Experimental Demonstration of Stray-Field Immunity beyond 5 mT for an Automotive-Grade Rotary Position Sensor ^†

Abstract

This paper experimentally demonstrates the stray-field robustness capability of a novel Hall-based rotary position sensor concept (Huber, S., et al, 2018). The sensor targets safety-related automotive applications, for example powertrain and power steering. In these applications, the safety requirement specifies a maximum stray-field induced error of 0.4°. Therefore, the robustness in corner cases needs to be assessed. We demonstrate the stray-field immunity in multiple corner cases for temperature from −40 °C up to 160 °C and over lifetime. The impact of a uniform 5 mT stray field over all conditions (3σ) is shown to be less than 0.25°. The fully-integrated automotive-qualified sensor is implemented in a 0.18-μm CMOS technology, and achieves 0.7° of angular accuracy.

1. Introduction

Magnetic stray fields are naturally present in modern vehicles, especially electrified cars. Large magnetic stray fields in the order of 1 mT have been measured in electrified cars [1]. The most critical location is naturally close to the high-voltage (HV) cables between the battery and the electrical motor. Legacy magnetic sensors, which are ubiquitous in modern cars [2], are corrupted by such a high stray field. Consider for example a legacy magnetic sensor operating with a signal amplitude of 40 mT near a wire carrying 400 A (a typical specification). The stray field drops slowly with the distance as 1/r (see Figure 1a). At a distance r = 30 cm, the field (B~0.3 mT) is still large enough to introduce a significant angular error (~0.4°). Practically, this creates an exclusion zone (see Figure 1b) of about 30 cm away from any HV cables for legacy Hall sensors severely limiting their practicality. Inductive sensors, on the other hand, are naturally insensitive to low-frequency stray fields. However, they require more space due to the overhead of the discrete coils. To address the stray-field challenge in Hall-based sensors, we developed a new gradiometric Hall sensor concept described in our companion paper [3].

2. Materials and Methods

The gradiometric concept is implemented in a CMOS chip [4] with an appropriate integrated magneto-concentrator (IMC). The sensor is sensitive to magnetic field gradients in the plane of the chip. In production tests, the sensors are trimmed for optimum stray-field rejection by adjusting the sensitivities of the individual Hall elements. The stray-field rejection is thus limited by residual mismatch. To assess the robustness of the concept, we selected corner parts from a production lot with the highest residual mismatch. The sensors were subjected to a uniform stray field of 5 mT generated by a Helmotz coil (corresponding to a magnetic field strength H of 4 kA/m). The signal of approximately 21 mT/mm was generated by a 4-pole magnet at nominal airgap. The stray-field induced error is calculated as the difference of the output angle reported by sensor in presence and absence of the stray field.

$${\mathsf{\theta }}_{\mathrm{err},\mathrm{stray}}={\mathsf{\theta }}_{\mathrm{out}}\left({\mathrm{B}}_{\mathrm{stray}}=5\mathrm{mT}\right)-{\mathsf{\theta }}_{\mathrm{out}}\left({\mathrm{B}}_{\mathrm{stray}}=0\right)$$

(1)

For evaluating the sensor response to a nearby current, the sensors were mounted on a rotating stage near a busbar carrying 400 A. This configuration generates an inhomogeneous stray field (see the stray-field gradient dB/dr in Figure 1a).

3. Results

3.1. Impact of Uniform Stray Fields

Figure 2a shows that the accuracy, limited by thermal drift, is 0.7° in the absence of stray field. Figure 2b shows the impact, for about 600 chips, of a 5 mT uniform stray field as per automotive standard (ISO 11452-8). The stray-field induced error is less than 0.25° for the whole population. To investigate the residual stray-field parasitic sensitivity, we characterized (Figure 3) a corner device from the extreme of the production distribution. According to the production test, this device exhibited the worst rejection of out-of-plane field. A stray field of 25 mT, 5× larger than the typical specification to magnify the effect, was rotated out-of-plane with respect to the sensor. As expected, the maximum error was measured when the stray field was almost orthogonal to the sensor surface. The in-plane case was more favourable (data not shown). For typical devices, the stray-field induced errors were below the noise level.

3.2. Impact of Current-Carrying Wire

Additionally, we investigated the more practical case of non-uniform stray-fields, with a field gradient similar to the useful signal. The sensors were mounted on a multi-axis rotating stage above a busbar carrying 400-A (see Figure 4a). The orientation shown in the picture is the worst case, in which the sensor is affected by the stray-field gradient in the sensor plane. The signal (21 mT/mm) is generated by a 4-pole magnet mounted on the backside of the PCB. The stray-field induced error was shown to be less than 0.15° at 30 mm of distance. The inset in Figure 4a shows the histogram of the measured errors of a corner device over roughly 100 different orientations while maintaining a 30-mm center-to-center distance. We derived an equation (Figure 4b) for the total stray-field induced error, with error terms arising from the average stray field and its gradient. With just 1-cm of clearance with respect to the wire, the error remains below 0.4° even in the worst case.

4. Conclusions

In summary, we demonstrated that the angular position signal remains robust even at just 1-cm of clearance with respect to a wire carrying 400 A of current. This is an improvement by more than an order of magnitude with respect to traditional Hall sensors. The fully-integrated automotive-qualified sensor is implemented in a 0.18-μm CMOS technology, and achieves 0.7° of accuracy. This extends the state of the art for Hall-based sensors. Moreover, it is competitive with inductive solutions, and even outperforms them in terms of compactness.