OpenTorq: The Bike of Theseus

OpenTorq is a ground-up reimagining of the electric bicycle drivetrain. Instead of a sealed hub motor or a proprietary mid-drive with bespoke gears, OpenTorq uses a steel-belt continuously variable transmission (CVT) paired with a high-torque brushless DC motor. The entire assembly bolts together with standard metric fasteners and can be disassembled with a socket set, hex keys, and a bearing puller. No welds, no glue, no single-sourced custom chips.

The CVT uses a pair of variable-pitch pulleys and a reinforced steel belt, controlled by a small auxiliary motor that adjusts the ratio seamlessly. Regenerative braking is handled by the same motor, feeding current into a modular battery pack built from standard 18650 cells in a welded-but-replaceable configuration. The controller is fully open-source, based on a STM32 microcontroller with field-oriented control, and all firmware is on GitHub.

The frame-interface is a set of four mounting points that can be adapted to almost any bike frame with simple brackets. The whole drivetrain weighs about 6 kg and delivers 250W nominal (500W peak), compliant with EU/AUS regulations but easily unlocked for off-road use. The design philosophy is 'no planned obsolescence' — every part that wears (belt, bearings, motor cables) can be replaced without special tools.

Mainstream e-bikes are becoming more powerful but also more disposable. Hub motors have heavy unsprung mass and flat-spot torque. Mid-drives rely on brittle plastic gears and single-use crank interfaces. Batteries are glued into enclosures. Controllers have locked firmware. When something breaks, you often need to replace a whole module — or the entire bike. Many owners can't even change a brake pad without a dealership visit.

OpenTorq exists to prove that e-bike drivetrains can be efficient, powerful, and maintainable. The itch: a bike that you can fix yourself, upgrade over time, and never throw away because of a dead battery or a broken gear.

• Design a CVT that fits within a standard bicycle bottom bracket area
• Achieve 90% efficiency across a 4:1 ratio range
• Use only off-the-shelf bearings, belts, and fasteners (M5/M6/M8)
• Develop an open-source motor controller with regenerative braking
• Create a battery pack that allows cell replacement without soldering
• Keep total drivetrain weight under 7 kg
• Publish all CAD files, BOM, and firmware under a permissive license
• Build and test one full prototype on a hybrid frame

• Not aiming for high volume manufacturing or retail sales
• Not designed for extreme off-road or high-speed (above 45 km/h legal limit)
• Will not include proprietary software or DRM of any kind
• Does not aim to be the lightest or cheapest; repairability and openness are the priorities
• Will not use carbon fiber or exotic materials (aluminum/steel only)

• Structure: 6061 aluminum extrusions, laser-cut steel brackets, standard M5/M6/M8 bolts, nylon lock nuts
• CVT: Comet-style steel belt (1 inch wide) and adjustable pulleys from McMaster-Carr, modified with sealed bearings
• Motor: Bafang BPM2 250W geared hub motor (used as a mid-drive, opened to access stator) — but rewound for lower Kv
• Controller: STM32F405 board (from robotis DYNAMIXEL?), actually a custom PCB on OSH Park with DRV8323 gate driver
• Battery: Samsung 30Q 18650 cells in 10S4P configuration, with a custom PCB-module that allows cell swap without soldering (using spring contacts)
• Sensors: AS5600 magnetic encoder for motor position, custom hall sensor board for crank RPM
• Fasteners: All hex drive, stainless steel where possible

The drivetrain mounts to the bike's bottom bracket and seat tube via a 4-point bracket system. The motor sits low behind the bottom bracket, driving a belt to the CVT input pulley. The CVT output pulley drives the chainring via a secondary belt. The CVT's ratio is adjusted by a small stepper motor that moves the moving sheave of the input pulley, controlled by the main CPU based on crank torque and motor speed.

Regenerative braking uses the motor as a generator: when the rider brakes (via a brake lever switch or automatic when torque is reversed), the controller switches to buck-mode and charges the battery. The energy is stored in the lithium pack. The controller also handles pedal assist: it reads crank torque via a strain gauge on the bracket, then commands motor torque to match a configurable multiplier.

All electronics are housed in a waterproof aluminum enclosure mounted on the down tube. The controller has a header for programming via USB, and all sensors use standard JST connectors. The battery pack is in a separate cam-locked box on the rear rack, with a dumb fuse and no BMS communication — just passive balancing and a low-voltage cutout in the controller.

• CVT belt wear is unknown at high torque; may require frequent replacement (every 1000 km?)
• The custom motor rewinding could fail if insulation isn't properly done
• Heat management: the motor and controller both generate heat in a compact space
• Regulatory compliance: the regenerative braking system must be safe (no short circuits)
• The CVT pulley alignment is critical; misalignment causes belt destruction
• Battery spring contacts may corrode or lose tension over time

• Is a steel belt CVT quiet enough for a bicycle? (Noise levels need testing)
• How much regenerative braking efficiency can we achieve? (Estimated 30-40%)
• Can the CVT handle the required 250W continuous without slipping?
• Should we design our own motor or modify an existing one? (Existing saves time but limits customization)
• What is the optimal gear ratio range? (5:1 with a low gear for hills?)
• How to make the bracket system universal for multiple frames? (Probably a modular jig for welding?)

The design was fully CAD'd and the BOM was sourced, but the cost of a custom motor and belt pulleys was around $800 per prototype — something I couldn't justify at the time. Then life moved on: a new job, a cross-country move, and the project went into deep hibernation. The idea is still worth keeping, and maybe one day I'll CNC the parts or find a sponsor.

The project name 'Theseus' refers to the ship of Theseus paradox — every part can be replaced, yet it remains the same bike. The open-source controller firmware (written in C) is about 60% complete on a private repo. There's a detailed design document with torque calculations and bearing life estimates.