Proximity sensors: inductive or capacitive? Using the MFIA to measure the sensitivity and response time of both sensing regimes

Introduction

A proximity sensor is a sensor that detects nearby objects without physical contact [1]. They are commonly used, in applications ranging from production lines to modern mobile devices, mainly thanks to their non-contact configuration and the accompanying long lifetime. The working principle is simple: generate a probe signal and detect the change of inductance or capacitance (to a broader extent, also current) in the sensor. In the case of the inductive sensor,  a primary magnetic field is generated and the conductive test object induces an eddy current which is subsequently detected using a secondary coil [2]. A capacitive sensor, on the other hand, functions by detecting the dielectric change of the surroundings when the test object moves closer. A secondary current detection scheme is not required in capacitive sensors, nor does the test object need to be conductive.

In this blog post, we will demonstrate these two sensing mechanisms using an MFIA impedance analyzer (or MFLI lock-in amplifier with MF-IA option) with a simple 10 mH unshielded inductor (TDK B82442T1106K050) as our sensor. Whether it operates in the inductive or capacitive sensing regime depends on the chosen frequency. We will see how flexible and versatile the MFIA is for the characterization of sensors and transducers.

 

Experimental setup

Before starting, we soldered the 10 mH inductor on a 4-terminal PCB carrier as our proximity sensor and plugged it into an MFITF fixture connected to the MFIA front panel, as shown in Fig. 1. Note that the sensor is tested in 4-terminal configuration by measuring current and voltage simultaneously, but we will look at the current and the reactance (inductance or capacitance). Such impedance parameters are readily displayed in LabOne, the Zurich Instruments control software, thanks to the real-time impedance calculation with the internal calibration of the MFIA.

Should you be interested in the characterization of the eddy current, a secondary coil can be used as a current collector [2]. Additional control of the primary current can also be established using the MF-PID option if needed [3].

Fig. 1. Photo showing the 10 mH inductor mounted on a PCB carrier and inserted into the MFITF connected to the MFIA. This is our proximity sensor for this demonstration. A red LED connected to auxiliary output 1 is lit as the warning indicator when a metal tip (probe) moves closer to the sensor but not yet in contact.

 

Explore the detailed circuit elements

We begin the measurement with a frequency sweep of 10000 data points from 1 Hz to 5 MHz using the LabOne Sweeper module on this ‘sensor’. The high density of data points is important as we want to have both the low-frequency plateau and high-frequency resonance displayed smoothly in the same plot. The result is shown in Fig. 2. We see a resistance plateau of around 100 Ohm at frequencies below 100 Hz. With an increasing frequency, the sensor becomes more and more inductive until reaching the resonance peak at 763.7 kHz.  After that, the sensor soon becomes capacitive, with -87 deg at 5 MHz.

Note that all circuit models in LabOne compose of two elements, which allow a fast real-time calculation of equivalent impedance parameters at a given frequency. However, a simple two-element model working well within a frequency range may not fit another range. Specifically on this sensor, when the frequency is higher than 763.7 kHz, the Rs+Ls model is NOT valid anymore, hence Ls is not shown in Fig. 2. Instead, one should toggle the circuit model into Rp||Cp to simply obtain Cp, or extract detailed circuit elements in post-processing.

To have a better overview, we export the sweeper data from LabOne in .csv format (or .m format for Matlab, .txt format for Z-view, and .hdf5 for Python for instance) for post-processing. We consider the following circuit model, shown in Fig. 3b, with circuit elements (Rs, Ls, Rp, and Cp) being frequency independent. As with many real-world inductors, this 10 mH sensor also deviates strongly from a purely inductive behavior at high frequencies, the main reason being the geometrical capacitance of winding, Cp, becomes dominating. To represent the core loss on the inductor, we also need a parallel (leakage) resistance, Rp [4]. And finally, in addition to the most obvious coil inductance, Ls, the wire also possesses a series resistance, Rs, which is important at low frequencies.

Regardless of the fitting algorithm, a suitable boundary condition is important whenever a resonance peak is present, otherwise, the fitting may not converge. A quick way to start is to refer to the LabOne two-element circuit model as an approximation. For instance, we can imagine Rs (fitted as 101.4 Ohm) close to the low-frequency impedance (measured as 100.5 Ohm at 100 Hz in Rs+Ls model), Ls (fitted as 9.3 mH) close to the middle-frequency series inductance (measured as 9.3 mH at 100 kHz in Rs+Ls model), and Cp (fitted as 4.7 pF) close to the high-frequency parallel capacitance (5.1 pF at 5 MHz in Rp||Cp model). The fitted curve is then plotted together with the measured one in the same Nyquist plot in Fig. 3a. And we see an almost perfect match between these two.

Fig. 2. A screenshot of LabOne Sweeper module showing the measured impedance (cyan trace), phase (orange trace), and current (green trace) in Bode plot. Note that the derived series inductance Ls from Rs+Ls model can also be added but it will become negative and meaningless at frequencies above resonance.

Fig. 3 (a) Nyquist plot of the sensor (cyan open dots) with fitting (orange line) using third-party software. (b) Equivalent circuit model of the sensor.

 

Inductive sensing below resonance

As seen above, at 100 kHz, the sensor shows a phase of 87.3 deg, behaving close to a perfect inductor. We fix the measurement at this frequency and look at the primary current amplitude (Demod 1 Sample R) and series inductance from Rs+Ls model in the LabOne Plotter module, as shown in Fig. 4. Here the test object has to be conductive (e.g. metal) to trigger a change in the primary current. Non-conductive objects (e.g. plastic, air bubbles) do not allow a high enough (secondary) eddy current hence almost no mutual inductance (compared with tens of uH for a metal object) will be added to the primary side. Note that the primary current may decrease (shown in Fig. 4) or even increase, depending on the magnetic permeability of the test object. Resolving these tiny current changes (e.g. 0.1 nA out of 100 uA) requires an instrument that has a large dynamic range (120 dB) and a flexible low-pass filtering (noise cancellation) such as the MFIA.

An inductive proximity sensor is often used in non-destructive testing (NDT) to selectively find defect locations in the DUT and has been mentioned in an earlier blog [2]. However, the response time tends to be longer as it is limited by the bandwidth of the low-pass filter.

Fig. 4 A screenshot of LabOne Plotter module showing the change of current (purple trace), inductance (red trace), and digital trigger output (brown trace) when a sharp metal tip moves closer to and away from the sensor.

 

Capacitive sensing above resonance

At 5 MHz, the sensor shows -87.2 deg and behaves almost like a pure capacitor. This happens as the parallel capacitance Cp becomes low in impedance and shunts the current along the capacitive path rather than the inductive path which exhibits a higher impedance. When a test object moves closer, regardless of its conductivity, its dielectric permittivity is different from the air where the unguarded electrical field is present outside the ‘capacitive’ sensor. Therefore both conductive or non-conductive objects are capable to introduce a change in the current as well as the capacitance.

Since the test signal is in a higher frequency, it is possible to increase the low-pass filter bandwidth from 100 Hz (used in the above inductive sensing example) to 10 kHz or even higher (up to 200 kHz on the MFIA). As seen in Fig. 5, the resulting signal will be noisier, but the response time also becomes much shorter. This suggests that a capacitive proximity sensor is preferable when speed is critical.

Fig. 5. A screenshot of LabOne Plotter module showing the change of current (purple trace), capacitance (red trace), and digital trigger output (brown trace) when a non-conductive object moves closer to and away from the sensor.

 

Triggering an external circuit

In a previous blog [3], we saw that the Data Acquisition (DAQ) Module in LabOne can be used to log events of current change. But to develop a complete proximity sensor, the sensor ‘head’ needs to be connected to a comparator circuit with digital or analog output. We will show here that this can also be easily done with the MFIA using its threshold unit. For example, in Fig. 6, we choose a proper threshold value of 36.3 uA, mimicking the situation where the sensor is brought closer to a metal tip. What is important here is the enable time. This means that the threshold will only enter the ‘1’ state when the condition is met continuously for 1 ms (arbitrarily chosen), which can be useful to prevent being falsely triggered by oscillations and unwanted spikes in the measured signal (as known as a glitch filter). When the metal tip moves away, we do not necessarily need this waiting time, and can simply set the disable time to 0. The resulting digital trigger can be visualized in Fig. 4 and Fig. 5. A similar way is to use the scope trigger and look at the raw signal before demodulation, which has already been published in an earlier blog on pass/fail testing [5].

For outputting an analog voltage, we can use either of the 4 auxiliary outputs on the MFIA, and set a proportional gain. This can be used to turn on LED indicators (Fig. 1) or drive motors to detach the sensor from the test object.

Fig. 6. A screenshot of DIO, Threshold Unit, and Auxilary Output settings in LabOne.

 

Conclusion

In this blog post, we have demonstrated mechanisms of both inductive and capacitive proximity sensors with an MFIA and a 10 mH inductor as our sensor. The high accuracy, the auto-ranging capability, and the real-time impedance display make the MFIA perfectly suitable for in-depth circuit analysis. With the characteristic frequency determined, the MFIA can sense a small change of current from the sensor and output digital or analog trigger signals based upon the change.

To briefly summarize the two mechanisms, inductive sensing is conductivity selective but slower due to a low working frequency. In contrast, capacitive sensing can be faster but is not conductivity selective and may suffer from a higher, unfiltered noise.

If you have any questions or suggestions, please get in touch.

 

References:

  1. Proximity sensor, Wikipedia entry. https://en.wikipedia.org/wiki/Proximity_sensor
  2. Eddy Current Testing with MFLI Lock-In Amplifier, Zurich Instruments blog. https://blogs.zhinst.com/marco/2015/08/05/eddy-current-testing-with-mfli-lock-in-amplifier/
  3. Sensor Characterization and Control, Zurich Instruments blog. https://blogs.zhinst.com/kivanc/2020/05/25/sensor-characterization-and-control/
  4. Analytical and experimental determination of the parasitic parameters in high-frequency inductor, BULLETIN OF THE POLISH ACADEMY OF SCIENCES, DOI: 10.1515/bpasts-2017-0013.
  5. Pass/Fail Tests for Failure Analysis, Zurich Instruments blog. https://blogs.zhinst.com/romain/pass-fail/

1 comment to Proximity sensors: inductive or capacitive? Using the MFIA to measure the sensitivity and response time of both sensing regimes