Posted on 05 July 2019

Precision Current Sensor with Exceptional Large Bandwidth

Free Bodo's Power Magazines!




With the information about the temperature, digital signal correction is improved remarkably

In the field of power efficient solutions, the precise knowledge about the flow of electrical power over a broad spectrum is indispensable. Without this knowledge, no decision could be taken about how to save electrical power without loss of functionality and comfort. A high quality low-cost solution for current measurement is therefore an important step towards electrical power efficiency and could play a significant role in photovoltaic, green building and green hospital applications.

By R. Weiss and K. Behringer, Senior Research Engineers Siemens AG, C. Bluemm and R. Weigel, Friedrich Alexander University/ Lehrstuhl für Technische Elektronik


Siemens engineers have developed a galvanic isolated current sensor that operates on the principle of the GMR-effect (giant magneto resistive effect). The sensor is designed for good accuracy (error less than 0.3 %) in the industrial temperature range. It is capable of measuring DC, AC and pulse currents up to 20 MHz. The whole measurement system can handle three different currents at the same time. It consists of three GMR-sensors, several ΔΣ modulators and one FPGA for digital signal conditioning. Due to the exceptional large bandwidth from DC to 20 MHz, it could also enable new applications in the field of renewable energies. Beside the accuracy, the low power consumption compared to a conventional Hall based closedloop current transducer makes the system very promising for applications where power efficiency is a key advantage to the customer.

The GMR-effect

For discovering the GMR-effect, the French researcher Albert Fert and the German researcher Peter Gruenberg were awarded with the Nobel Prize in 2007. The first commercial application of the GMR technology as read head sensor for hard drives addresses the multi billion dollar market for information technology and pushed the GMR technology to an extraordinary level during the last decade. Due to the improvements in GMR-material quality and the cost competitive availability of large scale production processes, the new technology is now ready to enable new applications in the field of industry and automotive sensing applications.

Giant magnetoresistance (GMR) is observed in structures which exist of alternating ferromagnetic and non-magnetic layers with a thickness of few nanometers. The effect allows to alter the electrical resistance of the entire layer structure through a change of the mutual magnetization orientation of the magnetic layers. If the layers are magnetized in opposite directions, the resistance is significantly higher than with magnetization in the same direction. The reason for that is the spin dependent scattering of electrons in ferromagnetic materials which is a well known quantum mechanical phenomenon.

Figure 1 gives an example of typical resistance alteration ΔR/R in a Fe/Cr structure, dependent on the mutual magnetization orientation of the Fe layers. To point out the enormous influence of variations in the layer thickness, different Cr layers from 9 to 18 A thickness are opposed.

Magneto resistance of a Fe-Cr layer structure with different Cr thickness at 4,2K

The distinctive relationship between the magnetization orientation of the magnetic layers and the electrical resistance is perfectly applicable for measuring external magnetic fields. The GMR sensor presented here is made up as a so-called spin valve system. It consists basically of two ferromagnetic layers, separated by a diamagnetic metallic layer. As shown in Figure 2, an additional anti-ferromagnetic layer is attached to one ferromagnetic layer (the reference layer) in order to pin its magnetization direction. In spite of that, the other ferromagnetic layer features a relatively free moving magnetization direction (the free layer). The spin valve system shows an approximately linear resistance change of typically 10%.

A Wheatstone bridge of four GMR elements, stimulated by a current-carrying U-shaped conductor

Measuring magnetic fields is of course not the only way to exploit the GMR-effect for sensor technology. The application spectrum covers any physical variable that can be technically associated with a magnetic field, such as position, velocity or current.

GMR-based current sensor

The set-up of a spin valve GMR element which is used for current measurements is shown in Figure 3. According to Ampère's circuital law, the primary current causes a magnetic field H around the conductor, it passes through. This conductor is attached to the GMR element with an isolation layer inbetween, made of PCB material. A change of the current alters the magnetic field, which in turn results in a rotating magnetization of the free layer of the spin valve GMR element.

Basic layer structure of a spin valve system (Current-inplane setting)

The GMR sensor presented here features four GMR elements. According to Figure 3, they are arranged in a Wheatstone configuration to enable differential measuring. This helps to compensate foreign magnetic fields with identical impact on all GMR elements and even temperature influence to some extent. As output, a single voltage signal UGMR is provided, which equals ideally zero when no current flows.

Adapting the GMR sensor for high accuracy current measurements

Compensating perturbations is not the only benefit of the Wheatstone configuration. The voltage drop over a simple shunt, to be connected in series with the Wheatstone bridge, allows to measure the temperature of the GMR elements. With the information about the temperature, digital signal correction is improved remarkably, as will be shown later.

A shunt is needed, since the total bridge resistance RWS, and therewith the feeding current IWS depends on temperature changes. On the other hand, RWS and IWS are independent of the stimulating magnetic field, due to the special arrangement of the GMR elements.

Consequently, two independent output signals are provided by the wheatstone/shunt combination: UGMR to represent the external magnetic field and UShunt to represent the temperature. As soon as these signals are digitalized, they can be conditioned with the FPGA. In the sensor system presented here, the ΔΣ modulators AD7401 from Analog Devices are used for digitalization.

Figure 4 depicts the circuit of a GMR sensor system, as used for a one-phase current. The capacitors and resistances, which were not mentioned before, work as follows:

R1/R2/C1: used for bandlimitation
R3/C2 : low-pass filter with 100 kHz cut-off frequency
C3, C4, C5, C6 : decoupling capacitors for power supply
R5/C7 , R6/C8, R4 : used for impedance matching, against overshoots

Actually, the original Siemens sensor system can handle three-phase instead of one-phase currents. Thus, all circuit elements of Figure 4 are tripled, except for the FPGA.

GMR sensor system for a one-phase current

Digital signal conditioning

Typical uncorrected output values of the Siemens current sensor are presented in Figure 5. Therefore, DC currents in the range of ±100 A at temperatures from -10°C to 70°C are applied. Compared to other current sensing elements, the sensor characteristics are already considerable. Yet, there is still room for digital improvement. For the three most critical sources of inaccuracy, namely the GMR sensor`s nonlinearity, the offset and the influence of temperature, solutions are provided. A less severe, but still noticeable problem is a certain hysteresis behavior.

Typical input-output interrelation of a Siemens GMR current sensor

For improving the accuracy of a measured value e.g. a current value, a digitally memorized model of the inversed sensor transfer characteristics can be exploited. This is a well-known technique, traditionally based on polynomials of an arbitrary degree for the sensor model. However, polynomials lack of accuracy, due to inevitable ripples, which increase with higher polynomial orders. Furthermore, it is very difficult to implement involved polynomials with low-cost hardware like DSPs or FPGAs.

For these reasons, a new innovative digital modeling method is implemented, which comes with the huge advantage of maximum flexibility in combination with minimal computational effort and memory usage: Derived from the computer science subfields of computeraided design and computer graphics, the numerical technique of Bspline interpolation can be perfectly adapted to correct sensor nonlinearities, offset and temperature.

The nature of interpolation is to construct new data points within the range of a discrete set of known data points. This set must be determined before measuring through a calibration process. Therefore, the response signals of the GMR sensor plus the response signals of the additional shunt to well-defined input stimuli are taken and analyzed. It is essential to stimulate just within those temperature and current ranges, which are important for later measurements. In order to use this data for B-spline interpolation, some post-processing is necessary, resulting in a grid of few, uniform spread data points. As soon as these points are hard-coded as constants in a look-up table, the FPGA is ready for measurements:

Based on the actual output variables of the GMR sensor and the shunt, the position within the grid of constants is localized and an according intermediate value (the measurement value) is gained through B-spline interpolation.

Since the memorized sensor model is of static nature, dynamic effects like hysteresis are not correctable with the B-spline approach. Nevertheless, an upgrade with additional signal correction techniques is conceivable, if further measurement improvement is desired.

Examples of measurements

Just as Figure 5 depicts typical raw sensor values, Figure 6 gives an impression of the same data after digital correction. Now, the axis of ordinate no longer displays voltage values, but the interpolated current.

Linearised measurement (with B-spline interpolation)

For a quantitative evaluation, Figure 7 gives the corresponding absolute measurement error. All measurement operations were run from the minimal to the maximal current value and back. In doing so, the hysteresis behavior of the GMR sensor shows its ultimate occurrence, clearly noticeable in Figure 7. This means that the resulting error plots are the worst-case scenario. In real measurements, however, less hysteresis impact can be expected and therewith higher accuracy.

Absolute error of the linearised measurement



VN:F [1.9.17_1161]
Rating: 6.0/6 (1 vote cast)
Precision Current Sensor with Exceptional Large Bandwidth, 6.0 out of 6 based on 1 rating

This post was written by:

- who has written 791 posts on PowerGuru - Power Electronics Information Portal.

Contact the author

Leave a Response

You must be logged in to post a comment.