Introduction of the problem
In some applications, linear motor drives are not used for precise position control, but rather for adjusting pressure, flow, and force in industrial processes, such as compressors. Position is only required at the lower-level control level, while the upper-level closed-loop control variable is force. In this specific application, driving a linear motor without a linear position sensor is almost impossible, but using standard sensors with an accuracy of 10µm to 50µm is very expensive. In this case, a linear sensor with an accuracy of 200µm to 1mm is perfectly adequate. A position sensor with similar accuracy can be designed using two low-cost Hall effect sensors, a mounting interface, and a small PCB board. To control the linear motor drive, a sinusoidal magnetic field must be provided outside the motor coil by a permanent magnet. In most cases, a sinusoidal magnetic field can be detected by placing a small position sensor within a certain distance. The sensor can be mounted near the magnet and coil and does not require the space occupied by the shaft of a traditional linear sensor.
Sensor Design
When measuring a magnetic field using two Hall effect sensors, the sensors must be installed within the effective distance of the magnet (as shown in Figure 1). This yields two sine waves with a 90° phase difference, as shown in Figure 2. These signals can be used to control the direction of motion.
The measurement utilizes an easy-to-install integrated sensor (see Figure 3). The Hall sensor output, including the current source, is amplified within the package. Only the two measurement signals from channels a and b require interface design with the converter input, which can be achieved using two operational amplifiers (a and b) with different outputs. This is done to reduce electromagnetic interfaces. Additionally, two op-amps are used to correct sensor offset. A constant voltage provided by another op-amp serves as the common-mode input for op-amps a and b, and this voltage is half the value of the sensor power supply. The complete circuit diagram is shown in Figure 3.
Sensor improvements
The sensor's performance can be tested by measuring the position of the magnetic field relative to the armature, as shown in Figure 4. Signal #1 is the measurement result of a sinusoidal signal (#2), and the difference between the two is shown by line #3 in the figure.
To verify the application of the sensor in the closed-loop control of the linear motor, the small sensor shown on the right side of Figure 10 was installed on the motor.
The obtained position signal was compared with a reference signal obtained from a photoelectric linear sensor with an accuracy of 10 μm, as shown in Figure 5. Lines 2# and 1# represent the two measured signals, and line 3# represents the error between them. Throughout the measurement process, the absolute error of the position was very small.
Sensor optimization
As shown in Figure 2, two large noise interferences will appear in the sinusoidal signal within one rotation of the electric field (Figure 6).
This is caused by a large current flowing through the coil closest to the sensor. To obtain a better position signal, the signal can be filtered to reduce interference.
If the noise embedded in the sinusoidal signal can be minimized, the sensor's performance can be optimized. This can be achieved in three steps. First, a copper-plated PCB enclosure can shield the entire sensor. Then, a twisted-pair cable is used instead of a regular cable to connect the sensor and the converter.
Figure 7 shows the sinusoidal signal measured by the shielded sensor. Clearly, signal interference is reduced. Compared to Figure 6, the amplitude of the superimposed noise is significantly reduced, but the signal is still not accurate enough for a vibration-free linear motor.
Because analog circuitry is a weak point in the interface, it was reduced in the second step of the optimization process and replaced with programmable devices. The analog devices mentioned above for sensor calibration and offset correction can be omitted. The block diagram of the new sensor is shown in Figure 8.
The sensor's advantage lies in its internal digital signal processor, which improves accuracy and reduces the impact of analog zero drift, temperature shifts, and mechanical stress on mathematical calculations. Additionally, the sensor features EEPROM memory and a serial communication interface for programming. This allows calibration without altering the resistors on the PCB. For already installed sensors, parameters can be modified without disassembly.
The final step in optimizing the new sensor was reducing the PCB size. A 4-layer PCB was used for this purpose. While slightly more expensive than a 2-layer PCB, a 4-layer PCB offers significant advantages in EMC compatibility, eliminating the need for an expensive sensor shielding box. Therefore, overall, a 4-layer PCB is more economical.
Figure 9 shows the functional diagram of the new sensor. Three of the original five operational amplifiers can be omitted without changing the performance. The eliminated components further reduce the PCB size to only 40% of the original. Figure 10 compares the two sensors.
Measurement
The measurement results of the optimized sensor are shown in Figure 11. Curve #1 represents the position, and curve #2 represents the reference position measured by the high-precision optical coupler position sensor; the two are indistinguishable. Unlike Figure 5, the position error, after being magnified, is expressed as a percentage, as shown by curve #3 in Figure 11. The position error of the optimized sensor is reduced to 1.6%, which is sufficient for application in the aforementioned scenarios.
Conclusion
The measurement results from the custom position sensor described in the paper demonstrate that it is possible to use a low-cost position sensor with sufficient accuracy in the underlying position control loop. This sensor can be installed inside the permanent magnet linear motor without occupying additional space. Compared to commercial optocoupler position sensors, this sensor is significantly less expensive.
