Ebike Controllers: The Component That Decides How the Bike Feels

A brushless motor cannot run on the current a battery produces. The battery makes direct current at a fixed polarity. The motor needs alternating current, in three phases, timed precisely to the position of a rotor that is spinning several thousand times a minute. Something has to sit between them and perform that translation thousands of times per second. That something is the controller, and it is the reason your bike accelerates the way it does.

Two identical bikes with the same 750W motor and the same 48V battery can feel like different machines because one has a 15A controller and the other has a 25A one. The spec sheets look almost the same. The bikes are not. This is the most consequential number in an ebike drivetrain that manufacturers routinely fail to print.

What a controller actually does

Inside the case are six power MOSFETs arranged as three half-bridges, one per motor phase. Each half-bridge can connect its motor wire to the battery positive rail, to the battery negative rail, or to neither. By stepping through a sequence of those connections, the controller creates a magnetic field in the stator that rotates, and the permanent magnets on the rotor chase it. That is the whole trick. The motor is a set of electromagnets being switched in a circle, and the controller is doing the switching.

The timing has to be right or nothing works. Energise a phase too early or too late relative to the rotor and you get less torque, more heat, or a violent stutter. So most ebike motors contain three hall-effect sensors spaced around the stator that report rotor position to the controller through a small five-wire cable. Sensorless controllers do exist, and they infer position by measuring the back-EMF voltage that the spinning magnets induce in whichever phase is not currently being driven. That works well at speed and badly at a standstill, because a stationary rotor generates no back-EMF, which is why sensorless setups are known for hesitating or kicking on takeoff.

Power level is set separately from timing, by pulse width modulation. The controller does not lower the voltage it sends to the motor; it switches the full battery voltage on and off at something like 10 to 20 kHz and varies the fraction of time it is on. At 25 percent duty cycle the motor sees an average of about a quarter of pack voltage. This is why controllers whine: if the switching frequency sits in or near the audible range, you hear it.

Everything else in the box exists to keep that process safe. There are big electrolytic capacitors across the input to absorb the current spikes the switching creates, a low voltage cutoff to stop the pack being over-discharged, current sensing to enforce the amp limit, and usually a temperature sensor to back things off before the transistors cook.

Why controller amps predict acceleration better than motor watts

Electrical power is voltage times current. On the battery side of the controller, that means peak power is roughly the pack voltage multiplied by the controller current limit. Some worked examples, using nominal voltages:

Now compare that with the number on the motor. A motor labelled 750W is being given a continuous rating, which is a statement about how much heat its windings and casing can shed indefinitely without cooking the magnets or the insulation. It is not a limit on what the motor will accept for thirty seconds. Feed the same 750W hub motor 1,200W and it will happily produce roughly 60 percent more torque, right up until thermal reality catches up on a long climb. That gap between rated and peak is why wattage figures on ebike listings are close to meaningless on their own, a subject we take apart in ebike motor wattage explained.

There is a second layer that explains why ebikes feel so strong from a standstill. The controller behaves like a step-down converter: when it is running at low duty cycle, the power going in roughly equals the power coming out, so if the motor is only seeing a quarter of pack voltage, it can be drawing close to four times the battery current in its windings. A controller with a 25A battery-side limit may specify a phase current limit of 50A or more. Torque follows phase current, not battery current, which is why a bike pulls hardest at the moment it is drawing least from the pack.

What tells the controller how much help to give

The controller does not decide on its own how much power you want. It reads inputs and maps them to a current command, and the choice of input is most of what people mean when they say a bike feels natural or feels like a scooter.

Throttle

A throttle is a hall-effect sensor producing a voltage that rises across roughly the low single digits as you twist or press it. The controller reads that voltage and commands current proportionally. It is a direct request for power, with no reference to what your legs are doing.

Cadence sensor

A cadence sensor is a ring of magnets by the crank and a pickup on the frame. It tells the controller one thing: the cranks are turning, and roughly how fast. The controller then applies whatever current the current assist level calls for. This is why cadence-sensor bikes feel like a switch. You start pedalling, nothing happens for part of a revolution while the controller collects enough pulses to be sure, and then the motor arrives all at once. It is also why they can surge away when you are pushing the bike backwards through a doorway with a hand on a pedal.

Torque sensor

A torque sensor measures actual force, usually with a strain gauge in the bottom bracket spindle or in the rear dropout, and reports it to the controller many times per second. The controller multiplies your effort by the assist factor, so pushing harder gets more help and easing off removes it immediately. That closed loop is what makes a torque-sensor bike feel like strong legs rather than a motor. It is the single upgrade that most changes how a bike rides, and it correlates strongly with price. Where the motor sits changes this too, which we cover in mid-drive versus hub motor.

The controller is also what enforces your class. The speed limit that stops assistance at 20 or 28 mph is a firmware setting in the controller comparing a wheel speed signal against a threshold, which is precisely why display menus can change it and why doing so has legal consequences. See ebike classes explained for what those thresholds mean where you ride.

Sine wave versus square wave

Both types do the same job with different waveforms, and the difference is genuinely audible and genuinely felt at low speed.

A square wave controller, more properly called trapezoidal or six-step, drives two of the three phases fully at any moment and leaves the third floating, then jumps to the next combination every 60 electrical degrees. Six discrete states per electrical revolution. Because the current steps rather than glides, torque is not perfectly constant through a revolution: there is ripple at every commutation, which you hear as a low growl or buzz at walking pace and feel as slight roughness.

A sine wave controller modulates all three phases continuously so the current in each follows a sine wave offset by 120 degrees from the others. The rotating field becomes smooth rather than stepped, torque ripple drops sharply, and the motor goes quiet. Starts from a standstill are cleaner because the controller can hold precise current at very low rotor speed. Most quality mid-drive systems use this approach, often in its more advanced form known as field oriented control, which continuously calculates the ideal current angle relative to the rotor.

Swipe sideways to see all columns →

Do not expect a sine wave controller to make a bike faster. Peak power is still volts times amps. What it changes is refinement, which matters most in exactly the situations budget bikes handle worst: pulling away in traffic, creeping in a crowd, and holding a steady low speed on a path.

Cutoffs, voltage sag, and thermal cutback

Three separate protections can make a bike suddenly feel weak, and riders tend to blame the battery for all of them.

Low voltage cutoff. The controller monitors pack voltage and stops drawing current below a set threshold, commonly around 41 to 42V on a nominal 48V pack, to protect the cells from over-discharge. Under load a tired pack sags below that threshold on a hill, the controller cuts, the bike coasts, the voltage recovers, and assistance returns. If your bike cuts out repeatedly on climbs when the display shows one or two bars, that is the low voltage cutoff arguing with a pack that can no longer hold voltage under load.

Current limiting. Whatever the controller is rated for is a ceiling it will not exceed even if the motor could use more. On a steep grade at low speed, the motor is asking for far more current than the controller will give, and the bike simply stops accelerating. Higher assist levels do not remove this ceiling; they only raise the fraction of it you are allowed to request.

Thermal cutback. Both controller and motor generate heat proportional to the square of current, and both have thermal sensors on better systems. On a sustained climb the controller progressively reduces current to protect itself and the motor. This is felt as the bike gradually going flat over several minutes rather than cutting out abruptly. Mid-drives are more prone to it than hub motors on very slow steep climbs because the motor is spinning fast internally but the bike is barely moving, so there is little airflow relative to the work being done.

None of these is a fault. They are the reason your motor still works. The failure modes below are the ones that are actually broken.

Failure modes and how to recognise them

MOSFET failure

Power transistors usually fail short rather than open. When one does, the symptom is abrupt: total loss of power, often with a blown inline fuse, sometimes accompanied by a sharp electrical smell that people describe as burnt plastic or fish. If two transistors short across a phase, the motor sees a short circuit and the wheel becomes noticeably hard to spin by hand with the power off, because the motor is now generating into a dead short and braking itself. That hand test is the most useful thing you can do in a driveway: lift the wheel, disconnect the controller from the battery, and spin. Free-spinning is good; a strong drag means a short in the controller or the phase wiring.

Hall sensor faults

Hall sensors sit inside the motor and fail from heat, vibration, or water reaching their connector. The signature is stuttering, cogging, or violent jerking at low speed, with the bike sometimes running acceptably once it is moving. The cheap thing to check first is the connector, not the motor: hall sensor cables use small multi-pin plugs that corrode, and a green or crusty pin is the whole diagnosis. Some controllers have a self-learn mode that retimes the phases, and some will run sensorless as a fallback with poor starts.

Water ingress

Controllers themselves are usually potted or well sealed. The weak points are the connectors and the point where the cable loom enters the case. Water wicks along the inside of a cable by capillary action, so a soaking that entered at a handlebar connector can end up in the controller weeks later. Symptoms are intermittent: cutting out over bumps, error codes that come and go, a display that resets. Repeated pressure washing at the motor and the controller area is the most common self-inflicted cause.

Reading the evidence

• Nothing at all, no display: start at the battery, its fuse, and the main power connector, not the controller.
• Display lights, motor dead: controller, phase wiring, or a safety input like a brake cutoff switch stuck closed. Brake levers with motor cutoffs are a surprisingly common culprit and cost almost nothing to replace.
• Works on throttle, not on pedal assist: cadence or torque sensor, or its cable, not the controller.
• Works on assist, not on throttle: throttle unit or its connector.
• Jerky at low speed only: hall sensors or their connector.
• Cuts out under load, recovers when you ease off: low voltage cutoff, which points at the pack rather than the controller. Our battery replacement guide covers how to tell a tired pack from a broken one.

Upgrading a controller, honestly

Fitting a higher-current controller is the cheapest way to make an ebike accelerate harder, which is why it is a popular modification and why forums are full of people who have done it. The mechanics are simple: more current means more phase current, which means proportionally more torque. It is also the modification with the most ways to go wrong.

Take those in order of seriousness. The legal point is the one that changes your status on the road: in the United States a low-speed electric bicycle is defined in part by a motor of less than 750W, and a bike that draws well past that is arguably no longer a bicycle, with knock-on effects for path access, insurance, and liability after a collision. We go through the tests in what makes an electric bike street legal.

The battery is the second problem. Every pack has a battery management system with a maximum continuous discharge current, and that limit is set by the cells and by the BMS transistors. Ask a 30A-rated BMS to deliver 40A and it will either shut off under load, which feels like the bike cutting out at exactly the moment you wanted power, or it will pass the current and run hot. Neither is good.

The third is thermal. Resistive heating in the motor windings rises with the square of current, so going from 20A to 30A does not increase heating by half, it roughly doubles it. A hub motor that was comfortable on a long climb at the original setting can reach temperatures that demagnetise magnets or melt winding insulation. If you do this, fit a motor with a temperature sensor, or be disciplined about long climbs.

A note on voltage. A controller sold as 48V has input capacitors and transistors chosen for that range. A 52V pack reaches 58.8V fully charged, which some 48V controllers tolerate and others do not. If you are moving up in voltage, buy a controller rated for the pack, not one you hope will cope.

What to check before you buy a bike or a replacement

For a complete bike, three questions get you most of the way:

1. What is the controller current limit? Multiply by pack voltage for the honest peak power figure. This is the number that determines whether the bike feels quick.
2. Sine wave or square wave? Brands that fit sine wave controllers usually say so, because it costs them money. Silence generally means square wave.
3. Is the controller a standard part or a proprietary one? A controller integrated into the motor casing or paired to a closed system is tidier and often better made, but when it fails in five years your only source is the brand. A separate controller in the downtube with common connectors can be replaced from many suppliers.

For a replacement controller, match the voltage range, match or slightly exceed the original current rating, confirm it supports the sensor type your motor uses, and check the connector standard before you order. Phase and hall connectors are not universal, and rewiring them by trial and error risks a short across the pack.

One last piece of advice that will save more money than any of this: when an ebike loses power, check the fuse, the main battery connector, and the brake cutoff switches before assuming the controller is dead. Those three account for a large share of the bikes that arrive at shops as controller failures.

If you are working out what to buy rather than what to fix, the controller is one of three numbers that matter together. Read battery voltage and capacity for the supply side, motor wattage for what the ratings really mean, and ebike frames and materials for what all that torque is being fed into.

Frequently asked questions