaded to engage with the rotor and require fluid provided by the pump to disengage the brakes. Fail-safe is provided by a needle valve linking the cylinder to the local reservoir which is engaged by a solenoid. When power is cut, the needle valve opens and allows the brakes to smoothly engage to full-stop. Rotation sensor on the output shaft to get accurate wheel speed, detect locked brake rotor, and try to detect wheel slippage.
Basically, a radiator made from 25mil Al-3003 sheet folded into fins and soldered to 0.25"OD Cu-122 tube, motor cooling blocks made from thicker Al-3003 plate and same Cu-122 tube, a cheap 12V diaphragm pump, and a small reservoir.
12S 15Ah LiFePO4 pack made from Headway 40152 cells with a service disconnect/fuse that splits the pack in half preventing any voltage differential appearing between the main power points.
Basically, two steel arms hard-mounted on a common shaft that swings between three positions (2P3T switch). The shaft is spring loaded to connect the outputs to an array of 120W power resistors to discharge the system when battery power is lost or the E-Stop is engaged. The middle position is a pre-charge state and the third position is full-on state (direct connection to battery). The ends of the arms have copper plates on top and bottom that slide through leaf-spring loaded contacts in the housing. The contacts are linked to the output connectors by grounding braid (1"x0.125"; ~1AWG). The breaker includes a MIDI fuse block, DOSA 1/16th brick 12V/3.5A power supply with its own ATC/ATO fuse, LEM HO 100-P current sensor, and control electronics for the breaker (ratchet mechanism, and optional motor, to 'charge'; solenoids to pre-charge close, full close, and open/discharge; bit like Systems Pow-R Breaker ala original Jurassic Park movie) with full-duplex serial connection and 12V power to rest of vehicle (provide power usage and pass along battery state, as well as one method for engaging E-Stop).
Big connectors are 55 strands of 25mil Bronze 510 wire crimped in a ~5mmID copper tube stuffed in a 8mmID/16mmOD 3D printed non-polarized housing. Basically, the 55 strands from each side are like two stiff-bristled brushes stabbing into each other to provide very large contact surface area and low insertion forces. For 55 strands of 25mil wire, the surface area is ~110mm^2 per mm of length and the cross-sectional area is ~17mm^2 or ~5AWG; packing efficiency of 110 strands in 8mmID tube is ~69%.
Smaller connectors are 19 (or 20) strands of 25mil Bronze 510 wire crimped in a ~3mmID copper tube stuffed in a 5mmID housing. For 19 strands of 25mil wire, the surface area is ~38mm^2 per mm of length and the cross-sectional area is ~6mm^2 or 9~10AWG; packing efficiency of 38 strands in 5mmID tube is ~61%.