Modelling and simulating magnetoresistive sensors in positioning and control systems

Sensor technologies gather essential data for electronic systems performing functions such as monitoring and control in diverse applications. Sensors based on the magnetoresistive effect, which was discovered by Lord Kelvin in 1857, enable accurate measurement of magnetic fields and can be used for contactless detection of position or proximity.

Anisotropic Magnetoresistive (AMR) sensors, produced using thin-film technology, achieve low jitter and high sensitivity, combined with high accuracy and low overall system cost. They are suitable for use in equipment such as industrial drives and controls, automotive engine and transmission systems, vehicle traction and stability controls, electronic compasses for navigation systems, vehicle detection for toll-road applications, and precision positioning systems.

In-depth analysis using simulation is essential before an overall system is implemented. As all components influence the way in which a system responds, a great deal of importance lies in the simulation of an overall system, especially during the planning stages and capture of system requirements. Here we look at the modelling and simulation of a system to measure rotational speed.

 

Fig. 1: AMR sensor systems consist of two packages.

Fig. 2: Anisotropic magnetoresistive effect.

Fig. 3: Configuration of AMR elements on die.

 

Signal detection

Modern sensor systems consist essentially of two components - an elementary sensor and a signal-processing ASIC (Figure 1). The AMR effect occurs in ferrous materials like permalloy. An alloy of 81% nickel and 19% iron, permalloy has been used as a sensing material since the early part of the 20th century. Figure 2 represents a thin film of permalloy with current flowing through it. When an external magnetic field is applied to permalloy, the change in its resistance is proportional to the square of the sine of the angle α. Ferromagnetic materials, such as permalloy, have magnetisation, which is a vector quantity defined at each point in the material. It is the rotation of this magnetisation vector from the direction of the current flow due to an external magnetic field which produces the change in resistance. The magnitude of the change in resistance depends on the properties of the permalloy. Permalloy’s properties cause it to change resistance by 2% to 3% in the presence of a magnetic field. All of NXP Semiconductors' sensor systems use permalloy for the elementary sensor.

The setup to determine speed consists of two components: an encoder wheel and the sensor system. The encoder wheel can be either active or passive. An active wheel is magnetised and an MR sensor detects the change between north and south poles. In the case of passive wheels, the magnetisation is replaced by a tooth structure. Passive encoder wheels usually have very small tolerances. When the sensor symmetrically faces a tooth or a gap between two teeth of a passive wheel, it produces no deflection in the magnetisation vector of AMR elements. Neglecting external noise fields, the output signal achieves a value of zero. If the sensor head is in front of a tooth edge, the magnetic input signal becomes extremal. The result, as a function of the type of change tooth/gap or gap/tooth, to a good approximation, is the minimum or maximum of a sinusoidal magnetic input signal.

 

Signal processing

To determine the speed, the magnetic input signal is encoded into an electrical pulse sequence and transmitted by a 7/14mA protocol. A comparator can be used to generate the pulse sequence. Hysteresis is usually added to the comparator circuit to eliminate the influence of lower levels of noise. Marked fluctuations in the gap between the sensor head and encoder wheel lead to fluctuations in the amplitude of the magnetic input signal.

Magnetic offset also endangers the working of the system. A noise field can boost or reduce the actual measured signal to such an extent that only one or neither of the thresholds of the Schmitt trigger is exceeded or under-run. Very high speeds of passive wheels can create eddy currents in the wheel that in turn produce a magnetic noise field. The resulting offset is a risk to operating reliability.

To eliminate the influence of this noise on the output signal, a signalprocessing ASIC is housed in another package. The latter also holds a line driver to provide the supply voltage for signal processing and a highvoltage interface (Figure 1). The central elements to eliminate malfunctions are an adjustable amplifier, offset cancellation, and smart comparator. Depending on the distance between the sensor and encoder wheel, the adjustable amplifier can match the signal level. For offset cancellation there is a control system that eliminates offsets and also maintains the 0Hz capability of the system. This helps to detect a standing encoder wheel. The thresholds of the smart comparator are variable and can be set so that hysteresis is between 20% and 45% of the signal amplitude. This ensures sufficient noise suppression.

The system described above was developed and validated with simulation support. The following outlines the development and also shows how the model can be used to assist design-in.

 

 

System simulation

To develop the sensor system, it is first necessary to gain an overview of the expected magnetic input signals. The starting point for this is to examine the specifications of the encoder wheel, the permanent magnet on the sensor head, and the expected dimensions and tolerances. Finite-Element Method (FEM) simulations using ANSYS software were conducted to determine the magnetic fields. The positions of the encoder wheel, sensor element, and the magnet determine the field strength. Figure 4 shows the magnetic input signal on the sensor bridge as a function of distance. In addition to the distance, deviations in position also cause the amplitude to reduce. Based on FEM simulation, mechanical specifications can thus be converted into expected magnetic variables. FEM simulation is also suitable for estimating their influence (Figure 5), and the results can be translated directly into admissible position tolerances.

 

 

Determining the magnetic fields is followed by simulation of the sensor system. The change in resistance of the AMR elements is a direct result of the AMR effect. The results of field simulation lead to a change in resistance that represents the signal delivered to the signal-processing stage. Simulink was used to model the analogue front-end. Each Simulink block corresponds to a component of the analogue signal processing, such as an amplifier or filter. A design is created in Hardware Description Language (HDL) to simulate the digitally implemented functionality, and is ready in final form after completing product development.

Simulation of the overall system is consequently a co-simulation of the behavioural model of the analogue components by Simulink and the HDL design by ModelSim . A Simulink reference model can gradually be replaced in a co-simulation by ready-implemented Verilog code in ModelSim. In this way the HDL design can be validated piece by piece. The process can be continued until the entire digital part is implemented in Verilog, while the analoguesystem parts remain as a Simulink model. This tool combination has also proved useful for evaluation of the IC.

 

 

Results

Modelling of this sort makes it possible to analyse the behaviour of the system as a function of the input signals. The first diagram in Figure 6 shows a magnetic input signal produced by varying the distance between the sensor and encoder wheel. This signal is the result of finite element simulation, which can be converted by the AMR effect into an electrical output signal of the bridge. The middle diagram is the result of analogue signal processing. The bottom diagram shows the output signal. Figure 7 shows examples of signal patterns in ModelSim.

Control of the simulations by MATLAB and combination with additional simulators create extra options. Simulations can be automated and there is the possibility of using extensive algorithms for signal evaluation in MATLAB. Through FEM simulators such as ANSYS, it is possible to extend the simulated system components as far as the MR sensor head and the associated encoder.