Electronic Energy Meter Based on a Tunnel Magnetoresistive Effect (TMR) Current Sensor

In the present work, the design and microfabrication of a tunneling magnetoresistance (TMR) electrical current sensor is presented. After its physical and electrical characterization, a wattmeter is developed to determine the active power delivered to a load from the AC 50/60 Hz mains line. Experimental results are shown up to 1000 W of power load. A relative uncertainty of less than 1.5% with resistive load and less than 1% with capacitive load was obtained. The described application is an example of how TMR sensing technology can play a relevant role in the management and control of electrical energy.


Introduction
Tunneling magnetoresistance (TMR) sensing is the fourth generation of magnetoresistive magnetic sensing technology after galvanomagnetic, anisotropic (AMR), and giant (GMR) technologies. The galvanomagnetic effect was achieved in semiconductors having a thickness greater than their length [1]. Better sensitivities were achieved using semiconductor materials such as InSb with Te as the preferred dopant atom [2]. AMR sensing technology was based on materials derived from binary and tertiary alloys of Fe, Ni, and Co, such as permalloy, deposited over a Si substrate. With these structures, AMR-based sensors achieved 2-4% of magnetoresistance variation, requiring a specific Barber pole geometry to have linearity and periodic flipping pulses to stabilize the internal initial magnetization over time [3]. With GMR sensing technology, 70% of magnetoresistance relative variation was reached [4], although in the most applications the variation was between 4% and 25%, depending on the microfabrication technology. The GMR basic structure is composed of two layers of ferromagnetic material (Fe, Co, and Ni alloys) separated by a non-magnetic material (like Cu). In practice, in order to have high magnetoresistance variation, new magnetic multilayers structures were fabricated by repetition of the basic structure (unpinned sandwiches, antiferromagnetic multilayers, and spin valves) [5][6][7][8][9][10]. TMR sensing technology is based on the magnetoresistive effect that occurs in a magnetic tunnel junction (MTJ) structure where the modulation of its resistivity is due to the spin-dependent tunneling effect (SDT). Once this effect was observed and technologically developed at room temperature, the design of new TMR-based sensors was made possible [11][12][13]. An MTJ element is composed of two ferromagnetic layers separated by an isolation layer. Usually, reference layer orientations. The full bridge configuration was obtained so that the test conducting wire (U-shaped current line) mounted below the chip was aligned with each MTJ array (Figure 1b), to enable opposing magnetic fields in each pair of MTJ elements and opposite sensitivities to the magnetic field [29].
Materials 2017, 10, 1134 3 of 11 to enable opposing magnetic fields in each pair of MTJ elements and opposite sensitivities to the magnetic field [29].

Physical and Electrical Characterization
The MTJ sensors were characterized individually, using the test structures indicated in Figure 1. Figure 3a shows the resistance-magnetic field characteristics of each MTJ sensor (360 MTJ elements in series), with magnetoresistance (TMR) values of 86%. The small coercivity and shift are the signature of the weekly pinned biasing at the free layer [31]. Figure 3b shows the bridge sensor output voltage under a current sweep between the interval −10 A to +10 A circulating through the U-shaped copper trace. A high electrical sensitivity of 15.5 mV/A was obtained, making the subsequent electronic signal amplification easier to achieve.

Physical and Electrical Characterization
The MTJ sensors were characterized individually, using the test structures indicated in Figure 1. Figure 3a shows the resistance-magnetic field characteristics of each MTJ sensor (360 MTJ elements in series), with magnetoresistance (TMR) values of 86%. The small coercivity and shift are the signature of the weekly pinned biasing at the free layer [31]. Figure 3b shows the bridge sensor output voltage under a current sweep between the interval −10 A to +10 A circulating through the U-shaped copper trace. A high electrical sensitivity of 15.5 mV/A was obtained, making the subsequent electronic signal amplification easier to achieve.

Practical Implementation
The sensor was designed to measure up to 30 A of electrical current in a switched-mode power converter. A printed circuit board was constructed to mount the sensor, the conditioning electronics, and the copper conductor. The conductor was placed under the printed circuit board, separated an appropriate distance from the TMR elements in order to obtain the optimum linear range within their R-H characteristic ( Figure 4).

Practical Implementation
The sensor was designed to measure up to 30 A of electrical current in a switched-mode power converter. A printed circuit board was constructed to mount the sensor, the conditioning electronics, and the copper conductor. The conductor was placed under the printed circuit board, separated an appropriate distance from the TMR elements in order to obtain the optimum linear range within their R-H characteristic ( Figure 4).

Physical and Electrical Characterization
The MTJ sensors were characterized individually, using the test structures indicated in Figure 1. Figure 3a shows the resistance-magnetic field characteristics of each MTJ sensor (360 MTJ elements in series), with magnetoresistance (TMR) values of 86%. The small coercivity and shift are the signature of the weekly pinned biasing at the free layer [31]. Figure 3b shows the bridge sensor output voltage under a current sweep between the interval −10 A to +10 A circulating through the U-shaped copper trace. A high electrical sensitivity of 15.5 mV/A was obtained, making the subsequent electronic signal amplification easier to achieve.

Practical Implementation
The sensor was designed to measure up to 30 A of electrical current in a switched-mode power converter. A printed circuit board was constructed to mount the sensor, the conditioning electronics, and the copper conductor. The conductor was placed under the printed circuit board, separated an appropriate distance from the TMR elements in order to obtain the optimum linear range within their R-H characteristic ( Figure 4).    Figure 5 shows a general block diagram of the proposed electronic energy meter. It comprises the TMR sensor (described previously), the signal conditioning electronics, and the metering section.  Figure 5 shows a general block diagram of the proposed electronic energy meter. It comprises the TMR sensor (described previously), the signal conditioning electronics, and the metering section.

Electronic Processing Unit (Metering Section)
There exists a wide range of electronic processors made by different manufacturers. They process electrical energy delivered to a load from the 50/60 Hz AC mains and offer at their output a variety of associated parameters like active and reactive powers or rms (root-mean-square) voltage and current [33,34]. In this work, the ADE7755 energy processor from Analog Devices (Norwood, MA, USA) was used [35]. This unit was designed to measure the active power supplied to an electric load by a monophasic AC line. Figure 6 shows its internal block diagram. It can be distinguished by three fundamental sections: analog signal processing and acquisition, digital signal processing, and conversion and power supply unit. Analog signal processing and acquisition: It is composed of two programmable gain amplifiers (with values selected between ×1, ×2, ×8, and ×16). Each amplifier is used as the front-end of the voltage and current channels. The inputs of both amplifiers must be a low-level voltage amplitude and the line voltage and load current must be previously attenuated and conditioned. The output of each amplifier is acquired by an analog-to-digital (A/D) converter based on a sigma-delta conversion supplying 16-bit resolution.
Digital signal processing and conversion: It is the main block of the energy processor. As the central part, there is a multiplier that processes the digitalized voltage and current signals. The current channel has an additional phase correction block that compensates for the phase when the load has a strong inductive component, and it has a digital high-pass filter to reject possible offsets. The output of the multiplier is proportional to the instantaneous power delivered to the load and,

Electronic Processing Unit (Metering Section)
There exists a wide range of electronic processors made by different manufacturers. They process electrical energy delivered to a load from the 50/60 Hz AC mains and offer at their output a variety of associated parameters like active and reactive powers or rms (root-mean-square) voltage and current [33,34]. In this work, the ADE7755 energy processor from Analog Devices (Norwood, MA, USA) was used [35]. This unit was designed to measure the active power supplied to an electric load by a monophasic AC line. Figure 6 shows its internal block diagram. It can be distinguished by three fundamental sections: analog signal processing and acquisition, digital signal processing, and conversion and power supply unit.  Figure 5 shows a general block diagram of the proposed electronic energy meter. It comprises the TMR sensor (described previously), the signal conditioning electronics, and the metering section.

Electronic Processing Unit (Metering Section)
There exists a wide range of electronic processors made by different manufacturers. They process electrical energy delivered to a load from the 50/60 Hz AC mains and offer at their output a variety of associated parameters like active and reactive powers or rms (root-mean-square) voltage and current [33,34]. In this work, the ADE7755 energy processor from Analog Devices (Norwood, MA, USA) was used [35]. This unit was designed to measure the active power supplied to an electric load by a monophasic AC line. Figure 6 shows its internal block diagram. It can be distinguished by three fundamental sections: analog signal processing and acquisition, digital signal processing, and conversion and power supply unit. Analog signal processing and acquisition: It is composed of two programmable gain amplifiers (with values selected between ×1, ×2, ×8, and ×16). Each amplifier is used as the front-end of the voltage and current channels. The inputs of both amplifiers must be a low-level voltage amplitude and the line voltage and load current must be previously attenuated and conditioned. The output of each amplifier is acquired by an analog-to-digital (A/D) converter based on a sigma-delta conversion supplying 16-bit resolution.
Digital signal processing and conversion: It is the main block of the energy processor. As the central part, there is a multiplier that processes the digitalized voltage and current signals. The current channel has an additional phase correction block that compensates for the phase when the load has a strong inductive component, and it has a digital high-pass filter to reject possible offsets. The output of the multiplier is proportional to the instantaneous power delivered to the load and, Analog signal processing and acquisition: It is composed of two programmable gain amplifiers (with values selected between ×1, ×2, ×8, and ×16). Each amplifier is used as the front-end of the voltage and current channels. The inputs of both amplifiers must be a low-level voltage amplitude and the line voltage and load current must be previously attenuated and conditioned. The output of each amplifier is acquired by an analog-to-digital (A/D) converter based on a sigma-delta conversion supplying 16-bit resolution.
Digital signal processing and conversion: It is the main block of the energy processor. As the central part, there is a multiplier that processes the digitalized voltage and current signals. The current channel has an additional phase correction block that compensates for the phase when the load has a strong inductive component, and it has a digital high-pass filter to reject possible offsets. The output of the multiplier is proportional to the instantaneous power delivered to the load and, once it has been low-pass filtered, a signal proportional to the active power is obtained. Finally, the digital-to-frequency converter outputs a signal with a frequency that indicates the active power and is prepared to interface easily with electromechanical or digital counters (energy registers).
Power supply unit: It is the part of the processor that provides the energy needed by the analog and digital units. Specifically, there is a 2.5 V precision reference voltage used by the A/D converters and, if needed, by the external conditioning electronics.

Signal Conditioning Electronics
Because the TMR current sensor is in a Wheatstone bridge topology, its output is a differential voltage signal v o (t) that requires differential-to-unipolar electronic conditioning. Figure 7 shows this electronic circuit in detail. A fully differential high-pass filter is connected at the current sensor output to reject DC offsets caused by the existent mismatching between the value of the four MR1 to MR4 active elements at zero current. Following the filtration, an instrumentation amplifier (INA118 model from Texas Instruments) is used to provide the necessary gain and to convert the differential signal in an unipolar V o output voltage easily acquired by the energy processor. At its output, a simple 3.3 µF capacitor is connected to reject the possible offset due to instrumentation amplifier. I supp is a DC constant that is the current source implemented to bias the Wheatstone bridge. once it has been low-pass filtered, a signal proportional to the active power is obtained. Finally, the digital-to-frequency converter outputs a signal with a frequency that indicates the active power and is prepared to interface easily with electromechanical or digital counters (energy registers). Power supply unit: It is the part of the processor that provides the energy needed by the analog and digital units. Specifically, there is a 2.5 V precision reference voltage used by the A/D converters and, if needed, by the external conditioning electronics.

Signal Conditioning Electronics
Because the TMR current sensor is in a Wheatstone bridge topology, its output is a differential voltage signal vo(t) that requires differential-to-unipolar electronic conditioning. Figure 7 shows this electronic circuit in detail. A fully differential high-pass filter is connected at the current sensor output to reject DC offsets caused by the existent mismatching between the value of the four MR1 to MR4 active elements at zero current. Following the filtration, an instrumentation amplifier (INA118 model from Texas Instruments) is used to provide the necessary gain and to convert the differential signal in an unipolar Vo output voltage easily acquired by the energy processor. At its output, a simple 3.3 µF capacitor is connected to reject the possible offset due to instrumentation amplifier. Isupp is a DC constant that is the current source implemented to bias the Wheatstone bridge. To adapt the voltage and current channels to the energy processor inputs, it is necessary to attenuate the output voltage of both channels. Both 250 mVrms and 15 mVrms voltages are present, respectively, at the current and voltage channel inputs of the ADE7755, using the values of the designed attenuators shown in Figure 8. The 33 nF capacitor prevents antialiasing for the time necessary for the processor to achieve the A/D conversion.  To adapt the voltage and current channels to the energy processor inputs, it is necessary to attenuate the output voltage of both channels. Both 250 mV rms and 15 mV rms voltages are present, respectively, at the current and voltage channel inputs of the ADE7755, using the values of the designed attenuators shown in Figure 8. The 33 nF capacitor prevents antialiasing for the time necessary for the processor to achieve the A/D conversion. once it has been low-pass filtered, a signal proportional to the active power is obtained. Finally, the digital-to-frequency converter outputs a signal with a frequency that indicates the active power and is prepared to interface easily with electromechanical or digital counters (energy registers). Power supply unit: It is the part of the processor that provides the energy needed by the analog and digital units. Specifically, there is a 2.5 V precision reference voltage used by the A/D converters and, if needed, by the external conditioning electronics.

Signal Conditioning Electronics
Because the TMR current sensor is in a Wheatstone bridge topology, its output is a differential voltage signal vo(t) that requires differential-to-unipolar electronic conditioning. Figure 7 shows this electronic circuit in detail. A fully differential high-pass filter is connected at the current sensor output to reject DC offsets caused by the existent mismatching between the value of the four MR1 to MR4 active elements at zero current. Following the filtration, an instrumentation amplifier (INA118 model from Texas Instruments) is used to provide the necessary gain and to convert the differential signal in an unipolar Vo output voltage easily acquired by the energy processor. At its output, a simple 3.3 µF capacitor is connected to reject the possible offset due to instrumentation amplifier. Isupp is a DC constant that is the current source implemented to bias the Wheatstone bridge. To adapt the voltage and current channels to the energy processor inputs, it is necessary to attenuate the output voltage of both channels. Both 250 mVrms and 15 mVrms voltages are present, respectively, at the current and voltage channel inputs of the ADE7755, using the values of the designed attenuators shown in Figure 8. The 33 nF capacitor prevents antialiasing for the time necessary for the processor to achieve the A/D conversion.

Experimental Results
After an adjustment process, the meter was submitted to load variations corresponding to different power consumptions. In the proposed wattmeter, a resistive variable load was built to change the power delivered to the load from 0 to 1000 W. Figure 9 schematically shows the variable load used to test the TMR wattmeter.

Experimental Results
After an adjustment process, the meter was submitted to load variations corresponding to different power consumptions. In the proposed wattmeter, a resistive variable load was built to change the power delivered to the load from 0 to 1000 W. Figure 9 schematically shows the variable load used to test the TMR wattmeter. The experimental procedure was to take the number of output pulses provided by the ADE7755 at its CF (calibration frequency) output corresponding to 200, 400, 600, 800, and 1000 W power consumptions of pure resistive load during a 24-h period for each power value. A reference wattmeter (2551 from Xitron) was used to select the different powers delivered to the load. The experimental active power obtained from the CF output was compared with the readings of the reference wattmeter; Figure 10a shows that comparison and Figure 10b depicts the experimental relationship between the active power and the number of counts at the CF output. A conversion factor of 11.47 W/number of counts was obtained with a very low value of residual power (0.7 W). Figure 10c shows that the maximum relative uncertainty was less than 1.5% in the specified active power interval. To check the wattmeter behavior with no pure resistive loads, a 30 µF capacitor was connected in series ( Figure 9) to shift the power factor to a cos ϕ = 0.75 value. Table 1 shows the relative uncertainty obtained with this test.
(a) (b) Figure 9. Variable load designed to test the TMR wattmeter.
The experimental procedure was to take the number of output pulses provided by the ADE7755 at its CF (calibration frequency) output corresponding to 200, 400, 600, 800, and 1000 W power consumptions of pure resistive load during a 24-h period for each power value. A reference wattmeter (2551 from Xitron) was used to select the different powers delivered to the load. The experimental active power obtained from the CF output was compared with the readings of the reference wattmeter; Figure 10a shows that comparison and Figure 10b depicts the experimental relationship between the active power and the number of counts at the CF output. A conversion factor of 11.47 W/number of counts was obtained with a very low value of residual power (0.7 W). Figure 10c shows that the maximum relative uncertainty was less than 1.5% in the specified active power interval. To check the wattmeter behavior with no pure resistive loads, a 30 µF capacitor was connected in series ( Figure 9) to shift the power factor to a cos ϕ = 0.75 value. Table 1 shows the relative uncertainty obtained with this test.

Experimental Results
After an adjustment process, the meter was submitted to load variations corresponding to different power consumptions. In the proposed wattmeter, a resistive variable load was built to change the power delivered to the load from 0 to 1000 W. Figure 9 schematically shows the variable load used to test the TMR wattmeter. The experimental procedure was to take the number of output pulses provided by the ADE7755 at its CF (calibration frequency) output corresponding to 200, 400, 600, 800, and 1000 W power consumptions of pure resistive load during a 24-h period for each power value. A reference wattmeter (2551 from Xitron) was used to select the different powers delivered to the load. The experimental active power obtained from the CF output was compared with the readings of the reference wattmeter; Figure 10a shows that comparison and Figure 10b depicts the experimental relationship between the active power and the number of counts at the CF output. A conversion factor of 11.47 W/number of counts was obtained with a very low value of residual power (0.7 W). Figure 10c shows that the maximum relative uncertainty was less than 1.5% in the specified active power interval. To check the wattmeter behavior with no pure resistive loads, a 30 µF capacitor was connected in series ( Figure 9) to shift the power factor to a cos ϕ = 0.75 value. Table 1 shows the relative uncertainty obtained with this test.

Pref (W) Pexp (W) εr(%)
R, 137.5 Ω 587.6 584.9 0.5 C, 30 µF cos ϕ = 0.75 As the final experimental results, Figure 11 shows the electronic schematic used to build the energy meter. That schematic was built in a practical prototype over a printed-circuit-board (PCB); Figure 12a shows such a practical implementation, and Figure 12b shows actual waveforms taken over the prototype.   As the final experimental results, Figure 11 shows the electronic schematic used to build the energy meter. That schematic was built in a practical prototype over a printed-circuit-board (PCB); Figure 12a shows such a practical implementation, and Figure 12b shows actual waveforms taken over the prototype.

Pref (W) Pexp (W) εr(%)
R, 137.5 Ω 587.6 584.9 0.5 C, 30 µF cos ϕ = 0.75 As the final experimental results, Figure 11 shows the electronic schematic used to build the energy meter. That schematic was built in a practical prototype over a printed-circuit-board (PCB); Figure 12a shows such a practical implementation, and Figure 12b shows actual waveforms taken over the prototype. Figure 11. Whole schematic circuit that constitutes the TMR-based energy meter. Figure 11. Whole schematic circuit that constitutes the TMR-based energy meter.

Conclusions
An energy meter is presented based on a TMR current sensor with good linearity and relative uncertainty. The proposed work is applied to measure the active power delivered to a load from the AC 50/60 Hz main line. To measure current, conventional energy meters are based on the resistive-shunt method or Hall current sensors. The shunt method does not provide electrical isolation and has some self-heating properties due to the power dissipation on it. The Hall method provides electrical isolation from the mains but the use of ferrite cores causes bulky and heavy meters. The TMR method provides both electrical isolation and low volume, showing the feasibility to design new TMR-based instruments. The wattmeter presented is a prototype based on a microfabricated TMR sensor combined with a PCB, but an actual proof of concept can be built by integrating the sensor and the electronics in the same package or PCB. The presented work is an example of how spintronic-based materials and electronics engineering can produce new electronic measurement instruments. Both disciplines converge in a new one that could be named spintronics engineering.

Conclusions
An energy meter is presented based on a TMR current sensor with good linearity and relative uncertainty. The proposed work is applied to measure the active power delivered to a load from the AC 50/60 Hz main line. To measure current, conventional energy meters are based on the resistive-shunt method or Hall current sensors. The shunt method does not provide electrical isolation and has some self-heating properties due to the power dissipation on it. The Hall method provides electrical isolation from the mains but the use of ferrite cores causes bulky and heavy meters. The TMR method provides both electrical isolation and low volume, showing the feasibility to design new TMR-based instruments. The wattmeter presented is a prototype based on a microfabricated TMR sensor combined with a PCB, but an actual proof of concept can be built by integrating the sensor and the electronics in the same package or PCB. The presented work is an example of how spintronic-based materials and electronics engineering can produce new electronic measurement instruments. Both disciplines converge in a new one that could be named spintronics engineering.