How To Solve Angle Drift of The Magnetic Code Disk in Robot Magnetic Encoder Sensors?

In high-precision motion control applications such as robot joints, servo motors, and gimbals, magnetic encoders are increasingly becoming the core position-feedback devices, thanks to their non-contact operation, immunity to contaminants, and compact size. However, many engineers encounter a frustrating problem in practice – angle drift: the angle reading slowly “wanders off” even when the motor is stationary, or accuracy degrades significantly with temperature changes. Where does magnetic code disk angle drift come from, and how can we fix it?

1. What Is Angle Drift and Where Does It Come From?

A magnetic encoder detects the rotating magnetic field of a permanent magnet attached to the motor shaft, outputs sine/cosine signals with a 90° phase difference, and then calculates the angle via an arctangent function. In theory, this is a perfect measurement chain. In reality, however, various non-ideal factors cause the angle reading to deviate from the true value – this is “angle drift”.

In summary, angle drift originates from three main aspects:

1.1 Mechanical Installation Errors

This is the most common culprit. Radial offset between the magnet and the sensor’s sensing centre (eccentricity), the angle between the rotation axis and the chip normal (tilt), and excessive or fluctuating air gap all directly distort the magnetic field signal. For example, an eccentricity of 0.3 mm can introduce an angle error of ±2.1°. In addition, non-uniform magnetisation or irregular magnet shape can cause the sensed field to deviate from the ideal sinusoidal waveform.

1.2 Ambient Temperature Variations

The characteristics of Hall sensors and signal-processing circuits change with temperature. As temperature rises, the magnet’s field strength decreases, the chip’s offset voltage drifts, and the ADC reference voltage fluctuates – these factors combine to produce significant temperature drift in the angle reading. Actual measurements show that with different magnet qualities, the zero-point temperature drift can worsen from +0.02 °/℃ to +0.15 °/℃.

1.3 External Interference and Dynamic Factors

A transverse interfering magnetic field shifts the centre of the Lissajous figure, causing errors at almost all angles except 0° and 180°. Moreover, dynamic response delays during high-speed motor operation, and signal noise from power-supply ripple, can make the angle reading jitter and become unstable.

2. Solutions: A Combination of Hardware, Calibration, and Algorithms

There is no single “magic bullet” for solving angle drift. A coordinated approach is needed at three levels: mechanical installation, chip calibration, and algorithmic compensation.

Level 1: Solid Hardware Foundation – Stop Errors at the Source

Precise installation is the first line of defence. Magnetic encoders are quite sensitive to mounting tolerances. Generally, the air gap should be kept between 1.0 mm and 3.0 mm (2.0 mm is optimal), coaxiality eccentricity ≤ 0.05 mm, and tilt angle ≤ 3°. During installation, use a dial gauge or laser alignment tool for calibration. Also, keep the encoder away from strong magnetic interference sources such as motor windings and power transistors; if necessary, add a Permalloy magnetic shield.

Magnet selection is also critical. Choose a single-pair axial magnetisation circular magnet with a surface field strength between 300 mT and 500 mT. Low-quality magnets not only have uneven fields but also poor temperature characteristics, which directly worsen drift.

Power supply and PCB design must follow good practice. Power ripple should be controlled within 10 mV, and SPI signal lines should be short, equal-length, and routed cleanly. Do not place copper pours or traces under the chip to reduce parasitic capacitance interference with the magnetic sensing elements.

Level 2: Make the Most of Chip Calibration – Realise the Factory Precision in Practice

Modern magnetic encoder chips typically incorporate multiple calibration stages, which are the most direct way to handle angle drift.

Factory calibration (chip level): During wafer testing, the chip manufacturer already performs DC offset compensation, amplitude imbalance correction, and quadrature error correction, with parameters stored in the on-chip EEPROM. This is the “foundation” of accuracy.

User self-calibration (system level): After installation, rotate the motor one full revolution at a constant speed. The built-in DSP collects angle data over the full turn and automatically identifies and compensates for installation eccentricity, air-gap variations, and other errors. Many chips support a one-click trigger without requiring an external host, which is very production-friendly.

Dynamic temperature drift compensation: The chip monitors temperature in real time. When the temperature deviates from the calibrated value, it automatically applies a temperature-compensation coefficient to correct the angle. For example, the Naxin Micro MT6835 uses a three-stage calibration system to keep full-temperature-range drift within ±0.08°.

Level 3: Algorithmic Compensation – Further Improve Accuracy

For applications demanding even higher precision, additional algorithmic compensation is needed.

Adaptive online calibration: Traditional off-line calibration fixes the compensation parameters and cannot adapt to long-term drift due to temperature changes or magnet aging. An adaptive algorithm based on Recursive Least Squares (RLS) can update compensation parameters in real time, keeping angle error within ±0.05° across the full temperature range (–40 ℃ to +125 ℃).

Kalman filtering for noise reduction: For high-frequency noise that causes angle jitter, a Kalman filter effectively suppresses observation noise and improves the stability of the angle output.

Neural network compensation: A neural network model can be built to fit the temperature-drift characteristics; after a limited number of training iterations, it achieves highly accurate adaptive temperature-drift correction.

3. Real-World Performance – How Good Can It Get?

With the combined approach of “hardware installation + chip calibration + algorithmic compensation”, modern magnetic encoders can achieve remarkably high accuracy:

• The MT6816 achieves system-level integrated angle error within ±0.03°.

• The MT6835, after calibration, improves its Integral Nonlinearity (INL) from ±1° to < ±0.07°, with full-temperature drift < ±0.08°.

• Some high-end solutions can even optimise INL to ±0.03°, meeting the ±0.05° level accuracy required for gimbals and servos.

4. Summary and Outlook

In essence, magnetic code disk angle drift is the result of superimposed disturbances from mechanical, thermal, electromagnetic, and other factors on the magnetic field signal. There is no single cure-all; the solution should follow the principle of “hardware as the base, calibration as the mainstay, and algorithms as the finisher”:

1. Installation accuracy is the foundation – air gap, coaxiality, and magnet selection cannot be overlooked.

2. Chip calibration is the primary tool – make full use of factory calibration and user self-calibration.

3. Algorithmic compensation is the advanced step – for demanding scenarios, introduce adaptive filtering and online calibration.

As chip integration continues to increase and AI algorithms are introduced, future magnetic encoders will evolve towards “self-sensing, self-calibrating, and self-compensating” capabilities. At that point, angle drift will no longer be a headache for engineers, and the precision and reliability of robot joints will reach new heights.