# Field-Oriented Control in Plain Language

> What field-oriented control does, the Clarke and Park idea with a worked example, why torque is current, and when simpler drive is enough.

Canonical: https://roboticsengineering.in/field-oriented-control-plain-language/ | Published: 2026-11-05 | Updated: 2026-11-05 | Author: Jignesh S Panchal

**Field-oriented control (FOC) drives a brushless motor with smooth sinusoidal currents and controls the torque-producing part of the current directly. It does this by converting the three phase currents into a frame that rotates with the rotor, where torque becomes a simple, almost DC quantity that ordinary PI loops can control.**

## Key takeaways

- FOC splits the stator current into a d-axis part, which lines up with the rotor magnet, and a q-axis part, which produces torque.
- For a typical permanent-magnet motor, torque is proportional to the q-axis current, so controlling that current controls torque.
- FOC needs the rotor angle and at least two phase current measurements, plus a fast control loop.
- For simple speed control with no demand for smooth low-speed torque, six-step drive can be enough and is simpler.

## Why Sinusoidal Drive Matters

Sinusoidal drive means feeding the three motor phases with smooth sine-shaped currents, rather than switching them on and off in six blocks. Six-step control, covered in [BLDC motors and ESCs in robotics](https://roboticsengineering.in/bldc-motor-esc-robotics/), produces torque ripple, which shows up as vibration and noise at low speed. FOC shapes the current so the stator field stays at a steady angle to the rotor, which gives smoother torque, and so it is common in gimbals, servo drives and robot joints. Choosing the right motor and drive is covered in [motor selection for robots](https://roboticsengineering.in/motor-selection-for-robots/).

## dq Frame Idea

The dq frame is a coordinate system that rotates at the same speed as the rotor, with the d axis along the rotor magnet and the q axis 90 electrical degrees ahead. Seen from this frame, a spinning current vector looks like two fixed numbers, Id and Iq. Getting there takes two steps.

The Clarke transform converts three phase currents into two stationary components. With ia + ib + ic = 0, it is `I_alpha = ia` and `I_beta = (ia + 2 x ib) / sqrt(3)`. The Park transform then rotates those by the rotor electrical angle theta: `Id = I_alpha x cos(theta) + I_beta x sin(theta)` and `Iq = -I_alpha x sin(theta) + I_beta x cos(theta)`.

Worked example, with assumed numbers. Let the stator current vector be 2 A in size at 120 degrees, with the rotor at theta = 30 degrees. The phase currents are ia = 2 x cos(120) = -1 A, ib = 2 x cos(0) = 2 A and ic = 2 x cos(-120) = -1 A, which sum to zero as required.

1. Clarke: I_alpha = -1 A. I_beta = (-1 + 2 x 2) / 1.7321 = 3 / 1.7321 = 1.7321 A.
2. Park with cos(30) = 0.8660 and sin(30) = 0.5: Id = (-1 x 0.8660) + (1.7321 x 0.5) = -0.8660 + 0.8660 = 0 A.
3. Iq = -(-1 x 0.5) + (1.7321 x 0.8660) = 0.5 + 1.5 = 2.0 A.

The current vector is 90 degrees ahead of the rotor (120 minus 30), so all of it is torque-producing current: Id is zero and Iq equals the full 2 A.

## Torque Is Current

For a typical surface-mounted permanent-magnet motor, torque is `T = 1.5 x p x lambda x Iq`, where p is the number of pole pairs and lambda is the permanent-magnet flux linkage. The product 1.5 x p x lambda is the motor’s torque constant, Kt, in newton metres per amp. With assumed values p = 7 and lambda = 0.005 Wb, Kt = 1.5 x 7 x 0.005 = 0.0525 N m per A. At Iq = 2 A, the torque is 0.0525 x 2 = 0.105 N m. Real motors publish Kt in the datasheet, so use that value and not these assumed numbers.

This is why FOC is attractive for robots: you ask for torque by setting a current, the same as a servo’s torque mode. The Id target is usually zero for these motors, because Id does not produce useful torque and only heats the windings. For the link between current, torque and power, you can cross-check with the [DC motor power and efficiency calculator](https://roboticsengineering.in/dc-motor-power-efficiency-calculator/).

## Current Loops

A current loop is a feedback controller that compares the measured current with the target and adjusts the voltage to close the gap. FOC runs two PI loops: one holds Id at zero and the other holds Iq at the torque you want. Their outputs, Vd and Vq, go through the inverse Park transform and a PWM generator to drive the three half-bridges. These loops commonly run at the PWM rate or a fraction of it, often several kilohertz to tens of kilohertz, but check your controller’s documentation. Speed and position loops sit outside as slower layers, and the outer loop tuning is covered in [PID tuning for robot motion](https://roboticsengineering.in/pid-tuning-robot-motion/). A simple speed loop with an encoder is shown in the [PID motor speed control project](https://roboticsengineering.in/projects/pid-motor-speed-control-encoder-arduino/).

## Sensor Needs

FOC needs to know the rotor electrical angle and the phase currents. Angle can come from a magnetic or optical encoder, from Hall sensors with interpolation, or from a sensorless observer based on back-EMF, with the usual weakness at very low speed. Current needs shunt resistors or current sensors on at least two phases, because the third follows from ia + ib + ic = 0. A three-phase gate driver such as the [DRV8313 three-phase driver](https://roboticsengineering.in/components/drv8313-three-phase-driver/) provides the power stage that these loops control; read its datasheet for current and thermal limits.

## Where FOC Is Overkill

FOC costs more in code, compute and sensing. If a wheel only needs to spin at a set speed with no demand for smooth torque near zero, six-step control with Hall sensors may do the job. A drone propeller or a fan usually runs well on a sensorless ESC. Many hobby brushed-motor robots, like the [ESP32 robot car](https://roboticsengineering.in/projects/esp32-robot-car/), do not need either. FOC pays off for slow, smooth, precise joints, gimbals and anywhere you want torque control.

## How to check this yourself

1. Take a motor datasheet and note the pole pair count and the torque constant Kt.
2. Pick a phase-current vector, compute ia, ib and ic, and confirm they sum to zero.
3. Apply the Clarke and Park equations by hand or in a spreadsheet, and confirm Iq equals the vector length when the vector is 90 degrees ahead of the rotor.
4. Multiply Iq by Kt and compare the torque with the motor’s rated torque in the datasheet.

## Common mistakes

- Using the wrong electrical angle, for example mechanical angle where pole pairs should multiply it.
- Measuring current at a bad moment in the PWM cycle, which fills the current signal with switching noise.
- Assuming FOC needs an encoder, when sensorless FOC exists but has low-speed limits.
- Applying FOC to a motor whose datasheet gives no usable constants.

## Frequently asked questions

### What is field-oriented control in simple words?

It is a way of driving a brushless motor by controlling the torque-producing current directly. The controller mathematically rotates the three phase currents into a frame that turns with the rotor, so torque becomes a simple number it can control with ordinary feedback loops, giving smooth, quiet motion.

### Is FOC better than six-step commutation?

FOC usually gives smoother torque and quieter low-speed motion, but it needs more computing, a rotor angle and current sensing. Six-step is simpler and good enough for fans, drone propellers and many wheels. Choose by what your robot needs, not by which is newer.

### Can an Arduino or ESP32 run FOC?

Open-source FOC libraries exist for both, and ESP32 has more computing headroom than a basic Arduino Uno. Whether it is enough depends on loop rate, motor count and sensor type, so check the library documentation. Test with a current-limited supply first.

## Related reading

- [Learn hub](https://roboticsengineering.in/learn/): more motor and control explainers.

## About the author and how this is kept accurate

Written, checked and maintained by [Jignesh Panchal](https://www.linkedin.com/in/jignesh-s-panchal/) of [Jaadu Digital](https://jaadudigital.com/). Figures in worked examples come from the stated assumptions, not from a bench test of our own. Read the [editorial standard](https://roboticsengineering.in/editorial-standard/), follow the author on [LinkedIn](https://www.linkedin.com/in/jignesh-s-panchal/), or [send a correction](https://roboticsengineering.in/contact/). Last reviewed: November 2026.
